Merge branch 'master' of ctdb into 'master' of samba

Signed-off-by: Stefan Metzmacher <metze@samba.org>
Signed-off-by: Michael Adam <obnox@samba.org>

277 files changed, 65250 insertions, 0 deletions

diff --git a/ctdb/lib/popt/CHANGES b/ctdb/lib/popt/CHANGES
new file mode 100644
index 0000000000..db16a5fdd0
--- /dev/null
+++ b/ctdb/lib/popt/CHANGES
@@ -0,0 +1,46 @@
+1.5 -> 1.6
+	- add ability to perform callbacks for every, not just first, match.
+
+1.3 -> 1.5
+	- heavy dose of const's
+	- poptParseArgvString() now NULL terminates the list
+
+1.2.3 -> 1.3
+	- added support for single -
+	- misc bug fixes
+	- portability improvements
+
+1.2.2 -> 1.2.3
+	- fixed memset() in help message generation (Dale Hawkins)
+	- added extern "C" stuff to popt.h for C++ compilers (Dale Hawkins)
+	- const'ified poptParseArgvString (Jeff Garzik)
+
+1.2.1 -> 1.2.2
+	- fixed bug in chaind alias happens which seems to have only
+	  affected --triggers in rpm
+	- added POPT_ARG_VAL
+	- popt.3 installed by default
+
+1.2 -> 1.2.1
+	- added POPT_ARG_INTL_DOMAIN (Elliot Lee)
+	- updated Makefile's to be more GNUish (Elliot Lee)
+
+1.1 -> 1.2
+	- added popt.3 man page (Robert Lynch)
+	- don't use mmap anymore (its lack of portability isn't worth the
+	  trouble)
+	- added test script
+	- added support for exec
+	- removed support for *_POPT_ALIASES env variable -- it was a bad
+	  idea
+	- reorganized into multiple source files
+	- added automatic help generation, POPT_AUTOHELP
+	- added table callbacks
+	- added table inclusion
+	- updated man page for new features
+	- added test scripts
+
+1.0 -> 1.1
+	- moved to autoconf (Fred Fish)
+	- added STRERROR replacement (Norbert Warmuth)
+	- added const keywords (Bruce Perens)
diff --git a/ctdb/lib/popt/COPYING b/ctdb/lib/popt/COPYING
new file mode 100644
index 0000000000..b4c7ca876c
--- /dev/null
+++ b/ctdb/lib/popt/COPYING
@@ -0,0 +1,22 @@
+Copyright (c) 1998  Red Hat Software
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE
+X CONSORTIUM BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+Except as contained in this notice, the name of the X Consortium shall not be
+used in advertising or otherwise to promote the sale, use or other dealings
+in this Software without prior written authorization from the X Consortium.
diff --git a/ctdb/lib/popt/README b/ctdb/lib/popt/README
new file mode 100644
index 0000000000..0b5205bfdd
--- /dev/null
+++ b/ctdb/lib/popt/README
@@ -0,0 +1,18 @@
+This is the popt command line option parsing library. While it is similiar
+to getopt(3), it contains a number of enhancements, including:
+
+	1) popt is fully reentrant
+	2) popt can parse arbitrary argv[] style arrays while 
+	   getopt(2) makes this quite difficult
+	3) popt allows users to alias command line arguments
+	4) popt provides convience functions for parsing strings
+	   into argv[] style arrays
+
+popt is used by rpm, the Red Hat install program, and many other Red Hat
+utilities, all of which provide excellent examples of how to use popt. 
+Complete documentation on popt is available in popt.ps (included in this
+tarball), which is excerpted with permission from the book "Linux
+Application Development" by Michael K. Johnson and Erik Troan (availble
+from Addison Wesley in May, 1998).
+
+Comments on popt should be addressed to ewt@redhat.com.
diff --git a/ctdb/lib/popt/findme.c b/ctdb/lib/popt/findme.c
new file mode 100644
index 0000000000..a950e50018
--- /dev/null
+++ b/ctdb/lib/popt/findme.c
@@ -0,0 +1,50 @@
+/** \ingroup popt
+ * \file popt/findme.c
+ */
+
+/* (C) 1998-2002 Red Hat, Inc. -- Licensing details are in the COPYING
+   file accompanying popt source distributions, available from 
+   ftp://ftp.rpm.org/pub/rpm/dist. */
+
+#include "system.h"
+#include "findme.h"
+
+const char * findProgramPath(const char * argv0) {
+    char * path = getenv("PATH");
+    char * pathbuf;
+    char * start, * chptr;
+    char * buf;
+
+    if (argv0 == NULL) return NULL;	/* XXX can't happen */
+    /* If there is a / in the argv[0], it has to be an absolute path */
+    if (strchr(argv0, '/'))
+	return xstrdup(argv0);
+
+    if (path == NULL) return NULL;
+
+    start = pathbuf = alloca(strlen(path) + 1);
+    buf = malloc(strlen(path) + strlen(argv0) + sizeof("/"));
+    if (buf == NULL) return NULL;	/* XXX can't happen */
+    strcpy(pathbuf, path);
+
+    chptr = NULL;
+    /*@-branchstate@*/
+    do {
+	if ((chptr = strchr(start, ':')))
+	    *chptr = '\0';
+	sprintf(buf, "%s/%s", start, argv0);
+
+	if (!access(buf, X_OK))
+	    return buf;
+
+	if (chptr) 
+	    start = chptr + 1;
+	else
+	    start = NULL;
+    } while (start && *start);
+    /*@=branchstate@*/
+
+    free(buf);
+
+    return NULL;
+}
diff --git a/ctdb/lib/popt/findme.h b/ctdb/lib/popt/findme.h
new file mode 100644
index 0000000000..a016b867ea
--- /dev/null
+++ b/ctdb/lib/popt/findme.h
@@ -0,0 +1,20 @@
+/** \ingroup popt
+ * \file popt/findme.h
+ */
+
+/* (C) 1998-2000 Red Hat, Inc. -- Licensing details are in the COPYING
+   file accompanying popt source distributions, available from 
+   ftp://ftp.rpm.org/pub/rpm/dist. */
+
+#ifndef H_FINDME
+#define H_FINDME
+
+/**
+ * Return absolute path to executable by searching PATH.
+ * @param argv0		name of executable
+ * @return		(malloc'd) absolute path to executable (or NULL)
+ */
+/*@null@*/ const char * findProgramPath(/*@null@*/ const char * argv0)
+	/*@*/;
+
+#endif
diff --git a/ctdb/lib/popt/libpopt.m4 b/ctdb/lib/popt/libpopt.m4
new file mode 100644
index 0000000000..79980d1d6b
--- /dev/null
+++ b/ctdb/lib/popt/libpopt.m4
@@ -0,0 +1,43 @@
+dnl Check to see if we should use the included popt
+
+INCLUDED_POPT=auto
+AC_ARG_WITH(included-popt,
+[  --with-included-popt    use bundled popt library, not from system],
+[ INCLUDED_POPT=$withval ])
+
+AC_SUBST(POPT_LIBS)
+AC_SUBST(POPT_CFLAGS)
+
+if test x"$INCLUDED_POPT" != x"yes"; then
+	AC_CHECK_HEADERS(popt.h)
+	AC_CHECK_LIB(popt, poptGetContext, [ POPT_LIBS="-lpopt" ])
+	if test x"$ac_cv_header_popt_h" = x"no" -o x"$ac_cv_lib_popt_poptGetContext" = x"no"; then
+		INCLUDED_POPT=yes
+		POPT_CFLAGS=""
+	else
+		INCLUDED_POPT=no
+	fi
+fi
+
+AC_MSG_CHECKING(whether to use included popt)
+AC_MSG_RESULT($INCLUDED_POPT)
+if test x"$INCLUDED_POPT" != x"no"; then
+	dnl find the popt sources. This is meant to work both for 
+	dnl popt standalone builds, and builds of packages using popt
+	poptdir=""
+	poptpaths="$srcdir $srcdir/lib/popt $srcdir/popt $srcdir/../popt"
+	for d in $poptpaths; do
+		if test -f "$d/popt.c"; then
+			poptdir="$d"		
+			POPT_CFLAGS="-I$d"
+			AC_SUBST(poptdir)
+			break
+		fi
+	done
+        if test x"$poptdir" = "x"; then
+		AC_MSG_ERROR([cannot find popt source in $poptpaths])
+	fi
+	POPT_OBJ="popt.o findme.o poptconfig.o popthelp.o poptparse.o"
+	AC_SUBST(POPT_OBJ)
+	AC_CHECK_HEADERS([float.h alloca.h])
+fi
diff --git a/ctdb/lib/popt/popt.c b/ctdb/lib/popt/popt.c
new file mode 100644
index 0000000000..33a5ec7d4f
--- /dev/null
+++ b/ctdb/lib/popt/popt.c
@@ -0,0 +1,1240 @@
+/** \ingroup popt
+ * \file popt/popt.c
+ */
+
+/* (C) 1998-2002 Red Hat, Inc. -- Licensing details are in the COPYING
+   file accompanying popt source distributions, available from
+   ftp://ftp.rpm.org/pub/rpm/dist */
+
+#undef	MYDEBUG
+
+#include "system.h"
+
+#if HAVE_MATH_H
+#include <math.h>
+#endif
+#if HAVE_FLOAT_H
+#include <float.h>
+#endif
+
+#include "findme.h"
+#include "poptint.h"
+
+#ifdef	MYDEBUG
+/*@unchecked@*/
+int _popt_debug = 0;
+#endif
+
+#ifndef HAVE_STRERROR
+static char * strerror(int errno) {
+    extern int sys_nerr;
+    extern char * sys_errlist[];
+
+    if ((0 <= errno) && (errno < sys_nerr))
+	return sys_errlist[errno];
+    else
+	return POPT_("unknown errno");
+}
+#endif
+
+#ifdef MYDEBUG
+/*@unused@*/ static void prtcon(const char *msg, poptContext con)
+{
+    if (msg) fprintf(stderr, "%s", msg);
+    fprintf(stderr, "\tcon %p os %p nextCharArg \"%s\" nextArg \"%s\" argv[%d] \"%s\"\n",
+	con, con->os,
+	(con->os->nextCharArg ? con->os->nextCharArg : ""),
+	(con->os->nextArg ? con->os->nextArg : ""),
+	con->os->next,
+	(con->os->argv && con->os->argv[con->os->next]
+		? con->os->argv[con->os->next] : ""));
+}
+#endif
+
+void poptSetExecPath(poptContext con, const char * path, int allowAbsolute)
+{
+    con->execPath = _free(con->execPath);
+    con->execPath = xstrdup(path);
+    con->execAbsolute = allowAbsolute;
+    /*@-nullstate@*/ /* LCL: con->execPath can be NULL? */
+    return;
+    /*@=nullstate@*/
+}
+
+static void invokeCallbacksPRE(poptContext con, const struct poptOption * opt)
+	/*@globals internalState@*/
+	/*@modifies internalState@*/
+{
+    if (opt != NULL)
+    for (; opt->longName || opt->shortName || opt->arg; opt++) {
+	if (opt->arg == NULL) continue;		/* XXX program error. */
+	if ((opt->argInfo & POPT_ARG_MASK) == POPT_ARG_INCLUDE_TABLE) {
+	    /* Recurse on included sub-tables. */
+	    invokeCallbacksPRE(con, opt->arg);
+	} else if ((opt->argInfo & POPT_ARG_MASK) == POPT_ARG_CALLBACK &&
+		   (opt->argInfo & POPT_CBFLAG_PRE))
+	{   /*@-castfcnptr@*/
+	    poptCallbackType cb = (poptCallbackType)opt->arg;
+	    /*@=castfcnptr@*/
+	    /* Perform callback. */
+	    /*@-moduncon -noeffectuncon @*/
+	    cb(con, POPT_CALLBACK_REASON_PRE, NULL, NULL, opt->descrip);
+	    /*@=moduncon =noeffectuncon @*/
+	}
+    }
+}
+
+static void invokeCallbacksPOST(poptContext con, const struct poptOption * opt)
+	/*@globals internalState@*/
+	/*@modifies internalState@*/
+{
+    if (opt != NULL)
+    for (; opt->longName || opt->shortName || opt->arg; opt++) {
+	if (opt->arg == NULL) continue;		/* XXX program error. */
+	if ((opt->argInfo & POPT_ARG_MASK) == POPT_ARG_INCLUDE_TABLE) {
+	    /* Recurse on included sub-tables. */
+	    invokeCallbacksPOST(con, opt->arg);
+	} else if ((opt->argInfo & POPT_ARG_MASK) == POPT_ARG_CALLBACK &&
+		   (opt->argInfo & POPT_CBFLAG_POST))
+	{   /*@-castfcnptr@*/
+	    poptCallbackType cb = (poptCallbackType)opt->arg;
+	    /*@=castfcnptr@*/
+	    /* Perform callback. */
+	    /*@-moduncon -noeffectuncon @*/
+	    cb(con, POPT_CALLBACK_REASON_POST, NULL, NULL, opt->descrip);
+	    /*@=moduncon =noeffectuncon @*/
+	}
+    }
+}
+
+static void invokeCallbacksOPTION(poptContext con,
+				  const struct poptOption * opt,
+				  const struct poptOption * myOpt,
+				  /*@null@*/ const void * myData, int shorty)
+	/*@globals internalState@*/
+	/*@modifies internalState@*/
+{
+    const struct poptOption * cbopt = NULL;
+
+    if (opt != NULL)
+    for (; opt->longName || opt->shortName || opt->arg; opt++) {
+	if ((opt->argInfo & POPT_ARG_MASK) == POPT_ARG_INCLUDE_TABLE) {
+	    /* Recurse on included sub-tables. */
+	    if (opt->arg != NULL)	/* XXX program error */
+		invokeCallbacksOPTION(con, opt->arg, myOpt, myData, shorty);
+	} else if ((opt->argInfo & POPT_ARG_MASK) == POPT_ARG_CALLBACK &&
+		  !(opt->argInfo & POPT_CBFLAG_SKIPOPTION)) {
+	    /* Save callback info. */
+	    cbopt = opt;
+	} else if (cbopt != NULL &&
+		   ((myOpt->shortName && opt->shortName && shorty &&
+			myOpt->shortName == opt->shortName) ||
+		    (myOpt->longName && opt->longName &&
+		/*@-nullpass@*/		/* LCL: opt->longName != NULL */
+			!strcmp(myOpt->longName, opt->longName)))
+		/*@=nullpass@*/
+		   )
+	{   /*@-castfcnptr@*/
+	    poptCallbackType cb = (poptCallbackType)cbopt->arg;
+	    /*@=castfcnptr@*/
+	    const void * cbData = (cbopt->descrip ? cbopt->descrip : myData);
+	    /* Perform callback. */
+	    if (cb != NULL) {	/* XXX program error */
+		/*@-moduncon -noeffectuncon @*/
+		cb(con, POPT_CALLBACK_REASON_OPTION, myOpt,
+			con->os->nextArg, cbData);
+		/*@=moduncon =noeffectuncon @*/
+	    }
+	    /* Terminate (unless explcitly continuing). */
+	    if (!(cbopt->argInfo & POPT_CBFLAG_CONTINUE))
+		return;
+	}
+    }
+}
+
+poptContext poptGetContext(const char * name, int argc, const char ** argv,
+			   const struct poptOption * options, int flags)
+{
+    poptContext con = malloc(sizeof(*con));
+
+    if (con == NULL) return NULL;	/* XXX can't happen */
+    memset(con, 0, sizeof(*con));
+
+    con->os = con->optionStack;
+    con->os->argc = argc;
+    /*@-dependenttrans -assignexpose@*/	/* FIX: W2DO? */
+    con->os->argv = argv;
+    /*@=dependenttrans =assignexpose@*/
+    con->os->argb = NULL;
+
+    if (!(flags & POPT_CONTEXT_KEEP_FIRST))
+	con->os->next = 1;			/* skip argv[0] */
+
+    con->leftovers = calloc( (argc + 1), sizeof(*con->leftovers) );
+    /*@-dependenttrans -assignexpose@*/	/* FIX: W2DO? */
+    con->options = options;
+    /*@=dependenttrans =assignexpose@*/
+    con->aliases = NULL;
+    con->numAliases = 0;
+    con->flags = flags;
+    con->execs = NULL;
+    con->numExecs = 0;
+    con->finalArgvAlloced = argc * 2;
+    con->finalArgv = calloc( con->finalArgvAlloced, sizeof(*con->finalArgv) );
+    con->execAbsolute = 1;
+    con->arg_strip = NULL;
+
+    if (getenv("POSIXLY_CORRECT") || getenv("POSIX_ME_HARDER"))
+	con->flags |= POPT_CONTEXT_POSIXMEHARDER;
+
+    if (name) {
+	char * t = malloc(strlen(name) + 1);
+	if (t) con->appName = strcpy(t, name);
+    }
+
+    /*@-internalglobs@*/
+    invokeCallbacksPRE(con, con->options);
+    /*@=internalglobs@*/
+
+    return con;
+}
+
+static void cleanOSE(/*@special@*/ struct optionStackEntry *os)
+	/*@uses os @*/
+	/*@releases os->nextArg, os->argv, os->argb @*/
+	/*@modifies os @*/
+{
+    os->nextArg = _free(os->nextArg);
+    os->argv = _free(os->argv);
+    os->argb = PBM_FREE(os->argb);
+}
+
+/*@-boundswrite@*/
+void poptResetContext(poptContext con)
+{
+    int i;
+
+    if (con == NULL) return;
+    while (con->os > con->optionStack) {
+	cleanOSE(con->os--);
+    }
+    con->os->argb = PBM_FREE(con->os->argb);
+    con->os->currAlias = NULL;
+    con->os->nextCharArg = NULL;
+    con->os->nextArg = NULL;
+    con->os->next = 1;			/* skip argv[0] */
+
+    con->numLeftovers = 0;
+    con->nextLeftover = 0;
+    con->restLeftover = 0;
+    con->doExec = NULL;
+
+    if (con->finalArgv != NULL)
+    for (i = 0; i < con->finalArgvCount; i++) {
+	/*@-unqualifiedtrans@*/		/* FIX: typedef double indirection. */
+	con->finalArgv[i] = _free(con->finalArgv[i]);
+	/*@=unqualifiedtrans@*/
+    }
+
+    con->finalArgvCount = 0;
+    con->arg_strip = PBM_FREE(con->arg_strip);
+    /*@-nullstate@*/	/* FIX: con->finalArgv != NULL */
+    return;
+    /*@=nullstate@*/
+}
+/*@=boundswrite@*/
+
+/* Only one of longName, shortName should be set, not both. */
+/*@-boundswrite@*/
+static int handleExec(/*@special@*/ poptContext con,
+		/*@null@*/ const char * longName, char shortName)
+	/*@uses con->execs, con->numExecs, con->flags, con->doExec,
+		con->finalArgv, con->finalArgvAlloced, con->finalArgvCount @*/
+	/*@modifies con @*/
+{
+    poptItem item;
+    int i;
+
+    if (con->execs == NULL || con->numExecs <= 0) /* XXX can't happen */
+	return 0;
+
+    for (i = con->numExecs - 1; i >= 0; i--) {
+	item = con->execs + i;
+	if (longName && !(item->option.longName &&
+			!strcmp(longName, item->option.longName)))
+	    continue;
+	else if (shortName != item->option.shortName)
+	    continue;
+	break;
+    }
+    if (i < 0) return 0;
+
+
+    if (con->flags & POPT_CONTEXT_NO_EXEC)
+	return 1;
+
+    if (con->doExec == NULL) {
+	con->doExec = con->execs + i;
+	return 1;
+    }
+
+    /* We already have an exec to do; remember this option for next
+       time 'round */
+    if ((con->finalArgvCount + 1) >= (con->finalArgvAlloced)) {
+	con->finalArgvAlloced += 10;
+	con->finalArgv = realloc(con->finalArgv,
+			sizeof(*con->finalArgv) * con->finalArgvAlloced);
+    }
+
+    i = con->finalArgvCount++;
+    if (con->finalArgv != NULL)	/* XXX can't happen */
+    {	char *s  = malloc((longName ? strlen(longName) : 0) + 3);
+	if (s != NULL) {	/* XXX can't happen */
+	    if (longName)
+		sprintf(s, "--%s", longName);
+	    else
+		sprintf(s, "-%c", shortName);
+	    con->finalArgv[i] = s;
+	} else
+	    con->finalArgv[i] = NULL;
+    }
+
+    /*@-nullstate@*/	/* FIX: con->finalArgv[] == NULL */
+    return 1;
+    /*@=nullstate@*/
+}
+/*@=boundswrite@*/
+
+/* Only one of longName, shortName may be set at a time */
+static int handleAlias(/*@special@*/ poptContext con,
+		/*@null@*/ const char * longName, char shortName,
+		/*@exposed@*/ /*@null@*/ const char * nextCharArg)
+	/*@uses con->aliases, con->numAliases, con->optionStack, con->os,
+		con->os->currAlias, con->os->currAlias->option.longName @*/
+	/*@modifies con @*/
+{
+    poptItem item = con->os->currAlias;
+    int rc;
+    int i;
+
+    if (item) {
+	if (longName && (item->option.longName &&
+		!strcmp(longName, item->option.longName)))
+	    return 0;
+	if (shortName && shortName == item->option.shortName)
+	    return 0;
+    }
+
+    if (con->aliases == NULL || con->numAliases <= 0) /* XXX can't happen */
+	return 0;
+
+    for (i = con->numAliases - 1; i >= 0; i--) {
+	item = con->aliases + i;
+	if (longName && !(item->option.longName &&
+			!strcmp(longName, item->option.longName)))
+	    continue;
+	else if (shortName != item->option.shortName)
+	    continue;
+	break;
+    }
+    if (i < 0) return 0;
+
+    if ((con->os - con->optionStack + 1) == POPT_OPTION_DEPTH)
+	return POPT_ERROR_OPTSTOODEEP;
+
+/*@-boundsread@*/
+    if (nextCharArg && *nextCharArg)
+	con->os->nextCharArg = nextCharArg;
+/*@=boundsread@*/
+
+    con->os++;
+    con->os->next = 0;
+    con->os->stuffed = 0;
+    con->os->nextArg = NULL;
+    con->os->nextCharArg = NULL;
+    con->os->currAlias = con->aliases + i;
+    rc = poptDupArgv(con->os->currAlias->argc, con->os->currAlias->argv,
+		&con->os->argc, &con->os->argv);
+    con->os->argb = NULL;
+
+    return (rc ? rc : 1);
+}
+
+/*@-bounds -boundswrite @*/
+static int execCommand(poptContext con)
+	/*@globals internalState @*/
+	/*@modifies internalState @*/
+{
+    poptItem item = con->doExec;
+    const char ** argv;
+    int argc = 0;
+#if defined(__hpux) || defined(HAVE_SETUID) || defined(HAVE_SETREUID)
+    int rc;
+#endif
+
+    if (item == NULL) /*XXX can't happen*/
+	return POPT_ERROR_NOARG;
+
+    if (item->argv == NULL || item->argc < 1 ||
+	(!con->execAbsolute && strchr(item->argv[0], '/')))
+	    return POPT_ERROR_NOARG;
+
+    argv = malloc(sizeof(*argv) *
+			(6 + item->argc + con->numLeftovers + con->finalArgvCount));
+    if (argv == NULL) return POPT_ERROR_MALLOC;	/* XXX can't happen */
+
+    if (!strchr(item->argv[0], '/') && con->execPath) {
+	char *s = alloca(strlen(con->execPath) + strlen(item->argv[0]) + sizeof("/"));
+	sprintf(s, "%s/%s", con->execPath, item->argv[0]);
+	argv[argc] = s;
+    } else {
+	argv[argc] = findProgramPath(item->argv[0]);
+    }
+    if (argv[argc++] == NULL) return POPT_ERROR_NOARG;
+
+    if (item->argc > 1) {
+	memcpy(argv + argc, item->argv + 1, sizeof(*argv) * (item->argc - 1));
+	argc += (item->argc - 1);
+    }
+
+    if (con->finalArgv != NULL && con->finalArgvCount > 0) {
+	memcpy(argv + argc, con->finalArgv,
+		sizeof(*argv) * con->finalArgvCount);
+	argc += con->finalArgvCount;
+    }
+
+    if (con->leftovers != NULL && con->numLeftovers > 0) {
+#if 0
+	argv[argc++] = "--";
+#endif
+	memcpy(argv + argc, con->leftovers, sizeof(*argv) * con->numLeftovers);
+	argc += con->numLeftovers;
+    }
+
+    argv[argc] = NULL;
+
+#ifdef __hpux
+    rc = setresuid(getuid(), getuid(),-1);
+    if (rc) return POPT_ERROR_ERRNO;
+#else
+/*
+ * XXX " ... on BSD systems setuid() should be preferred over setreuid()"
+ * XXX 	sez' Timur Bakeyev <mc@bat.ru>
+ * XXX	from Norbert Warmuth <nwarmuth@privat.circular.de>
+ */
+#if defined(HAVE_SETUID)
+    rc = setuid(getuid());
+    if (rc) return POPT_ERROR_ERRNO;
+#elif defined (HAVE_SETREUID)
+    rc = setreuid(getuid(), getuid()); /*hlauer: not portable to hpux9.01 */
+    if (rc) return POPT_ERROR_ERRNO;
+#else
+    ; /* Can't drop privileges */
+#endif
+#endif
+
+    if (argv[0] == NULL)
+	return POPT_ERROR_NOARG;
+
+#ifdef	MYDEBUG
+if (_popt_debug)
+    {	const char ** avp;
+	fprintf(stderr, "==> execvp(%s) argv[%d]:", argv[0], argc);
+	for (avp = argv; *avp; avp++)
+	    fprintf(stderr, " '%s'", *avp);
+	fprintf(stderr, "\n");
+    }
+#endif
+
+    (void) execvp(argv[0], (char *const *)(intptr_t)argv);
+
+    return POPT_ERROR_ERRNO;
+}
+/*@=bounds =boundswrite @*/
+
+/*@-boundswrite@*/
+/*@observer@*/ /*@null@*/ static const struct poptOption *
+findOption(const struct poptOption * opt, /*@null@*/ const char * longName,
+		char shortName,
+		/*@null@*/ /*@out@*/ poptCallbackType * callback,
+		/*@null@*/ /*@out@*/ const void ** callbackData,
+		int singleDash)
+	/*@modifies *callback, *callbackData */
+{
+    const struct poptOption * cb = NULL;
+
+    /* This happens when a single - is given */
+    if (singleDash && !shortName && (longName && *longName == '\0'))
+	shortName = '-';
+
+    for (; opt->longName || opt->shortName || opt->arg; opt++) {
+
+	if ((opt->argInfo & POPT_ARG_MASK) == POPT_ARG_INCLUDE_TABLE) {
+	    const struct poptOption * opt2;
+
+	    /* Recurse on included sub-tables. */
+	    if (opt->arg == NULL) continue;	/* XXX program error */
+	    opt2 = findOption(opt->arg, longName, shortName, callback,
+			      callbackData, singleDash);
+	    if (opt2 == NULL) continue;
+	    /* Sub-table data will be inheirited if no data yet. */
+	    if (!(callback && *callback)) return opt2;
+	    if (!(callbackData && *callbackData == NULL)) return opt2;
+	    /*@-observertrans -dependenttrans @*/
+	    *callbackData = opt->descrip;
+	    /*@=observertrans =dependenttrans @*/
+	    return opt2;
+	} else if ((opt->argInfo & POPT_ARG_MASK) == POPT_ARG_CALLBACK) {
+	    cb = opt;
+	} else if (longName && opt->longName &&
+		   (!singleDash || (opt->argInfo & POPT_ARGFLAG_ONEDASH)) &&
+		/*@-nullpass@*/		/* LCL: opt->longName != NULL */
+		   !strcmp(longName, opt->longName))
+		/*@=nullpass@*/
+	{
+	    break;
+	} else if (shortName && shortName == opt->shortName) {
+	    break;
+	}
+    }
+
+    if (!opt->longName && !opt->shortName)
+	return NULL;
+    /*@-modobserver -mods @*/
+    if (callback) *callback = NULL;
+    if (callbackData) *callbackData = NULL;
+    if (cb) {
+	if (callback)
+	/*@-castfcnptr@*/
+	    *callback = (poptCallbackType)cb->arg;
+	/*@=castfcnptr@*/
+	if (!(cb->argInfo & POPT_CBFLAG_INC_DATA)) {
+	    if (callbackData)
+		/*@-observertrans@*/	/* FIX: typedef double indirection. */
+		*callbackData = cb->descrip;
+		/*@=observertrans@*/
+	}
+    }
+    /*@=modobserver =mods @*/
+
+    return opt;
+}
+/*@=boundswrite@*/
+
+static const char * findNextArg(/*@special@*/ poptContext con,
+		unsigned argx, int delete_arg)
+	/*@uses con->optionStack, con->os,
+		con->os->next, con->os->argb, con->os->argc, con->os->argv @*/
+	/*@modifies con @*/
+{
+    struct optionStackEntry * os = con->os;
+    const char * arg;
+
+    do {
+	int i;
+	arg = NULL;
+	while (os->next == os->argc && os > con->optionStack) os--;
+	if (os->next == os->argc && os == con->optionStack) break;
+	if (os->argv != NULL)
+	for (i = os->next; i < os->argc; i++) {
+	    /*@-sizeoftype@*/
+	    if (os->argb && PBM_ISSET(i, os->argb))
+		/*@innercontinue@*/ continue;
+	    if (*os->argv[i] == '-')
+		/*@innercontinue@*/ continue;
+	    if (--argx > 0)
+		/*@innercontinue@*/ continue;
+	    arg = os->argv[i];
+	    if (delete_arg) {
+		if (os->argb == NULL) os->argb = PBM_ALLOC(os->argc);
+		if (os->argb != NULL)	/* XXX can't happen */
+		PBM_SET(i, os->argb);
+	    }
+	    /*@innerbreak@*/ break;
+	    /*@=sizeoftype@*/
+	}
+	if (os > con->optionStack) os--;
+    } while (arg == NULL);
+    return arg;
+}
+
+/*@-boundswrite@*/
+static /*@only@*/ /*@null@*/ const char *
+expandNextArg(/*@special@*/ poptContext con, const char * s)
+	/*@uses con->optionStack, con->os,
+		con->os->next, con->os->argb, con->os->argc, con->os->argv @*/
+	/*@modifies con @*/
+{
+    const char * a = NULL;
+    size_t alen;
+    char *t, *te;
+    size_t tn = strlen(s) + 1;
+    char c;
+
+    te = t = malloc(tn);;
+    if (t == NULL) return NULL;		/* XXX can't happen */
+    while ((c = *s++) != '\0') {
+	switch (c) {
+#if 0	/* XXX can't do this */
+	case '\\':	/* escape */
+	    c = *s++;
+	    /*@switchbreak@*/ break;
+#endif
+	case '!':
+	    if (!(s[0] == '#' && s[1] == ':' && s[2] == '+'))
+		/*@switchbreak@*/ break;
+	    /* XXX Make sure that findNextArg deletes only next arg. */
+	    if (a == NULL) {
+		if ((a = findNextArg(con, 1, 1)) == NULL)
+		    /*@switchbreak@*/ break;
+	    }
+	    s += 3;
+
+	    alen = strlen(a);
+	    tn += alen;
+	    *te = '\0';
+	    t = realloc(t, tn);
+	    te = t + strlen(t);
+	    strncpy(te, a, alen); te += alen;
+	    continue;
+	    /*@notreached@*/ /*@switchbreak@*/ break;
+	default:
+	    /*@switchbreak@*/ break;
+	}
+	*te++ = c;
+    }
+    *te = '\0';
+    t = realloc(t, strlen(t) + 1);	/* XXX memory leak, hard to plug */
+    return t;
+}
+/*@=boundswrite@*/
+
+static void poptStripArg(/*@special@*/ poptContext con, int which)
+	/*@uses con->arg_strip, con->optionStack @*/
+	/*@defines con->arg_strip @*/
+	/*@modifies con @*/
+{
+    /*@-sizeoftype@*/
+    if (con->arg_strip == NULL)
+	con->arg_strip = PBM_ALLOC(con->optionStack[0].argc);
+    if (con->arg_strip != NULL)		/* XXX can't happen */
+    PBM_SET(which, con->arg_strip);
+    /*@=sizeoftype@*/
+    /*@-compdef@*/ /* LCL: con->arg_strip udefined? */
+    return;
+    /*@=compdef@*/
+}
+
+int poptSaveLong(long * arg, int argInfo, long aLong)
+{
+    /* XXX Check alignment, may fail on funky platforms. */
+    if (arg == NULL || (((unsigned long)arg) & (sizeof(*arg)-1)))
+	return POPT_ERROR_NULLARG;
+
+    if (argInfo & POPT_ARGFLAG_NOT)
+	aLong = ~aLong;
+    switch (argInfo & POPT_ARGFLAG_LOGICALOPS) {
+    case 0:
+	*arg = aLong;
+	break;
+    case POPT_ARGFLAG_OR:
+	*arg |= aLong;
+	break;
+    case POPT_ARGFLAG_AND:
+	*arg &= aLong;
+	break;
+    case POPT_ARGFLAG_XOR:
+	*arg ^= aLong;
+	break;
+    default:
+	return POPT_ERROR_BADOPERATION;
+	/*@notreached@*/ break;
+    }
+    return 0;
+}
+
+int poptSaveInt(/*@null@*/ int * arg, int argInfo, long aLong)
+{
+    /* XXX Check alignment, may fail on funky platforms. */
+    if (arg == NULL || (((unsigned long)arg) & (sizeof(*arg)-1)))
+	return POPT_ERROR_NULLARG;
+
+    if (argInfo & POPT_ARGFLAG_NOT)
+	aLong = ~aLong;
+    switch (argInfo & POPT_ARGFLAG_LOGICALOPS) {
+    case 0:
+	*arg = aLong;
+	break;
+    case POPT_ARGFLAG_OR:
+	*arg |= aLong;
+	break;
+    case POPT_ARGFLAG_AND:
+	*arg &= aLong;
+	break;
+    case POPT_ARGFLAG_XOR:
+	*arg ^= aLong;
+	break;
+    default:
+	return POPT_ERROR_BADOPERATION;
+	/*@notreached@*/ break;
+    }
+    return 0;
+}
+
+/*@-boundswrite@*/
+/* returns 'val' element, -1 on last item, POPT_ERROR_* on error */
+int poptGetNextOpt(poptContext con)
+{
+    const struct poptOption * opt = NULL;
+    int done = 0;
+
+    if (con == NULL)
+	return -1;
+    while (!done) {
+	const char * origOptString = NULL;
+	poptCallbackType cb = NULL;
+	const void * cbData = NULL;
+	const char * longArg = NULL;
+	int canstrip = 0;
+	int shorty = 0;
+
+	while (!con->os->nextCharArg && con->os->next == con->os->argc
+		&& con->os > con->optionStack) {
+	    cleanOSE(con->os--);
+	}
+	if (!con->os->nextCharArg && con->os->next == con->os->argc) {
+	    /*@-internalglobs@*/
+	    invokeCallbacksPOST(con, con->options);
+	    /*@=internalglobs@*/
+	    if (con->doExec) return execCommand(con);
+	    return -1;
+	}
+
+	/* Process next long option */
+	if (!con->os->nextCharArg) {
+	    char * localOptString, * optString;
+	    int thisopt;
+
+	    /*@-sizeoftype@*/
+	    if (con->os->argb && PBM_ISSET(con->os->next, con->os->argb)) {
+		con->os->next++;
+		continue;
+	    }
+	    /*@=sizeoftype@*/
+	    thisopt = con->os->next;
+	    if (con->os->argv != NULL)	/* XXX can't happen */
+	    origOptString = con->os->argv[con->os->next++];
+
+	    if (origOptString == NULL)	/* XXX can't happen */
+		return POPT_ERROR_BADOPT;
+
+	    if (con->restLeftover || *origOptString != '-') {
+		if (con->flags & POPT_CONTEXT_POSIXMEHARDER)
+		    con->restLeftover = 1;
+		if (con->flags & POPT_CONTEXT_ARG_OPTS) {
+		    con->os->nextArg = xstrdup(origOptString);
+		    return 0;
+		}
+		if (con->leftovers != NULL)	/* XXX can't happen */
+		    con->leftovers[con->numLeftovers++] = origOptString;
+		continue;
+	    }
+
+	    /* Make a copy we can hack at */
+	    localOptString = optString =
+		strcpy(alloca(strlen(origOptString) + 1), origOptString);
+
+	    if (optString[0] == '\0')
+		return POPT_ERROR_BADOPT;
+
+	    if (optString[1] == '-' && !optString[2]) {
+		con->restLeftover = 1;
+		continue;
+	    } else {
+		char *oe;
+		int singleDash;
+
+		optString++;
+		if (*optString == '-')
+		    singleDash = 0, optString++;
+		else
+		    singleDash = 1;
+
+		/* XXX aliases with arg substitution need "--alias=arg" */
+		if (handleAlias(con, optString, '\0', NULL))
+		    continue;
+
+		if (handleExec(con, optString, '\0'))
+		    continue;
+
+		/* Check for "--long=arg" option. */
+		for (oe = optString; *oe && *oe != '='; oe++)
+		    {};
+		if (*oe == '=') {
+		    *oe++ = '\0';
+		    /* XXX longArg is mapped back to persistent storage. */
+		    longArg = origOptString + (oe - localOptString);
+		}
+
+		opt = findOption(con->options, optString, '\0', &cb, &cbData,
+				 singleDash);
+		if (!opt && !singleDash)
+		    return POPT_ERROR_BADOPT;
+	    }
+
+	    if (!opt) {
+		con->os->nextCharArg = origOptString + 1;
+	    } else {
+		if (con->os == con->optionStack &&
+		   opt->argInfo & POPT_ARGFLAG_STRIP)
+		{
+		    canstrip = 1;
+		    poptStripArg(con, thisopt);
+		}
+		shorty = 0;
+	    }
+	}
+
+	/* Process next short option */
+	/*@-branchstate@*/		/* FIX: W2DO? */
+	if (con->os->nextCharArg) {
+	    origOptString = con->os->nextCharArg;
+
+	    con->os->nextCharArg = NULL;
+
+	    if (handleAlias(con, NULL, *origOptString, origOptString + 1))
+		continue;
+
+	    if (handleExec(con, NULL, *origOptString)) {
+		/* Restore rest of short options for further processing */
+		origOptString++;
+		if (*origOptString != '\0')
+		    con->os->nextCharArg = origOptString;
+		continue;
+	    }
+
+	    opt = findOption(con->options, NULL, *origOptString, &cb,
+			     &cbData, 0);
+	    if (!opt)
+		return POPT_ERROR_BADOPT;
+	    shorty = 1;
+
+	    origOptString++;
+	    if (*origOptString != '\0')
+		con->os->nextCharArg = origOptString;
+	}
+	/*@=branchstate@*/
+
+	if (opt == NULL) return POPT_ERROR_BADOPT;	/* XXX can't happen */
+	if (opt->arg && (opt->argInfo & POPT_ARG_MASK) == POPT_ARG_NONE) {
+	    if (poptSaveInt((int *)opt->arg, opt->argInfo, 1L))
+		return POPT_ERROR_BADOPERATION;
+	} else if ((opt->argInfo & POPT_ARG_MASK) == POPT_ARG_VAL) {
+	    if (opt->arg) {
+		if (poptSaveInt((int *)opt->arg, opt->argInfo, (long)opt->val))
+		    return POPT_ERROR_BADOPERATION;
+	    }
+	} else if ((opt->argInfo & POPT_ARG_MASK) != POPT_ARG_NONE) {
+	    con->os->nextArg = _free(con->os->nextArg);
+	    /*@-usedef@*/	/* FIX: W2DO? */
+	    if (longArg) {
+	    /*@=usedef@*/
+		longArg = expandNextArg(con, longArg);
+		con->os->nextArg = longArg;
+	    } else if (con->os->nextCharArg) {
+		longArg = expandNextArg(con, con->os->nextCharArg);
+		con->os->nextArg = longArg;
+		con->os->nextCharArg = NULL;
+	    } else {
+		while (con->os->next == con->os->argc &&
+		       con->os > con->optionStack) {
+		    cleanOSE(con->os--);
+		}
+		if (con->os->next == con->os->argc) {
+		    if (!(opt->argInfo & POPT_ARGFLAG_OPTIONAL))
+			/*@-compdef@*/	/* FIX: con->os->argv not defined */
+			return POPT_ERROR_NOARG;
+			/*@=compdef@*/
+		    con->os->nextArg = NULL;
+		} else {
+
+		    /*
+		     * Make sure this isn't part of a short arg or the
+		     * result of an alias expansion.
+		     */
+		    if (con->os == con->optionStack &&
+			(opt->argInfo & POPT_ARGFLAG_STRIP) &&
+			canstrip) {
+			poptStripArg(con, con->os->next);
+		    }
+		
+		    if (con->os->argv != NULL) {	/* XXX can't happen */
+			/* XXX watchout: subtle side-effects live here. */
+			longArg = con->os->argv[con->os->next++];
+			longArg = expandNextArg(con, longArg);
+			con->os->nextArg = longArg;
+		    }
+		}
+	    }
+	    longArg = NULL;
+
+	    if (opt->arg) {
+		switch (opt->argInfo & POPT_ARG_MASK) {
+		case POPT_ARG_STRING:
+		    /* XXX memory leak, hard to plug */
+		    *((const char **) opt->arg) = (con->os->nextArg)
+			? xstrdup(con->os->nextArg) : NULL;
+		    /*@switchbreak@*/ break;
+
+		case POPT_ARG_INT:
+		case POPT_ARG_LONG:
+		{   long aLong = 0;
+		    char *end;
+
+		    if (con->os->nextArg) {
+			aLong = strtol(con->os->nextArg, &end, 0);
+			if (!(end && *end == '\0'))
+			    return POPT_ERROR_BADNUMBER;
+		    }
+
+		    if ((opt->argInfo & POPT_ARG_MASK) == POPT_ARG_LONG) {
+			if (aLong == LONG_MIN || aLong == LONG_MAX)
+			    return POPT_ERROR_OVERFLOW;
+			if (poptSaveLong((long *)opt->arg, opt->argInfo, aLong))
+			    return POPT_ERROR_BADOPERATION;
+		    } else {
+			if (aLong > INT_MAX || aLong < INT_MIN)
+			    return POPT_ERROR_OVERFLOW;
+			if (poptSaveInt((int *)opt->arg, opt->argInfo, aLong))
+			    return POPT_ERROR_BADOPERATION;
+		    }
+		}   /*@switchbreak@*/ break;
+
+		case POPT_ARG_FLOAT:
+		case POPT_ARG_DOUBLE:
+		{   double aDouble = 0.0;
+		    char *end;
+
+		    if (con->os->nextArg) {
+			/*@-mods@*/
+			int saveerrno = errno;
+			errno = 0;
+			aDouble = strtod(con->os->nextArg, &end);
+			if (errno == ERANGE)
+			    return POPT_ERROR_OVERFLOW;
+			errno = saveerrno;
+			/*@=mods@*/
+			if (*end != '\0')
+			    return POPT_ERROR_BADNUMBER;
+		    }
+
+		    if ((opt->argInfo & POPT_ARG_MASK) == POPT_ARG_DOUBLE) {
+			*((double *) opt->arg) = aDouble;
+		    } else {
+#define _ABS(a)	((((a) - 0.0) < DBL_EPSILON) ? -(a) : (a))
+			if ((_ABS(aDouble) - FLT_MAX) > DBL_EPSILON)
+			    return POPT_ERROR_OVERFLOW;
+			if ((FLT_MIN - _ABS(aDouble)) > DBL_EPSILON)
+			    return POPT_ERROR_OVERFLOW;
+			*((float *) opt->arg) = aDouble;
+		    }
+		}   /*@switchbreak@*/ break;
+		default:
+		    fprintf(stdout,
+			POPT_("option type (%d) not implemented in popt\n"),
+			(opt->argInfo & POPT_ARG_MASK));
+		    exit(EXIT_FAILURE);
+		    /*@notreached@*/ /*@switchbreak@*/ break;
+		}
+	    }
+	}
+
+	if (cb) {
+	    /*@-internalglobs@*/
+	    invokeCallbacksOPTION(con, con->options, opt, cbData, shorty);
+	    /*@=internalglobs@*/
+	} else if (opt->val && ((opt->argInfo & POPT_ARG_MASK) != POPT_ARG_VAL))
+	    done = 1;
+
+	if ((con->finalArgvCount + 2) >= (con->finalArgvAlloced)) {
+	    con->finalArgvAlloced += 10;
+	    con->finalArgv = realloc(con->finalArgv,
+			    sizeof(*con->finalArgv) * con->finalArgvAlloced);
+	}
+
+	if (con->finalArgv != NULL)
+	{   char *s = malloc((opt->longName ? strlen(opt->longName) : 0) + 3);
+	    if (s != NULL) {	/* XXX can't happen */
+		if (opt->longName)
+		    sprintf(s, "%s%s",
+			((opt->argInfo & POPT_ARGFLAG_ONEDASH) ? "-" : "--"),
+			opt->longName);
+		else
+		    sprintf(s, "-%c", opt->shortName);
+		con->finalArgv[con->finalArgvCount++] = s;
+	    } else
+		con->finalArgv[con->finalArgvCount++] = NULL;
+	}
+
+	if (opt->arg && (opt->argInfo & POPT_ARG_MASK) == POPT_ARG_NONE)
+	    /*@-ifempty@*/ ; /*@=ifempty@*/
+	else if ((opt->argInfo & POPT_ARG_MASK) == POPT_ARG_VAL)
+	    /*@-ifempty@*/ ; /*@=ifempty@*/
+	else if ((opt->argInfo & POPT_ARG_MASK) != POPT_ARG_NONE) {
+	    if (con->finalArgv != NULL && con->os->nextArg)
+	        con->finalArgv[con->finalArgvCount++] =
+			/*@-nullpass@*/	/* LCL: con->os->nextArg != NULL */
+			xstrdup(con->os->nextArg);
+			/*@=nullpass@*/
+	}
+    }
+
+    return (opt ? opt->val : -1);	/* XXX can't happen */
+}
+/*@=boundswrite@*/
+
+const char * poptGetOptArg(poptContext con)
+{
+    const char * ret = NULL;
+    /*@-branchstate@*/
+    if (con) {
+	ret = con->os->nextArg;
+	con->os->nextArg = NULL;
+    }
+    /*@=branchstate@*/
+    return ret;
+}
+
+const char * poptGetArg(poptContext con)
+{
+    const char * ret = NULL;
+    if (con && con->leftovers != NULL && con->nextLeftover < con->numLeftovers)
+	ret = con->leftovers[con->nextLeftover++];
+    return ret;
+}
+
+const char * poptPeekArg(poptContext con)
+{
+    const char * ret = NULL;
+    if (con && con->leftovers != NULL && con->nextLeftover < con->numLeftovers)
+	ret = con->leftovers[con->nextLeftover];
+    return ret;
+}
+
+/*@-boundswrite@*/
+const char ** poptGetArgs(poptContext con)
+{
+    if (con == NULL ||
+	con->leftovers == NULL || con->numLeftovers == con->nextLeftover)
+	return NULL;
+
+    /* some apps like [like RPM ;-) ] need this NULL terminated */
+    con->leftovers[con->numLeftovers] = NULL;
+
+    /*@-nullret -nullstate @*/	/* FIX: typedef double indirection. */
+    return (con->leftovers + con->nextLeftover);
+    /*@=nullret =nullstate @*/
+}
+/*@=boundswrite@*/
+
+poptContext poptFreeContext(poptContext con)
+{
+    poptItem item;
+    int i;
+
+    if (con == NULL) return con;
+    poptResetContext(con);
+    con->os->argb = _free(con->os->argb);
+
+    if (con->aliases != NULL)
+    for (i = 0; i < con->numAliases; i++) {
+	item = con->aliases + i;
+	/*@-modobserver -observertrans -dependenttrans@*/
+	item->option.longName = _free(item->option.longName);
+	item->option.descrip = _free(item->option.descrip);
+	item->option.argDescrip = _free(item->option.argDescrip);
+	/*@=modobserver =observertrans =dependenttrans@*/
+	item->argv = _free(item->argv);
+    }
+    con->aliases = _free(con->aliases);
+
+    if (con->execs != NULL)
+    for (i = 0; i < con->numExecs; i++) {
+	item = con->execs + i;
+	/*@-modobserver -observertrans -dependenttrans@*/
+	item->option.longName = _free(item->option.longName);
+	item->option.descrip = _free(item->option.descrip);
+	item->option.argDescrip = _free(item->option.argDescrip);
+	/*@=modobserver =observertrans =dependenttrans@*/
+	item->argv = _free(item->argv);
+    }
+    con->execs = _free(con->execs);
+
+    con->leftovers = _free(con->leftovers);
+    con->finalArgv = _free(con->finalArgv);
+    con->appName = _free(con->appName);
+    con->otherHelp = _free(con->otherHelp);
+    con->execPath = _free(con->execPath);
+    con->arg_strip = PBM_FREE(con->arg_strip);
+    
+    con = _free(con);
+    return con;
+}
+
+int poptAddAlias(poptContext con, struct poptAlias alias,
+		/*@unused@*/ int flags)
+{
+    poptItem item = alloca(sizeof(*item));
+    memset(item, 0, sizeof(*item));
+    item->option.longName = alias.longName;
+    item->option.shortName = alias.shortName;
+    item->option.argInfo = POPT_ARGFLAG_DOC_HIDDEN;
+    item->option.arg = 0;
+    item->option.val = 0;
+    item->option.descrip = NULL;
+    item->option.argDescrip = NULL;
+    item->argc = alias.argc;
+    item->argv = alias.argv;
+    return poptAddItem(con, item, 0);
+}
+
+/*@-boundswrite@*/
+/*@-mustmod@*/ /* LCL: con not modified? */
+int poptAddItem(poptContext con, poptItem newItem, int flags)
+{
+    poptItem * items, item;
+    int * nitems;
+
+    switch (flags) {
+    case 1:
+	items = &con->execs;
+	nitems = &con->numExecs;
+	break;
+    case 0:
+	items = &con->aliases;
+	nitems = &con->numAliases;
+	break;
+    default:
+	return 1;
+	/*@notreached@*/ break;
+    }
+
+    *items = realloc((*items), ((*nitems) + 1) * sizeof(**items));
+    if ((*items) == NULL)
+	return 1;
+
+    item = (*items) + (*nitems);
+
+    item->option.longName =
+	(newItem->option.longName ? xstrdup(newItem->option.longName) : NULL);
+    item->option.shortName = newItem->option.shortName;
+    item->option.argInfo = newItem->option.argInfo;
+    item->option.arg = newItem->option.arg;
+    item->option.val = newItem->option.val;
+    item->option.descrip =
+	(newItem->option.descrip ? xstrdup(newItem->option.descrip) : NULL);
+    item->option.argDescrip =
+       (newItem->option.argDescrip ? xstrdup(newItem->option.argDescrip) : NULL);
+    item->argc = newItem->argc;
+    item->argv = newItem->argv;
+
+    (*nitems)++;
+
+    return 0;
+}
+/*@=mustmod@*/
+/*@=boundswrite@*/
+
+const char * poptBadOption(poptContext con, int flags)
+{
+    struct optionStackEntry * os = NULL;
+
+    if (con != NULL)
+	os = (flags & POPT_BADOPTION_NOALIAS) ? con->optionStack : con->os;
+
+    /*@-nullderef@*/	/* LCL: os->argv != NULL */
+    return (os && os->argv ? os->argv[os->next - 1] : NULL);
+    /*@=nullderef@*/
+}
+
+const char *poptStrerror(const int error)
+{
+    switch (error) {
+      case POPT_ERROR_NOARG:
+	return POPT_("missing argument");
+      case POPT_ERROR_BADOPT:
+	return POPT_("unknown option");
+      case POPT_ERROR_BADOPERATION:
+	return POPT_("mutually exclusive logical operations requested");
+      case POPT_ERROR_NULLARG:
+	return POPT_("opt->arg should not be NULL");
+      case POPT_ERROR_OPTSTOODEEP:
+	return POPT_("aliases nested too deeply");
+      case POPT_ERROR_BADQUOTE:
+	return POPT_("error in parameter quoting");
+      case POPT_ERROR_BADNUMBER:
+	return POPT_("invalid numeric value");
+      case POPT_ERROR_OVERFLOW:
+	return POPT_("number too large or too small");
+      case POPT_ERROR_MALLOC:
+	return POPT_("memory allocation failed");
+      case POPT_ERROR_ERRNO:
+	return strerror(errno);
+      default:
+	return POPT_("unknown error");
+    }
+}
+
+int poptStuffArgs(poptContext con, const char ** argv)
+{
+    int argc;
+    int rc;
+
+    if ((con->os - con->optionStack) == POPT_OPTION_DEPTH)
+	return POPT_ERROR_OPTSTOODEEP;
+
+    for (argc = 0; argv[argc]; argc++)
+	{};
+
+    con->os++;
+    con->os->next = 0;
+    con->os->nextArg = NULL;
+    con->os->nextCharArg = NULL;
+    con->os->currAlias = NULL;
+    rc = poptDupArgv(argc, argv, &con->os->argc, &con->os->argv);
+    con->os->argb = NULL;
+    con->os->stuffed = 1;
+
+    return rc;
+}
+
+const char * poptGetInvocationName(poptContext con)
+{
+    return (con->os->argv ? con->os->argv[0] : "");
+}
+
+/*@-boundswrite@*/
+int poptStrippedArgv(poptContext con, int argc, char ** argv)
+{
+    int numargs = argc;
+    int j = 1;
+    int i;
+    
+    /*@-sizeoftype@*/
+    if (con->arg_strip)
+    for (i = 1; i < argc; i++) {
+	if (PBM_ISSET(i, con->arg_strip))
+	    numargs--;
+    }
+    
+    for (i = 1; i < argc; i++) {
+	if (con->arg_strip && PBM_ISSET(i, con->arg_strip))
+	    continue;
+	argv[j] = (j < numargs) ? argv[i] : NULL;
+	j++;
+    }
+    /*@=sizeoftype@*/
+    
+    return numargs;
+}
+/*@=boundswrite@*/
diff --git a/ctdb/lib/popt/popt.h b/ctdb/lib/popt/popt.h
new file mode 100644
index 0000000000..04c9f655ab
--- /dev/null
+++ b/ctdb/lib/popt/popt.h
@@ -0,0 +1,541 @@
+/** \file popt/popt.h
+ * \ingroup popt
+ */
+
+/* (C) 1998-2000 Red Hat, Inc. -- Licensing details are in the COPYING
+   file accompanying popt source distributions, available from 
+   ftp://ftp.rpm.org/pub/rpm/dist. */
+
+#ifndef H_POPT
+#define H_POPT
+
+#include <stdio.h>			/* for FILE * */
+
+#define POPT_OPTION_DEPTH	10
+
+/** \ingroup popt
+ * \name Arg type identifiers
+ */
+/*@{*/
+#define POPT_ARG_NONE		0	/*!< no arg */
+#define POPT_ARG_STRING		1	/*!< arg will be saved as string */
+#define POPT_ARG_INT		2	/*!< arg will be converted to int */
+#define POPT_ARG_LONG		3	/*!< arg will be converted to long */
+#define POPT_ARG_INCLUDE_TABLE	4	/*!< arg points to table */
+#define POPT_ARG_CALLBACK	5	/*!< table-wide callback... must be
+					   set first in table; arg points 
+					   to callback, descrip points to 
+					   callback data to pass */
+#define POPT_ARG_INTL_DOMAIN    6       /*!< set the translation domain
+					   for this table and any
+					   included tables; arg points
+					   to the domain string */
+#define POPT_ARG_VAL		7	/*!< arg should take value val */
+#define	POPT_ARG_FLOAT		8	/*!< arg will be converted to float */
+#define	POPT_ARG_DOUBLE		9	/*!< arg will be converted to double */
+
+#define POPT_ARG_MASK		0x0000FFFF
+/*@}*/
+
+/** \ingroup popt
+ * \name Arg modifiers
+ */
+/*@{*/
+#define POPT_ARGFLAG_ONEDASH	0x80000000  /*!< allow -longoption */
+#define POPT_ARGFLAG_DOC_HIDDEN 0x40000000  /*!< don't show in help/usage */
+#define POPT_ARGFLAG_STRIP	0x20000000  /*!< strip this arg from argv(only applies to long args) */
+#define	POPT_ARGFLAG_OPTIONAL	0x10000000  /*!< arg may be missing */
+
+#define	POPT_ARGFLAG_OR		0x08000000  /*!< arg will be or'ed */
+#define	POPT_ARGFLAG_NOR	0x09000000  /*!< arg will be nor'ed */
+#define	POPT_ARGFLAG_AND	0x04000000  /*!< arg will be and'ed */
+#define	POPT_ARGFLAG_NAND	0x05000000  /*!< arg will be nand'ed */
+#define	POPT_ARGFLAG_XOR	0x02000000  /*!< arg will be xor'ed */
+#define	POPT_ARGFLAG_NOT	0x01000000  /*!< arg will be negated */
+#define POPT_ARGFLAG_LOGICALOPS \
+        (POPT_ARGFLAG_OR|POPT_ARGFLAG_AND|POPT_ARGFLAG_XOR)
+
+#define	POPT_BIT_SET	(POPT_ARG_VAL|POPT_ARGFLAG_OR)
+					/*!< set arg bit(s) */
+#define	POPT_BIT_CLR	(POPT_ARG_VAL|POPT_ARGFLAG_NAND)
+					/*!< clear arg bit(s) */
+
+#define	POPT_ARGFLAG_SHOW_DEFAULT 0x00800000 /*!< show default value in --help */
+
+/*@}*/
+
+/** \ingroup popt
+ * \name Callback modifiers
+ */
+/*@{*/
+#define POPT_CBFLAG_PRE		0x80000000  /*!< call the callback before parse */
+#define POPT_CBFLAG_POST	0x40000000  /*!< call the callback after parse */
+#define POPT_CBFLAG_INC_DATA	0x20000000  /*!< use data from the include line,
+					       not the subtable */
+#define POPT_CBFLAG_SKIPOPTION	0x10000000  /*!< don't callback with option */
+#define POPT_CBFLAG_CONTINUE	0x08000000  /*!< continue callbacks with option */
+/*@}*/
+
+/** \ingroup popt
+ * \name Error return values
+ */
+/*@{*/
+#define POPT_ERROR_NOARG	-10	/*!< missing argument */
+#define POPT_ERROR_BADOPT	-11	/*!< unknown option */
+#define POPT_ERROR_OPTSTOODEEP	-13	/*!< aliases nested too deeply */
+#define POPT_ERROR_BADQUOTE	-15	/*!< error in paramter quoting */
+#define POPT_ERROR_ERRNO	-16	/*!< errno set, use strerror(errno) */
+#define POPT_ERROR_BADNUMBER	-17	/*!< invalid numeric value */
+#define POPT_ERROR_OVERFLOW	-18	/*!< number too large or too small */
+#define	POPT_ERROR_BADOPERATION	-19	/*!< mutually exclusive logical operations requested */
+#define	POPT_ERROR_NULLARG	-20	/*!< opt->arg should not be NULL */
+#define	POPT_ERROR_MALLOC	-21	/*!< memory allocation failed */
+/*@}*/
+
+/** \ingroup popt
+ * \name poptBadOption() flags
+ */
+/*@{*/
+#define POPT_BADOPTION_NOALIAS  (1 << 0)  /*!< don't go into an alias */
+/*@}*/
+
+/** \ingroup popt
+ * \name poptGetContext() flags
+ */
+/*@{*/
+#define POPT_CONTEXT_NO_EXEC	(1 << 0)  /*!< ignore exec expansions */
+#define POPT_CONTEXT_KEEP_FIRST	(1 << 1)  /*!< pay attention to argv[0] */
+#define POPT_CONTEXT_POSIXMEHARDER (1 << 2) /*!< options can't follow args */
+#define POPT_CONTEXT_ARG_OPTS	(1 << 4) /*!< return args as options with value 0 */
+/*@}*/
+
+/** \ingroup popt
+ */
+struct poptOption {
+/*@observer@*/ /*@null@*/ const char * longName; /*!< may be NULL */
+    char shortName;			/*!< may be '\0' */
+    int argInfo;
+/*@shared@*/ /*@null@*/ void * arg;	/*!< depends on argInfo */
+    int val;			/*!< 0 means don't return, just update flag */
+/*@observer@*/ /*@null@*/ const char * descrip;	/*!< description for autohelp -- may be NULL */
+/*@observer@*/ /*@null@*/ const char * argDescrip; /*!< argument description for autohelp */
+};
+
+/** \ingroup popt
+ * A popt alias argument for poptAddAlias().
+ */
+struct poptAlias {
+/*@owned@*/ /*@null@*/ const char * longName;	/*!< may be NULL */
+    char shortName;		/*!< may be '\0' */
+    int argc;
+/*@owned@*/ const char ** argv;	/*!< must be free()able */
+};
+
+/** \ingroup popt
+ * A popt alias or exec argument for poptAddItem().
+ */
+/*@-exporttype@*/
+typedef struct poptItem_s {
+    struct poptOption option;	/*!< alias/exec name(s) and description. */
+    int argc;			/*!< (alias) no. of args. */
+/*@owned@*/ const char ** argv;	/*!< (alias) args, must be free()able. */
+} * poptItem;
+/*@=exporttype@*/
+
+/** \ingroup popt
+ * \name Auto-generated help/usage
+ */
+/*@{*/
+
+/**
+ * Empty table marker to enable displaying popt alias/exec options.
+ */
+/*@-exportvar@*/
+/*@unchecked@*/ /*@observer@*/
+extern struct poptOption poptAliasOptions[];
+/*@=exportvar@*/
+#define POPT_AUTOALIAS { NULL, '\0', POPT_ARG_INCLUDE_TABLE, poptAliasOptions, \
+			0, "Options implemented via popt alias/exec:", NULL },
+
+/**
+ * Auto help table options.
+ */
+/*@-exportvar@*/
+/*@unchecked@*/ /*@observer@*/
+extern struct poptOption poptHelpOptions[];
+/*@=exportvar@*/
+#define POPT_AUTOHELP { NULL, '\0', POPT_ARG_INCLUDE_TABLE, poptHelpOptions, \
+			0, "Help options:", NULL },
+
+#define POPT_TABLEEND { NULL, '\0', 0, 0, 0, NULL, NULL }
+/*@}*/
+
+/** \ingroup popt
+ */
+/*@-exporttype@*/
+typedef /*@abstract@*/ struct poptContext_s * poptContext;
+/*@=exporttype@*/
+
+/** \ingroup popt
+ */
+#ifndef __cplusplus
+/*@-exporttype -typeuse@*/
+typedef struct poptOption * poptOption;
+/*@=exporttype =typeuse@*/
+#endif
+
+/*@-exportconst@*/
+enum poptCallbackReason {
+    POPT_CALLBACK_REASON_PRE	= 0, 
+    POPT_CALLBACK_REASON_POST	= 1,
+    POPT_CALLBACK_REASON_OPTION = 2
+};
+/*@=exportconst@*/
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+/*@-type@*/
+
+/** \ingroup popt
+ * Table callback prototype.
+ * @param con		context
+ * @param reason	reason for callback
+ * @param opt		option that triggered callback
+ * @param arg		@todo Document.
+ * @param data		@todo Document.
+ */
+typedef void (*poptCallbackType) (poptContext con, 
+		enum poptCallbackReason reason,
+		/*@null@*/ const struct poptOption * opt,
+		/*@null@*/ const char * arg,
+		/*@null@*/ const void * data)
+	/*@*/;
+
+/** \ingroup popt
+ * Initialize popt context.
+ * @param name
+ * @param argc		no. of arguments
+ * @param argv		argument array
+ * @param options	address of popt option table
+ * @param flags		or'd POPT_CONTEXT_* bits
+ * @return		initialized popt context
+ */
+/*@only@*/ /*@null@*/ poptContext poptGetContext(
+		/*@dependent@*/ /*@keep@*/ const char * name,
+		int argc, /*@dependent@*/ /*@keep@*/ const char ** argv,
+		/*@dependent@*/ /*@keep@*/ const struct poptOption * options,
+		int flags)
+	/*@*/;
+
+/** \ingroup popt
+ * Reinitialize popt context.
+ * @param con		context
+ */
+/*@-exportlocal@*/
+void poptResetContext(/*@null@*/poptContext con)
+	/*@modifies con @*/;
+/*@=exportlocal@*/
+
+/** \ingroup popt
+ * Return value of next option found.
+ * @param con		context
+ * @return		next option val, -1 on last item, POPT_ERROR_* on error
+ */
+int poptGetNextOpt(/*@null@*/poptContext con)
+	/*@globals fileSystem, internalState @*/
+	/*@modifies con, fileSystem, internalState @*/;
+
+/*@-redecl@*/
+/** \ingroup popt
+ * Return next option argument (if any).
+ * @param con		context
+ * @return		option argument, NULL if no more options are available
+ */
+/*@observer@*/ /*@null@*/ const char * poptGetOptArg(/*@null@*/poptContext con)
+	/*@modifies con @*/;
+
+/** \ingroup popt
+ * Return current option's argument.
+ * @param con		context
+ * @return		option argument, NULL if no more options are available
+ */
+/*@observer@*/ /*@null@*/ const char * poptGetArg(/*@null@*/poptContext con)
+	/*@modifies con @*/;
+
+/** \ingroup popt
+ * Peek at current option's argument.
+ * @param con		context
+ * @return		option argument
+ */
+/*@observer@*/ /*@null@*/ const char * poptPeekArg(/*@null@*/poptContext con)
+	/*@*/;
+
+/** \ingroup popt
+ * Return remaining arguments.
+ * @param con		context
+ * @return		argument array, terminated with NULL
+ */
+/*@observer@*/ /*@null@*/ const char ** poptGetArgs(/*@null@*/poptContext con)
+	/*@modifies con @*/;
+
+/** \ingroup popt
+ * Return the option which caused the most recent error.
+ * @param con		context
+ * @param flags
+ * @return		offending option
+ */
+/*@observer@*/ const char * poptBadOption(/*@null@*/poptContext con, int flags)
+	/*@*/;
+/*@=redecl@*/
+
+/** \ingroup popt
+ * Destroy context.
+ * @param con		context
+ * @return		NULL always
+ */
+/*@null@*/ poptContext poptFreeContext( /*@only@*/ /*@null@*/ poptContext con)
+	/*@modifies con @*/;
+
+/** \ingroup popt
+ * Add arguments to context.
+ * @param con		context
+ * @param argv		argument array, NULL terminated
+ * @return		0 on success, POPT_ERROR_OPTSTOODEEP on failure
+ */
+int poptStuffArgs(poptContext con, /*@keep@*/ const char ** argv)
+	/*@modifies con @*/;
+
+/** \ingroup popt
+ * Add alias to context.
+ * @todo Pass alias by reference, not value.
+ * @deprecated Use poptAddItem instead.
+ * @param con		context
+ * @param alias		alias to add
+ * @param flags		(unused)
+ * @return		0 on success
+ */
+/*@unused@*/
+int poptAddAlias(poptContext con, struct poptAlias alias, int flags)
+	/*@modifies con @*/;
+
+/** \ingroup popt
+ * Add alias/exec item to context.
+ * @param con		context
+ * @param newItem	alias/exec item to add
+ * @param flags		0 for alias, 1 for exec
+ * @return		0 on success
+ */
+int poptAddItem(poptContext con, poptItem newItem, int flags)
+	/*@modifies con @*/;
+
+/** \ingroup popt
+ * Read configuration file.
+ * @param con		context
+ * @param fn		file name to read
+ * @return		0 on success, POPT_ERROR_ERRNO on failure
+ */
+int poptReadConfigFile(poptContext con, const char * fn)
+	/*@globals fileSystem, internalState @*/
+	/*@modifies con->execs, con->numExecs,
+		fileSystem, internalState @*/;
+
+/** \ingroup popt
+ * Read default configuration from /etc/popt and $HOME/.popt.
+ * @param con		context
+ * @param useEnv	(unused)
+ * @return		0 on success, POPT_ERROR_ERRNO on failure
+ */
+int poptReadDefaultConfig(poptContext con, /*@unused@*/ int useEnv)
+	/*@globals fileSystem, internalState @*/
+	/*@modifies con->execs, con->numExecs,
+		fileSystem, internalState @*/;
+
+/** \ingroup popt
+ * Duplicate an argument array.
+ * @note: The argument array is malloc'd as a single area, so only argv must
+ * be free'd.
+ *
+ * @param argc		no. of arguments
+ * @param argv		argument array
+ * @retval argcPtr	address of returned no. of arguments
+ * @retval argvPtr	address of returned argument array
+ * @return		0 on success, POPT_ERROR_NOARG on failure
+ */
+int poptDupArgv(int argc, /*@null@*/ const char **argv,
+		/*@null@*/ /*@out@*/ int * argcPtr,
+		/*@null@*/ /*@out@*/ const char *** argvPtr)
+	/*@modifies *argcPtr, *argvPtr @*/;
+
+/** \ingroup popt
+ * Parse a string into an argument array.
+ * The parse allows ', ", and \ quoting, but ' is treated the same as " and
+ * both may include \ quotes.
+ * @note: The argument array is malloc'd as a single area, so only argv must
+ * be free'd.
+ *
+ * @param s		string to parse
+ * @retval argcPtr	address of returned no. of arguments
+ * @retval argvPtr	address of returned argument array
+ */
+int poptParseArgvString(const char * s,
+		/*@out@*/ int * argcPtr, /*@out@*/ const char *** argvPtr)
+	/*@modifies *argcPtr, *argvPtr @*/;
+
+/** \ingroup popt
+ * Parses an input configuration file and returns an string that is a 
+ * command line.  For use with popt.  You must free the return value when done.
+ *
+ * Given the file:
+\verbatim
+# this line is ignored
+    #   this one too
+aaa
+  bbb
+    ccc   
+bla=bla
+
+this_is   =   fdsafdas
+     bad_line=        
+  reall bad line  
+  reall bad line  = again
+5555=   55555   
+  test = with lots of spaces
+\endverbatim
+*
+* The result is:
+\verbatim
+--aaa --bbb --ccc --bla="bla" --this_is="fdsafdas" --5555="55555" --test="with lots of spaces"
+\endverbatim
+*
+* Passing this to poptParseArgvString() yields an argv of:
+\verbatim
+'--aaa'
+'--bbb' 
+'--ccc' 
+'--bla=bla' 
+'--this_is=fdsafdas' 
+'--5555=55555' 
+'--test=with lots of spaces' 
+\endverbatim
+ *
+ * @bug NULL is returned if file line is too long.
+ * @bug Silently ignores invalid lines.
+ *
+ * @param fp		file handle to read
+ * @param *argstrp	return string of options (malloc'd)
+ * @param flags		unused
+ * @return		0 on success
+ * @see			poptParseArgvString
+ */
+/*@-fcnuse@*/
+int poptConfigFileToString(FILE *fp, /*@out@*/ char ** argstrp, int flags)
+	/*@globals fileSystem @*/
+	/*@modifies *fp, *argstrp, fileSystem @*/;
+/*@=fcnuse@*/
+
+/** \ingroup popt
+ * Return formatted error string for popt failure.
+ * @param error		popt error
+ * @return		error string
+ */
+/*@-redecl@*/
+/*@observer@*/ const char *poptStrerror(const int error)
+	/*@*/;
+/*@=redecl@*/
+
+/** \ingroup popt
+ * Limit search for executables.
+ * @param con		context
+ * @param path		single path to search for executables
+ * @param allowAbsolute	absolute paths only?
+ */
+void poptSetExecPath(poptContext con, const char * path, int allowAbsolute)
+	/*@modifies con @*/;
+
+/** \ingroup popt
+ * Print detailed description of options.
+ * @param con		context
+ * @param fp		ouput file handle
+ * @param flags		(unused)
+ */
+void poptPrintHelp(poptContext con, FILE * fp, /*@unused@*/ int flags)
+	/*@globals fileSystem @*/
+	/*@modifies *fp, fileSystem @*/;
+
+/** \ingroup popt
+ * Print terse description of options.
+ * @param con		context
+ * @param fp		ouput file handle
+ * @param flags		(unused)
+ */
+void poptPrintUsage(poptContext con, FILE * fp, /*@unused@*/ int flags)
+	/*@globals fileSystem @*/
+	/*@modifies *fp, fileSystem @*/;
+
+/** \ingroup popt
+ * Provide text to replace default "[OPTION...]" in help/usage output.
+ * @param con		context
+ * @param text		replacement text
+ */
+/*@-fcnuse@*/
+void poptSetOtherOptionHelp(poptContext con, const char * text)
+	/*@modifies con @*/;
+/*@=fcnuse@*/
+
+/** \ingroup popt
+ * Return argv[0] from context.
+ * @param con		context
+ * @return		argv[0]
+ */
+/*@-redecl -fcnuse@*/
+/*@observer@*/ const char * poptGetInvocationName(poptContext con)
+	/*@*/;
+/*@=redecl =fcnuse@*/
+
+/** \ingroup popt
+ * Shuffle argv pointers to remove stripped args, returns new argc.
+ * @param con		context
+ * @param argc		no. of args
+ * @param argv		arg vector
+ * @return		new argc
+ */
+/*@-fcnuse@*/
+int poptStrippedArgv(poptContext con, int argc, char ** argv)
+	/*@modifies *argv @*/;
+/*@=fcnuse@*/
+
+/**
+ * Save a long, performing logical operation with value.
+ * @warning Alignment check may be too strict on certain platorms.
+ * @param arg		integer pointer, aligned on int boundary.
+ * @param argInfo	logical operation (see POPT_ARGFLAG_*)
+ * @param aLong		value to use
+ * @return		0 on success, POPT_ERROR_NULLARG/POPT_ERROR_BADOPERATION
+ */
+/*@-incondefs@*/
+int poptSaveLong(/*@null@*/ long * arg, int argInfo, long aLong)
+	/*@modifies *arg @*/
+	/*@requires maxSet(arg) >= 0 /\ maxRead(arg) == 0 @*/;
+/*@=incondefs@*/
+
+/**
+ * Save an integer, performing logical operation with value.
+ * @warning Alignment check may be too strict on certain platorms.
+ * @param arg		integer pointer, aligned on int boundary.
+ * @param argInfo	logical operation (see POPT_ARGFLAG_*)
+ * @param aLong		value to use
+ * @return		0 on success, POPT_ERROR_NULLARG/POPT_ERROR_BADOPERATION
+ */
+/*@-incondefs@*/
+int poptSaveInt(/*@null@*/ int * arg, int argInfo, long aLong)
+	/*@modifies *arg @*/
+	/*@requires maxSet(arg) >= 0 /\ maxRead(arg) == 0 @*/;
+/*@=incondefs@*/
+
+/*@=type@*/
+#ifdef  __cplusplus
+}
+#endif
+
+#endif
diff --git a/ctdb/lib/popt/poptconfig.c b/ctdb/lib/popt/poptconfig.c
new file mode 100644
index 0000000000..5edf916452
--- /dev/null
+++ b/ctdb/lib/popt/poptconfig.c
@@ -0,0 +1,191 @@
+/** \ingroup popt
+ * \file popt/poptconfig.c
+ */
+
+/* (C) 1998-2002 Red Hat, Inc. -- Licensing details are in the COPYING
+   file accompanying popt source distributions, available from 
+   ftp://ftp.rpm.org/pub/rpm/dist. */
+
+#include "system.h"
+#include "poptint.h"
+
+/*@-compmempass@*/	/* FIX: item->option.longName kept, not dependent. */
+static void configLine(poptContext con, char * line)
+	/*@modifies con @*/
+{
+    /*@-type@*/
+    int nameLength = strlen(con->appName);
+    /*@=type@*/
+    const char * entryType;
+    const char * opt;
+    poptItem item = alloca(sizeof(*item));
+    int i, j;
+    
+/*@-boundswrite@*/
+    memset(item, 0, sizeof(*item));
+
+    /*@-type@*/
+    if (strncmp(line, con->appName, nameLength)) return;
+    /*@=type@*/
+
+    line += nameLength;
+    if (*line == '\0' || !isspace(*line)) return;
+
+    while (*line != '\0' && isspace(*line)) line++;
+    entryType = line;
+    while (*line == '\0' || !isspace(*line)) line++;
+    *line++ = '\0';
+
+    while (*line != '\0' && isspace(*line)) line++;
+    if (*line == '\0') return;
+    opt = line;
+    while (*line == '\0' || !isspace(*line)) line++;
+    *line++ = '\0';
+
+    while (*line != '\0' && isspace(*line)) line++;
+    if (*line == '\0') return;
+
+    /*@-temptrans@*/ /* FIX: line alias is saved */
+    if (opt[0] == '-' && opt[1] == '-')
+	item->option.longName = opt + 2;
+    else if (opt[0] == '-' && opt[2] == '\0')
+	item->option.shortName = opt[1];
+    /*@=temptrans@*/
+
+    if (poptParseArgvString(line, &item->argc, &item->argv)) return;
+
+    /*@-modobserver@*/
+    item->option.argInfo = POPT_ARGFLAG_DOC_HIDDEN;
+    for (i = 0, j = 0; i < item->argc; i++, j++) {
+	const char * f;
+	if (!strncmp(item->argv[i], "--POPTdesc=", sizeof("--POPTdesc=")-1)) {
+	    f = item->argv[i] + sizeof("--POPTdesc=");
+	    if (f[0] == '$' && f[1] == '"') f++;
+	    item->option.descrip = f;
+	    item->option.argInfo &= ~POPT_ARGFLAG_DOC_HIDDEN;
+	    j--;
+	} else
+	if (!strncmp(item->argv[i], "--POPTargs=", sizeof("--POPTargs=")-1)) {
+	    f = item->argv[i] + sizeof("--POPTargs=");
+	    if (f[0] == '$' && f[1] == '"') f++;
+	    item->option.argDescrip = f;
+	    item->option.argInfo &= ~POPT_ARGFLAG_DOC_HIDDEN;
+	    item->option.argInfo |= POPT_ARG_STRING;
+	    j--;
+	} else
+	if (j != i)
+	    item->argv[j] = item->argv[i];
+    }
+    if (j != i) {
+	item->argv[j] = NULL;
+	item->argc = j;
+    }
+    /*@=modobserver@*/
+/*@=boundswrite@*/
+	
+    /*@-nullstate@*/ /* FIX: item->argv[] may be NULL */
+    if (!strcmp(entryType, "alias"))
+	(void) poptAddItem(con, item, 0);
+    else if (!strcmp(entryType, "exec"))
+	(void) poptAddItem(con, item, 1);
+    /*@=nullstate@*/
+}
+/*@=compmempass@*/
+
+int poptReadConfigFile(poptContext con, const char * fn)
+{
+    char * file;
+    const char * chptr, * end;
+    char * buf;
+/*@dependent@*/ char * dst;
+    int fd, rc;
+    off_t fileLength;
+
+    fd = open(fn, O_RDONLY);
+    if (fd < 0)
+	return (errno == ENOENT ? 0 : POPT_ERROR_ERRNO);
+
+    fileLength = lseek(fd, 0, SEEK_END);
+    if (fileLength == -1 || lseek(fd, 0, 0) == -1) {
+	rc = errno;
+	(void) close(fd);
+	/*@-mods@*/
+	errno = rc;
+	/*@=mods@*/
+	return POPT_ERROR_ERRNO;
+    }
+
+    file = alloca(fileLength + 1);
+    if (read(fd, (char *)file, fileLength) != fileLength) {
+	rc = errno;
+	(void) close(fd);
+	/*@-mods@*/
+	errno = rc;
+	/*@=mods@*/
+	return POPT_ERROR_ERRNO;
+    }
+    if (close(fd) == -1)
+	return POPT_ERROR_ERRNO;
+
+/*@-boundswrite@*/
+    dst = buf = alloca(fileLength + 1);
+
+    chptr = file;
+    end = (file + fileLength);
+    /*@-infloops@*/	/* LCL: can't detect chptr++ */
+    while (chptr < end) {
+	switch (*chptr) {
+	  case '\n':
+	    *dst = '\0';
+	    dst = buf;
+	    while (*dst && isspace(*dst)) dst++;
+	    if (*dst && *dst != '#')
+		configLine(con, dst);
+	    chptr++;
+	    /*@switchbreak@*/ break;
+	  case '\\':
+	    *dst++ = *chptr++;
+	    if (chptr < end) {
+		if (*chptr == '\n') 
+		    dst--, chptr++;	
+		    /* \ at the end of a line does not insert a \n */
+		else
+		    *dst++ = *chptr++;
+	    }
+	    /*@switchbreak@*/ break;
+	  default:
+	    *dst++ = *chptr++;
+	    /*@switchbreak@*/ break;
+	}
+    }
+    /*@=infloops@*/
+/*@=boundswrite@*/
+
+    return 0;
+}
+
+int poptReadDefaultConfig(poptContext con, /*@unused@*/ int useEnv)
+{
+    char * fn, * home;
+    int rc;
+
+    /*@-type@*/
+    if (!con->appName) return 0;
+    /*@=type@*/
+
+    rc = poptReadConfigFile(con, "/etc/popt");
+    if (rc) return rc;
+#if defined(HAVE_GETUID) && defined(HAVE_GETEUID)
+    if (getuid() != geteuid()) return 0;
+#endif
+
+    if ((home = getenv("HOME"))) {
+	fn = alloca(strlen(home) + 20);
+	strcpy(fn, home);
+	strcat(fn, "/.popt");
+	rc = poptReadConfigFile(con, fn);
+	if (rc) return rc;
+    }
+
+    return 0;
+}
diff --git a/ctdb/lib/popt/popthelp.c b/ctdb/lib/popt/popthelp.c
new file mode 100644
index 0000000000..7ae3de7fb5
--- /dev/null
+++ b/ctdb/lib/popt/popthelp.c
@@ -0,0 +1,742 @@
+/* -*- Mode: C; tab-width: 8; indent-tabs-mode: t; c-basic-offset: 4 -*- */
+
+/*@-type@*/
+/** \ingroup popt
+ * \file popt/popthelp.c
+ */
+
+/* (C) 1998-2002 Red Hat, Inc. -- Licensing details are in the COPYING
+   file accompanying popt source distributions, available from 
+   ftp://ftp.rpm.org/pub/rpm/dist. */
+
+#include "system.h"
+#include "poptint.h"
+
+/**
+ * Display arguments.
+ * @param con		context
+ * @param foo		(unused)
+ * @param key		option(s)
+ * @param arg		(unused)
+ * @param data		(unused)
+ */
+static void displayArgs(poptContext con,
+		/*@unused@*/ enum poptCallbackReason foo,
+		struct poptOption * key, 
+		/*@unused@*/ const char * arg, /*@unused@*/ void * data)
+	/*@globals fileSystem@*/
+	/*@modifies fileSystem@*/
+{
+    if (key->shortName == '?')
+	poptPrintHelp(con, stdout, 0);
+    else
+	poptPrintUsage(con, stdout, 0);
+    exit(0);
+}
+
+#ifdef	NOTYET
+/*@unchecked@*/
+static int show_option_defaults = 0;
+#endif
+
+/**
+ * Empty table marker to enable displaying popt alias/exec options.
+ */
+/*@observer@*/ /*@unchecked@*/
+struct poptOption poptAliasOptions[] = {
+    POPT_TABLEEND
+};
+
+/**
+ * Auto help table options.
+ */
+/*@-castfcnptr@*/
+/*@observer@*/ /*@unchecked@*/
+struct poptOption poptHelpOptions[] = {
+  { NULL, '\0', POPT_ARG_CALLBACK, (void *)&displayArgs, '\0', NULL, NULL },
+  { "help", '?', 0, NULL, '?', N_("Show this help message"), NULL },
+  { "usage", '\0', 0, NULL, 'u', N_("Display brief usage message"), NULL },
+#ifdef	NOTYET
+  { "defaults", '\0', POPT_ARG_NONE, &show_option_defaults, 0,
+	N_("Display option defaults in message"), NULL },
+#endif
+    POPT_TABLEEND
+} ;
+/*@=castfcnptr@*/
+
+/**
+ * @param table		option(s)
+ */
+/*@observer@*/ /*@null@*/ static const char *const
+getTableTranslationDomain(/*@null@*/ const struct poptOption *table)
+	/*@*/
+{
+    const struct poptOption *opt;
+
+    if (table != NULL)
+    for (opt = table; opt->longName || opt->shortName || opt->arg; opt++) {
+	if (opt->argInfo == POPT_ARG_INTL_DOMAIN)
+	    return opt->arg;
+    }
+    return NULL;
+}
+
+/**
+ * @param opt		option(s)
+ * @param translation_domain	translation domain
+ */
+/*@observer@*/ /*@null@*/ static const char *const
+getArgDescrip(const struct poptOption * opt,
+		/*@-paramuse@*/ /* FIX: i18n macros disabled with lclint */
+		/*@null@*/ const char * translation_domain)
+		/*@=paramuse@*/
+	/*@*/
+{
+    if (!(opt->argInfo & POPT_ARG_MASK)) return NULL;
+
+    if (opt == (poptHelpOptions + 1) || opt == (poptHelpOptions + 2))
+	if (opt->argDescrip) return POPT_(opt->argDescrip);
+
+    if (opt->argDescrip) return D_(translation_domain, opt->argDescrip);
+
+    switch (opt->argInfo & POPT_ARG_MASK) {
+    case POPT_ARG_NONE:		return POPT_("NONE");
+#ifdef	DYING
+    case POPT_ARG_VAL:		return POPT_("VAL");
+#else
+    case POPT_ARG_VAL:		return NULL;
+#endif
+    case POPT_ARG_INT:		return POPT_("INT");
+    case POPT_ARG_LONG:		return POPT_("LONG");
+    case POPT_ARG_STRING:	return POPT_("STRING");
+    case POPT_ARG_FLOAT:	return POPT_("FLOAT");
+    case POPT_ARG_DOUBLE:	return POPT_("DOUBLE");
+    default:			return POPT_("ARG");
+    }
+}
+
+/**
+ * Display default value for an option.
+ * @param lineLength
+ * @param opt		option(s)
+ * @param translation_domain	translation domain
+ * @return
+ */
+static /*@only@*/ /*@null@*/ char *
+singleOptionDefaultValue(int lineLength,
+		const struct poptOption * opt,
+		/*@-paramuse@*/ /* FIX: i18n macros disabled with lclint */
+		/*@null@*/ const char * translation_domain)
+		/*@=paramuse@*/
+	/*@*/
+{
+    const char * defstr = D_(translation_domain, "default");
+    char * le = malloc(4*lineLength + 1);
+    char * l = le;
+
+    if (le == NULL) return NULL;	/* XXX can't happen */
+/*@-boundswrite@*/
+    *le = '\0';
+    *le++ = '(';
+    strcpy(le, defstr);	le += strlen(le);
+    *le++ = ':';
+    *le++ = ' ';
+    if (opt->arg)	/* XXX programmer error */
+    switch (opt->argInfo & POPT_ARG_MASK) {
+    case POPT_ARG_VAL:
+    case POPT_ARG_INT:
+    {	long aLong = *((int *)opt->arg);
+	le += sprintf(le, "%ld", aLong);
+    }	break;
+    case POPT_ARG_LONG:
+    {	long aLong = *((long *)opt->arg);
+	le += sprintf(le, "%ld", aLong);
+    }	break;
+    case POPT_ARG_FLOAT:
+    {	double aDouble = *((float *)opt->arg);
+	le += sprintf(le, "%g", aDouble);
+    }	break;
+    case POPT_ARG_DOUBLE:
+    {	double aDouble = *((double *)opt->arg);
+	le += sprintf(le, "%g", aDouble);
+    }	break;
+    case POPT_ARG_STRING:
+    {	const char * s = *(const char **)opt->arg;
+	if (s == NULL) {
+	    strcpy(le, "null");	le += strlen(le);
+	} else {
+	    size_t slen = 4*lineLength - (le - l) - sizeof("\"...\")");
+	    *le++ = '"';
+	    strncpy(le, s, slen); le[slen] = '\0'; le += strlen(le);	
+	    if (slen < strlen(s)) {
+		strcpy(le, "...");	le += strlen(le);
+	    }
+	    *le++ = '"';
+	}
+    }	break;
+    case POPT_ARG_NONE:
+    default:
+	l = _free(l);
+	return NULL;
+	/*@notreached@*/ break;
+    }
+    *le++ = ')';
+    *le = '\0';
+/*@=boundswrite@*/
+
+    return l;
+}
+
+/**
+ * Display help text for an option.
+ * @param fp		output file handle
+ * @param maxLeftCol
+ * @param opt		option(s)
+ * @param translation_domain	translation domain
+ */
+static void singleOptionHelp(FILE * fp, int maxLeftCol, 
+		const struct poptOption * opt,
+		/*@null@*/ const char * translation_domain)
+	/*@globals fileSystem @*/
+	/*@modifies *fp, fileSystem @*/
+{
+    int indentLength = maxLeftCol + 5;
+    int lineLength = 79 - indentLength;
+    const char * help = D_(translation_domain, opt->descrip);
+    const char * argDescrip = getArgDescrip(opt, translation_domain);
+    int helpLength;
+    char * defs = NULL;
+    char * left;
+    int nb = maxLeftCol + 1;
+
+    /* Make sure there's more than enough room in target buffer. */
+    if (opt->longName)	nb += strlen(opt->longName);
+    if (argDescrip)	nb += strlen(argDescrip);
+
+/*@-boundswrite@*/
+    left = malloc(nb);
+    if (left == NULL) return;	/* XXX can't happen */
+    left[0] = '\0';
+    left[maxLeftCol] = '\0';
+
+    if (opt->longName && opt->shortName)
+	sprintf(left, "-%c, %s%s", opt->shortName,
+		((opt->argInfo & POPT_ARGFLAG_ONEDASH) ? "-" : "--"),
+		opt->longName);
+    else if (opt->shortName != '\0') 
+	sprintf(left, "-%c", opt->shortName);
+    else if (opt->longName)
+	sprintf(left, "%s%s",
+		((opt->argInfo & POPT_ARGFLAG_ONEDASH) ? "-" : "--"),
+		opt->longName);
+    if (!*left) goto out;
+
+    if (argDescrip) {
+	char * le = left + strlen(left);
+
+	if (opt->argInfo & POPT_ARGFLAG_OPTIONAL)
+	    *le++ = '[';
+
+	/* Choose type of output */
+	/*@-branchstate@*/
+	if (opt->argInfo & POPT_ARGFLAG_SHOW_DEFAULT) {
+	    defs = singleOptionDefaultValue(lineLength, opt, translation_domain);
+	    if (defs) {
+		char * t = malloc((help ? strlen(help) : 0) +
+				strlen(defs) + sizeof(" "));
+		if (t) {
+		    char * te = t;
+		    *te = '\0';
+		    if (help) {
+			strcpy(te, help);	te += strlen(te);
+		    }
+		    *te++ = ' ';
+		    strcpy(te, defs);
+		    defs = _free(defs);
+		}
+		defs = t;
+	    }
+	}
+	/*@=branchstate@*/
+
+	if (opt->argDescrip == NULL) {
+	    switch (opt->argInfo & POPT_ARG_MASK) {
+	    case POPT_ARG_NONE:
+		break;
+	    case POPT_ARG_VAL:
+#ifdef	NOTNOW	/* XXX pug ugly nerdy output */
+	    {	long aLong = opt->val;
+		int ops = (opt->argInfo & POPT_ARGFLAG_LOGICALOPS);
+		int negate = (opt->argInfo & POPT_ARGFLAG_NOT);
+
+		/* Don't bother displaying typical values */
+		if (!ops && (aLong == 0L || aLong == 1L || aLong == -1L))
+		    break;
+		*le++ = '[';
+		switch (ops) {
+		case POPT_ARGFLAG_OR:
+		    *le++ = '|';
+		    /*@innerbreak@*/ break;
+		case POPT_ARGFLAG_AND:
+		    *le++ = '&';
+		    /*@innerbreak@*/ break;
+		case POPT_ARGFLAG_XOR:
+		    *le++ = '^';
+		    /*@innerbreak@*/ break;
+		default:
+		    /*@innerbreak@*/ break;
+		}
+		*le++ = '=';
+		if (negate) *le++ = '~';
+		/*@-formatconst@*/
+		le += sprintf(le, (ops ? "0x%lx" : "%ld"), aLong);
+		/*@=formatconst@*/
+		*le++ = ']';
+	    }
+#endif
+		break;
+	    case POPT_ARG_INT:
+	    case POPT_ARG_LONG:
+	    case POPT_ARG_FLOAT:
+	    case POPT_ARG_DOUBLE:
+	    case POPT_ARG_STRING:
+		*le++ = '=';
+		strcpy(le, argDescrip);		le += strlen(le);
+		break;
+	    default:
+		break;
+	    }
+	} else {
+	    *le++ = '=';
+	    strcpy(le, argDescrip);		le += strlen(le);
+	}
+	if (opt->argInfo & POPT_ARGFLAG_OPTIONAL)
+	    *le++ = ']';
+	*le = '\0';
+    }
+/*@=boundswrite@*/
+
+    if (help)
+	fprintf(fp,"  %-*s   ", maxLeftCol, left);
+    else {
+	fprintf(fp,"  %s\n", left); 
+	goto out;
+    }
+
+    left = _free(left);
+    if (defs) {
+	help = defs; defs = NULL;
+    }
+
+    helpLength = strlen(help);
+/*@-boundsread@*/
+    while (helpLength > lineLength) {
+	const char * ch;
+	char format[16];
+
+	ch = help + lineLength - 1;
+	while (ch > help && !isspace(*ch)) ch--;
+	if (ch == help) break;		/* give up */
+	while (ch > (help + 1) && isspace(*ch)) ch--;
+	ch++;
+
+	sprintf(format, "%%.%ds\n%%%ds", (int) (ch - help), indentLength);
+	/*@-formatconst@*/
+	fprintf(fp, format, help, " ");
+	/*@=formatconst@*/
+	help = ch;
+	while (isspace(*help) && *help) help++;
+	helpLength = strlen(help);
+    }
+/*@=boundsread@*/
+
+    if (helpLength) fprintf(fp, "%s\n", help);
+
+out:
+    /*@-dependenttrans@*/
+    defs = _free(defs);
+    /*@=dependenttrans@*/
+    left = _free(left);
+}
+
+/**
+ * @param opt		option(s)
+ * @param translation_domain	translation domain
+ */
+static int maxArgWidth(const struct poptOption * opt,
+		       /*@null@*/ const char * translation_domain)
+	/*@*/
+{
+    int max = 0;
+    int len = 0;
+    const char * s;
+    
+    if (opt != NULL)
+    while (opt->longName || opt->shortName || opt->arg) {
+	if ((opt->argInfo & POPT_ARG_MASK) == POPT_ARG_INCLUDE_TABLE) {
+	    if (opt->arg)	/* XXX program error */
+	    len = maxArgWidth(opt->arg, translation_domain);
+	    if (len > max) max = len;
+	} else if (!(opt->argInfo & POPT_ARGFLAG_DOC_HIDDEN)) {
+	    len = sizeof("  ")-1;
+	    if (opt->shortName != '\0') len += sizeof("-X")-1;
+	    if (opt->shortName != '\0' && opt->longName) len += sizeof(", ")-1;
+	    if (opt->longName) {
+		len += ((opt->argInfo & POPT_ARGFLAG_ONEDASH)
+			? sizeof("-")-1 : sizeof("--")-1);
+		len += strlen(opt->longName);
+	    }
+
+	    s = getArgDescrip(opt, translation_domain);
+	    if (s)
+		len += sizeof("=")-1 + strlen(s);
+	    if (opt->argInfo & POPT_ARGFLAG_OPTIONAL) len += sizeof("[]")-1;
+	    if (len > max) max = len;
+	}
+
+	opt++;
+    }
+    
+    return max;
+}
+
+/**
+ * Display popt alias and exec help.
+ * @param fp		output file handle
+ * @param items		alias/exec array
+ * @param nitems	no. of alias/exec entries
+ * @param left
+ * @param translation_domain	translation domain
+ */
+static void itemHelp(FILE * fp,
+		/*@null@*/ poptItem items, int nitems, int left,
+		/*@null@*/ const char * translation_domain)
+	/*@globals fileSystem @*/
+	/*@modifies *fp, fileSystem @*/
+{
+    poptItem item;
+    int i;
+
+    if (items != NULL)
+    for (i = 0, item = items; i < nitems; i++, item++) {
+	const struct poptOption * opt;
+	opt = &item->option;
+	if ((opt->longName || opt->shortName) && 
+	    !(opt->argInfo & POPT_ARGFLAG_DOC_HIDDEN))
+	    singleOptionHelp(fp, left, opt, translation_domain);
+    }
+}
+
+/**
+ * Display help text for a table of options.
+ * @param con		context
+ * @param fp		output file handle
+ * @param table		option(s)
+ * @param left
+ * @param translation_domain	translation domain
+ */
+static void singleTableHelp(poptContext con, FILE * fp,
+		/*@null@*/ const struct poptOption * table, int left,
+		/*@null@*/ const char * translation_domain)
+	/*@globals fileSystem @*/
+	/*@modifies *fp, fileSystem @*/
+{
+    const struct poptOption * opt;
+    const char *sub_transdom;
+
+    if (table == poptAliasOptions) {
+	itemHelp(fp, con->aliases, con->numAliases, left, NULL);
+	itemHelp(fp, con->execs, con->numExecs, left, NULL);
+	return;
+    }
+
+    if (table != NULL)
+    for (opt = table; (opt->longName || opt->shortName || opt->arg); opt++) {
+	if ((opt->longName || opt->shortName) && 
+	    !(opt->argInfo & POPT_ARGFLAG_DOC_HIDDEN))
+	    singleOptionHelp(fp, left, opt, translation_domain);
+    }
+
+    if (table != NULL)
+    for (opt = table; (opt->longName || opt->shortName || opt->arg); opt++) {
+	if ((opt->argInfo & POPT_ARG_MASK) != POPT_ARG_INCLUDE_TABLE)
+	    continue;
+	sub_transdom = getTableTranslationDomain(opt->arg);
+	if (sub_transdom == NULL)
+	    sub_transdom = translation_domain;
+	    
+	if (opt->descrip)
+	    fprintf(fp, "\n%s\n", D_(sub_transdom, opt->descrip));
+
+	singleTableHelp(con, fp, opt->arg, left, sub_transdom);
+    }
+}
+
+/**
+ * @param con		context
+ * @param fp		output file handle
+ */
+static int showHelpIntro(poptContext con, FILE * fp)
+	/*@globals fileSystem @*/
+	/*@modifies *fp, fileSystem @*/
+{
+    int len = 6;
+    const char * fn;
+
+    fprintf(fp, POPT_("Usage:"));
+    if (!(con->flags & POPT_CONTEXT_KEEP_FIRST)) {
+/*@-boundsread@*/
+	/*@-nullderef@*/	/* LCL: wazzup? */
+	fn = con->optionStack->argv[0];
+	/*@=nullderef@*/
+/*@=boundsread@*/
+	if (fn == NULL) return len;
+	if (strchr(fn, '/')) fn = strrchr(fn, '/') + 1;
+	fprintf(fp, " %s", fn);
+	len += strlen(fn) + 1;
+    }
+
+    return len;
+}
+
+void poptPrintHelp(poptContext con, FILE * fp, /*@unused@*/ int flags)
+{
+    int leftColWidth;
+
+    (void) showHelpIntro(con, fp);
+    if (con->otherHelp)
+	fprintf(fp, " %s\n", con->otherHelp);
+    else
+	fprintf(fp, " %s\n", POPT_("[OPTION...]"));
+
+    leftColWidth = maxArgWidth(con->options, NULL);
+    singleTableHelp(con, fp, con->options, leftColWidth, NULL);
+}
+
+/**
+ * @param fp		output file handle
+ * @param cursor
+ * @param opt		option(s)
+ * @param translation_domain	translation domain
+ */
+static int singleOptionUsage(FILE * fp, int cursor, 
+		const struct poptOption * opt,
+		/*@null@*/ const char *translation_domain)
+	/*@globals fileSystem @*/
+	/*@modifies *fp, fileSystem @*/
+{
+    int len = 4;
+    char shortStr[2] = { '\0', '\0' };
+    const char * item = shortStr;
+    const char * argDescrip = getArgDescrip(opt, translation_domain);
+
+    if (opt->shortName != '\0' && opt->longName != NULL) {
+	len += 2;
+	if (!(opt->argInfo & POPT_ARGFLAG_ONEDASH)) len++;
+	len += strlen(opt->longName);
+    } else if (opt->shortName != '\0') {
+	len++;
+	shortStr[0] = opt->shortName;
+	shortStr[1] = '\0';
+    } else if (opt->longName) {
+	len += strlen(opt->longName);
+	if (!(opt->argInfo & POPT_ARGFLAG_ONEDASH)) len++;
+	item = opt->longName;
+    }
+
+    if (len == 4) return cursor;
+
+    if (argDescrip) 
+	len += strlen(argDescrip) + 1;
+
+    if ((cursor + len) > 79) {
+	fprintf(fp, "\n       ");
+	cursor = 7;
+    } 
+
+    if (opt->longName && opt->shortName) {
+	fprintf(fp, " [-%c|-%s%s%s%s]",
+	    opt->shortName, ((opt->argInfo & POPT_ARGFLAG_ONEDASH) ? "" : "-"),
+	    opt->longName,
+	    (argDescrip ? " " : ""),
+	    (argDescrip ? argDescrip : ""));
+    } else {
+	fprintf(fp, " [-%s%s%s%s]",
+	    ((opt->shortName || (opt->argInfo & POPT_ARGFLAG_ONEDASH)) ? "" : "-"),
+	    item,
+	    (argDescrip ? (opt->shortName != '\0' ? " " : "=") : ""),
+	    (argDescrip ? argDescrip : ""));
+    }
+
+    return cursor + len + 1;
+}
+
+/**
+ * Display popt alias and exec usage.
+ * @param fp		output file handle
+ * @param cursor
+ * @param item		alias/exec array
+ * @param nitems	no. of ara/exec entries
+ * @param translation_domain	translation domain
+ */
+static int itemUsage(FILE * fp, int cursor, poptItem item, int nitems,
+		/*@null@*/ const char * translation_domain)
+	/*@globals fileSystem @*/
+	/*@modifies *fp, fileSystem @*/
+{
+    int i;
+
+    /*@-branchstate@*/		/* FIX: W2DO? */
+    if (item != NULL)
+    for (i = 0; i < nitems; i++, item++) {
+	const struct poptOption * opt;
+	opt = &item->option;
+        if ((opt->argInfo & POPT_ARG_MASK) == POPT_ARG_INTL_DOMAIN) {
+	    translation_domain = (const char *)opt->arg;
+	} else if ((opt->longName || opt->shortName) &&
+		 !(opt->argInfo & POPT_ARGFLAG_DOC_HIDDEN)) {
+	    cursor = singleOptionUsage(fp, cursor, opt, translation_domain);
+	}
+    }
+    /*@=branchstate@*/
+
+    return cursor;
+}
+
+/**
+ * Keep track of option tables already processed.
+ */
+typedef struct poptDone_s {
+    int nopts;
+    int maxopts;
+    const void ** opts;
+} * poptDone;
+
+/**
+ * Display usage text for a table of options.
+ * @param con		context
+ * @param fp		output file handle
+ * @param cursor
+ * @param opt		option(s)
+ * @param translation_domain	translation domain
+ * @param done		tables already processed
+ * @return
+ */
+static int singleTableUsage(poptContext con, FILE * fp, int cursor,
+		/*@null@*/ const struct poptOption * opt,
+		/*@null@*/ const char * translation_domain,
+		/*@null@*/ poptDone done)
+	/*@globals fileSystem @*/
+	/*@modifies *fp, done, fileSystem @*/
+{
+    /*@-branchstate@*/		/* FIX: W2DO? */
+    if (opt != NULL)
+    for (; (opt->longName || opt->shortName || opt->arg) ; opt++) {
+        if ((opt->argInfo & POPT_ARG_MASK) == POPT_ARG_INTL_DOMAIN) {
+	    translation_domain = (const char *)opt->arg;
+	} else if ((opt->argInfo & POPT_ARG_MASK) == POPT_ARG_INCLUDE_TABLE) {
+	    if (done) {
+		int i = 0;
+		for (i = 0; i < done->nopts; i++) {
+/*@-boundsread@*/
+		    const void * that = done->opts[i];
+/*@=boundsread@*/
+		    if (that == NULL || that != opt->arg)
+			/*@innercontinue@*/ continue;
+		    /*@innerbreak@*/ break;
+		}
+		/* Skip if this table has already been processed. */
+		if (opt->arg == NULL || i < done->nopts)
+		    continue;
+/*@-boundswrite@*/
+		if (done->nopts < done->maxopts)
+		    done->opts[done->nopts++] = (const void *) opt->arg;
+/*@=boundswrite@*/
+	    }
+	    cursor = singleTableUsage(con, fp, cursor, opt->arg,
+			translation_domain, done);
+	} else if ((opt->longName || opt->shortName) &&
+		 !(opt->argInfo & POPT_ARGFLAG_DOC_HIDDEN)) {
+	    cursor = singleOptionUsage(fp, cursor, opt, translation_domain);
+	}
+    }
+    /*@=branchstate@*/
+
+    return cursor;
+}
+
+/**
+ * Return concatenated short options for display.
+ * @todo Sub-tables should be recursed.
+ * @param opt		option(s)
+ * @param fp		output file handle
+ * @retval str		concatenation of short options
+ * @return		length of display string
+ */
+static int showShortOptions(const struct poptOption * opt, FILE * fp,
+		/*@null@*/ char * str)
+	/*@globals fileSystem @*/
+	/*@modifies *str, *fp, fileSystem @*/
+{
+    char * s = alloca(300);	/* larger than the ascii set */
+
+    s[0] = '\0';
+    /*@-branchstate@*/		/* FIX: W2DO? */
+    if (str == NULL) {
+	memset(s, 0, sizeof(s));
+	str = s;
+    }
+    /*@=branchstate@*/
+
+/*@-boundswrite@*/
+    if (opt != NULL)
+    for (; (opt->longName || opt->shortName || opt->arg); opt++) {
+	if (opt->shortName && !(opt->argInfo & POPT_ARG_MASK))
+	    str[strlen(str)] = opt->shortName;
+	else if ((opt->argInfo & POPT_ARG_MASK) == POPT_ARG_INCLUDE_TABLE)
+	    if (opt->arg)	/* XXX program error */
+		(void) showShortOptions(opt->arg, fp, str);
+    } 
+/*@=boundswrite@*/
+
+    if (s != str || *s != '\0')
+	return 0;
+
+    fprintf(fp, " [-%s]", s);
+    return strlen(s) + 4;
+}
+
+void poptPrintUsage(poptContext con, FILE * fp, /*@unused@*/ int flags)
+{
+    poptDone done = memset(alloca(sizeof(*done)), 0, sizeof(*done));
+    int cursor;
+
+    done->nopts = 0;
+    done->maxopts = 64;
+    cursor = done->maxopts * sizeof(*done->opts);
+/*@-boundswrite@*/
+    done->opts = memset(alloca(cursor), 0, cursor);
+    done->opts[done->nopts++] = (const void *) con->options;
+/*@=boundswrite@*/
+
+    cursor = showHelpIntro(con, fp);
+    cursor += showShortOptions(con->options, fp, NULL);
+    cursor = singleTableUsage(con, fp, cursor, con->options, NULL, done);
+    cursor = itemUsage(fp, cursor, con->aliases, con->numAliases, NULL);
+    cursor = itemUsage(fp, cursor, con->execs, con->numExecs, NULL);
+
+    if (con->otherHelp) {
+	cursor += strlen(con->otherHelp) + 1;
+	if (cursor > 79) fprintf(fp, "\n       ");
+	fprintf(fp, " %s", con->otherHelp);
+    }
+
+    fprintf(fp, "\n");
+}
+
+void poptSetOtherOptionHelp(poptContext con, const char * text)
+{
+    con->otherHelp = _free(con->otherHelp);
+    con->otherHelp = xstrdup(text);
+}
+/*@=type@*/
diff --git a/ctdb/lib/popt/poptint.h b/ctdb/lib/popt/poptint.h
new file mode 100644
index 0000000000..d0789971bd
--- /dev/null
+++ b/ctdb/lib/popt/poptint.h
@@ -0,0 +1,116 @@
+/** \ingroup popt
+ * \file popt/poptint.h
+ */
+
+/* (C) 1998-2000 Red Hat, Inc. -- Licensing details are in the COPYING
+   file accompanying popt source distributions, available from 
+   ftp://ftp.rpm.org/pub/rpm/dist. */
+
+#ifndef H_POPTINT
+#define H_POPTINT
+
+/**
+ * Wrapper to free(3), hides const compilation noise, permit NULL, return NULL.
+ * @param p		memory to free
+ * @retval		NULL always
+ */
+/*@unused@*/ static inline /*@null@*/ void *
+_free(/*@only@*/ /*@null@*/ const void * p)
+	/*@modifies p @*/
+{
+    if (p != NULL)	free((void *)(intptr_t)p);
+    return NULL;
+}
+
+/* Bit mask macros. */
+/*@-exporttype -redef @*/
+typedef	unsigned int __pbm_bits;
+/*@=exporttype =redef @*/
+#define	__PBM_NBITS		(8 * sizeof (__pbm_bits))
+#define	__PBM_IX(d)		((d) / __PBM_NBITS)
+#define __PBM_MASK(d)		((__pbm_bits) 1 << (((unsigned)(d)) % __PBM_NBITS))
+/*@-exporttype -redef @*/
+typedef struct {
+    __pbm_bits bits[1];
+} pbm_set;
+/*@=exporttype =redef @*/
+#define	__PBM_BITS(set)	((set)->bits)
+
+#define	PBM_ALLOC(d)	calloc(__PBM_IX (d) + 1, sizeof(__pbm_bits))
+#define	PBM_FREE(s)	_free(s);
+#define PBM_SET(d, s)   (__PBM_BITS (s)[__PBM_IX (d)] |= __PBM_MASK (d))
+#define PBM_CLR(d, s)   (__PBM_BITS (s)[__PBM_IX (d)] &= ~__PBM_MASK (d))
+#define PBM_ISSET(d, s) ((__PBM_BITS (s)[__PBM_IX (d)] & __PBM_MASK (d)) != 0)
+
+struct optionStackEntry {
+    int argc;
+/*@only@*/ /*@null@*/
+    const char ** argv;
+/*@only@*/ /*@null@*/
+    pbm_set * argb;
+    int next;
+/*@only@*/ /*@null@*/
+    const char * nextArg;
+/*@observer@*/ /*@null@*/
+    const char * nextCharArg;
+/*@dependent@*/ /*@null@*/
+    poptItem currAlias;
+    int stuffed;
+};
+
+struct poptContext_s {
+    struct optionStackEntry optionStack[POPT_OPTION_DEPTH];
+/*@dependent@*/
+    struct optionStackEntry * os;
+/*@owned@*/ /*@null@*/
+    const char ** leftovers;
+    int numLeftovers;
+    int nextLeftover;
+/*@keep@*/
+    const struct poptOption * options;
+    int restLeftover;
+/*@only@*/ /*@null@*/
+    const char * appName;
+/*@only@*/ /*@null@*/
+    poptItem aliases;
+    int numAliases;
+    int flags;
+/*@owned@*/ /*@null@*/
+    poptItem execs;
+    int numExecs;
+/*@only@*/ /*@null@*/
+    const char ** finalArgv;
+    int finalArgvCount;
+    int finalArgvAlloced;
+/*@dependent@*/ /*@null@*/
+    poptItem doExec;
+/*@only@*/
+    const char * execPath;
+    int execAbsolute;
+/*@only@*/
+    const char * otherHelp;
+/*@null@*/
+    pbm_set * arg_strip;
+};
+
+#ifdef HAVE_LIBINTL_H
+#include <libintl.h>
+#endif
+
+#if defined(HAVE_GETTEXT) && !defined(__LCLINT__)
+#define _(foo) gettext(foo)
+#else
+#define _(foo) foo
+#endif
+
+#if defined(HAVE_DCGETTEXT) && !defined(__LCLINT__)
+#define D_(dom, str) dgettext(dom, str)
+#define POPT_(foo) D_("popt", foo)
+#else
+#define D_(dom, str) str
+#define POPT_(foo) foo
+#endif
+
+#define N_(foo) foo
+
+#endif
diff --git a/ctdb/lib/popt/poptparse.c b/ctdb/lib/popt/poptparse.c
new file mode 100644
index 0000000000..a0dea80041
--- /dev/null
+++ b/ctdb/lib/popt/poptparse.c
@@ -0,0 +1,227 @@
+/** \ingroup popt
+ * \file popt/poptparse.c
+ */
+
+/* (C) 1998-2002 Red Hat, Inc. -- Licensing details are in the COPYING
+   file accompanying popt source distributions, available from 
+   ftp://ftp.rpm.org/pub/rpm/dist. */
+
+#include "system.h"
+
+#define POPT_ARGV_ARRAY_GROW_DELTA 5
+
+/*@-boundswrite@*/
+int poptDupArgv(int argc, const char **argv,
+		int * argcPtr, const char *** argvPtr)
+{
+    size_t nb = (argc + 1) * sizeof(*argv);
+    const char ** argv2;
+    char * dst;
+    int i;
+
+    if (argc <= 0 || argv == NULL)	/* XXX can't happen */
+	return POPT_ERROR_NOARG;
+    for (i = 0; i < argc; i++) {
+	if (argv[i] == NULL)
+	    return POPT_ERROR_NOARG;
+	nb += strlen(argv[i]) + 1;
+    }
+	
+    dst = malloc(nb);
+    if (dst == NULL)			/* XXX can't happen */
+	return POPT_ERROR_MALLOC;
+    argv2 = (void *) dst;
+    dst += (argc + 1) * sizeof(*argv);
+
+    /*@-branchstate@*/
+    for (i = 0; i < argc; i++) {
+	argv2[i] = dst;
+	dst += strlen(strcpy(dst, argv[i])) + 1;
+    }
+    /*@=branchstate@*/
+    argv2[argc] = NULL;
+
+    if (argvPtr) {
+	*argvPtr = argv2;
+    } else {
+	free(argv2);
+	argv2 = NULL;
+    }
+    if (argcPtr)
+	*argcPtr = argc;
+    return 0;
+}
+/*@=boundswrite@*/
+
+/*@-bounds@*/
+int poptParseArgvString(const char * s, int * argcPtr, const char *** argvPtr)
+{
+    const char * src;
+    char quote = '\0';
+    int argvAlloced = POPT_ARGV_ARRAY_GROW_DELTA;
+    const char ** argv = malloc(sizeof(*argv) * argvAlloced);
+    int argc = 0;
+    int buflen = strlen(s) + 1;
+    char * buf = memset(alloca(buflen), 0, buflen);
+    int rc = POPT_ERROR_MALLOC;
+
+    if (argv == NULL) return rc;
+    argv[argc] = buf;
+
+    for (src = s; *src != '\0'; src++) {
+	if (quote == *src) {
+	    quote = '\0';
+	} else if (quote != '\0') {
+	    if (*src == '\\') {
+		src++;
+		if (!*src) {
+		    rc = POPT_ERROR_BADQUOTE;
+		    goto exit;
+		}
+		if (*src != quote) *buf++ = '\\';
+	    }
+	    *buf++ = *src;
+	} else if (isspace(*src)) {
+	    if (*argv[argc] != '\0') {
+		buf++, argc++;
+		if (argc == argvAlloced) {
+		    argvAlloced += POPT_ARGV_ARRAY_GROW_DELTA;
+		    argv = realloc(argv, sizeof(*argv) * argvAlloced);
+		    if (argv == NULL) goto exit;
+		}
+		argv[argc] = buf;
+	    }
+	} else switch (*src) {
+	  case '"':
+	  case '\'':
+	    quote = *src;
+	    /*@switchbreak@*/ break;
+	  case '\\':
+	    src++;
+	    if (!*src) {
+		rc = POPT_ERROR_BADQUOTE;
+		goto exit;
+	    }
+	    /*@fallthrough@*/
+	  default:
+	    *buf++ = *src;
+	    /*@switchbreak@*/ break;
+	}
+    }
+
+    if (strlen(argv[argc])) {
+	argc++, buf++;
+    }
+
+    rc = poptDupArgv(argc, argv, argcPtr, argvPtr);
+
+exit:
+    if (argv) free(argv);
+    return rc;
+}
+/*@=bounds@*/
+
+/* still in the dev stage.
+ * return values, perhaps 1== file erro
+ * 2== line to long
+ * 3== umm.... more?
+ */
+int poptConfigFileToString(FILE *fp, char ** argstrp, /*@unused@*/ int flags)
+{
+    char line[999];
+    char * argstr;
+    char * p;
+    char * q;
+    char * x;
+    int t;
+    int argvlen = 0;
+    size_t maxlinelen = sizeof(line);
+    size_t linelen;
+    int maxargvlen = 480;
+    int linenum = 0;
+
+    *argstrp = NULL;
+
+    /*   |   this_is   =   our_line
+     *	     p             q      x
+     */
+
+    if (fp == NULL)
+	return POPT_ERROR_NULLARG;
+
+    argstr = calloc(maxargvlen, sizeof(*argstr));
+    if (argstr == NULL) return POPT_ERROR_MALLOC;
+
+    while (fgets(line, (int)maxlinelen, fp) != NULL) {
+	linenum++;
+	p = line;
+
+	/* loop until first non-space char or EOL */
+	while( *p != '\0' && isspace(*p) )
+	    p++;
+
+	linelen = strlen(p);
+	if (linelen >= maxlinelen-1)
+	    return POPT_ERROR_OVERFLOW;	/* XXX line too long */
+
+	if (*p == '\0' || *p == '\n') continue;	/* line is empty */
+	if (*p == '#') continue;		/* comment line */
+
+	q = p;
+
+	while (*q != '\0' && (!isspace(*q)) && *q != '=')
+	    q++;
+
+	if (isspace(*q)) {
+	    /* a space after the name, find next non space */
+	    *q++='\0';
+	    while( *q != '\0' && isspace((int)*q) ) q++;
+	}
+	if (*q == '\0') {
+	    /* single command line option (ie, no name=val, just name) */
+	    q[-1] = '\0';		/* kill off newline from fgets() call */
+	    argvlen += (t = q - p) + (sizeof(" --")-1);
+	    if (argvlen >= maxargvlen) {
+		maxargvlen = (t > maxargvlen) ? t*2 : maxargvlen*2;
+		argstr = realloc(argstr, maxargvlen);
+		if (argstr == NULL) return POPT_ERROR_MALLOC;
+	    }
+	    strcat(argstr, " --");
+	    strcat(argstr, p);
+	    continue;
+	}
+	if (*q != '=')
+	    continue;	/* XXX for now, silently ignore bogus line */
+		
+	/* *q is an equal sign. */
+	*q++ = '\0';
+
+	/* find next non-space letter of value */
+	while (*q != '\0' && isspace(*q))
+	    q++;
+	if (*q == '\0')
+	    continue;	/* XXX silently ignore missing value */
+
+	/* now, loop and strip all ending whitespace */
+	x = p + linelen;
+	while (isspace(*--x))
+	    *x = 0;	/* null out last char if space (including fgets() NL) */
+
+	/* rest of line accept */
+	t = x - p;
+	argvlen += t + (sizeof("' --='")-1);
+	if (argvlen >= maxargvlen) {
+	    maxargvlen = (t > maxargvlen) ? t*2 : maxargvlen*2;
+	    argstr = realloc(argstr, maxargvlen);
+	    if (argstr == NULL) return POPT_ERROR_MALLOC;
+	}
+	strcat(argstr, " --");
+	strcat(argstr, p);
+	strcat(argstr, "=\"");
+	strcat(argstr, q);
+	strcat(argstr, "\"");
+    }
+
+    *argstrp = argstr;
+    return 0;
+}
diff --git a/ctdb/lib/popt/samba.m4 b/ctdb/lib/popt/samba.m4
new file mode 100644
index 0000000000..b6e939f5a0
--- /dev/null
+++ b/ctdb/lib/popt/samba.m4
@@ -0,0 +1,10 @@
+m4_include(lib/popt/libpopt.m4)
+
+if test x"$POPT_OBJ" = "x"; then
+	SMB_EXT_LIB(LIBPOPT, [${POPT_LIBS}])
+	SMB_ENABLE(LIBPOPT,YES)
+else
+	SMB_SUBSYSTEM(LIBPOPT,
+	[lib/popt/findme.o lib/popt/popt.o lib/popt/poptconfig.o lib/popt/popthelp.o lib/popt/poptparse.o], [], [-Ilib/popt])
+fi
+
diff --git a/ctdb/lib/popt/system.h b/ctdb/lib/popt/system.h
new file mode 100644
index 0000000000..000e23d484
--- /dev/null
+++ b/ctdb/lib/popt/system.h
@@ -0,0 +1,74 @@
+#include "config.h"
+
+#if defined (__GLIBC__) && defined(__LCLINT__)
+/*@-declundef@*/
+/*@unchecked@*/
+extern __const __int32_t *__ctype_tolower;
+/*@unchecked@*/
+extern __const __int32_t *__ctype_toupper;
+/*@=declundef@*/
+#endif
+
+#include <ctype.h>
+
+#include <errno.h>
+#include <fcntl.h>
+#include <limits.h>
+
+#if HAVE_MCHECK_H 
+#include <mcheck.h>
+#endif
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#if HAVE_UNISTD_H
+#include <unistd.h>
+#endif
+
+#ifdef __NeXT
+/* access macros are not declared in non posix mode in unistd.h -
+ don't try to use posix on NeXTstep 3.3 ! */
+#include <libc.h>
+#endif
+
+#if defined(__LCLINT__)
+/*@-declundef -incondefs -redecl@*/ /* LCL: missing annotation */
+/*@only@*/ void * alloca (size_t __size)
+	/*@ensures MaxSet(result) == (__size - 1) @*/
+	/*@*/;
+/*@=declundef =incondefs =redecl@*/
+#endif
+
+/* AIX requires this to be the first thing in the file.  */ 
+#ifndef __GNUC__
+# if HAVE_ALLOCA_H
+#  include <alloca.h>
+# else
+#  ifdef _AIX
+#pragma alloca
+#  else
+#   ifndef alloca /* predefined by HP cc +Olibcalls */
+char *alloca ();
+#   endif
+#  endif
+# endif
+#elif defined(__GNUC__) && defined(__STRICT_ANSI__)
+#define alloca __builtin_alloca
+#endif
+
+/*@-redecl -redef@*/
+/*@mayexit@*/ /*@only@*/ char * xstrdup (const char *str)
+	/*@*/;
+/*@=redecl =redef@*/
+
+#if HAVE_MCHECK_H && defined(__GNUC__)
+#define	vmefail()	(fprintf(stderr, "virtual memory exhausted.\n"), exit(EXIT_FAILURE), NULL)
+#define xstrdup(_str)   (strcpy((malloc(strlen(_str)+1) ? : vmefail()), (_str)))
+#else
+#define	xstrdup(_str)	strdup(_str)
+#endif  /* HAVE_MCHECK_H && defined(__GNUC__) */
+
+
+#include "popt.h"
diff --git a/ctdb/lib/replace/.checker_innocent b/ctdb/lib/replace/.checker_innocent
new file mode 100644
index 0000000000..e619176540
--- /dev/null
+++ b/ctdb/lib/replace/.checker_innocent
@@ -0,0 +1,4 @@
+>>>MISTAKE21_create_files_6a9e68ada99a97cb
+>>>MISTAKE21_os2_delete_9b2bfa7f38711d09
+>>>MISTAKE21_os2_delete_2fcc29aaa99a97cb
+>>>SECURITY2_os2_delete_9b2bfa7f1c9396ca
diff --git a/ctdb/lib/replace/Makefile b/ctdb/lib/replace/Makefile
new file mode 100644
index 0000000000..364990153e
--- /dev/null
+++ b/ctdb/lib/replace/Makefile
@@ -0,0 +1,63 @@
+# simple makefile wrapper to run waf
+
+WAF=WAF_MAKE=1 PATH=buildtools/bin:../../buildtools/bin:$$PATH waf
+
+all:
+	$(WAF) build
+
+install:
+	$(WAF) install
+
+uninstall:
+	$(WAF) uninstall
+
+test:
+	$(WAF) test $(TEST_OPTIONS)
+
+testenv:
+	$(WAF) test --testenv $(TEST_OPTIONS)
+
+quicktest:
+	$(WAF) test --quick $(TEST_OPTIONS)
+
+dist:
+	touch .tmplock
+	WAFLOCK=.tmplock $(WAF) dist
+
+distcheck:
+	touch .tmplock
+	WAFLOCK=.tmplock $(WAF) distcheck
+
+clean:
+	$(WAF) clean
+
+distclean:
+	$(WAF) distclean
+
+reconfigure: configure
+	$(WAF) reconfigure
+
+show_waf_options:
+	$(WAF) --help
+
+# some compatibility make targets
+everything: all
+
+testsuite: all
+
+check: test
+
+torture: all
+
+# this should do an install as well, once install is finished
+installcheck: test
+
+etags:
+	$(WAF) etags
+
+ctags:
+	$(WAF) ctags
+
+bin/%:: FORCE
+	$(WAF) --targets=`basename $@`
+FORCE:
diff --git a/ctdb/lib/replace/README b/ctdb/lib/replace/README
new file mode 100644
index 0000000000..9dd4f7305f
--- /dev/null
+++ b/ctdb/lib/replace/README
@@ -0,0 +1,127 @@
+This subsystem ensures that we can always use a certain core set of 
+functions and types, that are either provided by the OS or by replacement 
+functions / definitions in this subsystem. The aim is to try to stick 
+to POSIX functions in here as much as possible. Convenience functions 
+that are available on no platform at all belong in other subsystems
+(such as LIBUTIL).
+
+The following functions are guaranteed:
+
+ftruncate
+strlcpy
+strlcat
+mktime
+rename
+initgroups
+memmove
+strdup
+setlinebuf
+vsyslog
+timegm
+setenv
+unsetenv
+strndup
+strnlen
+waitpid
+seteuid
+setegid
+asprintf
+snprintf
+vasprintf
+vsnprintf
+opendir
+readdir
+telldir
+seekdir
+clock_gettime
+closedir
+dlopen
+dlclose
+dlsym
+dlerror
+chroot
+bzero
+strerror
+errno
+mkdtemp
+mkstemp (a secure one!)
+pread
+pwrite
+chown
+lchown
+readline (the library)
+inet_ntoa
+inet_ntop
+inet_pton
+inet_aton
+strtoll
+strtoull
+socketpair
+strptime
+getaddrinfo
+freeaddrinfo
+getnameinfo
+gai_strerror
+getifaddrs
+freeifaddrs
+utime
+utimes
+dup2
+link
+readlink
+symlink
+realpath
+poll
+setproctitle
+
+Types:
+bool
+socklen_t
+uint{8,16,32,64}_t
+int{8,16,32,64}_t
+intptr_t
+sig_atomic_t
+blksize_t
+blkcnt_t
+
+Constants:
+PATH_NAME_MAX
+UINT{16,32,64}_MAX
+INT32_MAX
+RTLD_LAZY
+HOST_NAME_MAX
+UINT16_MAX
+UINT32_MAX
+UINT64_MAX
+CHAR_BIT
+
+Macros:
+va_copy
+__FUNCTION__
+__FILE__
+__LINE__
+__LINESTR__
+__location__
+__STRING
+__STRINGSTRING
+MIN
+MAX
+QSORT_CAST
+ZERO_STRUCT
+ZERO_STRUCTP
+ZERO_STRUCTPN
+ZERO_ARRAY
+ARRAY_SIZE
+PTR_DIFF
+
+Headers:
+stdint.h
+stdbool.h
+
+Optional C keywords:
+volatile
+
+Prerequisites:
+memset (for bzero)
+syslog (for vsyslog)
+mktemp (for mkstemp and mkdtemp)
diff --git a/ctdb/lib/replace/autoconf-2.60.m4 b/ctdb/lib/replace/autoconf-2.60.m4
new file mode 100644
index 0000000000..b2694fde62
--- /dev/null
+++ b/ctdb/lib/replace/autoconf-2.60.m4
@@ -0,0 +1,236 @@
+# AC_GNU_SOURCE
+# --------------
+AC_DEFUN([AC_GNU_SOURCE],
+[AH_VERBATIM([_GNU_SOURCE],
+[/* Enable GNU extensions on systems that have them.  */
+#ifndef _GNU_SOURCE
+# undef _GNU_SOURCE
+#endif])dnl
+AC_BEFORE([$0], [AC_COMPILE_IFELSE])dnl
+AC_BEFORE([$0], [AC_RUN_IFELSE])dnl
+AC_DEFINE([_GNU_SOURCE])
+])
+
+# _AC_C_STD_TRY(STANDARD, TEST-PROLOGUE, TEST-BODY, OPTION-LIST,
+#		ACTION-IF-AVAILABLE, ACTION-IF-UNAVAILABLE)
+# --------------------------------------------------------------
+# Check whether the C compiler accepts features of STANDARD (e.g `c89', `c99')
+# by trying to compile a program of TEST-PROLOGUE and TEST-BODY.  If this fails,
+# try again with each compiler option in the space-separated OPTION-LIST; if one
+# helps, append it to CC.  If eventually successful, run ACTION-IF-AVAILABLE,
+# else ACTION-IF-UNAVAILABLE.
+AC_DEFUN([_AC_C_STD_TRY],
+[AC_MSG_CHECKING([for $CC option to accept ISO ]m4_translit($1, [c], [C]))
+AC_CACHE_VAL(ac_cv_prog_cc_$1,
+[ac_cv_prog_cc_$1=no
+ac_save_CC=$CC
+AC_LANG_CONFTEST([AC_LANG_PROGRAM([$2], [$3])])
+for ac_arg in '' $4
+do
+  CC="$ac_save_CC $ac_arg"
+  _AC_COMPILE_IFELSE([], [ac_cv_prog_cc_$1=$ac_arg])
+  test "x$ac_cv_prog_cc_$1" != "xno" && break
+done
+rm -f conftest.$ac_ext
+CC=$ac_save_CC
+])# AC_CACHE_VAL
+case "x$ac_cv_prog_cc_$1" in
+  x)
+    AC_MSG_RESULT([none needed]) ;;
+  xno)
+    AC_MSG_RESULT([unsupported]) ;;
+  *)
+    CC="$CC $ac_cv_prog_cc_$1"
+    AC_MSG_RESULT([$ac_cv_prog_cc_$1]) ;;
+esac
+AS_IF([test "x$ac_cv_prog_cc_$1" != xno], [$5], [$6])
+])# _AC_C_STD_TRY
+
+# _AC_PROG_CC_C99 ([ACTION-IF-AVAILABLE], [ACTION-IF-UNAVAILABLE])
+# ----------------------------------------------------------------
+# If the C compiler is not in ISO C99 mode by default, try to add an
+# option to output variable CC to make it so.  This macro tries
+# various options that select ISO C99 on some system or another.  It
+# considers the compiler to be in ISO C99 mode if it handles mixed
+# code and declarations, _Bool, inline and restrict.
+AC_DEFUN([_AC_PROG_CC_C99],
+[_AC_C_STD_TRY([c99],
+[[#include <stdarg.h>
+#include <stdbool.h>
+#include <stdlib.h>
+#include <wchar.h>
+#include <stdio.h>
+
+struct incomplete_array
+{
+  int datasize;
+  double data[];
+};
+
+struct named_init {
+  int number;
+  const wchar_t *name;
+  double average;
+};
+
+typedef const char *ccp;
+
+static inline int
+test_restrict(ccp restrict text)
+{
+  // See if C++-style comments work.
+  // Iterate through items via the restricted pointer.
+  // Also check for declarations in for loops.
+  for (unsigned int i = 0; *(text+i) != '\0'; ++i)
+    continue;
+  return 0;
+}
+
+// Check varargs and va_copy work.
+static void
+test_varargs(const char *format, ...)
+{
+  va_list args;
+  va_start(args, format);
+  va_list args_copy;
+  va_copy(args_copy, args);
+
+  const char *str;
+  int number;
+  float fnumber;
+
+  while (*format)
+    {
+      switch (*format++)
+	{
+	case 's': // string
+	  str = va_arg(args_copy, const char *);
+	  break;
+	case 'd': // int
+	  number = va_arg(args_copy, int);
+	  break;
+	case 'f': // float
+	  fnumber = (float) va_arg(args_copy, double);
+	  break;
+	default:
+	  break;
+	}
+    }
+  va_end(args_copy);
+  va_end(args);
+}
+]],
+[[
+  // Check bool and long long datatypes.
+  _Bool success = false;
+  long long int bignum = -1234567890LL;
+  unsigned long long int ubignum = 1234567890uLL;
+
+  // Check restrict.
+  if (test_restrict("String literal") != 0)
+    success = true;
+  char *restrict newvar = "Another string";
+
+  // Check varargs.
+  test_varargs("s, d' f .", "string", 65, 34.234);
+
+  // Check incomplete arrays work.
+  struct incomplete_array *ia =
+    malloc(sizeof(struct incomplete_array) + (sizeof(double) * 10));
+  ia->datasize = 10;
+  for (int i = 0; i < ia->datasize; ++i)
+    ia->data[i] = (double) i * 1.234;
+
+  // Check named initialisers.
+  struct named_init ni = {
+    .number = 34,
+    .name = L"Test wide string",
+    .average = 543.34343,
+  };
+
+  ni.number = 58;
+
+  int dynamic_array[ni.number];
+  dynamic_array[43] = 543;
+
+  // work around unused variable warnings
+  return  bignum == 0LL || ubignum == 0uLL || newvar[0] == 'x';
+]],
+dnl Try
+dnl GCC		-std=gnu99 (unused restrictive modes: -std=c99 -std=iso9899:1999)
+dnl AIX		-qlanglvl=extc99 (unused restrictive mode: -qlanglvl=stdc99)
+dnl Intel ICC	-c99
+dnl IRIX	-c99
+dnl Solaris	(unused because it causes the compiler to assume C99 semantics for
+dnl		library functions, and this is invalid before Solaris 10: -xc99)
+dnl Tru64	-c99
+dnl with extended modes being tried first.
+[[-std=gnu99 -c99 -qlanglvl=extc99]], [$1], [$2])[]dnl
+])# _AC_PROG_CC_C99
+
+# AC_PROG_CC_C99
+# --------------
+AC_DEFUN([AC_PROG_CC_C99],
+[ AC_REQUIRE([AC_PROG_CC])dnl
+  _AC_PROG_CC_C99
+])
+
+# AC_USE_SYSTEM_EXTENSIONS
+# ------------------------
+# Enable extensions on systems that normally disable them,
+# typically due to standards-conformance issues.
+m4_ifndef([AC_USE_SYSTEM_EXTENSIONS],[
+AC_DEFUN([AC_USE_SYSTEM_EXTENSIONS],
+[AC_BEFORE([$0], [AC_COMPILE_IFELSE])dnl
+AC_BEFORE([$0], [AC_RUN_IFELSE])dnl
+
+  AC_CHECK_HEADER([minix/config.h], [MINIX=yes], [MINIX=])
+  if test "$MINIX" = yes; then
+    AC_DEFINE([_POSIX_SOURCE], [1],
+      [Define to 1 if you need to in order for `stat' and other
+       things to work.])
+    AC_DEFINE([_POSIX_1_SOURCE], [2],
+      [Define to 2 if the system does not provide POSIX.1 features
+       except with this defined.])
+    AC_DEFINE([_MINIX], [1],
+      [Define to 1 if on MINIX.])
+  fi
+
+  AH_VERBATIM([__EXTENSIONS__],
+[/* Enable extensions on AIX 3, Interix.  */
+#ifndef _ALL_SOURCE
+# undef _ALL_SOURCE
+#endif
+/* Enable GNU extensions on systems that have them.  */
+#ifndef _GNU_SOURCE
+# undef _GNU_SOURCE
+#endif
+/* Enable threading extensions on Solaris.  */
+#ifndef _POSIX_PTHREAD_SEMANTICS
+# undef _POSIX_PTHREAD_SEMANTICS
+#endif
+/* Enable extensions on HP NonStop.  */
+#ifndef _TANDEM_SOURCE
+# undef _TANDEM_SOURCE
+#endif
+/* Enable general extensions on Solaris.  */
+#ifndef __EXTENSIONS__
+# undef __EXTENSIONS__
+#endif
+])
+  AC_CACHE_CHECK([whether it is safe to define __EXTENSIONS__],
+    [ac_cv_safe_to_define___extensions__],
+    [AC_COMPILE_IFELSE(
+       [AC_LANG_PROGRAM([[
+#	  define __EXTENSIONS__ 1
+	  ]AC_INCLUDES_DEFAULT])],
+       [ac_cv_safe_to_define___extensions__=yes],
+       [ac_cv_safe_to_define___extensions__=no])])
+  test $ac_cv_safe_to_define___extensions__ = yes &&
+    AC_DEFINE([__EXTENSIONS__])
+  AC_DEFINE([_ALL_SOURCE])
+  AC_DEFINE([_GNU_SOURCE])
+  AC_DEFINE([_POSIX_PTHREAD_SEMANTICS])
+  AC_DEFINE([_TANDEM_SOURCE])
+])# AC_USE_SYSTEM_EXTENSIONS
+])
diff --git a/ctdb/lib/replace/configure b/ctdb/lib/replace/configure
new file mode 100755
index 0000000000..6a9f875511
--- /dev/null
+++ b/ctdb/lib/replace/configure
@@ -0,0 +1,21 @@
+#!/bin/sh
+
+PREVPATH=`dirname $0`
+
+if [ -f $PREVPATH/../../buildtools/bin/waf ]; then
+	WAF=../../buildtools/bin/waf
+elif [ -f $PREVPATH/buildtools/bin/waf ]; then
+	WAF=./buildtools/bin/waf
+else
+	echo "replace: Unable to find waf"
+	exit 1
+fi
+
+# using JOBS=1 gives maximum compatibility with
+# systems like AIX which have broken threading in python
+JOBS=1
+export JOBS
+
+cd . || exit 1
+$WAF configure "$@" || exit 1
+cd $PREVPATH
diff --git a/ctdb/lib/replace/crypt.c b/ctdb/lib/replace/crypt.c
new file mode 100644
index 0000000000..3a067bcc77
--- /dev/null
+++ b/ctdb/lib/replace/crypt.c
@@ -0,0 +1,770 @@
+/*
+   This bit of code was derived from the UFC-crypt package which
+   carries the following copyright 
+   
+   Modified for use by Samba by Andrew Tridgell, October 1994
+
+   Note that this routine is only faster on some machines. Under Linux 1.1.51 
+   libc 4.5.26 I actually found this routine to be slightly slower.
+
+   Under SunOS I found a huge speedup by using these routines 
+   (a factor of 20 or so)
+
+   Warning: I've had a report from Steve Kennedy <steve@gbnet.org>
+   that this crypt routine may sometimes get the wrong answer. Only
+   use UFC_CRYT if you really need it.
+
+*/
+
+#include "replace.h"
+
+#ifndef HAVE_CRYPT
+
+/*
+ * UFC-crypt: ultra fast crypt(3) implementation
+ *
+ * Copyright (C) 1991-1998, Free Software Foundation, Inc.
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 3 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Library General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, see <http://www.gnu.org/licenses/>.
+ *
+ * @(#)crypt_util.c	2.31 02/08/92
+ *
+ * Support routines
+ *
+ */
+
+
+#ifndef long32
+#define long32 int32_t
+#endif
+
+#ifndef long64
+#define long64 int64_t
+#endif
+
+#ifndef ufc_long
+#define ufc_long unsigned
+#endif
+
+#ifndef _UFC_64_
+#define _UFC_32_
+#endif
+
+/* 
+ * Permutation done once on the 56 bit 
+ *  key derived from the original 8 byte ASCII key.
+ */
+static int pc1[56] = { 
+  57, 49, 41, 33, 25, 17,  9,  1, 58, 50, 42, 34, 26, 18,
+  10,  2, 59, 51, 43, 35, 27, 19, 11,  3, 60, 52, 44, 36,
+  63, 55, 47, 39, 31, 23, 15,  7, 62, 54, 46, 38, 30, 22,
+  14,  6, 61, 53, 45, 37, 29, 21, 13,  5, 28, 20, 12,  4
+};
+
+/*
+ * How much to rotate each 28 bit half of the pc1 permutated
+ *  56 bit key before using pc2 to give the i' key
+ */
+static int rots[16] = { 
+  1, 1, 2, 2, 2, 2, 2, 2, 1, 2, 2, 2, 2, 2, 2, 1 
+};
+
+/* 
+ * Permutation giving the key 
+ * of the i' DES round 
+ */
+static int pc2[48] = { 
+  14, 17, 11, 24,  1,  5,  3, 28, 15,  6, 21, 10,
+  23, 19, 12,  4, 26,  8, 16,  7, 27, 20, 13,  2,
+  41, 52, 31, 37, 47, 55, 30, 40, 51, 45, 33, 48,
+  44, 49, 39, 56, 34, 53, 46, 42, 50, 36, 29, 32
+};
+
+/*
+ * The E expansion table which selects
+ * bits from the 32 bit intermediate result.
+ */
+static int esel[48] = { 
+  32,  1,  2,  3,  4,  5,  4,  5,  6,  7,  8,  9,
+   8,  9, 10, 11, 12, 13, 12, 13, 14, 15, 16, 17,
+  16, 17, 18, 19, 20, 21, 20, 21, 22, 23, 24, 25,
+  24, 25, 26, 27, 28, 29, 28, 29, 30, 31, 32,  1
+};
+static int e_inverse[64];
+
+/* 
+ * Permutation done on the 
+ * result of sbox lookups 
+ */
+static int perm32[32] = {
+  16,  7, 20, 21, 29, 12, 28, 17,  1, 15, 23, 26,  5, 18, 31, 10,
+  2,   8, 24, 14, 32, 27,  3,  9, 19, 13, 30,  6, 22, 11,  4, 25
+};
+
+/* 
+ * The sboxes
+ */
+static int sbox[8][4][16]= {
+        { { 14,  4, 13,  1,  2, 15, 11,  8,  3, 10,  6, 12,  5,  9,  0,  7 },
+          {  0, 15,  7,  4, 14,  2, 13,  1, 10,  6, 12, 11,  9,  5,  3,  8 },
+          {  4,  1, 14,  8, 13,  6,  2, 11, 15, 12,  9,  7,  3, 10,  5,  0 },
+          { 15, 12,  8,  2,  4,  9,  1,  7,  5, 11,  3, 14, 10,  0,  6, 13 }
+        },
+
+        { { 15,  1,  8, 14,  6, 11,  3,  4,  9,  7,  2, 13, 12,  0,  5, 10 },
+          {  3, 13,  4,  7, 15,  2,  8, 14, 12,  0,  1, 10,  6,  9, 11,  5 },
+          {  0, 14,  7, 11, 10,  4, 13,  1,  5,  8, 12,  6,  9,  3,  2, 15 },
+          { 13,  8, 10,  1,  3, 15,  4,  2, 11,  6,  7, 12,  0,  5, 14,  9 }
+        },
+
+        { { 10,  0,  9, 14,  6,  3, 15,  5,  1, 13, 12,  7, 11,  4,  2,  8 },
+          { 13,  7,  0,  9,  3,  4,  6, 10,  2,  8,  5, 14, 12, 11, 15,  1 },
+          { 13,  6,  4,  9,  8, 15,  3,  0, 11,  1,  2, 12,  5, 10, 14,  7 },
+          {  1, 10, 13,  0,  6,  9,  8,  7,  4, 15, 14,  3, 11,  5,  2, 12 }
+        },
+
+        { {  7, 13, 14,  3,  0,  6,  9, 10,  1,  2,  8,  5, 11, 12,  4, 15 },
+          { 13,  8, 11,  5,  6, 15,  0,  3,  4,  7,  2, 12,  1, 10, 14,  9 },
+          { 10,  6,  9,  0, 12, 11,  7, 13, 15,  1,  3, 14,  5,  2,  8,  4 },
+          {  3, 15,  0,  6, 10,  1, 13,  8,  9,  4,  5, 11, 12,  7,  2, 14 }
+        },
+
+        { {  2, 12,  4,  1,  7, 10, 11,  6,  8,  5,  3, 15, 13,  0, 14,  9 },
+          { 14, 11,  2, 12,  4,  7, 13,  1,  5,  0, 15, 10,  3,  9,  8,  6 },
+          {  4,  2,  1, 11, 10, 13,  7,  8, 15,  9, 12,  5,  6,  3,  0, 14 },
+          { 11,  8, 12,  7,  1, 14,  2, 13,  6, 15,  0,  9, 10,  4,  5,  3 }
+        },
+
+        { { 12,  1, 10, 15,  9,  2,  6,  8,  0, 13,  3,  4, 14,  7,  5, 11 },
+          { 10, 15,  4,  2,  7, 12,  9,  5,  6,  1, 13, 14,  0, 11,  3,  8 },
+          {  9, 14, 15,  5,  2,  8, 12,  3,  7,  0,  4, 10,  1, 13, 11,  6 },
+          {  4,  3,  2, 12,  9,  5, 15, 10, 11, 14,  1,  7,  6,  0,  8, 13 }
+        },
+
+        { {  4, 11,  2, 14, 15,  0,  8, 13,  3, 12,  9,  7,  5, 10,  6,  1 },
+          { 13,  0, 11,  7,  4,  9,  1, 10, 14,  3,  5, 12,  2, 15,  8,  6 },
+          {  1,  4, 11, 13, 12,  3,  7, 14, 10, 15,  6,  8,  0,  5,  9,  2 },
+          {  6, 11, 13,  8,  1,  4, 10,  7,  9,  5,  0, 15, 14,  2,  3, 12 }
+        },
+
+        { { 13,  2,  8,  4,  6, 15, 11,  1, 10,  9,  3, 14,  5,  0, 12,  7 },
+          {  1, 15, 13,  8, 10,  3,  7,  4, 12,  5,  6, 11,  0, 14,  9,  2 },
+          {  7, 11,  4,  1,  9, 12, 14,  2,  0,  6, 10, 13, 15,  3,  5,  8 },
+          {  2,  1, 14,  7,  4, 10,  8, 13, 15, 12,  9,  0,  3,  5,  6, 11 }
+        }
+};
+
+/* 
+ * This is the final 
+ * permutation matrix
+ */
+static int final_perm[64] = {
+  40,  8, 48, 16, 56, 24, 64, 32, 39,  7, 47, 15, 55, 23, 63, 31,
+  38,  6, 46, 14, 54, 22, 62, 30, 37,  5, 45, 13, 53, 21, 61, 29,
+  36,  4, 44, 12, 52, 20, 60, 28, 35,  3, 43, 11, 51, 19, 59, 27,
+  34,  2, 42, 10, 50, 18, 58, 26, 33,  1, 41,  9, 49, 17, 57, 25
+};
+
+/* 
+ * The 16 DES keys in BITMASK format 
+ */
+#ifdef _UFC_32_
+long32 _ufc_keytab[16][2];
+#endif
+
+#ifdef _UFC_64_
+long64 _ufc_keytab[16];
+#endif
+
+
+#define ascii_to_bin(c) ((c)>='a'?(c-59):(c)>='A'?((c)-53):(c)-'.')
+#define bin_to_ascii(c) ((c)>=38?((c)-38+'a'):(c)>=12?((c)-12+'A'):(c)+'.')
+
+/* Macro to set a bit (0..23) */
+#define BITMASK(i) ( (1<<(11-(i)%12+3)) << ((i)<12?16:0) )
+
+/*
+ * sb arrays:
+ *
+ * Workhorses of the inner loop of the DES implementation.
+ * They do sbox lookup, shifting of this  value, 32 bit
+ * permutation and E permutation for the next round.
+ *
+ * Kept in 'BITMASK' format.
+ */
+
+#ifdef _UFC_32_
+long32 _ufc_sb0[8192], _ufc_sb1[8192], _ufc_sb2[8192], _ufc_sb3[8192];
+static long32 *sb[4] = {_ufc_sb0, _ufc_sb1, _ufc_sb2, _ufc_sb3}; 
+#endif
+
+#ifdef _UFC_64_
+long64 _ufc_sb0[4096], _ufc_sb1[4096], _ufc_sb2[4096], _ufc_sb3[4096];
+static long64 *sb[4] = {_ufc_sb0, _ufc_sb1, _ufc_sb2, _ufc_sb3}; 
+#endif
+
+/* 
+ * eperm32tab: do 32 bit permutation and E selection
+ *
+ * The first index is the byte number in the 32 bit value to be permuted
+ *  -  second  -   is the value of this byte
+ *  -  third   -   selects the two 32 bit values
+ *
+ * The table is used and generated internally in init_des to speed it up
+ */
+static ufc_long eperm32tab[4][256][2];
+
+/* 
+ * do_pc1: permform pc1 permutation in the key schedule generation.
+ *
+ * The first   index is the byte number in the 8 byte ASCII key
+ *  -  second    -      -    the two 28 bits halfs of the result
+ *  -  third     -   selects the 7 bits actually used of each byte
+ *
+ * The result is kept with 28 bit per 32 bit with the 4 most significant
+ * bits zero.
+ */
+static ufc_long do_pc1[8][2][128];
+
+/*
+ * do_pc2: permform pc2 permutation in the key schedule generation.
+ *
+ * The first   index is the septet number in the two 28 bit intermediate values
+ *  -  second    -    -  -  septet values
+ *
+ * Knowledge of the structure of the pc2 permutation is used.
+ *
+ * The result is kept with 28 bit per 32 bit with the 4 most significant
+ * bits zero.
+ */
+static ufc_long do_pc2[8][128];
+
+/*
+ * efp: undo an extra e selection and do final
+ *      permutation giving the DES result.
+ * 
+ *      Invoked 6 bit a time on two 48 bit values
+ *      giving two 32 bit longs.
+ */
+static ufc_long efp[16][64][2];
+
+static unsigned char bytemask[8]  = {
+  0x80, 0x40, 0x20, 0x10, 0x08, 0x04, 0x02, 0x01
+};
+
+static ufc_long longmask[32] = {
+  0x80000000, 0x40000000, 0x20000000, 0x10000000,
+  0x08000000, 0x04000000, 0x02000000, 0x01000000,
+  0x00800000, 0x00400000, 0x00200000, 0x00100000,
+  0x00080000, 0x00040000, 0x00020000, 0x00010000,
+  0x00008000, 0x00004000, 0x00002000, 0x00001000,
+  0x00000800, 0x00000400, 0x00000200, 0x00000100,
+  0x00000080, 0x00000040, 0x00000020, 0x00000010,
+  0x00000008, 0x00000004, 0x00000002, 0x00000001
+};
+
+
+/*
+ * Silly rewrite of 'bzero'. I do so
+ * because some machines don't have
+ * bzero and some don't have memset.
+ */
+
+static void clearmem(char *start, int cnt)
+  { while(cnt--)
+      *start++ = '\0';
+  }
+
+static int initialized = 0;
+
+/* lookup a 6 bit value in sbox */
+
+#define s_lookup(i,s) sbox[(i)][(((s)>>4) & 0x2)|((s) & 0x1)][((s)>>1) & 0xf];
+
+/*
+ * Initialize unit - may be invoked directly
+ * by fcrypt users.
+ */
+
+static void ufc_init_des(void)
+  { int comes_from_bit;
+    int bit, sg;
+    ufc_long j;
+    ufc_long mask1, mask2;
+
+    /*
+     * Create the do_pc1 table used
+     * to affect pc1 permutation
+     * when generating keys
+     */
+    for(bit = 0; bit < 56; bit++) {
+      comes_from_bit  = pc1[bit] - 1;
+      mask1 = bytemask[comes_from_bit % 8 + 1];
+      mask2 = longmask[bit % 28 + 4];
+      for(j = 0; j < 128; j++) {
+	if(j & mask1) 
+	  do_pc1[comes_from_bit / 8][bit / 28][j] |= mask2;
+      }
+    }
+
+    /*
+     * Create the do_pc2 table used
+     * to affect pc2 permutation when
+     * generating keys
+     */
+    for(bit = 0; bit < 48; bit++) {
+      comes_from_bit  = pc2[bit] - 1;
+      mask1 = bytemask[comes_from_bit % 7 + 1];
+      mask2 = BITMASK(bit % 24);
+      for(j = 0; j < 128; j++) {
+	if(j & mask1)
+	  do_pc2[comes_from_bit / 7][j] |= mask2;
+      }
+    }
+
+    /* 
+     * Now generate the table used to do combined
+     * 32 bit permutation and e expansion
+     *
+     * We use it because we have to permute 16384 32 bit
+     * longs into 48 bit in order to initialize sb.
+     *
+     * Looping 48 rounds per permutation becomes 
+     * just too slow...
+     *
+     */
+
+    clearmem((char*)eperm32tab, sizeof(eperm32tab));
+
+    for(bit = 0; bit < 48; bit++) {
+      ufc_long inner_mask1,comes_from;
+	
+      comes_from = perm32[esel[bit]-1]-1;
+      inner_mask1      = bytemask[comes_from % 8];
+	
+      for(j = 256; j--;) {
+	if(j & inner_mask1)
+	  eperm32tab[comes_from / 8][j][bit / 24] |= BITMASK(bit % 24);
+      }
+    }
+    
+    /* 
+     * Create the sb tables:
+     *
+     * For each 12 bit segment of an 48 bit intermediate
+     * result, the sb table precomputes the two 4 bit
+     * values of the sbox lookups done with the two 6
+     * bit halves, shifts them to their proper place,
+     * sends them through perm32 and finally E expands
+     * them so that they are ready for the next
+     * DES round.
+     *
+     */
+    for(sg = 0; sg < 4; sg++) {
+      int j1, j2;
+      int s1, s2;
+    
+      for(j1 = 0; j1 < 64; j1++) {
+	s1 = s_lookup(2 * sg, j1);
+	for(j2 = 0; j2 < 64; j2++) {
+	  ufc_long to_permute, inx;
+    
+	  s2         = s_lookup(2 * sg + 1, j2);
+	  to_permute = ((s1 << 4)  | s2) << (24 - 8 * sg);
+
+#ifdef _UFC_32_
+	  inx = ((j1 << 6)  | j2) << 1;
+	  sb[sg][inx  ]  = eperm32tab[0][(to_permute >> 24) & 0xff][0];
+	  sb[sg][inx+1]  = eperm32tab[0][(to_permute >> 24) & 0xff][1];
+	  sb[sg][inx  ] |= eperm32tab[1][(to_permute >> 16) & 0xff][0];
+	  sb[sg][inx+1] |= eperm32tab[1][(to_permute >> 16) & 0xff][1];
+  	  sb[sg][inx  ] |= eperm32tab[2][(to_permute >>  8) & 0xff][0];
+	  sb[sg][inx+1] |= eperm32tab[2][(to_permute >>  8) & 0xff][1];
+	  sb[sg][inx  ] |= eperm32tab[3][(to_permute)       & 0xff][0];
+	  sb[sg][inx+1] |= eperm32tab[3][(to_permute)       & 0xff][1];
+#endif
+#ifdef _UFC_64_
+	  inx = ((j1 << 6)  | j2);
+	  sb[sg][inx]  = 
+	    ((long64)eperm32tab[0][(to_permute >> 24) & 0xff][0] << 32) |
+	     (long64)eperm32tab[0][(to_permute >> 24) & 0xff][1];
+	  sb[sg][inx] |=
+	    ((long64)eperm32tab[1][(to_permute >> 16) & 0xff][0] << 32) |
+	     (long64)eperm32tab[1][(to_permute >> 16) & 0xff][1];
+  	  sb[sg][inx] |= 
+	    ((long64)eperm32tab[2][(to_permute >>  8) & 0xff][0] << 32) |
+	     (long64)eperm32tab[2][(to_permute >>  8) & 0xff][1];
+	  sb[sg][inx] |=
+	    ((long64)eperm32tab[3][(to_permute)       & 0xff][0] << 32) |
+	     (long64)eperm32tab[3][(to_permute)       & 0xff][1];
+#endif
+	}
+      }
+    }  
+
+    /* 
+     * Create an inverse matrix for esel telling
+     * where to plug out bits if undoing it
+     */
+    for(bit=48; bit--;) {
+      e_inverse[esel[bit] - 1     ] = bit;
+      e_inverse[esel[bit] - 1 + 32] = bit + 48;
+    }
+
+    /* 
+     * create efp: the matrix used to
+     * undo the E expansion and effect final permutation
+     */
+    clearmem((char*)efp, sizeof efp);
+    for(bit = 0; bit < 64; bit++) {
+      int o_bit, o_long;
+      ufc_long word_value, inner_mask1, inner_mask2;
+      int comes_from_f_bit, comes_from_e_bit;
+      int comes_from_word, bit_within_word;
+
+      /* See where bit i belongs in the two 32 bit long's */
+      o_long = bit / 32; /* 0..1  */
+      o_bit  = bit % 32; /* 0..31 */
+
+      /* 
+       * And find a bit in the e permutated value setting this bit.
+       *
+       * Note: the e selection may have selected the same bit several
+       * times. By the initialization of e_inverse, we only look
+       * for one specific instance.
+       */
+      comes_from_f_bit = final_perm[bit] - 1;         /* 0..63 */
+      comes_from_e_bit = e_inverse[comes_from_f_bit]; /* 0..95 */
+      comes_from_word  = comes_from_e_bit / 6;        /* 0..15 */
+      bit_within_word  = comes_from_e_bit % 6;        /* 0..5  */
+
+      inner_mask1 = longmask[bit_within_word + 26];
+      inner_mask2 = longmask[o_bit];
+
+      for(word_value = 64; word_value--;) {
+	if(word_value & inner_mask1)
+	  efp[comes_from_word][word_value][o_long] |= inner_mask2;
+      }
+    }
+    initialized++;
+  }
+
+/* 
+ * Process the elements of the sb table permuting the
+ * bits swapped in the expansion by the current salt.
+ */
+
+#ifdef _UFC_32_
+static void shuffle_sb(long32 *k, ufc_long saltbits)
+  { ufc_long j;
+    long32 x;
+    for(j=4096; j--;) {
+      x = (k[0] ^ k[1]) & (long32)saltbits;
+      *k++ ^= x;
+      *k++ ^= x;
+    }
+  }
+#endif
+
+#ifdef _UFC_64_
+static void shuffle_sb(long64 *k, ufc_long saltbits)
+  { ufc_long j;
+    long64 x;
+    for(j=4096; j--;) {
+      x = ((*k >> 32) ^ *k) & (long64)saltbits;
+      *k++ ^= (x << 32) | x;
+    }
+  }
+#endif
+
+/* 
+ * Setup the unit for a new salt
+ * Hopefully we'll not see a new salt in each crypt call.
+ */
+
+static unsigned char current_salt[3] = "&&"; /* invalid value */
+static ufc_long current_saltbits = 0;
+static int direction = 0;
+
+static void setup_salt(const char *s1)
+  { ufc_long i, j, saltbits;
+    const unsigned char *s2 = (const unsigned char *)s1;
+
+    if(!initialized)
+      ufc_init_des();
+
+    if(s2[0] == current_salt[0] && s2[1] == current_salt[1])
+      return;
+    current_salt[0] = s2[0]; current_salt[1] = s2[1];
+
+    /* 
+     * This is the only crypt change to DES:
+     * entries are swapped in the expansion table
+     * according to the bits set in the salt.
+     */
+    saltbits = 0;
+    for(i = 0; i < 2; i++) {
+      long c=ascii_to_bin(s2[i]);
+      if(c < 0 || c > 63)
+	c = 0;
+      for(j = 0; j < 6; j++) {
+	if((c >> j) & 0x1)
+	  saltbits |= BITMASK(6 * i + j);
+      }
+    }
+
+    /*
+     * Permute the sb table values
+     * to reflect the changed e
+     * selection table
+     */
+    shuffle_sb(_ufc_sb0, current_saltbits ^ saltbits); 
+    shuffle_sb(_ufc_sb1, current_saltbits ^ saltbits);
+    shuffle_sb(_ufc_sb2, current_saltbits ^ saltbits);
+    shuffle_sb(_ufc_sb3, current_saltbits ^ saltbits);
+
+    current_saltbits = saltbits;
+  }
+
+static void ufc_mk_keytab(char *key)
+  { ufc_long v1, v2, *k1;
+    int i;
+#ifdef _UFC_32_
+    long32 v, *k2 = &_ufc_keytab[0][0];
+#endif
+#ifdef _UFC_64_
+    long64 v, *k2 = &_ufc_keytab[0];
+#endif
+
+    v1 = v2 = 0; k1 = &do_pc1[0][0][0];
+    for(i = 8; i--;) {
+      v1 |= k1[*key   & 0x7f]; k1 += 128;
+      v2 |= k1[*key++ & 0x7f]; k1 += 128;
+    }
+
+    for(i = 0; i < 16; i++) {
+      k1 = &do_pc2[0][0];
+
+      v1 = (v1 << rots[i]) | (v1 >> (28 - rots[i]));
+      v  = k1[(v1 >> 21) & 0x7f]; k1 += 128;
+      v |= k1[(v1 >> 14) & 0x7f]; k1 += 128;
+      v |= k1[(v1 >>  7) & 0x7f]; k1 += 128;
+      v |= k1[(v1      ) & 0x7f]; k1 += 128;
+
+#ifdef _UFC_32_
+      *k2++ = v;
+      v = 0;
+#endif
+#ifdef _UFC_64_
+      v <<= 32;
+#endif
+
+      v2 = (v2 << rots[i]) | (v2 >> (28 - rots[i]));
+      v |= k1[(v2 >> 21) & 0x7f]; k1 += 128;
+      v |= k1[(v2 >> 14) & 0x7f]; k1 += 128;
+      v |= k1[(v2 >>  7) & 0x7f]; k1 += 128;
+      v |= k1[(v2      ) & 0x7f];
+
+      *k2++ = v;
+    }
+
+    direction = 0;
+  }
+
+/* 
+ * Undo an extra E selection and do final permutations
+ */
+
+ufc_long *_ufc_dofinalperm(ufc_long l1, ufc_long l2, ufc_long r1, ufc_long r2)
+  { ufc_long v1, v2, x;
+    static ufc_long ary[2];
+
+    x = (l1 ^ l2) & current_saltbits; l1 ^= x; l2 ^= x;
+    x = (r1 ^ r2) & current_saltbits; r1 ^= x; r2 ^= x;
+
+    v1=v2=0; l1 >>= 3; l2 >>= 3; r1 >>= 3; r2 >>= 3;
+
+    v1 |= efp[15][ r2         & 0x3f][0]; v2 |= efp[15][ r2 & 0x3f][1];
+    v1 |= efp[14][(r2 >>= 6)  & 0x3f][0]; v2 |= efp[14][ r2 & 0x3f][1];
+    v1 |= efp[13][(r2 >>= 10) & 0x3f][0]; v2 |= efp[13][ r2 & 0x3f][1];
+    v1 |= efp[12][(r2 >>= 6)  & 0x3f][0]; v2 |= efp[12][ r2 & 0x3f][1];
+
+    v1 |= efp[11][ r1         & 0x3f][0]; v2 |= efp[11][ r1 & 0x3f][1];
+    v1 |= efp[10][(r1 >>= 6)  & 0x3f][0]; v2 |= efp[10][ r1 & 0x3f][1];
+    v1 |= efp[ 9][(r1 >>= 10) & 0x3f][0]; v2 |= efp[ 9][ r1 & 0x3f][1];
+    v1 |= efp[ 8][(r1 >>= 6)  & 0x3f][0]; v2 |= efp[ 8][ r1 & 0x3f][1];
+
+    v1 |= efp[ 7][ l2         & 0x3f][0]; v2 |= efp[ 7][ l2 & 0x3f][1];
+    v1 |= efp[ 6][(l2 >>= 6)  & 0x3f][0]; v2 |= efp[ 6][ l2 & 0x3f][1];
+    v1 |= efp[ 5][(l2 >>= 10) & 0x3f][0]; v2 |= efp[ 5][ l2 & 0x3f][1];
+    v1 |= efp[ 4][(l2 >>= 6)  & 0x3f][0]; v2 |= efp[ 4][ l2 & 0x3f][1];
+
+    v1 |= efp[ 3][ l1         & 0x3f][0]; v2 |= efp[ 3][ l1 & 0x3f][1];
+    v1 |= efp[ 2][(l1 >>= 6)  & 0x3f][0]; v2 |= efp[ 2][ l1 & 0x3f][1];
+    v1 |= efp[ 1][(l1 >>= 10) & 0x3f][0]; v2 |= efp[ 1][ l1 & 0x3f][1];
+    v1 |= efp[ 0][(l1 >>= 6)  & 0x3f][0]; v2 |= efp[ 0][ l1 & 0x3f][1];
+
+    ary[0] = v1; ary[1] = v2;
+    return ary;
+  }
+
+/* 
+ * crypt only: convert from 64 bit to 11 bit ASCII 
+ * prefixing with the salt
+ */
+
+static char *output_conversion(ufc_long v1, ufc_long v2, const char *salt)
+  { static char outbuf[14];
+    int i, s;
+
+    outbuf[0] = salt[0];
+    outbuf[1] = salt[1] ? salt[1] : salt[0];
+
+    for(i = 0; i < 5; i++)
+      outbuf[i + 2] = bin_to_ascii((v1 >> (26 - 6 * i)) & 0x3f);
+
+    s  = (v2 & 0xf) << 2;
+    v2 = (v2 >> 2) | ((v1 & 0x3) << 30);
+
+    for(i = 5; i < 10; i++)
+      outbuf[i + 2] = bin_to_ascii((v2 >> (56 - 6 * i)) & 0x3f);
+
+    outbuf[12] = bin_to_ascii(s);
+    outbuf[13] = 0;
+
+    return outbuf;
+  }
+
+/* 
+ * UNIX crypt function
+ */
+
+static ufc_long *_ufc_doit(ufc_long , ufc_long, ufc_long, ufc_long, ufc_long);
+   
+char *ufc_crypt(const char *key,const char *salt)
+  { ufc_long *s;
+    char ktab[9];
+
+    /*
+     * Hack DES tables according to salt
+     */
+    setup_salt(salt);
+
+    /*
+     * Setup key schedule
+     */
+    clearmem(ktab, sizeof ktab);
+    strncpy(ktab, key, 8);
+    ufc_mk_keytab(ktab);
+
+    /*
+     * Go for the 25 DES encryptions
+     */
+    s = _ufc_doit((ufc_long)0, (ufc_long)0, 
+		  (ufc_long)0, (ufc_long)0, (ufc_long)25);
+
+    /*
+     * And convert back to 6 bit ASCII
+     */
+    return output_conversion(s[0], s[1], salt);
+  }
+
+
+#ifdef _UFC_32_
+
+/*
+ * 32 bit version
+ */
+
+extern long32 _ufc_keytab[16][2];
+extern long32 _ufc_sb0[], _ufc_sb1[], _ufc_sb2[], _ufc_sb3[];
+
+#define SBA(sb, v) (*(long32*)((char*)(sb)+(v)))
+
+static ufc_long *_ufc_doit(ufc_long l1, ufc_long l2, ufc_long r1, ufc_long r2, ufc_long itr)
+  { int i;
+    long32 s, *k;
+
+    while(itr--) {
+      k = &_ufc_keytab[0][0];
+      for(i=8; i--; ) {
+	s = *k++ ^ r1;
+	l1 ^= SBA(_ufc_sb1, s & 0xffff); l2 ^= SBA(_ufc_sb1, (s & 0xffff)+4);  
+        l1 ^= SBA(_ufc_sb0, s >>= 16);   l2 ^= SBA(_ufc_sb0, (s)         +4); 
+        s = *k++ ^ r2; 
+        l1 ^= SBA(_ufc_sb3, s & 0xffff); l2 ^= SBA(_ufc_sb3, (s & 0xffff)+4);
+        l1 ^= SBA(_ufc_sb2, s >>= 16);   l2 ^= SBA(_ufc_sb2, (s)         +4);
+
+        s = *k++ ^ l1; 
+        r1 ^= SBA(_ufc_sb1, s & 0xffff); r2 ^= SBA(_ufc_sb1, (s & 0xffff)+4);  
+        r1 ^= SBA(_ufc_sb0, s >>= 16);   r2 ^= SBA(_ufc_sb0, (s)         +4); 
+        s = *k++ ^ l2; 
+        r1 ^= SBA(_ufc_sb3, s & 0xffff); r2 ^= SBA(_ufc_sb3, (s & 0xffff)+4);  
+        r1 ^= SBA(_ufc_sb2, s >>= 16);   r2 ^= SBA(_ufc_sb2, (s)         +4);
+      } 
+      s=l1; l1=r1; r1=s; s=l2; l2=r2; r2=s;
+    }
+    return _ufc_dofinalperm(l1, l2, r1, r2);
+  }
+
+#endif
+
+#ifdef _UFC_64_
+
+/*
+ * 64 bit version
+ */
+
+extern long64 _ufc_keytab[16];
+extern long64 _ufc_sb0[], _ufc_sb1[], _ufc_sb2[], _ufc_sb3[];
+
+#define SBA(sb, v) (*(long64*)((char*)(sb)+(v)))
+
+static ufc_long *_ufc_doit(ufc_long l1, ufc_long l2, ufc_long r1, ufc_long r2, ufc_long itr)
+  { int i;
+    long64 l, r, s, *k;
+
+    l = (((long64)l1) << 32) | ((long64)l2);
+    r = (((long64)r1) << 32) | ((long64)r2);
+
+    while(itr--) {
+      k = &_ufc_keytab[0];
+      for(i=8; i--; ) {
+	s = *k++ ^ r;
+	l ^= SBA(_ufc_sb3, (s >>  0) & 0xffff);
+        l ^= SBA(_ufc_sb2, (s >> 16) & 0xffff);
+        l ^= SBA(_ufc_sb1, (s >> 32) & 0xffff);
+        l ^= SBA(_ufc_sb0, (s >> 48) & 0xffff);
+
+	s = *k++ ^ l;
+	r ^= SBA(_ufc_sb3, (s >>  0) & 0xffff);
+        r ^= SBA(_ufc_sb2, (s >> 16) & 0xffff);
+        r ^= SBA(_ufc_sb1, (s >> 32) & 0xffff);
+        r ^= SBA(_ufc_sb0, (s >> 48) & 0xffff);
+      } 
+      s=l; l=r; r=s;
+    }
+
+    l1 = l >> 32; l2 = l & 0xffffffff;
+    r1 = r >> 32; r2 = r & 0xffffffff;
+    return _ufc_dofinalperm(l1, l2, r1, r2);
+  }
+
+#endif
+
+
+#else
+ int ufc_dummy_procedure(void);
+ int ufc_dummy_procedure(void) {return 0;}
+#endif
diff --git a/ctdb/lib/replace/crypt.m4 b/ctdb/lib/replace/crypt.m4
new file mode 100644
index 0000000000..fae2a58685
--- /dev/null
+++ b/ctdb/lib/replace/crypt.m4
@@ -0,0 +1,6 @@
+###############################################
+# test for where we get crypt() from
+AC_CHECK_HEADERS(crypt.h)
+AC_SEARCH_LIBS_EXT(crypt, [crypt], CRYPT_LIBS,
+  [ AC_DEFINE(HAVE_CRYPT,1,[Whether the system has the crypt() function]) ],
+  [ LIBREPLACEOBJ="${LIBREPLACEOBJ} $libreplacedir/crypt.o" ])
diff --git a/ctdb/lib/replace/dlfcn.c b/ctdb/lib/replace/dlfcn.c
new file mode 100644
index 0000000000..88431ed576
--- /dev/null
+++ b/ctdb/lib/replace/dlfcn.c
@@ -0,0 +1,76 @@
+/* 
+   Unix SMB/CIFS implementation.
+   Samba system utilities
+   Copyright (C) Andrew Tridgell 1992-1998
+   Copyright (C) Jeremy Allison 1998-2002
+   Copyright (C) Jelmer Vernooij 2006
+
+     ** NOTE! The following LGPL license applies to the replace
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "replace.h"
+#ifdef HAVE_DL_H
+#include <dl.h>
+#endif
+
+#ifndef HAVE_DLOPEN
+#ifdef DLOPEN_TAKES_UNSIGNED_FLAGS
+void *rep_dlopen(const char *name, unsigned int flags)
+#else
+void *rep_dlopen(const char *name, int flags)
+#endif
+{
+#ifdef HAVE_SHL_LOAD
+	if (name == NULL)
+		return PROG_HANDLE;
+	return (void *)shl_load(name, flags, 0);
+#else
+	return NULL;
+#endif
+}
+#endif
+
+#ifndef HAVE_DLSYM
+void *rep_dlsym(void *handle, const char *symbol)
+{
+#ifdef HAVE_SHL_FINDSYM
+	void *sym_addr;
+	if (!shl_findsym((shl_t *)&handle, symbol, TYPE_UNDEFINED, &sym_addr))
+		return sym_addr;
+#endif
+    return NULL;
+}
+#endif
+
+#ifndef HAVE_DLERROR
+char *rep_dlerror(void)
+{
+	return "dynamic loading of objects not supported on this platform";
+}
+#endif
+
+#ifndef HAVE_DLCLOSE
+int rep_dlclose(void *handle)
+{
+#ifdef HAVE_SHL_CLOSE
+	return shl_unload((shl_t)handle);
+#else
+	return 0;
+#endif
+}
+#endif
diff --git a/ctdb/lib/replace/dlfcn.m4 b/ctdb/lib/replace/dlfcn.m4
new file mode 100644
index 0000000000..722e0246b7
--- /dev/null
+++ b/ctdb/lib/replace/dlfcn.m4
@@ -0,0 +1,31 @@
+dnl dummies provided by dlfcn.c if not available
+save_LIBS="$LIBS"
+LIBS=""
+
+libreplace_cv_dlfcn=no
+AC_SEARCH_LIBS(dlopen, dl)
+
+AC_CHECK_HEADERS(dlfcn.h)
+AC_CHECK_FUNCS([dlopen dlsym dlerror dlclose],[],[libreplace_cv_dlfcn=yes])
+
+libreplace_cv_shl=no
+AC_SEARCH_LIBS(shl_load, sl)
+AC_CHECK_HEADERS(dl.h)
+AC_CHECK_FUNCS([shl_load shl_unload shl_findsym],[],[libreplace_cv_shl=yes])
+
+AC_VERIFY_C_PROTOTYPE([void *dlopen(const char* filename, unsigned int flags)],
+	[
+	return 0;
+	],[
+	AC_DEFINE(DLOPEN_TAKES_UNSIGNED_FLAGS, 1, [Whether dlopen takes unsigned int flags])
+	],[],[
+	#include <dlfcn.h>
+	])
+
+if test x"${libreplace_cv_dlfcn}" = x"yes";then
+	LIBREPLACEOBJ="${LIBREPLACEOBJ} $libreplacedir/dlfcn.o"
+fi
+
+LIBDL="$LIBS"
+AC_SUBST(LIBDL)
+LIBS="$save_LIBS"
diff --git a/ctdb/lib/replace/getaddrinfo.c b/ctdb/lib/replace/getaddrinfo.c
new file mode 100644
index 0000000000..c5cd52be93
--- /dev/null
+++ b/ctdb/lib/replace/getaddrinfo.c
@@ -0,0 +1,497 @@
+/*
+PostgreSQL Database Management System
+(formerly known as Postgres, then as Postgres95)
+
+Portions Copyright (c) 1996-2005, The PostgreSQL Global Development Group
+
+Portions Copyright (c) 1994, The Regents of the University of California
+
+Permission to use, copy, modify, and distribute this software and its
+documentation for any purpose, without fee, and without a written agreement
+is hereby granted, provided that the above copyright notice and this paragraph
+and the following two paragraphs appear in all copies.
+
+IN NO EVENT SHALL THE UNIVERSITY OF CALIFORNIA BE LIABLE TO ANY PARTY FOR
+DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, INCLUDING
+LOST PROFITS, ARISING OUT OF THE USE OF THIS SOFTWARE AND ITS DOCUMENTATION,
+EVEN IF THE UNIVERSITY OF CALIFORNIA HAS BEEN ADVISED OF THE POSSIBILITY OF
+SUCH DAMAGE.
+
+THE UNIVERSITY OF CALIFORNIA SPECIFICALLY DISCLAIMS ANY WARRANTIES,
+INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY
+AND FITNESS FOR A PARTICULAR PURPOSE. THE SOFTWARE PROVIDED HEREUNDER IS
+ON AN "AS IS" BASIS, AND THE UNIVERSITY OF CALIFORNIA HAS NO OBLIGATIONS
+TO PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS.
+
+*/
+
+/*-------------------------------------------------------------------------
+ *
+ * getaddrinfo.c
+ *	  Support getaddrinfo() on platforms that don't have it.
+ *
+ * We also supply getnameinfo() here, assuming that the platform will have
+ * it if and only if it has getaddrinfo().	If this proves false on some
+ * platform, we'll need to split this file and provide a separate configure
+ * test for getnameinfo().
+ *
+ * Copyright (c) 2003-2007, PostgreSQL Global Development Group
+ *
+ * Copyright (C) 2007 Jeremy Allison.
+ * Modified to return multiple IPv4 addresses for Samba.
+ *
+ *-------------------------------------------------------------------------
+ */
+
+#include "replace.h"
+#include "system/network.h"
+
+#ifndef SMB_MALLOC
+#define SMB_MALLOC(s) malloc(s)
+#endif
+
+#ifndef SMB_STRDUP
+#define SMB_STRDUP(s) strdup(s)
+#endif
+
+static int check_hostent_err(struct hostent *hp)
+{
+	if (!hp) {
+		switch (h_errno) {
+			case HOST_NOT_FOUND:
+			case NO_DATA:
+				return EAI_NONAME;
+			case TRY_AGAIN:
+				return EAI_AGAIN;
+			case NO_RECOVERY:
+			default:
+				return EAI_FAIL;
+		}
+	}
+	if (!hp->h_name || hp->h_addrtype != AF_INET) {
+		return EAI_FAIL;
+	}
+	return 0;
+}
+
+static char *canon_name_from_hostent(struct hostent *hp,
+				int *perr)
+{
+	char *ret = NULL;
+
+	*perr = check_hostent_err(hp);
+	if (*perr) {
+		return NULL;
+	}
+	ret = SMB_STRDUP(hp->h_name);
+	if (!ret) {
+		*perr = EAI_MEMORY;
+	}
+	return ret;
+}
+
+static char *get_my_canon_name(int *perr)
+{
+	char name[HOST_NAME_MAX+1];
+
+	if (gethostname(name, HOST_NAME_MAX) == -1) {
+		*perr = EAI_FAIL;
+		return NULL;
+	}
+	/* Ensure null termination. */
+	name[HOST_NAME_MAX] = '\0';
+	return canon_name_from_hostent(gethostbyname(name), perr);
+}
+
+static char *get_canon_name_from_addr(struct in_addr ip,
+				int *perr)
+{
+	return canon_name_from_hostent(
+			gethostbyaddr(&ip, sizeof(ip), AF_INET),
+			perr);
+}
+
+static struct addrinfo *alloc_entry(const struct addrinfo *hints,
+				struct in_addr ip,
+				unsigned short port)
+{
+	struct sockaddr_in *psin = NULL;
+	struct addrinfo *ai = SMB_MALLOC(sizeof(*ai));
+
+	if (!ai) {
+		return NULL;
+	}
+	memset(ai, '\0', sizeof(*ai));
+
+	psin = SMB_MALLOC(sizeof(*psin));
+	if (!psin) {
+		free(ai);
+		return NULL;
+	}
+
+	memset(psin, '\0', sizeof(*psin));
+
+	psin->sin_family = AF_INET;
+	psin->sin_port = htons(port);
+	psin->sin_addr = ip;
+
+	ai->ai_flags = 0;
+	ai->ai_family = AF_INET;
+	ai->ai_socktype = hints->ai_socktype;
+	ai->ai_protocol = hints->ai_protocol;
+	ai->ai_addrlen = sizeof(*psin);
+	ai->ai_addr = (struct sockaddr *) psin;
+	ai->ai_canonname = NULL;
+	ai->ai_next = NULL;
+
+	return ai;
+}
+
+/*
+ * get address info for a single ipv4 address.
+ *
+ *	Bugs:	- servname can only be a number, not text.
+ */
+
+static int getaddr_info_single_addr(const char *service,
+				uint32_t addr,
+				const struct addrinfo *hints,
+				struct addrinfo **res)
+{
+
+	struct addrinfo *ai = NULL;
+	struct in_addr ip;
+	unsigned short port = 0;
+
+	if (service) {
+		port = (unsigned short)atoi(service);
+	}
+	ip.s_addr = htonl(addr);
+
+	ai = alloc_entry(hints, ip, port);
+	if (!ai) {
+		return EAI_MEMORY;
+	}
+
+	/* If we're asked for the canonical name,
+	 * make sure it returns correctly. */
+	if (!(hints->ai_flags & AI_NUMERICSERV) &&
+			hints->ai_flags & AI_CANONNAME) {
+		int err;
+		if (addr == INADDR_LOOPBACK || addr == INADDR_ANY) {
+			ai->ai_canonname = get_my_canon_name(&err);
+		} else {
+			ai->ai_canonname =
+			get_canon_name_from_addr(ip,&err);
+		}
+		if (ai->ai_canonname == NULL) {
+			freeaddrinfo(ai);
+			return err;
+		}
+	}
+
+	*res = ai;
+	return 0;
+}
+
+/*
+ * get address info for multiple ipv4 addresses.
+ *
+ *	Bugs:	- servname can only be a number, not text.
+ */
+
+static int getaddr_info_name(const char *node,
+				const char *service,
+				const struct addrinfo *hints,
+				struct addrinfo **res)
+{
+	struct addrinfo *listp = NULL, *prevp = NULL;
+	char **pptr = NULL;
+	int err;
+	struct hostent *hp = NULL;
+	unsigned short port = 0;
+
+	if (service) {
+		port = (unsigned short)atoi(service);
+	}
+
+	hp = gethostbyname(node);
+	err = check_hostent_err(hp);
+	if (err) {
+		return err;
+	}
+
+	for(pptr = hp->h_addr_list; *pptr; pptr++) {
+		struct in_addr ip = *(struct in_addr *)*pptr;
+		struct addrinfo *ai = alloc_entry(hints, ip, port);
+
+		if (!ai) {
+			freeaddrinfo(listp);
+			return EAI_MEMORY;
+		}
+
+		if (!listp) {
+			listp = ai;
+			prevp = ai;
+			ai->ai_canonname = SMB_STRDUP(hp->h_name);
+			if (!ai->ai_canonname) {
+				freeaddrinfo(listp);
+				return EAI_MEMORY;
+			}
+		} else {
+			prevp->ai_next = ai;
+			prevp = ai;
+		}
+	}
+	*res = listp;
+	return 0;
+}
+
+/*
+ * get address info for ipv4 sockets.
+ *
+ *	Bugs:	- servname can only be a number, not text.
+ */
+
+int rep_getaddrinfo(const char *node,
+		const char *service,
+		const struct addrinfo * hintp,
+		struct addrinfo ** res)
+{
+	struct addrinfo hints;
+
+	/* Setup the hints struct. */
+	if (hintp == NULL) {
+		memset(&hints, 0, sizeof(hints));
+		hints.ai_family = AF_INET;
+		hints.ai_socktype = SOCK_STREAM;
+	} else {
+		memcpy(&hints, hintp, sizeof(hints));
+	}
+
+	if (hints.ai_family != AF_INET && hints.ai_family != AF_UNSPEC) {
+		return EAI_FAMILY;
+	}
+
+	if (hints.ai_socktype == 0) {
+		hints.ai_socktype = SOCK_STREAM;
+	}
+
+	if (!node && !service) {
+		return EAI_NONAME;
+	}
+
+	if (node) {
+		if (node[0] == '\0') {
+			return getaddr_info_single_addr(service,
+					INADDR_ANY,
+					&hints,
+					res);
+		} else if (hints.ai_flags & AI_NUMERICHOST) {
+			struct in_addr ip;
+			if (!inet_aton(node, &ip)) {
+				return EAI_FAIL;
+			}
+			return getaddr_info_single_addr(service,
+					ntohl(ip.s_addr),
+					&hints,
+					res);
+		} else {
+			return getaddr_info_name(node,
+						service,
+						&hints,
+						res);
+		}
+	} else if (hints.ai_flags & AI_PASSIVE) {
+		return getaddr_info_single_addr(service,
+					INADDR_ANY,
+					&hints,
+					res);
+	}
+	return getaddr_info_single_addr(service,
+					INADDR_LOOPBACK,
+					&hints,
+					res);
+}
+
+
+void rep_freeaddrinfo(struct addrinfo *res)
+{
+	struct addrinfo *next = NULL;
+
+	for (;res; res = next) {
+		next = res->ai_next;
+		if (res->ai_canonname) {
+			free(res->ai_canonname);
+		}
+		if (res->ai_addr) {
+			free(res->ai_addr);
+		}
+		free(res);
+	}
+}
+
+
+const char *rep_gai_strerror(int errcode)
+{
+#ifdef HAVE_HSTRERROR
+	int			hcode;
+
+	switch (errcode)
+	{
+		case EAI_NONAME:
+			hcode = HOST_NOT_FOUND;
+			break;
+		case EAI_AGAIN:
+			hcode = TRY_AGAIN;
+			break;
+		case EAI_FAIL:
+		default:
+			hcode = NO_RECOVERY;
+			break;
+	}
+
+	return hstrerror(hcode);
+#else							/* !HAVE_HSTRERROR */
+
+	switch (errcode)
+	{
+		case EAI_NONAME:
+			return "Unknown host";
+		case EAI_AGAIN:
+			return "Host name lookup failure";
+#ifdef EAI_BADFLAGS
+		case EAI_BADFLAGS:
+			return "Invalid argument";
+#endif
+#ifdef EAI_FAMILY
+		case EAI_FAMILY:
+			return "Address family not supported";
+#endif
+#ifdef EAI_MEMORY
+		case EAI_MEMORY:
+			return "Not enough memory";
+#endif
+#ifdef EAI_NODATA
+		case EAI_NODATA:
+			return "No host data of that type was found";
+#endif
+#ifdef EAI_SERVICE
+		case EAI_SERVICE:
+			return "Class type not found";
+#endif
+#ifdef EAI_SOCKTYPE
+		case EAI_SOCKTYPE:
+			return "Socket type not supported";
+#endif
+		default:
+			return "Unknown server error";
+	}
+#endif   /* HAVE_HSTRERROR */
+}
+
+static int gethostnameinfo(const struct sockaddr *sa,
+			char *node,
+			size_t nodelen,
+			int flags)
+{
+	int ret = -1;
+	char *p = NULL;
+
+	if (!(flags & NI_NUMERICHOST)) {
+		struct hostent *hp = gethostbyaddr(
+				&((struct sockaddr_in *)sa)->sin_addr,
+				sizeof(struct in_addr),
+				sa->sa_family);
+		ret = check_hostent_err(hp);
+		if (ret == 0) {
+			/* Name looked up successfully. */
+			ret = snprintf(node, nodelen, "%s", hp->h_name);
+			if (ret < 0 || (size_t)ret >= nodelen) {
+				return EAI_MEMORY;
+			}
+			if (flags & NI_NOFQDN) {
+				p = strchr(node,'.');
+				if (p) {
+					*p = '\0';
+				}
+			}
+			return 0;
+		}
+
+		if (flags & NI_NAMEREQD) {
+			/* If we require a name and didn't get one,
+			 * automatically fail. */
+			return ret;
+		}
+		/* Otherwise just fall into the numeric host code... */
+	}
+	p = inet_ntoa(((struct sockaddr_in *)sa)->sin_addr);
+	ret = snprintf(node, nodelen, "%s", p);
+	if (ret < 0 || (size_t)ret >= nodelen) {
+		return EAI_MEMORY;
+	}
+	return 0;
+}
+
+static int getservicenameinfo(const struct sockaddr *sa,
+			char *service,
+			size_t servicelen,
+			int flags)
+{
+	int ret = -1;
+	int port = ntohs(((struct sockaddr_in *)sa)->sin_port);
+
+	if (!(flags & NI_NUMERICSERV)) {
+		struct servent *se = getservbyport(
+				port,
+				(flags & NI_DGRAM) ? "udp" : "tcp");
+		if (se && se->s_name) {
+			/* Service name looked up successfully. */
+			ret = snprintf(service, servicelen, "%s", se->s_name);
+			if (ret < 0 || (size_t)ret >= servicelen) {
+				return EAI_MEMORY;
+			}
+			return 0;
+		}
+		/* Otherwise just fall into the numeric service code... */
+	}
+	ret = snprintf(service, servicelen, "%d", port);
+	if (ret < 0 || (size_t)ret >= servicelen) {
+		return EAI_MEMORY;
+	}
+	return 0;
+}
+
+/*
+ * Convert an ipv4 address to a hostname.
+ *
+ * Bugs:	- No IPv6 support.
+ */
+int rep_getnameinfo(const struct sockaddr *sa, socklen_t salen,
+			char *node, size_t nodelen,
+			char *service, size_t servicelen, int flags)
+{
+
+	/* Invalid arguments. */
+	if (sa == NULL || (node == NULL && service == NULL)) {
+		return EAI_FAIL;
+	}
+
+	if (sa->sa_family != AF_INET) {
+		return EAI_FAIL;
+	}
+
+	if (salen < sizeof(struct sockaddr_in)) {
+		return EAI_FAIL;
+	}
+
+	if (node) {
+		return gethostnameinfo(sa, node, nodelen, flags);
+	}
+
+	if (service) {
+		return getservicenameinfo(sa, service, servicelen, flags);
+	}
+	return 0;
+}
diff --git a/ctdb/lib/replace/getaddrinfo.h b/ctdb/lib/replace/getaddrinfo.h
new file mode 100644
index 0000000000..cf040da2db
--- /dev/null
+++ b/ctdb/lib/replace/getaddrinfo.h
@@ -0,0 +1,91 @@
+/*
+PostgreSQL Database Management System
+(formerly known as Postgres, then as Postgres95)
+
+Portions Copyright (c) 1996-2005, The PostgreSQL Global Development Group
+
+Portions Copyright (c) 1994, The Regents of the University of California
+
+Permission to use, copy, modify, and distribute this software and its
+documentation for any purpose, without fee, and without a written agreement
+is hereby granted, provided that the above copyright notice and this paragraph
+and the following two paragraphs appear in all copies.
+
+IN NO EVENT SHALL THE UNIVERSITY OF CALIFORNIA BE LIABLE TO ANY PARTY FOR
+DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, INCLUDING
+LOST PROFITS, ARISING OUT OF THE USE OF THIS SOFTWARE AND ITS DOCUMENTATION,
+EVEN IF THE UNIVERSITY OF CALIFORNIA HAS BEEN ADVISED OF THE POSSIBILITY OF
+SUCH DAMAGE.
+
+THE UNIVERSITY OF CALIFORNIA SPECIFICALLY DISCLAIMS ANY WARRANTIES,
+INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY
+AND FITNESS FOR A PARTICULAR PURPOSE. THE SOFTWARE PROVIDED HEREUNDER IS
+ON AN "AS IS" BASIS, AND THE UNIVERSITY OF CALIFORNIA HAS NO OBLIGATIONS
+TO PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS.
+
+*/
+
+/*-------------------------------------------------------------------------
+ *
+ * getaddrinfo.h
+ *	  Support getaddrinfo() on platforms that don't have it.
+ *
+ * Note: we use our own routines on platforms that don't HAVE_STRUCT_ADDRINFO,
+ * whether or not the library routine getaddrinfo() can be found.  This
+ * policy is needed because on some platforms a manually installed libbind.a
+ * may provide getaddrinfo(), yet the system headers may not provide the
+ * struct definitions needed to call it.  To avoid conflict with the libbind
+ * definition in such cases, we rename our routines to pg_xxx() via macros.
+ *
+
+in lib/replace we use rep_xxx()
+
+ * This code will also work on platforms where struct addrinfo is defined
+ * in the system headers but no getaddrinfo() can be located.
+ *
+ * Copyright (c) 2003-2007, PostgreSQL Global Development Group
+ *
+ *-------------------------------------------------------------------------
+ */
+#ifndef GETADDRINFO_H
+#define GETADDRINFO_H
+
+#ifndef HAVE_GETADDRINFO
+
+/* Rename private copies per comments above */
+#ifdef getaddrinfo
+#undef getaddrinfo
+#endif
+#define getaddrinfo rep_getaddrinfo
+#define HAVE_GETADDRINFO
+
+#ifdef freeaddrinfo
+#undef freeaddrinfo
+#endif
+#define freeaddrinfo rep_freeaddrinfo
+#define HAVE_FREEADDRINFO
+
+#ifdef gai_strerror
+#undef gai_strerror
+#endif
+#define gai_strerror rep_gai_strerror
+#define HAVE_GAI_STRERROR
+
+#ifdef getnameinfo
+#undef getnameinfo
+#endif
+#define getnameinfo rep_getnameinfo
+#ifndef HAVE_GETNAMEINFO
+#define HAVE_GETNAMEINFO
+#endif
+
+extern int rep_getaddrinfo(const char *node, const char *service,
+			const struct addrinfo * hints, struct addrinfo ** res);
+extern void rep_freeaddrinfo(struct addrinfo * res);
+extern const char *rep_gai_strerror(int errcode);
+extern int rep_getnameinfo(const struct sockaddr * sa, socklen_t salen,
+			char *node, size_t nodelen,
+			char *service, size_t servicelen, int flags);
+#endif   /* HAVE_GETADDRINFO */
+
+#endif   /* GETADDRINFO_H */
diff --git a/ctdb/lib/replace/getifaddrs.c b/ctdb/lib/replace/getifaddrs.c
new file mode 100644
index 0000000000..84d790689d
--- /dev/null
+++ b/ctdb/lib/replace/getifaddrs.c
@@ -0,0 +1,352 @@
+/* 
+   Unix SMB/CIFS implementation.
+   Samba utility functions
+   Copyright (C) Andrew Tridgell 1998
+   Copyright (C) Jeremy Allison 2007
+   Copyright (C) Jelmer Vernooij <jelmer@samba.org> 2007
+   
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.
+*/
+
+#define SOCKET_WRAPPER_NOT_REPLACE
+
+#include "replace.h"
+#include "system/network.h"
+
+#include <unistd.h>
+#include <stdio.h>
+#include <sys/types.h>
+
+#ifdef HAVE_SYS_TIME_H
+#include <sys/time.h>
+#endif
+
+#ifndef SIOCGIFCONF
+#ifdef HAVE_SYS_SOCKIO_H
+#include <sys/sockio.h>
+#endif
+#endif
+
+#ifdef HAVE_IFACE_GETIFADDRS
+#define _FOUND_IFACE_ANY
+#else
+
+void rep_freeifaddrs(struct ifaddrs *ifp)
+{
+	if (ifp != NULL) {
+		free(ifp->ifa_name);
+		free(ifp->ifa_addr);
+		free(ifp->ifa_netmask);
+		free(ifp->ifa_dstaddr);
+		freeifaddrs(ifp->ifa_next);
+		free(ifp);
+	}
+}
+
+static struct sockaddr *sockaddr_dup(struct sockaddr *sa)
+{
+	struct sockaddr *ret;
+	socklen_t socklen;
+#ifdef HAVE_SOCKADDR_SA_LEN
+	socklen = sa->sa_len;
+#else
+	socklen = sizeof(struct sockaddr_storage);
+#endif
+	ret = calloc(1, socklen);
+	if (ret == NULL)
+		return NULL;
+	memcpy(ret, sa, socklen);
+	return ret;
+}
+#endif
+
+#if HAVE_IFACE_IFCONF
+
+/* this works for Linux 2.2, Solaris 2.5, SunOS4, HPUX 10.20, OSF1
+   V4.0, Ultrix 4.4, SCO Unix 3.2, IRIX 6.4 and FreeBSD 3.2.
+
+   It probably also works on any BSD style system.  */
+
+int rep_getifaddrs(struct ifaddrs **ifap)
+{
+	struct ifconf ifc;
+	char buff[8192];
+	int fd, i, n;
+	struct ifreq *ifr=NULL;
+	struct ifaddrs *curif;
+	struct ifaddrs *lastif = NULL;
+
+	*ifap = NULL;
+
+	if ((fd = socket(AF_INET, SOCK_DGRAM, 0)) == -1) {
+		return -1;
+	}
+  
+	ifc.ifc_len = sizeof(buff);
+	ifc.ifc_buf = buff;
+
+	if (ioctl(fd, SIOCGIFCONF, &ifc) != 0) {
+		close(fd);
+		return -1;
+	} 
+
+	ifr = ifc.ifc_req;
+  
+	n = ifc.ifc_len / sizeof(struct ifreq);
+
+	/* Loop through interfaces, looking for given IP address */
+	for (i=n-1; i>=0; i--) {
+		if (ioctl(fd, SIOCGIFFLAGS, &ifr[i]) == -1) {
+			freeifaddrs(*ifap);
+			return -1;
+		}
+
+		curif = calloc(1, sizeof(struct ifaddrs));
+		curif->ifa_name = strdup(ifr[i].ifr_name);
+		curif->ifa_flags = ifr[i].ifr_flags;
+		curif->ifa_dstaddr = NULL;
+		curif->ifa_data = NULL;
+		curif->ifa_next = NULL;
+
+		curif->ifa_addr = NULL;
+		if (ioctl(fd, SIOCGIFADDR, &ifr[i]) != -1) {
+			curif->ifa_addr = sockaddr_dup(&ifr[i].ifr_addr);
+		}
+
+		curif->ifa_netmask = NULL;
+		if (ioctl(fd, SIOCGIFNETMASK, &ifr[i]) != -1) {
+			curif->ifa_netmask = sockaddr_dup(&ifr[i].ifr_addr);
+		}
+
+		if (lastif == NULL) {
+			*ifap = curif;
+		} else {
+			lastif->ifa_next = curif;
+		}
+		lastif = curif;
+	}
+
+	close(fd);
+
+	return 0;
+}  
+
+#define _FOUND_IFACE_ANY
+#endif /* HAVE_IFACE_IFCONF */
+#ifdef HAVE_IFACE_IFREQ
+
+#ifndef I_STR
+#include <sys/stropts.h>
+#endif
+
+/****************************************************************************
+this should cover most of the streams based systems
+Thanks to Andrej.Borsenkow@mow.siemens.ru for several ideas in this code
+****************************************************************************/
+int rep_getifaddrs(struct ifaddrs **ifap)
+{
+	struct ifreq ifreq;
+	struct strioctl strioctl;
+	char buff[8192];
+	int fd, i, n;
+	struct ifreq *ifr=NULL;
+	struct ifaddrs *curif;
+	struct ifaddrs *lastif = NULL;
+
+	*ifap = NULL;
+
+	if ((fd = socket(AF_INET, SOCK_DGRAM, 0)) == -1) {
+		return -1;
+	}
+  
+	strioctl.ic_cmd = SIOCGIFCONF;
+	strioctl.ic_dp  = buff;
+	strioctl.ic_len = sizeof(buff);
+	if (ioctl(fd, I_STR, &strioctl) < 0) {
+		close(fd);
+		return -1;
+	} 
+
+	/* we can ignore the possible sizeof(int) here as the resulting
+	   number of interface structures won't change */
+	n = strioctl.ic_len / sizeof(struct ifreq);
+
+	/* we will assume that the kernel returns the length as an int
+           at the start of the buffer if the offered size is a
+           multiple of the structure size plus an int */
+	if (n*sizeof(struct ifreq) + sizeof(int) == strioctl.ic_len) {
+		ifr = (struct ifreq *)(buff + sizeof(int));  
+	} else {
+		ifr = (struct ifreq *)buff;  
+	}
+
+	/* Loop through interfaces */
+
+	for (i = 0; i<n; i++) {
+		ifreq = ifr[i];
+  
+		curif = calloc(1, sizeof(struct ifaddrs));
+		if (lastif == NULL) {
+			*ifap = curif;
+		} else {
+			lastif->ifa_next = curif;
+		}
+
+		strioctl.ic_cmd = SIOCGIFFLAGS;
+		strioctl.ic_dp  = (char *)&ifreq;
+		strioctl.ic_len = sizeof(struct ifreq);
+		if (ioctl(fd, I_STR, &strioctl) != 0) {
+			freeifaddrs(*ifap);
+			return -1;
+		}
+
+		curif->ifa_flags = ifreq.ifr_flags;
+		
+		strioctl.ic_cmd = SIOCGIFADDR;
+		strioctl.ic_dp  = (char *)&ifreq;
+		strioctl.ic_len = sizeof(struct ifreq);
+		if (ioctl(fd, I_STR, &strioctl) != 0) {
+			freeifaddrs(*ifap);
+			return -1;
+		}
+
+		curif->ifa_name = strdup(ifreq.ifr_name);
+		curif->ifa_addr = sockaddr_dup(&ifreq.ifr_addr);
+		curif->ifa_dstaddr = NULL;
+		curif->ifa_data = NULL;
+		curif->ifa_next = NULL;
+		curif->ifa_netmask = NULL;
+
+		strioctl.ic_cmd = SIOCGIFNETMASK;
+		strioctl.ic_dp  = (char *)&ifreq;
+		strioctl.ic_len = sizeof(struct ifreq);
+		if (ioctl(fd, I_STR, &strioctl) != 0) {
+			freeifaddrs(*ifap);
+			return -1;
+		}
+
+		curif->ifa_netmask = sockaddr_dup(&ifreq.ifr_addr);
+
+		lastif = curif;
+	}
+
+	close(fd);
+
+	return 0;
+}
+
+#define _FOUND_IFACE_ANY
+#endif /* HAVE_IFACE_IFREQ */
+#ifdef HAVE_IFACE_AIX
+
+/****************************************************************************
+this one is for AIX (tested on 4.2)
+****************************************************************************/
+int rep_getifaddrs(struct ifaddrs **ifap)
+{
+	char buff[8192];
+	int fd, i;
+	struct ifconf ifc;
+	struct ifreq *ifr=NULL;
+	struct ifaddrs *curif;
+	struct ifaddrs *lastif = NULL;
+
+	*ifap = NULL;
+
+	if ((fd = socket(AF_INET, SOCK_DGRAM, 0)) == -1) {
+		return -1;
+	}
+
+	ifc.ifc_len = sizeof(buff);
+	ifc.ifc_buf = buff;
+
+	if (ioctl(fd, SIOCGIFCONF, &ifc) != 0) {
+		close(fd);
+		return -1;
+	}
+
+	ifr = ifc.ifc_req;
+
+	/* Loop through interfaces */
+	i = ifc.ifc_len;
+
+	while (i > 0) {
+		unsigned int inc;
+
+		inc = ifr->ifr_addr.sa_len;
+
+		if (ioctl(fd, SIOCGIFADDR, ifr) != 0) {
+			freeaddrinfo(*ifap);
+			return -1;
+		}
+
+		curif = calloc(1, sizeof(struct ifaddrs));
+		if (lastif == NULL) {
+			*ifap = curif;
+		} else {
+			lastif->ifa_next = curif;
+		}
+
+		curif->ifa_name = strdup(ifr->ifr_name);
+		curif->ifa_addr = sockaddr_dup(&ifr->ifr_addr);
+		curif->ifa_dstaddr = NULL;
+		curif->ifa_data = NULL;
+		curif->ifa_netmask = NULL;
+		curif->ifa_next = NULL;
+
+		if (ioctl(fd, SIOCGIFFLAGS, ifr) != 0) {
+			freeaddrinfo(*ifap);
+			return -1;
+		}
+
+		curif->ifa_flags = ifr->ifr_flags;
+
+		if (ioctl(fd, SIOCGIFNETMASK, ifr) != 0) {
+			freeaddrinfo(*ifap);
+			return -1;
+		}
+
+		curif->ifa_netmask = sockaddr_dup(&ifr->ifr_addr);
+
+		lastif = curif;
+
+	next:
+		/*
+		 * Patch from Archie Cobbs (archie@whistle.com).  The
+		 * addresses in the SIOCGIFCONF interface list have a
+		 * minimum size. Usually this doesn't matter, but if
+		 * your machine has tunnel interfaces, etc. that have
+		 * a zero length "link address", this does matter.  */
+
+		if (inc < sizeof(ifr->ifr_addr))
+			inc = sizeof(ifr->ifr_addr);
+		inc += IFNAMSIZ;
+
+		ifr = (struct ifreq*) (((char*) ifr) + inc);
+		i -= inc;
+	}
+
+	close(fd);
+	return 0;
+}
+
+#define _FOUND_IFACE_ANY
+#endif /* HAVE_IFACE_AIX */
+#ifndef _FOUND_IFACE_ANY
+int rep_getifaddrs(struct ifaddrs **ifap)
+{
+	errno = ENOSYS;
+	return -1;
+}
+#endif
diff --git a/ctdb/lib/replace/hdr_replace.h b/ctdb/lib/replace/hdr_replace.h
new file mode 100644
index 0000000000..6cfa50f809
--- /dev/null
+++ b/ctdb/lib/replace/hdr_replace.h
@@ -0,0 +1,2 @@
+/* this is a replacement header for a missing system header */
+#include "replace.h"
diff --git a/ctdb/lib/replace/inet_aton.c b/ctdb/lib/replace/inet_aton.c
new file mode 100644
index 0000000000..c6b3bb11a7
--- /dev/null
+++ b/ctdb/lib/replace/inet_aton.c
@@ -0,0 +1,33 @@
+/*
+ * Unix SMB/CIFS implementation.
+ * replacement functions
+ * Copyright (C) Michael Adam <obnox@samba.org> 2008
+ *
+ *  ** NOTE! The following LGPL license applies to the replace
+ *  ** library. This does NOT imply that all of Samba is released
+ *  ** under the LGPL
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 3 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "replace.h"
+#include "system/network.h"
+
+/**
+ * We know that we have inet_pton from earlier libreplace checks.
+ */
+int rep_inet_aton(const char *src, struct in_addr *dst)
+{
+	return (inet_pton(AF_INET, src, dst) > 0) ? 1 : 0;
+}
diff --git a/ctdb/lib/replace/inet_ntoa.c b/ctdb/lib/replace/inet_ntoa.c
new file mode 100644
index 0000000000..e3b80ebef8
--- /dev/null
+++ b/ctdb/lib/replace/inet_ntoa.c
@@ -0,0 +1,39 @@
+/*
+ * Unix SMB/CIFS implementation.
+ * replacement routines for broken systems
+ * Copyright (C) Andrew Tridgell 2003
+ * Copyright (C) Michael Adam 2008
+ *
+ *  ** NOTE! The following LGPL license applies to the replace
+ *  ** library. This does NOT imply that all of Samba is released
+ *  ** under the LGPL
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 3 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "replace.h"
+#include "system/network.h"
+
+/**
+ * NOTE: this is not thread safe, but it can't be, either
+ * since it returns a pointer to static memory.
+ */
+char *rep_inet_ntoa(struct in_addr ip)
+{
+	uint8_t *p = (uint8_t *)&ip.s_addr;
+	static char buf[18];
+	slprintf(buf, 17, "%d.%d.%d.%d",
+		 (int)p[0], (int)p[1], (int)p[2], (int)p[3]);
+	return buf;
+}
diff --git a/ctdb/lib/replace/inet_ntop.c b/ctdb/lib/replace/inet_ntop.c
new file mode 100644
index 0000000000..fb3d8e90c8
--- /dev/null
+++ b/ctdb/lib/replace/inet_ntop.c
@@ -0,0 +1,191 @@
+/*
+ * Copyright (C) 1996-2001  Internet Software Consortium.
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND INTERNET SOFTWARE CONSORTIUM
+ * DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE INCLUDING ALL
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL
+ * INTERNET SOFTWARE CONSORTIUM BE LIABLE FOR ANY SPECIAL, DIRECT,
+ * INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING
+ * FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT,
+ * NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION
+ * WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+
+
+#include "replace.h"
+#include "system/network.h"
+
+#define NS_INT16SZ	 2
+#define NS_IN6ADDRSZ	16
+
+/*
+ * WARNING: Don't even consider trying to compile this on a system where
+ * sizeof(int) < 4.  sizeof(int) > 4 is fine; all the world's not a VAX.
+ */
+
+static const char *inet_ntop4(const unsigned char *src, char *dst,
+			      socklen_t size);
+
+#ifdef AF_INET6
+static const char *inet_ntop6(const unsigned char *src, char *dst,
+			      socklen_t size);
+#endif
+
+/* char *
+ * isc_net_ntop(af, src, dst, size)
+ *	convert a network format address to presentation format.
+ * return:
+ *	pointer to presentation format address (`dst'), or NULL (see errno).
+ * author:
+ *	Paul Vixie, 1996.
+ */
+const char *
+rep_inet_ntop(int af, const void *src, char *dst, socklen_t size)
+{
+	switch (af) {
+	case AF_INET:
+		return (inet_ntop4(src, dst, size));
+#ifdef AF_INET6
+	case AF_INET6:
+		return (inet_ntop6(src, dst, size));
+#endif
+	default:
+		errno = EAFNOSUPPORT;
+		return (NULL);
+	}
+	/* NOTREACHED */
+}
+
+/* const char *
+ * inet_ntop4(src, dst, size)
+ *	format an IPv4 address
+ * return:
+ *	`dst' (as a const)
+ * notes:
+ *	(1) uses no statics
+ *	(2) takes a unsigned char* not an in_addr as input
+ * author:
+ *	Paul Vixie, 1996.
+ */
+static const char *
+inet_ntop4(const unsigned char *src, char *dst, socklen_t size)
+{
+	static const char *fmt = "%u.%u.%u.%u";
+	char tmp[sizeof "255.255.255.255"];
+	size_t len;
+
+	len = snprintf(tmp, sizeof tmp, fmt, src[0], src[1], src[2], src[3]);
+	if (len >= size) {
+		errno = ENOSPC;
+		return (NULL);
+	}
+	memcpy(dst, tmp, len + 1);
+
+	return (dst);
+}
+
+/* const char *
+ * isc_inet_ntop6(src, dst, size)
+ *	convert IPv6 binary address into presentation (printable) format
+ * author:
+ *	Paul Vixie, 1996.
+ */
+#ifdef AF_INET6
+static const char *
+inet_ntop6(const unsigned char *src, char *dst, socklen_t size)
+{
+	/*
+	 * Note that int32_t and int16_t need only be "at least" large enough
+	 * to contain a value of the specified size.  On some systems, like
+	 * Crays, there is no such thing as an integer variable with 16 bits.
+	 * Keep this in mind if you think this function should have been coded
+	 * to use pointer overlays.  All the world's not a VAX.
+	 */
+	char tmp[sizeof "ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255"], *tp;
+	struct { int base, len; } best, cur;
+	unsigned int words[NS_IN6ADDRSZ / NS_INT16SZ];
+	int i, inc;
+
+	/*
+	 * Preprocess:
+	 *	Copy the input (bytewise) array into a wordwise array.
+	 *	Find the longest run of 0x00's in src[] for :: shorthanding.
+	 */
+	memset(words, '\0', sizeof words);
+	for (i = 0; i < NS_IN6ADDRSZ; i++)
+		words[i / 2] |= (src[i] << ((1 - (i % 2)) << 3));
+	best.base = -1;
+	best.len = 0;
+	cur.base = -1;
+	cur.len = 0;
+	for (i = 0; i < (NS_IN6ADDRSZ / NS_INT16SZ); i++) {
+		if (words[i] == 0) {
+			if (cur.base == -1)
+				cur.base = i, cur.len = 1;
+			else
+				cur.len++;
+		} else {
+			if (cur.base != -1) {
+				if (best.base == -1 || cur.len > best.len)
+					best = cur;
+				cur.base = -1;
+			}
+		}
+	}
+	if (cur.base != -1) {
+		if (best.base == -1 || cur.len > best.len)
+			best = cur;
+	}
+	if (best.base != -1 && best.len < 2)
+		best.base = -1;
+
+	/*
+	 * Format the result.
+	 */
+	tp = tmp;
+	for (i = 0; i < (NS_IN6ADDRSZ / NS_INT16SZ); i++) {
+		/* Are we inside the best run of 0x00's? */
+		if (best.base != -1 && i >= best.base &&
+		    i < (best.base + best.len)) {
+			if (i == best.base)
+				*tp++ = ':';
+			continue;
+		}
+		/* Are we following an initial run of 0x00s or any real hex? */
+		if (i != 0)
+			*tp++ = ':';
+		/* Is this address an encapsulated IPv4? */
+		if (i == 6 && best.base == 0 &&
+		    (best.len == 6 || (best.len == 5 && words[5] == 0xffff))) {
+			if (!inet_ntop4(src+12, tp, sizeof tmp - (tp - tmp)))
+				return (NULL);
+			tp += strlen(tp);
+			break;
+		}
+		inc = snprintf(tp, 5, "%x", words[i]);
+		if (inc >= 5) {
+			abort();
+		}
+		tp += inc;
+	}
+	/* Was it a trailing run of 0x00's? */
+	if (best.base != -1 && (best.base + best.len) ==
+	    (NS_IN6ADDRSZ / NS_INT16SZ))
+		*tp++ = ':';
+	*tp++ = '\0';
+
+	/*
+	 * Check for overflow, copy, and we're done.
+	 */
+	if ((size_t)(tp - tmp) > size) {
+		errno = ENOSPC;
+		return (NULL);
+	}
+	memcpy(dst, tmp, tp - tmp);
+	return (dst);
+}
+#endif /* AF_INET6 */
diff --git a/ctdb/lib/replace/inet_pton.c b/ctdb/lib/replace/inet_pton.c
new file mode 100644
index 0000000000..80e4865ef4
--- /dev/null
+++ b/ctdb/lib/replace/inet_pton.c
@@ -0,0 +1,213 @@
+/*
+ * Copyright (C) 1996-2001  Internet Software Consortium.
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND INTERNET SOFTWARE CONSORTIUM
+ * DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE INCLUDING ALL
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL
+ * INTERNET SOFTWARE CONSORTIUM BE LIABLE FOR ANY SPECIAL, DIRECT,
+ * INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING
+ * FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT,
+ * NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION
+ * WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+
+#include "replace.h"
+#include "system/network.h"
+
+#define NS_INT16SZ	 2
+#define NS_INADDRSZ	 4
+#define NS_IN6ADDRSZ	16
+
+/*
+ * WARNING: Don't even consider trying to compile this on a system where
+ * sizeof(int) < 4.  sizeof(int) > 4 is fine; all the world's not a VAX.
+ */
+
+static int inet_pton4(const char *src, unsigned char *dst);
+#ifdef AF_INET6
+static int inet_pton6(const char *src, unsigned char *dst);
+#endif
+
+/* int
+ * inet_pton(af, src, dst)
+ *	convert from presentation format (which usually means ASCII printable)
+ *	to network format (which is usually some kind of binary format).
+ * return:
+ *	1 if the address was valid for the specified address family
+ *	0 if the address wasn't valid (`dst' is untouched in this case)
+ *	-1 if some other error occurred (`dst' is untouched in this case, too)
+ * author:
+ *	Paul Vixie, 1996.
+ */
+int
+rep_inet_pton(int af,
+	  const char *src,
+	  void *dst)
+{
+	switch (af) {
+	case AF_INET:
+		return (inet_pton4(src, dst));
+#ifdef AF_INET6
+	case AF_INET6:
+		return (inet_pton6(src, dst));
+#endif
+	default:
+		errno = EAFNOSUPPORT;
+		return (-1);
+	}
+	/* NOTREACHED */
+}
+
+/* int
+ * inet_pton4(src, dst)
+ *	like inet_aton() but without all the hexadecimal and shorthand.
+ * return:
+ *	1 if `src' is a valid dotted quad, else 0.
+ * notice:
+ *	does not touch `dst' unless it's returning 1.
+ * author:
+ *	Paul Vixie, 1996.
+ */
+static int
+inet_pton4(src, dst)
+	const char *src;
+	unsigned char *dst;
+{
+	static const char digits[] = "0123456789";
+	int saw_digit, octets, ch;
+	unsigned char tmp[NS_INADDRSZ], *tp;
+
+	saw_digit = 0;
+	octets = 0;
+	*(tp = tmp) = 0;
+	while ((ch = *src++) != '\0') {
+		const char *pch;
+
+		if ((pch = strchr(digits, ch)) != NULL) {
+			unsigned int new = *tp * 10 + (pch - digits);
+
+			if (new > 255)
+				return (0);
+			*tp = new;
+			if (! saw_digit) {
+				if (++octets > 4)
+					return (0);
+				saw_digit = 1;
+			}
+		} else if (ch == '.' && saw_digit) {
+			if (octets == 4)
+				return (0);
+			*++tp = 0;
+			saw_digit = 0;
+		} else
+			return (0);
+	}
+	if (octets < 4)
+		return (0);
+	memcpy(dst, tmp, NS_INADDRSZ);
+	return (1);
+}
+
+/* int
+ * inet_pton6(src, dst)
+ *	convert presentation level address to network order binary form.
+ * return:
+ *	1 if `src' is a valid [RFC1884 2.2] address, else 0.
+ * notice:
+ *	(1) does not touch `dst' unless it's returning 1.
+ *	(2) :: in a full address is silently ignored.
+ * credit:
+ *	inspired by Mark Andrews.
+ * author:
+ *	Paul Vixie, 1996.
+ */
+#ifdef AF_INET6
+static int
+inet_pton6(src, dst)
+	const char *src;
+	unsigned char *dst;
+{
+	static const char xdigits_l[] = "0123456789abcdef",
+			  xdigits_u[] = "0123456789ABCDEF";
+	unsigned char tmp[NS_IN6ADDRSZ], *tp, *endp, *colonp;
+	const char *xdigits, *curtok;
+	int ch, saw_xdigit;
+	unsigned int val;
+
+	memset((tp = tmp), '\0', NS_IN6ADDRSZ);
+	endp = tp + NS_IN6ADDRSZ;
+	colonp = NULL;
+	/* Leading :: requires some special handling. */
+	if (*src == ':')
+		if (*++src != ':')
+			return (0);
+	curtok = src;
+	saw_xdigit = 0;
+	val = 0;
+	while ((ch = *src++) != '\0') {
+		const char *pch;
+
+		if ((pch = strchr((xdigits = xdigits_l), ch)) == NULL)
+			pch = strchr((xdigits = xdigits_u), ch);
+		if (pch != NULL) {
+			val <<= 4;
+			val |= (pch - xdigits);
+			if (val > 0xffff)
+				return (0);
+			saw_xdigit = 1;
+			continue;
+		}
+		if (ch == ':') {
+			curtok = src;
+			if (!saw_xdigit) {
+				if (colonp)
+					return (0);
+				colonp = tp;
+				continue;
+			}
+			if (tp + NS_INT16SZ > endp)
+				return (0);
+			*tp++ = (unsigned char) (val >> 8) & 0xff;
+			*tp++ = (unsigned char) val & 0xff;
+			saw_xdigit = 0;
+			val = 0;
+			continue;
+		}
+		if (ch == '.' && ((tp + NS_INADDRSZ) <= endp) &&
+		    inet_pton4(curtok, tp) > 0) {
+			tp += NS_INADDRSZ;
+			saw_xdigit = 0;
+			break;	/* '\0' was seen by inet_pton4(). */
+		}
+		return (0);
+	}
+	if (saw_xdigit) {
+		if (tp + NS_INT16SZ > endp)
+			return (0);
+		*tp++ = (unsigned char) (val >> 8) & 0xff;
+		*tp++ = (unsigned char) val & 0xff;
+	}
+	if (colonp != NULL) {
+		/*
+		 * Since some memmove()'s erroneously fail to handle
+		 * overlapping regions, we'll do the shift by hand.
+		 */
+		const int n = tp - colonp;
+		int i;
+
+		for (i = 1; i <= n; i++) {
+			endp[- i] = colonp[n - i];
+			colonp[n - i] = 0;
+		}
+		tp = endp;
+	}
+	if (tp != endp)
+		return (0);
+	memcpy(dst, tmp, NS_IN6ADDRSZ);
+	return (1);
+}
+#endif
diff --git a/ctdb/lib/replace/install-sh b/ctdb/lib/replace/install-sh
new file mode 100755
index 0000000000..58719246f0
--- /dev/null
+++ b/ctdb/lib/replace/install-sh
@@ -0,0 +1,238 @@
+#! /bin/sh
+#
+# install - install a program, script, or datafile
+# This comes from X11R5.
+#
+# Calling this script install-sh is preferred over install.sh, to prevent
+# `make' implicit rules from creating a file called install from it
+# when there is no Makefile.
+#
+# This script is compatible with the BSD install script, but was written
+# from scratch.
+#
+
+
+# set DOITPROG to echo to test this script
+
+# Don't use :- since 4.3BSD and earlier shells don't like it.
+doit="${DOITPROG-}"
+
+
+# put in absolute paths if you don't have them in your path; or use env. vars.
+
+mvprog="${MVPROG-mv}"
+cpprog="${CPPROG-cp}"
+chmodprog="${CHMODPROG-chmod}"
+chownprog="${CHOWNPROG-chown}"
+chgrpprog="${CHGRPPROG-chgrp}"
+stripprog="${STRIPPROG-strip}"
+rmprog="${RMPROG-rm}"
+mkdirprog="${MKDIRPROG-mkdir}"
+
+transformbasename=""
+transform_arg=""
+instcmd="$mvprog"
+chmodcmd="$chmodprog 0755"
+chowncmd=""
+chgrpcmd=""
+stripcmd=""
+rmcmd="$rmprog -f"
+mvcmd="$mvprog"
+src=""
+dst=""
+dir_arg=""
+
+while [ x"$1" != x ]; do
+    case $1 in
+	-c) instcmd="$cpprog"
+	    shift
+	    continue;;
+
+	-d) dir_arg=true
+	    shift
+	    continue;;
+
+	-m) chmodcmd="$chmodprog $2"
+	    shift
+	    shift
+	    continue;;
+
+	-o) chowncmd="$chownprog $2"
+	    shift
+	    shift
+	    continue;;
+
+	-g) chgrpcmd="$chgrpprog $2"
+	    shift
+	    shift
+	    continue;;
+
+	-s) stripcmd="$stripprog"
+	    shift
+	    continue;;
+
+	-t=*) transformarg=`echo $1 | sed 's/-t=//'`
+	    shift
+	    continue;;
+
+	-b=*) transformbasename=`echo $1 | sed 's/-b=//'`
+	    shift
+	    continue;;
+
+	*)  if [ x"$src" = x ]
+	    then
+		src=$1
+	    else
+		# this colon is to work around a 386BSD /bin/sh bug
+		:
+		dst=$1
+	    fi
+	    shift
+	    continue;;
+    esac
+done
+
+if [ x"$src" = x ]
+then
+	echo "install:	no input file specified"
+	exit 1
+else
+	true
+fi
+
+if [ x"$dir_arg" != x ]; then
+	dst=$src
+	src=""
+	
+	if [ -d $dst ]; then
+		instcmd=:
+	else
+		instcmd=mkdir
+	fi
+else
+
+# Waiting for this to be detected by the "$instcmd $src $dsttmp" command
+# might cause directories to be created, which would be especially bad 
+# if $src (and thus $dsttmp) contains '*'.
+
+	if [ -f $src -o -d $src ]
+	then
+		true
+	else
+		echo "install:  $src does not exist"
+		exit 1
+	fi
+	
+	if [ x"$dst" = x ]
+	then
+		echo "install:	no destination specified"
+		exit 1
+	else
+		true
+	fi
+
+# If destination is a directory, append the input filename; if your system
+# does not like double slashes in filenames, you may need to add some logic
+
+	if [ -d $dst ]
+	then
+		dst="$dst"/`basename $src`
+	else
+		true
+	fi
+fi
+
+## this sed command emulates the dirname command
+dstdir=`echo $dst | sed -e 's,[^/]*$,,;s,/$,,;s,^$,.,'`
+
+# Make sure that the destination directory exists.
+#  this part is taken from Noah Friedman's mkinstalldirs script
+
+# Skip lots of stat calls in the usual case.
+if [ ! -d "$dstdir" ]; then
+defaultIFS='	
+'
+IFS="${IFS-${defaultIFS}}"
+
+oIFS="${IFS}"
+# Some sh's can't handle IFS=/ for some reason.
+IFS='%'
+set - `echo ${dstdir} | sed -e 's@/@%@g' -e 's@^%@/@'`
+IFS="${oIFS}"
+
+pathcomp=''
+
+while [ $# -ne 0 ] ; do
+	pathcomp="${pathcomp}${1}"
+	shift
+
+	if [ ! -d "${pathcomp}" ] ;
+        then
+		$mkdirprog "${pathcomp}"
+	else
+		true
+	fi
+
+	pathcomp="${pathcomp}/"
+done
+fi
+
+if [ x"$dir_arg" != x ]
+then
+	$doit $instcmd $dst &&
+
+	if [ x"$chowncmd" != x ]; then $doit $chowncmd $dst; else true ; fi &&
+	if [ x"$chgrpcmd" != x ]; then $doit $chgrpcmd $dst; else true ; fi &&
+	if [ x"$stripcmd" != x ]; then $doit $stripcmd $dst; else true ; fi &&
+	if [ x"$chmodcmd" != x ]; then $doit $chmodcmd $dst; else true ; fi
+else
+
+# If we're going to rename the final executable, determine the name now.
+
+	if [ x"$transformarg" = x ] 
+	then
+		dstfile=`basename $dst`
+	else
+		dstfile=`basename $dst $transformbasename | 
+			sed $transformarg`$transformbasename
+	fi
+
+# don't allow the sed command to completely eliminate the filename
+
+	if [ x"$dstfile" = x ] 
+	then
+		dstfile=`basename $dst`
+	else
+		true
+	fi
+
+# Make a temp file name in the proper directory.
+
+	dsttmp=$dstdir/#inst.$$#
+
+# Move or copy the file name to the temp name
+
+	$doit $instcmd $src $dsttmp &&
+
+	trap "rm -f ${dsttmp}" 0 &&
+
+# and set any options; do chmod last to preserve setuid bits
+
+# If any of these fail, we abort the whole thing.  If we want to
+# ignore errors from any of these, just make sure not to ignore
+# errors from the above "$doit $instcmd $src $dsttmp" command.
+
+	if [ x"$chowncmd" != x ]; then $doit $chowncmd $dsttmp; else true;fi &&
+	if [ x"$chgrpcmd" != x ]; then $doit $chgrpcmd $dsttmp; else true;fi &&
+	if [ x"$stripcmd" != x ]; then $doit $stripcmd $dsttmp; else true;fi &&
+	if [ x"$chmodcmd" != x ]; then $doit $chmodcmd $dsttmp; else true;fi &&
+
+# Now rename the file to the real destination.
+
+	$doit $rmcmd -f $dstdir/$dstfile &&
+	$doit $mvcmd $dsttmp $dstdir/$dstfile 
+
+fi &&
+
+
+exit 0
diff --git a/ctdb/lib/replace/libreplace.m4 b/ctdb/lib/replace/libreplace.m4
new file mode 100644
index 0000000000..f4ae715340
--- /dev/null
+++ b/ctdb/lib/replace/libreplace.m4
@@ -0,0 +1,490 @@
+AC_DEFUN_ONCE(AC_LIBREPLACE_LOCATION_CHECKS,
+[
+echo "LIBREPLACE_LOCATION_CHECKS: START"
+
+dnl find the libreplace sources. This is meant to work both for 
+dnl libreplace standalone builds, and builds of packages using libreplace
+libreplacedir=""
+libreplacepaths="$srcdir $srcdir/lib/replace $srcdir/libreplace $srcdir/../libreplace $srcdir/../replace $srcdir/../lib/replace $srcdir/../../../lib/replace"
+for d in $libreplacepaths; do
+	if test -f "$d/replace.c"; then
+		libreplacedir="$d"		
+		AC_SUBST(libreplacedir)
+		break;
+	fi
+done
+if test x"$libreplacedir" = "x"; then
+	AC_MSG_ERROR([cannot find libreplace in $libreplacepaths])
+fi
+LIBREPLACEOBJ="$libreplacedir/replace.o"
+AC_SUBST(LIBREPLACEOBJ)
+
+AC_CANONICAL_BUILD
+AC_CANONICAL_HOST
+AC_CANONICAL_TARGET
+
+echo "LIBREPLACE_LOCATION_CHECKS: END"
+]) dnl end AC_LIBREPLACE_LOCATION_CHECKS
+
+
+AC_DEFUN_ONCE(AC_LIBREPLACE_BROKEN_CHECKS,
+[
+echo "LIBREPLACE_BROKEN_CHECKS: START"
+
+dnl find the libreplace sources. This is meant to work both for 
+dnl libreplace standalone builds, and builds of packages using libreplace
+libreplacedir=""
+libreplacepaths="$srcdir $srcdir/lib/replace $srcdir/libreplace $srcdir/../libreplace $srcdir/../replace $srcdir/../lib/replace $srcdir/../../../lib/replace"
+for d in $libreplacepaths; do
+	if test -f "$d/replace.c"; then
+		libreplacedir="$d"		
+		AC_SUBST(libreplacedir)
+		break;
+	fi
+done
+if test x"$libreplacedir" = "x"; then
+	AC_MSG_ERROR([cannot find libreplace in $libreplacepaths])
+fi
+
+LIBREPLACEOBJ="$libreplacedir/replace.o"
+AC_SUBST(LIBREPLACEOBJ)
+
+LIBREPLACEOBJ="${LIBREPLACEOBJ} $libreplacedir/snprintf.o  $libreplacedir/xattr.o"
+
+AC_TYPE_UID_T
+AC_TYPE_MODE_T
+AC_TYPE_OFF_T
+AC_TYPE_SIZE_T
+AC_TYPE_PID_T
+AC_STRUCT_ST_RDEV
+AC_CHECK_TYPE(ino_t,unsigned)
+AC_CHECK_TYPE(loff_t,off_t)
+AC_CHECK_TYPE(offset_t,loff_t)
+AC_CHECK_TYPE(blksize_t,long)
+AC_CHECK_TYPE(blkcnt_t,long)
+
+AC_FUNC_MEMCMP
+
+AC_CHECK_FUNCS([pipe strftime srandom random srand rand usleep setbuffer lstat getpgrp utime utimes])
+
+AC_CHECK_HEADERS(stdbool.h stdint.h sys/select.h)
+AC_CHECK_HEADERS(setjmp.h utime.h sys/wait.h)
+
+LIBREPLACE_PROVIDE_HEADER([stdint.h])
+LIBREPLACE_PROVIDE_HEADER([stdbool.h])
+
+AC_DEFINE(HAVE_LIBREPLACE, 1, [We have libreplace])
+
+AC_CHECK_TYPE(bool, 
+[AC_DEFINE(HAVE_BOOL, 1, [Whether the bool type is available])],,
+[
+AC_INCLUDES_DEFAULT
+#ifdef HAVE_STDBOOL_H
+#include <stdbool.h>
+#endif]
+)
+
+AC_CHECK_TYPE(_Bool, 
+[AC_DEFINE(HAVE__Bool, 1, [Whether the _Bool type is available])],,
+[
+AC_INCLUDES_DEFAULT
+#ifdef HAVE_STDBOOL_H
+#include <stdbool.h>
+#endif]
+)
+
+AC_CHECK_HEADERS(linux/types.h)
+
+AC_CACHE_CHECK([for working mmap],libreplace_cv_HAVE_MMAP,[
+AC_TRY_RUN([#include "$libreplacedir/test/shared_mmap.c"],
+           libreplace_cv_HAVE_MMAP=yes,libreplace_cv_HAVE_MMAP=no,libreplace_cv_HAVE_MMAP=cross)])
+if test x"$libreplace_cv_HAVE_MMAP" = x"yes"; then
+    AC_DEFINE(HAVE_MMAP,1,[Whether mmap works])
+fi
+
+AC_CACHE_CHECK([for working mremap],libreplace_cv_HAVE_MREMAP,[
+AC_TRY_RUN([#include "$libreplacedir/test/shared_mremap.c"],
+           libreplace_cv_HAVE_MREMAP=yes,libreplace_cv_HAVE_MREMAP=no,libreplace_cv_HAVE_MREMAP=cross)])
+if test x"$libreplace_cv_HAVE_MREMAP" = x"yes"; then
+    AC_DEFINE(HAVE_MREMAP,1,[Whether mremap works])
+fi
+
+AC_CACHE_CHECK([for incoherent mmap],libreplace_cv_HAVE_INCOHERENT_MMAP,[
+AC_TRY_RUN([#include "$libreplacedir/test/incoherent_mmap.c"],
+           libreplace_cv_HAVE_INCOHERENT_MMAP=yes,libreplace_cv_HAVE_INCOHERENT_MMAP=no,libreplace_cv_HAVE_INCOHERENT_MMAP=cross)])
+if test x"$libreplace_cv_HAVE_INCOHERENT_MMAP" = x"yes"; then
+    AC_DEFINE(HAVE_INCOHERENT_MMAP,1,[Whether mmap is incoherent against write])
+fi
+
+
+AC_CHECK_HEADERS(sys/syslog.h syslog.h)
+AC_CHECK_HEADERS(sys/time.h time.h)
+AC_CHECK_HEADERS(stdarg.h vararg.h)
+AC_CHECK_HEADERS(sys/mount.h mntent.h)
+AC_CHECK_HEADERS(stropts.h)
+AC_CHECK_HEADERS(unix.h)
+AC_CHECK_HEADERS(malloc.h)
+AC_CHECK_HEADERS(syscall.h)
+AC_CHECK_HEADERS(sys/syscall.h)
+AC_CHECK_HEADERS(sys/ucontext.h)
+
+AC_CHECK_FUNCS(syscall setuid seteuid setreuid setresuid setgid setegid setregid setresgid setgroups)
+AC_CHECK_FUNCS(chroot bzero strerror strerror_r memalign posix_memalign getpagesize)
+AC_CHECK_FUNCS(vsyslog setlinebuf mktime ftruncate chsize rename)
+AC_CHECK_FUNCS(waitpid wait4 strlcpy strlcat initgroups memmove strdup)
+AC_CHECK_FUNCS(pread pwrite strndup strcasestr strtok_r mkdtemp dup2 dprintf vdprintf)
+AC_CHECK_FUNCS(isatty chown lchown link readlink symlink realpath)
+AC_CHECK_FUNCS(fdatasync,,[
+	# if we didn't find it, look in librt (Solaris hides it there...)
+	AC_CHECK_LIB(rt, fdatasync,
+		[libreplace_cv_HAVE_FDATASYNC_IN_LIBRT=yes
+		AC_DEFINE(HAVE_FDATASYNC, 1, Define to 1 if there is support for fdatasync)])
+])
+AC_HAVE_DECL(fdatasync, [#include <unistd.h>])
+AC_CHECK_FUNCS(clock_gettime,libreplace_cv_have_clock_gettime=yes,[
+	AC_CHECK_LIB(rt, clock_gettime,
+		[libreplace_cv_HAVE_CLOCK_GETTIME_IN_LIBRT=yes
+		libreplace_cv_have_clock_gettime=yes
+		AC_DEFINE(HAVE_CLOCK_GETTIME, 1, Define to 1 if there is support for clock_gettime)])
+])
+
+AC_CHECK_HEADERS(sys/attributes.h attr/xattr.h sys/xattr.h sys/extattr.h sys/uio.h)
+AC_CHECK_HEADERS(sys/ea.h sys/proplist.h)
+
+LIBREPLACE_FILESYS_LIBS=""
+
+############################################
+# Check for EA implementations
+case "$host_os" in
+  *freebsd4* | *dragonfly* )
+	AC_DEFINE(BROKEN_EXTATTR, 1, [Does extattr API work])
+  ;;
+  *)
+	AC_SEARCH_LIBS(getxattr, [attr])
+	AC_CHECK_FUNCS(attr_get attr_getf attr_list attr_listf attropen attr_remove)
+	AC_CHECK_FUNCS(attr_removef attr_set attr_setf extattr_delete_fd extattr_delete_file)
+	AC_CHECK_FUNCS(extattr_get_fd extattr_get_file extattr_list_fd extattr_list_file)
+	AC_CHECK_FUNCS(extattr_set_fd extattr_set_file fgetea fgetxattr flistea flistxattr)
+	AC_CHECK_FUNCS(fremoveea fremovexattr fsetea fsetxattr getea getxattr listea)
+	AC_CHECK_FUNCS(listxattr removeea removexattr setea setxattr)
+
+	AC_CHECK_LIB_EXT(attr, LIBREPLACE_FILESYS_LIBS, flistea)
+	AC_CHECK_LIB_EXT(attr, LIBREPLACE_FILESYS_LIBS, flistxattr)
+	AC_CHECK_LIB_EXT(attr, LIBREPLACE_FILESYS_LIBS, attr_listf)
+	AC_CHECK_LIB_EXT(attr, LIBREPLACE_FILESYS_LIBS, extattr_list_fd)
+
+  ;;
+esac
+
+
+########################################################
+# Do xattr functions take additional options like on Darwin?
+if test x"$ac_cv_func_getxattr" = x"yes" ; then
+	AC_CACHE_CHECK([whether xattr interface takes additional options], smb_attr_cv_xattr_add_opt, [
+		old_LIBS=$LIBS
+		LIBS="$LIBS $LIBREPLACE_FILESYS_LIBS"
+		AC_TRY_COMPILE([
+			#include <sys/types.h>
+			#if HAVE_ATTR_XATTR_H
+			#include <attr/xattr.h>
+			#elif HAVE_SYS_XATTR_H
+			#include <sys/xattr.h>
+			#endif
+		],[
+			getxattr(0, 0, 0, 0, 0, 0);
+		],
+	        [smb_attr_cv_xattr_add_opt=yes],
+		[smb_attr_cv_xattr_add_opt=no;LIBS=$old_LIBS])
+	])
+	if test x"$smb_attr_cv_xattr_add_opt" = x"yes"; then
+		AC_DEFINE(XATTR_ADDITIONAL_OPTIONS, 1, [xattr functions have additional options])
+	fi
+fi
+
+AC_CHECK_FUNCS(get_current_dir_name)
+AC_HAVE_DECL(setresuid, [#include <unistd.h>])
+AC_HAVE_DECL(setresgid, [#include <unistd.h>])
+AC_HAVE_DECL(errno, [#include <errno.h>])
+
+AC_CACHE_CHECK([for secure mkstemp],libreplace_cv_HAVE_SECURE_MKSTEMP,[
+AC_TRY_RUN([#include <stdlib.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <unistd.h>
+main() { 
+  struct stat st;
+  char tpl[20]="/tmp/test.XXXXXX";
+  char tpl2[20]="/tmp/test.XXXXXX";
+  int fd = mkstemp(tpl);
+  int fd2 = mkstemp(tpl2);
+  if (fd == -1) {
+        if (fd2 != -1) {
+                unlink(tpl2);
+        }
+        exit(1);
+  }
+  if (fd2 == -1) exit(1);
+  unlink(tpl);
+  unlink(tpl2);
+  if (fstat(fd, &st) != 0) exit(1);
+  if ((st.st_mode & 0777) != 0600) exit(1);
+  if (strcmp(tpl, "/tmp/test.XXXXXX") == 0) {
+        exit(1);
+  }
+  if (strcmp(tpl, tpl2) == 0) {
+        exit(1);
+  }
+  exit(0);
+}],
+libreplace_cv_HAVE_SECURE_MKSTEMP=yes,
+libreplace_cv_HAVE_SECURE_MKSTEMP=no,
+libreplace_cv_HAVE_SECURE_MKSTEMP=cross)])
+if test x"$libreplace_cv_HAVE_SECURE_MKSTEMP" = x"yes"; then
+    AC_DEFINE(HAVE_SECURE_MKSTEMP,1,[Whether mkstemp is secure])
+fi
+
+dnl Provided by snprintf.c:
+AC_CHECK_HEADERS(stdio.h strings.h)
+AC_CHECK_DECLS([snprintf, vsnprintf, asprintf, vasprintf])
+AC_CHECK_FUNCS(snprintf vsnprintf asprintf vasprintf)
+
+AC_CACHE_CHECK([for C99 vsnprintf],libreplace_cv_HAVE_C99_VSNPRINTF,[
+AC_TRY_RUN([
+#include <sys/types.h>
+#include <stdio.h>
+#include <stdarg.h>
+#include <stdlib.h>
+void foo(const char *format, ...) { 
+       va_list ap;
+       int len;
+       char buf[20];
+       long long l = 1234567890;
+       l *= 100;
+
+       va_start(ap, format);
+       len = vsnprintf(buf, 0, format, ap);
+       va_end(ap);
+       if (len != 5) exit(1);
+
+       va_start(ap, format);
+       len = vsnprintf(0, 0, format, ap);
+       va_end(ap);
+       if (len != 5) exit(2);
+
+       if (snprintf(buf, 3, "hello") != 5 || strcmp(buf, "he") != 0) exit(3);
+
+       if (snprintf(buf, 20, "%lld", l) != 12 || strcmp(buf, "123456789000") != 0) exit(4);
+       if (snprintf(buf, 20, "%zu", 123456789) != 9 || strcmp(buf, "123456789") != 0) exit(5);
+       if (snprintf(buf, 20, "%2\$d %1\$d", 3, 4) != 3 || strcmp(buf, "4 3") != 0) exit(6);
+       if (snprintf(buf, 20, "%s", 0) < 3) exit(7);
+
+       exit(0);
+}
+main() { foo("hello"); }
+],
+libreplace_cv_HAVE_C99_VSNPRINTF=yes,libreplace_cv_HAVE_C99_VSNPRINTF=no,libreplace_cv_HAVE_C99_VSNPRINTF=cross)])
+if test x"$libreplace_cv_HAVE_C99_VSNPRINTF" = x"yes"; then
+    AC_DEFINE(HAVE_C99_VSNPRINTF,1,[Whether there is a C99 compliant vsnprintf])
+fi
+
+
+dnl VA_COPY
+AC_CACHE_CHECK([for va_copy],libreplace_cv_HAVE_VA_COPY,[
+AC_TRY_LINK([#include <stdarg.h>
+va_list ap1,ap2;], [va_copy(ap1,ap2);],
+libreplace_cv_HAVE_VA_COPY=yes,libreplace_cv_HAVE_VA_COPY=no)])
+if test x"$libreplace_cv_HAVE_VA_COPY" = x"yes"; then
+    AC_DEFINE(HAVE_VA_COPY,1,[Whether va_copy() is available])
+fi
+
+if test x"$libreplace_cv_HAVE_VA_COPY" != x"yes"; then
+AC_CACHE_CHECK([for __va_copy],libreplace_cv_HAVE___VA_COPY,[
+AC_TRY_LINK([#include <stdarg.h>
+va_list ap1,ap2;], [__va_copy(ap1,ap2);],
+libreplace_cv_HAVE___VA_COPY=yes,libreplace_cv_HAVE___VA_COPY=no)])
+if test x"$libreplace_cv_HAVE___VA_COPY" = x"yes"; then
+    AC_DEFINE(HAVE___VA_COPY,1,[Whether __va_copy() is available])
+fi
+fi
+
+dnl __FUNCTION__ macro
+AC_CACHE_CHECK([for __FUNCTION__ macro],libreplace_cv_HAVE_FUNCTION_MACRO,[
+AC_TRY_COMPILE([#include <stdio.h>], [printf("%s\n", __FUNCTION__);],
+libreplace_cv_HAVE_FUNCTION_MACRO=yes,libreplace_cv_HAVE_FUNCTION_MACRO=no)])
+if test x"$libreplace_cv_HAVE_FUNCTION_MACRO" = x"yes"; then
+    AC_DEFINE(HAVE_FUNCTION_MACRO,1,[Whether there is a __FUNCTION__ macro])
+else
+    dnl __func__ macro
+    AC_CACHE_CHECK([for __func__ macro],libreplace_cv_HAVE_func_MACRO,[
+    AC_TRY_COMPILE([#include <stdio.h>], [printf("%s\n", __func__);],
+    libreplace_cv_HAVE_func_MACRO=yes,libreplace_cv_HAVE_func_MACRO=no)])
+    if test x"$libreplace_cv_HAVE_func_MACRO" = x"yes"; then
+       AC_DEFINE(HAVE_func_MACRO,1,[Whether there is a __func__ macro])
+    fi
+fi
+
+AC_CHECK_HEADERS([sys/param.h limits.h])
+
+AC_CHECK_TYPE(comparison_fn_t, 
+[AC_DEFINE(HAVE_COMPARISON_FN_T, 1,[Whether or not we have comparison_fn_t])])
+
+AC_HAVE_DECL(setenv, [#include <stdlib.h>])
+AC_CHECK_FUNCS(setenv unsetenv)
+AC_HAVE_DECL(environ, [#include <unistd.h>])
+
+AC_CHECK_FUNCS(strnlen)
+AC_CHECK_FUNCS(strtoull __strtoull strtouq strtoll __strtoll strtoq)
+
+AC_CHECK_FUNCS(memmem)
+
+# this test disabled as we don't actually need __VA_ARGS__ yet
+AC_TRY_CPP([
+#define eprintf(...) fprintf(stderr, __VA_ARGS__)
+eprintf("bla", "bar");
+], AC_DEFINE(HAVE__VA_ARGS__MACRO, 1, [Whether the __VA_ARGS__ macro is available]))
+
+
+AC_CACHE_CHECK([for sig_atomic_t type],libreplace_cv_sig_atomic_t, [
+    AC_TRY_COMPILE([
+#include <sys/types.h>
+#if STDC_HEADERS
+#include <stdlib.h>
+#include <stddef.h>
+#endif
+#include <signal.h>],[sig_atomic_t i = 0],
+	libreplace_cv_sig_atomic_t=yes,libreplace_cv_sig_atomic_t=no)])
+if test x"$libreplace_cv_sig_atomic_t" = x"yes"; then
+   AC_DEFINE(HAVE_SIG_ATOMIC_T_TYPE,1,[Whether we have the atomic_t variable type])
+fi
+
+
+dnl Check if the C compiler understands volatile (it should, being ANSI).
+AC_CACHE_CHECK([that the C compiler understands volatile],libreplace_cv_volatile, [
+	AC_TRY_COMPILE([#include <sys/types.h>],[volatile int i = 0],
+		libreplace_cv_volatile=yes,libreplace_cv_volatile=no)])
+if test x"$libreplace_cv_volatile" = x"yes"; then
+	AC_DEFINE(HAVE_VOLATILE, 1, [Whether the C compiler understands volatile])
+fi
+
+m4_include(system/config.m4)
+
+AC_CACHE_CHECK([for O_DIRECT flag to open(2)],libreplace_cv_HAVE_OPEN_O_DIRECT,[
+AC_TRY_COMPILE([
+#include <unistd.h>
+#ifdef HAVE_FCNTL_H
+#include <fcntl.h>
+#endif],
+[int fd = open("/dev/null", O_DIRECT);],
+libreplace_cv_HAVE_OPEN_O_DIRECT=yes,libreplace_cv_HAVE_OPEN_O_DIRECT=no)])
+if test x"$libreplace_cv_HAVE_OPEN_O_DIRECT" = x"yes"; then
+    AC_DEFINE(HAVE_OPEN_O_DIRECT,1,[Whether the open(2) accepts O_DIRECT])
+fi
+
+m4_include(dlfcn.m4)
+m4_include(strptime.m4)
+m4_include(win32.m4)
+m4_include(timegm.m4)
+m4_include(repdir.m4)
+m4_include(crypt.m4)
+
+if test x$libreplace_cv_have_clock_gettime = xyes ; then
+	SMB_CHECK_CLOCK_ID(CLOCK_MONOTONIC)
+	SMB_CHECK_CLOCK_ID(CLOCK_PROCESS_CPUTIME_ID)
+	SMB_CHECK_CLOCK_ID(CLOCK_REALTIME)
+fi
+
+AC_CACHE_CHECK([for struct timespec type],libreplace_cv_struct_timespec, [
+    AC_TRY_COMPILE([
+#include <sys/types.h>
+#if STDC_HEADERS
+#include <stdlib.h>
+#include <stddef.h>
+#endif
+#if TIME_WITH_SYS_TIME
+# include <sys/time.h>
+# include <time.h>
+#else
+# if HAVE_SYS_TIME_H
+#  include <sys/time.h>
+# else
+#  include <time.h>
+# endif
+#endif
+],[struct timespec ts;],
+	libreplace_cv_struct_timespec=yes,libreplace_cv_struct_timespec=no)])
+if test x"$libreplace_cv_struct_timespec" = x"yes"; then
+   AC_DEFINE(HAVE_STRUCT_TIMESPEC,1,[Whether we have struct timespec])
+fi
+
+AC_CACHE_CHECK([for ucontext_t type],libreplace_cv_ucontext_t, [
+    AC_TRY_COMPILE([
+#include <signal.h>
+#if HAVE_SYS_UCONTEXT_H
+#include <sys/ucontext.h>
+# endif
+],[ucontext_t uc; sigaddset(&uc.uc_sigmask, SIGUSR1);],
+	libreplace_cv_ucontext_t=yes,libreplace_cv_ucontext_t=no)])
+if test x"$libreplace_cv_ucontext_t" = x"yes"; then
+   AC_DEFINE(HAVE_UCONTEXT_T,1,[Whether we have ucontext_t])
+fi
+
+AC_CHECK_FUNCS([printf memset memcpy],,[AC_MSG_ERROR([Required function not found])])
+
+echo "LIBREPLACE_BROKEN_CHECKS: END"
+]) dnl end AC_LIBREPLACE_BROKEN_CHECKS
+
+AC_DEFUN_ONCE(AC__LIBREPLACE_ALL_CHECKS_START,
+[
+#LIBREPLACE_ALL_CHECKS: START"
+])
+AC_DEFUN_ONCE(AC__LIBREPLACE_ALL_CHECKS_END,
+[
+#LIBREPLACE_ALL_CHECKS: END"
+])
+m4_define(AC_LIBREPLACE_ALL_CHECKS,
+[
+AC__LIBREPLACE_ALL_CHECKS_START
+AC_LIBREPLACE_LOCATION_CHECKS
+AC_LIBREPLACE_CC_CHECKS
+AC_LIBREPLACE_BROKEN_CHECKS
+AC__LIBREPLACE_ALL_CHECKS_END
+CFLAGS="$CFLAGS -I$libreplacedir"
+])
+
+m4_include(libreplace_cc.m4)
+m4_include(libreplace_ld.m4)
+m4_include(libreplace_network.m4)
+m4_include(libreplace_macros.m4)
+
+
+dnl SMB_CHECK_CLOCK_ID(clockid)
+dnl Test whether the specified clock_gettime clock ID is available. If it
+dnl is, we define HAVE_clockid
+AC_DEFUN([SMB_CHECK_CLOCK_ID],
+[
+    AC_MSG_CHECKING(for $1)
+    AC_TRY_LINK([
+#if TIME_WITH_SYS_TIME
+# include <sys/time.h>
+# include <time.h>
+#else
+# if HAVE_SYS_TIME_H
+#  include <sys/time.h>
+# else
+#  include <time.h>
+# endif
+#endif
+    ],
+    [
+clockid_t clk = $1;
+    ],
+    [
+	AC_MSG_RESULT(yes)
+	AC_DEFINE(HAVE_$1, 1,
+	    [Whether the clock_gettime clock ID $1 is available])
+    ],
+    [
+	AC_MSG_RESULT(no)
+    ])
+])
+m4_ifndef([AC_USE_SYSTEM_EXTENSIONS],[m4_include(autoconf-2.60.m4)])
diff --git a/ctdb/lib/replace/libreplace_cc.m4 b/ctdb/lib/replace/libreplace_cc.m4
new file mode 100644
index 0000000000..50cb73583b
--- /dev/null
+++ b/ctdb/lib/replace/libreplace_cc.m4
@@ -0,0 +1,197 @@
+
+AC_DEFUN_ONCE(AC__LIBREPLACE_ONLY_CC_CHECKS_START,
+[
+echo "LIBREPLACE_CC_CHECKS: START"
+])
+
+AC_DEFUN_ONCE(AC__LIBREPLACE_ONLY_CC_CHECKS_END,
+[
+echo "LIBREPLACE_CC_CHECKS: END"
+])
+
+dnl
+dnl
+dnl AC_LIBREPLACE_CC_CHECKS
+dnl
+dnl Note: we need to use m4_define instead of AC_DEFUN because
+dnl       of the ordering of tests
+dnl       
+dnl 
+m4_define(AC_LIBREPLACE_CC_CHECKS,
+[
+AC__LIBREPLACE_ONLY_CC_CHECKS_START
+
+dnl stop the C89 attempt by autoconf - if autoconf detects -Ae it will enable it
+dnl which conflicts with C99 on HPUX
+ac_cv_prog_cc_Ae=no
+
+savedCFLAGS=$CFLAGS
+AC_PROG_CC
+CFLAGS=$savedCFLAGS
+
+dnl don't try for C99 if we are using gcc, as otherwise we 
+dnl lose immediate structure constants
+if test x"$GCC" != x"yes" ; then
+AC_PROG_CC_C99
+fi
+
+if test x"$GCC" = x"yes" ; then
+	AC_MSG_CHECKING([for version of gcc])
+	GCC_VERSION=`$CC -dumpversion`
+	AC_MSG_RESULT(${GCC_VERSION})
+fi
+AC_USE_SYSTEM_EXTENSIONS
+AC_C_BIGENDIAN
+AC_C_INLINE
+LIBREPLACE_C99_STRUCT_INIT([],[AC_MSG_WARN([c99 structure initializer are not supported])])
+
+AC_PROG_INSTALL
+
+AC_ISC_POSIX
+AC_N_DEFINE(_XOPEN_SOURCE_EXTENDED)
+
+AC_MSG_CHECKING(checking getconf LFS_CFLAGS for large file support flags)
+LFS_CFLAGS=`(getconf LFS_CFLAGS) 2>/dev/null` || LFS_CFLAGS=""
+
+AC_MSG_RESULT(${LFS_CFLAGS})
+if test "x$LFS_CFLAGS" != 'x-1' || test "x$LFS_CFLAGS" != "xundefined"; then
+   CFLAGS="$CFLAGS $LFS_CFLAGS"
+fi
+
+AC_SYS_LARGEFILE
+
+dnl Add #include for broken IRIX header files
+case "$host_os" in
+	*irix6*) AC_ADD_INCLUDE(<standards.h>)
+		AC_N_DEFINE(_XOPEN_SOURCE,600)
+		AC_N_DEFINE(_BSD_TYPES)
+		;;
+	*hpux*)
+		# mmap on HPUX is completely broken...
+		AC_DEFINE(MMAP_BLACKLIST, 1, [Whether MMAP is broken])
+		if test "`uname -r`" = "B.11.00" -o "`uname -r`" = "B.11.11"; then
+			AC_MSG_WARN([Enabling HPUX 11.00/11.11 header bug workaround])
+			CFLAGS="$CFLAGS -Dpread=pread64 -Dpwrite=pwrite64"
+		fi
+		if test "`uname -r`" = "B.11.23"; then
+			AC_MSG_WARN([Enabling HPUX 11.23 machine/sys/getppdp.h bug workaround])
+			CFLAGS="$CFLAGS -D_MACHINE_SYS_GETPPDP_INCLUDED"
+		fi
+		;;
+	*aix*)
+		AC_DEFINE(BROKEN_STRNDUP, 1, [Whether strndup is broken])
+		AC_DEFINE(BROKEN_STRNLEN, 1, [Whether strnlen is broken])
+		if test "${GCC}" != "yes"; then
+			## for funky AIX compiler using strncpy()
+			CFLAGS="$CFLAGS -D_LINUX_SOURCE_COMPAT -qmaxmem=32000"
+		fi
+		;;
+	*osf*)
+		# this brings in socklen_t
+		AC_N_DEFINE(_XOPEN_SOURCE,600)
+		AC_N_DEFINE(_OSF_SOURCE)
+		;;
+	#
+	# VOS may need to have POSIX support and System V compatibility enabled.
+	#
+	*vos*)
+		case "$CFLAGS" in
+			*-D_POSIX_C_SOURCE*);;
+			*)
+				CFLAGS="$CFLAGS -D_POSIX_C_SOURCE=200112L"
+				AC_DEFINE(_POSIX_C_SOURCE, 200112L, [Whether to enable POSIX support])
+				;;
+		esac
+		case "$CFLAGS" in
+			*-D_SYSV*|*-D_SVID_SOURCE*);;
+			*)
+				CFLAGS="$CFLAGS -D_SYSV"
+				AC_DEFINE(_SYSV, 1, [Whether to enable System V compatibility])
+				;;
+		esac
+		;;
+esac
+
+# Do not check for standards.h on darwin, we get nasty warnings on
+# OS/X Lion. Probably a positive-list of OS'es like IRIX and AIX
+# would be the better choice, but this seems to work fine
+
+case "$host_os" in
+     *darwin*)
+	;;
+     *)
+        AC_CHECK_HEADERS([standards.h])
+	;;
+esac
+
+# Solaris needs HAVE_LONG_LONG defined
+AC_CHECK_TYPES(long long)
+
+AC_CHECK_SIZEOF(int)
+AC_CHECK_SIZEOF(char)
+AC_CHECK_SIZEOF(short)
+AC_CHECK_SIZEOF(long)
+AC_CHECK_SIZEOF(long long)
+
+AC_CHECK_TYPE(int8_t, char)
+AC_CHECK_TYPE(uint8_t, unsigned char)
+AC_CHECK_TYPE(int16_t, short)
+AC_CHECK_TYPE(uint16_t, unsigned short)
+
+if test $ac_cv_sizeof_int -eq 4 ; then
+AC_CHECK_TYPE(int32_t, int)
+AC_CHECK_TYPE(uint32_t, unsigned int)
+elif test $ac_cv_size_long -eq 4 ; then
+AC_CHECK_TYPE(int32_t, long)
+AC_CHECK_TYPE(uint32_t, unsigned long)
+else
+AC_MSG_ERROR([LIBREPLACE no 32-bit type found])
+fi
+
+AC_CHECK_TYPE(int64_t, long long)
+AC_CHECK_TYPE(uint64_t, unsigned long long)
+
+AC_CHECK_TYPE(size_t, unsigned int)
+AC_CHECK_TYPE(ssize_t, int)
+
+AC_CHECK_SIZEOF(off_t)
+AC_CHECK_SIZEOF(size_t)
+AC_CHECK_SIZEOF(ssize_t)
+
+AC_CHECK_TYPES([intptr_t, uintptr_t, ptrdiff_t])
+
+if test x"$ac_cv_type_long_long" != x"yes";then
+	AC_MSG_ERROR([LIBREPLACE needs type 'long long'])
+fi
+if test $ac_cv_sizeof_long_long -lt 8;then
+	AC_MSG_ERROR([LIBREPLACE needs sizeof(long long) >= 8])
+fi
+
+############################################
+# check if the compiler can do immediate structures
+AC_SUBST(libreplace_cv_immediate_structures)
+AC_CACHE_CHECK([for immediate structures],libreplace_cv_immediate_structures,[
+	AC_TRY_COMPILE([
+		#include <stdio.h>
+	],[
+		typedef struct {unsigned x;} FOOBAR;
+		#define X_FOOBAR(x) ((FOOBAR) { x })
+		#define FOO_ONE X_FOOBAR(1)
+		FOOBAR f = FOO_ONE;   
+		static const struct {
+			FOOBAR y; 
+		} f2[] = {
+			{FOO_ONE}
+		};
+		static const FOOBAR f3[] = {FOO_ONE};
+	],
+	libreplace_cv_immediate_structures=yes,
+	libreplace_cv_immediate_structures=no,
+	libreplace_cv_immediate_structures=cross)
+])
+if test x"$libreplace_cv_immediate_structures" = x"yes"; then
+	AC_DEFINE(HAVE_IMMEDIATE_STRUCTURES,1,[Whether the compiler supports immediate structures])
+fi
+
+AC__LIBREPLACE_ONLY_CC_CHECKS_END
+]) dnl end AC_LIBREPLACE_CC_CHECKS
diff --git a/ctdb/lib/replace/libreplace_ld.m4 b/ctdb/lib/replace/libreplace_ld.m4
new file mode 100644
index 0000000000..bf0df6186d
--- /dev/null
+++ b/ctdb/lib/replace/libreplace_ld.m4
@@ -0,0 +1,337 @@
+#
+# This offers a nice overview how to build shared libraries on all platforms
+#        http://www.fortran-2000.com/ArnaudRecipes/sharedlib.html
+#
+
+AC_DEFUN([AC_LIBREPLACE_STLD],
+[
+	AC_PATH_PROG(PROG_AR, ar)
+
+	STLD=${PROG_AR}
+
+	AC_SUBST(STLD)
+])
+
+AC_DEFUN([AC_LIBREPLACE_STLD_FLAGS],
+[
+	STLD_FLAGS="-rcs"
+	AC_SUBST(STLD_FLAGS)
+])
+
+AC_DEFUN([AC_LD_EXPORT_DYNAMIC],
+[
+saved_LDFLAGS="$LDFLAGS"
+if AC_TRY_COMMAND([${CC-cc} $CFLAGS -Wl,--version 2>&1 | grep "GNU ld" >/dev/null]); then
+	LD_EXPORT_DYNAMIC="-Wl,-export-dynamic"
+else
+	case "$host_os" in
+		hpux* )
+		  LD_EXPORT_DYNAMIC="-Wl,-E"
+		  ;;
+		*)
+		  LD_EXPORT_DYNAMIC=""
+		  ;;
+	  esac
+fi
+AC_SUBST(LD_EXPORT_DYNAMIC)
+LDFLAGS="$saved_LDFLAGS"
+])
+
+AC_DEFUN([AC_LD_PICFLAG],
+[
+case "$host_os" in
+	*linux*|*gnu*)
+		PICFLAG="-fPIC" 
+		;;
+	*solaris*)
+		if test "${GCC}" = "yes"; then
+			PICFLAG="-fPIC"
+		else
+			PICFLAG="-KPIC"
+		fi
+		;;
+	*sunos*)
+		PICFLAG="-KPIC"   # Is this correct for SunOS
+		;;
+	*netbsd* | *freebsd* | *dragonfly* )  
+		PICFLAG="-fPIC -DPIC"
+		;;
+	*openbsd*)
+		PICFLAG="-fPIC"
+		;;
+	*irix*)
+		if test "${GCC}" = "yes"; then
+			PICFLAG="-fPIC"
+		else 
+			PICFLAG="-KPIC"
+		fi
+		;;
+	*aix*)
+		# as AIX code is always position independent...
+		PICFLAG="-O2"
+		;;
+	*hpux*)
+		if test "${GCC}" = "yes"; then
+			PICFLAG="-fPIC"
+		elif test "$host_cpu" = "ia64"; then
+		        PICFLAG="+z"
+		elif test $ac_cv_prog_cc_Ae = yes; then
+			PICFLAG="+z +ESnolit"
+		fi
+		;;
+	*osf*)
+		PICFLAG="-fPIC"
+		;;
+	*unixware*)
+		PICFLAG="-KPIC"
+		;;
+	*darwin*)
+		PICFLAG="-fno-common"
+		;;
+esac
+AC_SUBST(PICFLAG)
+])
+
+AC_DEFUN([AC_LIBREPLACE_LD_SHLIB_LINKER],
+[
+	LD_SHLIB_LINKER="${CC}"
+
+	case "$host_os" in
+		*irix*)
+			LD_SHLIB_LINKER="${PROG_LD}"
+			;;
+	esac
+
+	AC_SUBST(LD_SHLIB_LINKER)
+])
+
+AC_DEFUN([AC_LIBREPLACE_LD_SHLIB_FLAGS],
+[
+	LD_SHLIB_FLAGS="-shared"
+
+	case "$host_os" in
+		*linux*|*gnu*)
+			LD_SHLIB_FLAGS="-shared -Wl,-Bsymbolic"
+			;;
+		*solaris*)
+			LD_SHLIB_FLAGS="-G"
+			if test "${GCC}" = "no"; then
+				## ${CFLAGS} added for building 64-bit shared 
+				## libs using Sun's Compiler
+				LD_SHLIB_FLAGS="-G \${CFLAGS}"
+			fi
+			;;
+		*sunos*)
+			LD_SHLIB_FLAGS="-G"
+			;;
+		*irix*)
+			LD_SHLIB_FLAGS="-shared"
+			;;
+		*aix*)
+			LD_SHLIB_FLAGS="-Wl,-G,-bexpall,-bbigtoc"
+			;;
+		*hpux*)
+			if test "${GCC}" = "yes"; then
+				LD_SHLIB_FLAGS="-shared"
+			else
+				LD_SHLIB_FLAGS="-b"
+			fi
+			;;
+		*osf*)
+			LD_SHLIB_FLAGS="-shared"
+			;;
+		*darwin*)
+			LD_SHLIB_FLAGS="-dynamiclib -Wl,-search_paths_first"
+			;;
+	esac
+
+	AC_SUBST(LD_SHLIB_FLAGS)
+])
+
+AC_DEFUN([AC_LIBREPLACE_LD_SHLIB_DISALLOW_UNDEF_FLAG],
+[
+	LD_SHLIB_DISALLOW_UNDEF_FLAG=""
+
+	#
+	# TODO: enforce error not only warnings
+	#
+	# NOTE: -Wl,--no-allow-shlib-undefined isn't what we want...
+	#       as it bails out on broken system libraries
+	#
+	case "$host_os" in
+		*osf*)
+			LD_SHLIB_DISALLOW_UNDEF_FLAG="-warning_unresolved"
+			;;
+		*darwin*)
+			LD_SHLIB_DISALLOW_UNDEF_FLAG="-undefined error"
+			;;
+	esac
+
+	AC_SUBST(LD_SHLIB_DISALLOW_UNDEF_FLAG)
+])
+
+AC_DEFUN([AC_LIBREPLACE_SHLD],
+[
+	AC_REQUIRE([AC_LIBREPLACE_LD_SHLIB_LINKER])
+	SHLD="$LD_SHLIB_LINKER"
+	AC_SUBST(SHLD)
+])
+
+AC_DEFUN([AC_LIBREPLACE_SHLD_FLAGS],
+[
+	AC_REQUIRE([AC_LIBREPLACE_LD_SHLIB_FLAGS])
+	AC_REQUIRE([AC_LIBREPLACE_LD_SHLIB_DISALLOW_UNDEF_FLAG])
+	SHLD_FLAGS="$LD_SHLIB_FLAGS $LD_SHLIB_DISALLOW_UNDEF_FLAG"
+	AC_SUBST(SHLD_FLAGS)
+])
+
+AC_DEFUN([AC_LD_SHLIBEXT],
+[
+	SHLIBEXT="so"
+	case "$host_os" in
+		*hpux*)
+			if test "$host_cpu" = "ia64"; then
+				SHLIBEXT="so"
+			else
+				SHLIBEXT="sl"
+			fi
+		;;
+		*darwin*)
+			SHLIBEXT="dylib"
+		;;
+	esac
+	AC_SUBST(SHLIBEXT)
+])
+
+AC_DEFUN([AC_LD_SONAMEFLAG],
+[
+	AC_SUBST(SONAMEFLAG)
+	SONAMEFLAG=""
+	case "$host_os" in 
+		*linux*|*gnu*|*qnx*)
+			SONAMEFLAG="-Wl,-soname="
+			;;
+		*solaris*)
+			SONAMEFLAG="-h "
+			if test "${GCC}" = "yes"; then
+				SONAMEFLAG="-Wl,-soname="
+			fi
+			;;
+		*sunos*)
+			SONAMEFLAG="-Wl,-h,"
+			;;
+		*netbsd* | *freebsd* | *dragonfly* )
+			SONAMEFLAG="-Wl,-soname,"
+			;;
+		*openbsd*)
+			SONAMEFLAG="-Wl,-soname,"
+			;;
+		*irix*)
+			SONAMEFLAG="-Wl,-soname,"
+			;;
+		*hpux*)
+			SONAMEFLAG="-Wl,+h,"
+			;;
+		*osf*)
+			SONAMEFLAG="-Wl,-soname,"
+			;;
+		*unixware*)
+			SONAMEFLAG="-Wl,-soname,"
+			;;
+		*darwin*)
+			SONAMEFLAG="#"
+			;;
+		*aix*)
+			# Not supported
+			SONAMEFLAG="#"
+			;;
+		esac
+])
+
+AC_DEFUN([AC_LD_VERSIONSCRIPT],
+[
+	AC_SUBST(VERSIONSCRIPT)
+	VERSIONSCRIPT=""
+	case "$host_os" in 
+		*linux*|*gnu*)
+			VERSIONSCRIPT="-Wl,--version-script"
+			;;
+		*solaris*)
+			if test "${GCC}" = "yes"; then
+				VERSIONSCRIPT="-Wl,--version-script"
+			fi
+			;;
+		esac
+])
+
+AC_DEFUN([AC_LIBREPLACE_MDLD],
+[
+	AC_REQUIRE([AC_LIBREPLACE_LD_SHLIB_LINKER])
+	MDLD="$LD_SHLIB_LINKER"
+	AC_SUBST(MDLD)
+])
+
+AC_DEFUN([AC_LIBREPLACE_LD_SHLIB_ALLOW_UNDEF_FLAG],
+[
+	LD_ALLOW_SHLIB_UNDEF_FLAG=""
+
+	case "$host_os" in
+		*linux*|*gnu*)
+			LD_SHLIB_ALLOW_UNDEF_FLAG="-Wl,--allow-shlib-undefined"
+			;;
+		*osf*)
+			LD_SHLIB_ALLOW_UNDEF_FLAG="-Wl,-expect_unresolved,\"*\""
+			;;
+		*darwin*)
+			LD_SHLIB_ALLOW_UNDEF_FLAG="-undefined dynamic_lookup"
+			;;
+		*aix*)
+			LD_SHLIB_ALLOW_UNDEF_FLAG="-Wl,-bnoentry"
+			;;
+	esac
+
+	AC_SUBST(LD_SHLIB_ALLOW_UNDEF_FLAG)
+])
+
+AC_DEFUN([AC_LIBREPLACE_MDLD_FLAGS],
+[
+	AC_REQUIRE([AC_LIBREPLACE_LD_SHLIB_FLAGS])
+	AC_REQUIRE([AC_LIBREPLACE_LD_SHLIB_ALLOW_UNDEF_FLAG])
+	MDLD_FLAGS="$LD_SHLIB_FLAGS $LD_SHLIB_ALLOW_UNDEF_FLAG"
+	AC_SUBST(MDLD_FLAGS)
+])
+
+AC_DEFUN([AC_LIBREPLACE_RUNTIME_LIB_PATH_VAR],
+[
+	case "$host_os" in
+		*linux*|*gnu*)
+			LIB_PATH_VAR=LD_LIBRARY_PATH
+		;;
+		*bsd*)
+			LIB_PATH_VAR=LD_LIBRARY_PATH
+		;;
+		*solaris*)
+			LIB_PATH_VAR=LD_LIBRARY_PATH
+		;;
+		*hpux*)
+			LIB_PATH_VAR=SHLIB_PATH
+		;;
+		*osf*)
+			LIB_PATH_VAR=LD_LIBRARY_PATH
+		;;
+		*aix*)
+			LIB_PATH_VAR=LIBPATH
+			;;
+		*irix*)
+			LIB_PATH_VAR=LD_LIBRARY_PATH
+			;;
+		*darwin*)
+			LIB_PATH_VAR=DYLD_LIBRARY_PATH
+			;;
+		*)
+			LIB_PATH_VAR=LD_LIBRARY_PATH
+			;;
+	esac
+
+	AC_SUBST(LIB_PATH_VAR)
+])
diff --git a/ctdb/lib/replace/libreplace_macros.m4 b/ctdb/lib/replace/libreplace_macros.m4
new file mode 100644
index 0000000000..f3753c47c2
--- /dev/null
+++ b/ctdb/lib/replace/libreplace_macros.m4
@@ -0,0 +1,347 @@
+#
+# This is a collection of useful autoconf macros
+#
+
+############################################
+# Check if the compiler handles c99 struct initialization, and if not try -AC99 and -c99 flags
+# Usage: LIBREPLACE_C99_STRUCT_INIT(success-action,failure-action)
+# changes CFLAGS to add -AC99 or -c99 if needed
+AC_DEFUN([LIBREPLACE_C99_STRUCT_INIT],
+[
+saved_CFLAGS="$CFLAGS";
+c99_init=no
+if test x"$c99_init" = x"no"; then
+    AC_MSG_CHECKING(for C99 designated initializers)
+    CFLAGS="$saved_CFLAGS";
+    AC_TRY_COMPILE([#include <stdio.h>],
+     [ struct foo {int x;char y;};
+       struct foo bar = { .y = 'X', .x = 1 };	 
+     ],
+     [AC_MSG_RESULT(yes); c99_init=yes],[AC_MSG_RESULT(no)])
+fi
+if test x"$c99_init" = x"no"; then
+    AC_MSG_CHECKING(for C99 designated initializers with -AC99)
+    CFLAGS="$saved_CFLAGS -AC99";
+    AC_TRY_COMPILE([#include <stdio.h>],
+     [ struct foo {int x;char y;};
+       struct foo bar = { .y = 'X', .x = 1 };	 
+     ],
+     [AC_MSG_RESULT(yes); c99_init=yes],[AC_MSG_RESULT(no)])
+fi
+if test x"$c99_init" = x"no"; then
+    AC_MSG_CHECKING(for C99 designated initializers with -qlanglvl=extc99)
+    CFLAGS="$saved_CFLAGS -qlanglvl=extc99";
+    AC_TRY_COMPILE([#include <stdio.h>],
+     [ struct foo {int x;char y;};
+       struct foo bar = { .y = 'X', .x = 1 };	 
+     ],
+     [AC_MSG_RESULT(yes); c99_init=yes],[AC_MSG_RESULT(no)])
+fi
+if test x"$c99_init" = x"no"; then
+    AC_MSG_CHECKING(for C99 designated initializers with -qlanglvl=stdc99)
+    CFLAGS="$saved_CFLAGS -qlanglvl=stdc99";
+    AC_TRY_COMPILE([#include <stdio.h>],
+     [ struct foo {int x;char y;};
+       struct foo bar = { .y = 'X', .x = 1 };	 
+     ],
+     [AC_MSG_RESULT(yes); c99_init=yes],[AC_MSG_RESULT(no)])
+fi
+if test x"$c99_init" = x"no"; then
+    AC_MSG_CHECKING(for C99 designated initializers with -c99)
+    CFLAGS="$saved_CFLAGS -c99"
+    AC_TRY_COMPILE([#include <stdio.h>],
+     [ struct foo {int x;char y;};
+       struct foo bar = { .y = 'X', .x = 1 };	 
+     ],
+     [AC_MSG_RESULT(yes); c99_init=yes],[AC_MSG_RESULT(no)])
+fi
+
+if test "`uname`" = "HP-UX"; then
+  if test "$ac_cv_c_compiler_gnu" = no; then
+	# special override for broken HP-UX compiler - I can't find a way to test
+	# this properly (its a compiler bug)
+	CFLAGS="$CFLAGS -AC99";
+	c99_init=yes;
+  fi
+fi
+
+if test x"$c99_init" = x"yes"; then
+    saved_CFLAGS=""
+    $1
+else
+    CFLAGS="$saved_CFLAGS"
+    saved_CFLAGS=""
+    $2
+fi
+])
+
+dnl AC_PROG_CC_FLAG(flag)
+AC_DEFUN(AC_PROG_CC_FLAG,
+[AC_CACHE_CHECK(whether ${CC-cc} accepts -$1, ac_cv_prog_cc_$1,
+[echo 'void f(){}' > conftest.c
+if test -z "`${CC-cc} -$1 -c conftest.c 2>&1`"; then
+  ac_cv_prog_cc_$1=yes
+else
+  ac_cv_prog_cc_$1=no
+fi
+rm -f conftest*
+])])
+
+dnl see if a declaration exists for a function or variable
+dnl defines HAVE_function_DECL if it exists
+dnl AC_HAVE_DECL(var, includes)
+AC_DEFUN(AC_HAVE_DECL,
+[
+ AC_CACHE_CHECK([for $1 declaration],ac_cv_have_$1_decl,[
+    AC_TRY_COMPILE([$2],[int i = (int)$1],
+        ac_cv_have_$1_decl=yes,ac_cv_have_$1_decl=no)])
+ if test x"$ac_cv_have_$1_decl" = x"yes"; then
+    AC_DEFINE([HAVE_]translit([$1], [a-z], [A-Z])[_DECL],1,[Whether $1() is available])
+ fi
+])
+
+
+# AC_CHECK_LIB_EXT(LIBRARY, [EXT_LIBS], [FUNCTION],
+#              [ACTION-IF-FOUND], [ACTION-IF-NOT-FOUND],
+#              [ADD-ACTION-IF-FOUND],[OTHER-LIBRARIES])
+# ------------------------------------------------------
+#
+# Use a cache variable name containing both the library and function name,
+# because the test really is for library $1 defining function $3, not
+# just for library $1.  Separate tests with the same $1 and different $3s
+# may have different results.
+#
+# Note that using directly AS_VAR_PUSHDEF([ac_Lib], [ac_cv_lib_$1_$3])
+# is asking for trouble, since AC_CHECK_LIB($lib, fun) would give
+# ac_cv_lib_$lib_fun, which is definitely not what was meant.  Hence
+# the AS_LITERAL_IF indirection.
+#
+# FIXME: This macro is extremely suspicious.  It DEFINEs unconditionally,
+# whatever the FUNCTION, in addition to not being a *S macro.  Note
+# that the cache does depend upon the function we are looking for.
+#
+# It is on purpose we used `ac_check_lib_ext_save_LIBS' and not just
+# `ac_save_LIBS': there are many macros which don't want to see `LIBS'
+# changed but still want to use AC_CHECK_LIB_EXT, so they save `LIBS'.
+# And ``ac_save_LIBS' is too tempting a name, so let's leave them some
+# freedom.
+AC_DEFUN([AC_CHECK_LIB_EXT],
+[
+AH_CHECK_LIB_EXT([$1])
+ac_check_lib_ext_save_LIBS=$LIBS
+LIBS="-l$1 $$2 $7 $LIBS"
+AS_LITERAL_IF([$1],
+      [AS_VAR_PUSHDEF([ac_Lib_ext], [ac_cv_lib_ext_$1])],
+      [AS_VAR_PUSHDEF([ac_Lib_ext], [ac_cv_lib_ext_$1''])])dnl
+
+m4_ifval([$3],
+ [
+    AH_CHECK_FUNC_EXT([$3])
+    AS_LITERAL_IF([$1],
+              [AS_VAR_PUSHDEF([ac_Lib_func], [ac_cv_lib_ext_$1_$3])],
+              [AS_VAR_PUSHDEF([ac_Lib_func], [ac_cv_lib_ext_$1''_$3])])dnl
+    AC_CACHE_CHECK([for $3 in -l$1], ac_Lib_func,
+	[AC_TRY_LINK_FUNC($3,
+                 [AS_VAR_SET(ac_Lib_func, yes);
+		  AS_VAR_SET(ac_Lib_ext, yes)],
+                 [AS_VAR_SET(ac_Lib_func, no);
+		  AS_VAR_SET(ac_Lib_ext, no)])
+	])
+    AS_IF([test AS_VAR_GET(ac_Lib_func) = yes],
+        [AC_DEFINE_UNQUOTED(AS_TR_CPP(HAVE_$3))])dnl
+    AS_VAR_POPDEF([ac_Lib_func])dnl
+ ],[
+    AC_CACHE_CHECK([for -l$1], ac_Lib_ext,
+	[AC_TRY_LINK_FUNC([main],
+                 [AS_VAR_SET(ac_Lib_ext, yes)],
+                 [AS_VAR_SET(ac_Lib_ext, no)])
+	])
+ ])
+LIBS=$ac_check_lib_ext_save_LIBS
+
+AS_IF([test AS_VAR_GET(ac_Lib_ext) = yes],
+    [m4_default([$4], 
+        [AC_DEFINE_UNQUOTED(AS_TR_CPP(HAVE_LIB$1))
+		case "$$2" in
+		    *-l$1*)
+			;;
+		    *)
+			$2="-l$1 $$2"
+			;;
+		esac])
+		[$6]
+	    ],
+	    [$5])dnl
+AS_VAR_POPDEF([ac_Lib_ext])dnl
+])# AC_CHECK_LIB_EXT
+
+# AH_CHECK_LIB_EXT(LIBNAME)
+# ---------------------
+m4_define([AH_CHECK_LIB_EXT],
+[AH_TEMPLATE(AS_TR_CPP(HAVE_LIB$1),
+             [Define to 1 if you have the `]$1[' library (-l]$1[).])])
+
+dnl AC_SEARCH_LIBS_EXT(FUNCTION, SEARCH-LIBS, EXT_LIBS,
+dnl                    [ACTION-IF-FOUND], [ACTION-IF-NOT-FOUND],
+dnl                    [OTHER-LIBRARIES])
+dnl --------------------------------------------------------
+dnl Search for a library defining FUNC, if it's not already available.
+AC_DEFUN([AC_SEARCH_LIBS_EXT],
+[AC_CACHE_CHECK([for library containing $1], [ac_cv_search_ext_$1],
+[
+ac_func_search_ext_save_LIBS=$LIBS
+ac_cv_search_ext_$1=no
+AC_LINK_IFELSE([AC_LANG_CALL([], [$1])],
+	       [ac_cv_search_ext_$1="none required"])
+if test "$ac_cv_search_ext_$1" = no; then
+  for ac_lib in $2; do
+    LIBS="-l$ac_lib $$3 $6 $ac_func_search_save_ext_LIBS"
+    AC_LINK_IFELSE([AC_LANG_CALL([], [$1])],
+		   [ac_cv_search_ext_$1="-l$ac_lib"
+break])
+  done
+fi
+LIBS=$ac_func_search_ext_save_LIBS])
+AS_IF([test "$ac_cv_search_ext_$1" != no],
+  [test "$ac_cv_search_ext_$1" = "none required" || $3="$ac_cv_search_ext_$1 $$3"
+  $4],
+      [$5])dnl
+])
+
+dnl check for a function in a $LIBS and $OTHER_LIBS libraries variable.
+dnl AC_CHECK_FUNC_EXT(func,OTHER_LIBS,IF-TRUE,IF-FALSE)
+AC_DEFUN([AC_CHECK_FUNC_EXT],
+[
+    AH_CHECK_FUNC_EXT($1)	
+    ac_check_func_ext_save_LIBS=$LIBS
+    LIBS="$2 $LIBS"
+    AS_VAR_PUSHDEF([ac_var], [ac_cv_func_ext_$1])dnl
+    AC_CACHE_CHECK([for $1], ac_var,
+	[AC_LINK_IFELSE([AC_LANG_FUNC_LINK_TRY([$1])],
+                [AS_VAR_SET(ac_var, yes)],
+                [AS_VAR_SET(ac_var, no)])])
+    LIBS=$ac_check_func_ext_save_LIBS
+    AS_IF([test AS_VAR_GET(ac_var) = yes], 
+	    [AC_DEFINE_UNQUOTED(AS_TR_CPP([HAVE_$1])) $3], 
+	    [$4])dnl
+AS_VAR_POPDEF([ac_var])dnl
+])# AC_CHECK_FUNC
+
+# AH_CHECK_FUNC_EXT(FUNCNAME)
+# ---------------------
+m4_define([AH_CHECK_FUNC_EXT],
+[AH_TEMPLATE(AS_TR_CPP(HAVE_$1),
+             [Define to 1 if you have the `]$1[' function.])])
+
+dnl Define an AC_DEFINE with ifndef guard.
+dnl AC_N_DEFINE(VARIABLE [, VALUE])
+AC_DEFUN([AC_N_DEFINE],
+[
+AH_VERBATIM([$1], [
+#ifndef $1
+# undef $1
+#endif
+])
+
+ cat >>confdefs.h <<\EOF
+#ifndef $1
+[#define] $1 m4_if($#, 1, 1, [$2])
+#endif
+EOF
+])
+
+dnl Add an #include
+dnl AC_ADD_INCLUDE(VARIABLE)
+define(AC_ADD_INCLUDE,
+[cat >> confdefs.h <<\EOF
+[#include] $1
+EOF
+])
+
+dnl remove an #include
+dnl AC_REMOVE_INCLUDE(VARIABLE)
+define(AC_REMOVE_INCLUDE,
+[
+grep -v '[#include] $1' confdefs.h >confdefs.h.tmp
+cat confdefs.h.tmp > confdefs.h
+rm confdefs.h.tmp
+])
+
+dnl remove an #define
+dnl AC_REMOVE_DEFINE(VARIABLE)
+define(AC_REMOVE_DEFINE,
+[
+grep -v '[#define] $1 ' confdefs.h |grep -v '[#define] $1[$]'>confdefs.h.tmp
+cat confdefs.h.tmp > confdefs.h
+rm confdefs.h.tmp
+])
+
+dnl AS_HELP_STRING is not available in autoconf 2.57, and AC_HELP_STRING is deprecated
+dnl in autoconf 2.59, so define AS_HELP_STRING to be AC_HELP_STRING unless it is already
+dnl defined.
+m4_ifdef([AS_HELP_STRING], , [m4_define([AS_HELP_STRING], m4_defn([AC_HELP_STRING]))])
+
+dnl check if the prototype in the header matches the given one
+dnl AC_VERIFY_C_PROTOTYPE(prototype,functionbody,[IF-TRUE].[IF-FALSE],[extraheaders])
+AC_DEFUN(AC_VERIFY_C_PROTOTYPE,
+[AC_CACHE_CHECK([for prototype $1], AS_TR_SH([ac_cv_c_prototype_$1]),
+	AC_COMPILE_IFELSE([AC_LANG_SOURCE([
+		AC_INCLUDES_DEFAULT
+		$5
+		$1
+		{
+			$2
+		}
+	])],[
+		eval AS_TR_SH([ac_cv_c_prototype_$1])=yes
+	],[
+		eval AS_TR_SH([ac_cv_c_prototype_$1])=no
+	])
+)
+AS_IF([eval test $AS_TR_SH([ac_cv_c_prototype_$1]) = yes],[$3],[$4])
+])
+
+AC_DEFUN(LIBREPLACE_PROVIDE_HEADER, 
+[AC_CHECK_HEADER([$1], 
+		[ AC_CONFIG_COMMANDS(rm-$1, [rm -f $libreplacedir/$1], [libreplacedir=$libreplacedir]) ],
+		[ AC_CONFIG_COMMANDS(mk-$1, [echo "#include \"replace.h\"" > $libreplacedir/$1], [libreplacedir=$libreplacedir]) ]
+	)
+])
+
+dnl AC_HAVE_TYPE(TYPE,INCLUDES)
+AC_DEFUN([AC_HAVE_TYPE], [
+AC_REQUIRE([AC_HEADER_STDC])
+cv=`echo "$1" | sed 'y%./+- %__p__%'`
+AC_MSG_CHECKING(for $1)
+AC_CACHE_VAL([ac_cv_type_$cv],
+AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[
+AC_INCLUDES_DEFAULT
+$2]],
+[[$1 foo;]])],
+[eval "ac_cv_type_$cv=yes"],
+[eval "ac_cv_type_$cv=no"]))dnl
+ac_foo=`eval echo \\$ac_cv_type_$cv`
+AC_MSG_RESULT($ac_foo)
+if test "$ac_foo" = yes; then
+  ac_tr_hdr=HAVE_`echo $1 | sed 'y%abcdefghijklmnopqrstuvwxyz./- %ABCDEFGHIJKLMNOPQRSTUVWXYZ____%'`
+if false; then
+	AC_CHECK_TYPES($1)
+fi
+  AC_DEFINE_UNQUOTED($ac_tr_hdr, 1, [Define if you have type `$1'])
+fi
+])
+
+# AC_CHECK_VALUEOF(TYPE, [INCLUDES = DEFAULT-INCLUDES])
+# ---------------------------------------------------------------
+AC_DEFUN([AC_CHECK_VALUEOF],
+[AS_LITERAL_IF(m4_translit([[$1]], [*], [p]), [],
+	       [m4_fatal([$0: requires literal arguments])])]dnl
+[
+_AC_CACHE_CHECK_INT([value of $1], [AS_TR_SH([ac_cv_valueof_$1])],
+  [(long int) ($1)],
+  [AC_INCLUDES_DEFAULT([$2])],
+  [])
+
+AC_DEFINE_UNQUOTED(AS_TR_CPP(valueof_$1), $AS_TR_SH([ac_cv_valueof_$1]),
+		   [The value of `$1'.])
+])# AC_CHECK_VALUEOF
diff --git a/ctdb/lib/replace/libreplace_network.m4 b/ctdb/lib/replace/libreplace_network.m4
new file mode 100644
index 0000000000..bb2a84324e
--- /dev/null
+++ b/ctdb/lib/replace/libreplace_network.m4
@@ -0,0 +1,503 @@
+AC_DEFUN_ONCE(AC_LIBREPLACE_NETWORK_CHECKS,
+[
+echo "LIBREPLACE_NETWORK_CHECKS: START"
+
+AC_DEFINE(LIBREPLACE_NETWORK_CHECKS, 1, [LIBREPLACE_NETWORK_CHECKS were used])
+LIBREPLACE_NETWORK_OBJS=""
+LIBREPLACE_NETWORK_LIBS=""
+
+AC_CHECK_HEADERS(sys/socket.h netinet/in.h netdb.h arpa/inet.h)
+AC_CHECK_HEADERS(netinet/in_systm.h)
+AC_CHECK_HEADERS([netinet/ip.h], [], [],[
+	#include <sys/types.h>
+	#ifdef HAVE_NETINET_IN_H
+	#include <netinet/in.h>
+	#endif
+	#ifdef HAVE_NETINET_IN_SYSTM_H
+	#include <netinet/in_systm.h>
+	#endif
+])
+AC_CHECK_HEADERS(netinet/tcp.h netinet/in_ip.h)
+AC_CHECK_HEADERS(sys/sockio.h sys/un.h)
+AC_CHECK_HEADERS(sys/uio.h)
+
+dnl we need to check that net/if.h really can be used, to cope with hpux
+dnl where including it always fails
+AC_CACHE_CHECK([for usable net/if.h],libreplace_cv_USABLE_NET_IF_H,[
+	AC_COMPILE_IFELSE([AC_LANG_SOURCE([
+		AC_INCLUDES_DEFAULT
+		#if HAVE_SYS_SOCKET_H
+		# include <sys/socket.h>
+		#endif
+		#include <net/if.h>
+		int main(void) {return 0;}])],
+		[libreplace_cv_USABLE_NET_IF_H=yes],
+		[libreplace_cv_USABLE_NET_IF_H=no]
+	)
+])
+if test x"$libreplace_cv_USABLE_NET_IF_H" = x"yes";then
+	AC_DEFINE(HAVE_NET_IF_H, 1, usability of net/if.h)
+fi
+
+AC_HAVE_TYPE([socklen_t],[#include <sys/socket.h>])
+AC_HAVE_TYPE([sa_family_t],[#include <sys/socket.h>])
+AC_HAVE_TYPE([struct addrinfo], [#include <netdb.h>])
+AC_HAVE_TYPE([struct sockaddr], [#include <sys/socket.h>])
+AC_HAVE_TYPE([struct sockaddr_storage], [
+#include <sys/socket.h>
+#include <sys/types.h>
+#include <netinet/in.h>
+])
+AC_HAVE_TYPE([struct sockaddr_in6], [
+#include <sys/socket.h>
+#include <sys/types.h>
+#include <netinet/in.h>
+])
+
+if test x"$ac_cv_type_struct_sockaddr_storage" = x"yes"; then
+AC_CHECK_MEMBER(struct sockaddr_storage.ss_family,
+                AC_DEFINE(HAVE_SS_FAMILY, 1, [Defined if struct sockaddr_storage has ss_family field]),,
+                [
+#include <sys/socket.h>
+#include <sys/types.h>
+#include <netinet/in.h>
+		])
+
+if test x"$ac_cv_member_struct_sockaddr_storage_ss_family" != x"yes"; then
+AC_CHECK_MEMBER(struct sockaddr_storage.__ss_family,
+                AC_DEFINE(HAVE___SS_FAMILY, 1, [Defined if struct sockaddr_storage has __ss_family field]),,
+                [
+#include <sys/socket.h>
+#include <sys/types.h>
+#include <netinet/in.h>
+		])
+fi
+fi
+
+AC_CACHE_CHECK([for sin_len in sock],libreplace_cv_HAVE_SOCK_SIN_LEN,[
+	AC_TRY_COMPILE(
+		[
+#include <sys/types.h>
+#include <sys/socket.h>
+#include <netinet/in.h>
+		],[
+struct sockaddr_in sock; sock.sin_len = sizeof(sock);
+		],[
+		libreplace_cv_HAVE_SOCK_SIN_LEN=yes
+		],[
+		libreplace_cv_HAVE_SOCK_SIN_LEN=no
+		])
+])
+if test x"$libreplace_cv_HAVE_SOCK_SIN_LEN" = x"yes"; then
+	AC_DEFINE(HAVE_SOCK_SIN_LEN,1,[Whether the sockaddr_in struct has a sin_len property])
+fi
+
+############################################
+# check for unix domain sockets
+AC_CACHE_CHECK([for unix domain sockets],libreplace_cv_HAVE_UNIXSOCKET,[
+	AC_TRY_COMPILE([
+#include <sys/types.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <sys/socket.h>
+#include <sys/un.h>
+		],[
+struct sockaddr_un sunaddr;
+sunaddr.sun_family = AF_UNIX;
+		],[
+		libreplace_cv_HAVE_UNIXSOCKET=yes
+		],[
+		libreplace_cv_HAVE_UNIXSOCKET=no
+		])
+])
+if test x"$libreplace_cv_HAVE_UNIXSOCKET" = x"yes"; then
+	AC_DEFINE(HAVE_UNIXSOCKET,1,[If we need to build with unixscoket support])
+fi
+
+dnl The following test is roughly taken from the cvs sources.
+dnl
+dnl If we can't find connect, try looking in -lsocket, -lnsl, and -linet.
+dnl The Irix 5 libc.so has connect and gethostbyname, but Irix 5 also has
+dnl libsocket.so which has a bad implementation of gethostbyname (it
+dnl only looks in /etc/hosts), so we only look for -lsocket if we need
+dnl it.
+AC_CHECK_FUNCS(connect)
+if test x"$ac_cv_func_connect" = x"no"; then
+	AC_CHECK_LIB_EXT(nsl_s, LIBREPLACE_NETWORK_LIBS, connect)
+	AC_CHECK_LIB_EXT(nsl, LIBREPLACE_NETWORK_LIBS, connect)
+	AC_CHECK_LIB_EXT(socket, LIBREPLACE_NETWORK_LIBS, connect)
+	AC_CHECK_LIB_EXT(inet, LIBREPLACE_NETWORK_LIBS, connect)
+	dnl We can't just call AC_CHECK_FUNCS(connect) here,
+	dnl because the value has been cached.
+	if test x"$ac_cv_lib_ext_nsl_s_connect" = x"yes" ||
+		test x"$ac_cv_lib_ext_nsl_connect" = x"yes" ||
+		test x"$ac_cv_lib_ext_socket_connect" = x"yes" ||
+		test x"$ac_cv_lib_ext_inet_connect" = x"yes"
+	then
+		AC_DEFINE(HAVE_CONNECT,1,[Whether the system has connect()])
+	fi
+fi
+
+AC_CHECK_FUNCS(gethostbyname)
+if test x"$ac_cv_func_gethostbyname" = x"no"; then
+	AC_CHECK_LIB_EXT(nsl_s, LIBREPLACE_NETWORK_LIBS, gethostbyname)
+	AC_CHECK_LIB_EXT(nsl, LIBREPLACE_NETWORK_LIBS, gethostbyname)
+	AC_CHECK_LIB_EXT(socket, LIBREPLACE_NETWORK_LIBS, gethostbyname)
+	dnl We can't just call AC_CHECK_FUNCS(gethostbyname) here,
+	dnl because the value has been cached.
+	if test x"$ac_cv_lib_ext_nsl_s_gethostbyname" = x"yes" ||
+		test x"$ac_cv_lib_ext_nsl_gethostbyname" = x"yes" ||
+		test x"$ac_cv_lib_ext_socket_gethostbyname" = x"yes"
+	then
+		AC_DEFINE(HAVE_GETHOSTBYNAME,1,
+			  [Whether the system has gethostbyname()])
+	fi
+fi
+
+dnl HP-UX has if_nametoindex in -lipv6
+AC_CHECK_FUNCS(if_nametoindex)
+if test x"$ac_cv_func_if_nametoindex" = x"no"; then
+	AC_CHECK_LIB_EXT(ipv6, LIBREPLACE_NETWORK_LIBS, if_nametoindex)
+	dnl We can't just call AC_CHECK_FUNCS(if_nametoindex) here,
+	dnl because the value has been cached.
+	if test x"$ac_cv_lib_ext_ipv6_if_nametoindex" = x"yes"
+	then
+		AC_DEFINE(HAVE_IF_NAMETOINDEX, 1,
+			  [Whether the system has if_nametoindex()])
+	fi
+fi
+
+# The following tests need LIBS="${LIBREPLACE_NETWORK_LIBS}"
+old_LIBS=$LIBS
+LIBS="${LIBREPLACE_NETWORK_LIBS}"
+libreplace_SAVE_CPPFLAGS="$CPPFLAGS"
+CPPFLAGS="$CPPFLAGS -I$libreplacedir"
+
+AC_CHECK_FUNCS(socketpair,[],[LIBREPLACE_NETWORK_OBJS="${LIBREPLACE_NETWORK_OBJS} $libreplacedir/socketpair.o"])
+
+AC_CACHE_CHECK([for broken inet_ntoa],libreplace_cv_REPLACE_INET_NTOA,[
+AC_TRY_RUN([
+#include <stdio.h>
+#include <unistd.h>
+#include <sys/types.h>
+#include <netinet/in.h>
+#ifdef HAVE_ARPA_INET_H
+#include <arpa/inet.h>
+#endif
+main() { struct in_addr ip; ip.s_addr = 0x12345678;
+if (strcmp(inet_ntoa(ip),"18.52.86.120") &&
+    strcmp(inet_ntoa(ip),"120.86.52.18")) { exit(0); }
+exit(1);}],
+           libreplace_cv_REPLACE_INET_NTOA=yes,libreplace_cv_REPLACE_INET_NTOA=no,libreplace_cv_REPLACE_INET_NTOA=cross)])
+
+AC_CHECK_FUNCS(inet_ntoa,[],[libreplace_cv_REPLACE_INET_NTOA=yes])
+if test x"$libreplace_cv_REPLACE_INET_NTOA" = x"yes"; then
+    AC_DEFINE(REPLACE_INET_NTOA,1,[Whether inet_ntoa should be replaced])
+    LIBREPLACE_NETWORK_OBJS="${LIBREPLACE_NETWORK_OBJS} $libreplacedir/inet_ntoa.o"
+fi
+
+AC_CHECK_FUNCS(inet_aton,[],[LIBREPLACE_NETWORK_OBJS="${LIBREPLACE_NETWORK_OBJS} $libreplacedir/inet_aton.o"])
+
+AC_CHECK_FUNCS(inet_ntop,[],[LIBREPLACE_NETWORK_OBJS="${LIBREPLACE_NETWORK_OBJS} $libreplacedir/inet_ntop.o"])
+
+AC_CHECK_FUNCS(inet_pton,[],[LIBREPLACE_NETWORK_OBJS="${LIBREPLACE_NETWORK_OBJS} $libreplacedir/inet_pton.o"])
+
+dnl test for getaddrinfo/getnameinfo
+AC_CACHE_CHECK([for getaddrinfo],libreplace_cv_HAVE_GETADDRINFO,[
+AC_TRY_LINK([
+#include <sys/types.h>
+#if STDC_HEADERS
+#include <stdlib.h>
+#include <stddef.h>
+#endif
+#include <sys/socket.h>
+#include <netdb.h>],
+[
+struct sockaddr sa;
+struct addrinfo *ai = NULL;
+int ret = getaddrinfo(NULL, NULL, NULL, &ai);
+if (ret != 0) {
+	const char *es = gai_strerror(ret);
+}
+freeaddrinfo(ai);
+ret = getnameinfo(&sa, sizeof(sa),
+		NULL, 0,
+		NULL, 0, 0);
+
+],
+libreplace_cv_HAVE_GETADDRINFO=yes,libreplace_cv_HAVE_GETADDRINFO=no)])
+
+if test x"$libreplace_cv_HAVE_GETADDRINFO" = x"yes"; then
+	# getaddrinfo is broken on some AIX systems
+	# see bug 5910, use our replacements if we detect
+	# a broken system.
+	AC_TRY_RUN([
+		#include <stddef.h>
+		#include <sys/types.h>
+		#include <sys/socket.h>
+		#include <netdb.h>
+		int main(int argc, const char *argv[])
+		{
+			struct addrinfo hints = {0,};
+			struct addrinfo *ppres;
+			const char hostname1[] = "0.0.0.0";
+			const char hostname2[] = "127.0.0.1";
+			const char hostname3[] = "::";
+			hints.ai_socktype = SOCK_STREAM;
+			hints.ai_family = AF_UNSPEC;
+			hints.ai_flags =
+				AI_NUMERICHOST|AI_PASSIVE|AI_ADDRCONFIG;
+			/* Test for broken flag combination on AIX. */
+			if (getaddrinfo(hostname1, NULL, &hints, &ppres) == EAI_BADFLAGS) {
+				/* This fails on an IPv6-only box, but not with
+				   the EAI_BADFLAGS error. */
+				return 1;
+			}
+			if (getaddrinfo(hostname2, NULL, &hints, &ppres) == 0) {
+				/* IPv4 lookup works - good enough. */
+				return 0;
+			}
+			/* Uh-oh, no IPv4. Are we IPv6-only ? */
+			return getaddrinfo(hostname3, NULL, &hints, &ppres) != 0 ? 1 : 0;
+		}],
+		libreplace_cv_HAVE_GETADDRINFO=yes,
+		libreplace_cv_HAVE_GETADDRINFO=no)
+fi
+
+if test x"$libreplace_cv_HAVE_GETADDRINFO" = x"yes"; then
+	AC_DEFINE(HAVE_GETADDRINFO,1,[Whether the system has getaddrinfo])
+	AC_DEFINE(HAVE_GETNAMEINFO,1,[Whether the system has getnameinfo])
+	AC_DEFINE(HAVE_FREEADDRINFO,1,[Whether the system has freeaddrinfo])
+	AC_DEFINE(HAVE_GAI_STRERROR,1,[Whether the system has gai_strerror])
+else
+	LIBREPLACE_NETWORK_OBJS="${LIBREPLACE_NETWORK_OBJS} $libreplacedir/getaddrinfo.o"
+fi
+
+AC_CHECK_HEADERS([ifaddrs.h])
+
+dnl Used when getifaddrs is not available
+AC_CHECK_MEMBERS([struct sockaddr.sa_len], 
+	 [AC_DEFINE(HAVE_SOCKADDR_SA_LEN, 1, [Whether struct sockaddr has a sa_len member])],
+	 [],
+	 [#include <sys/socket.h>])
+
+dnl test for getifaddrs and freeifaddrs
+AC_CACHE_CHECK([for getifaddrs and freeifaddrs],libreplace_cv_HAVE_GETIFADDRS,[
+AC_TRY_LINK([
+#include <sys/types.h>
+#if STDC_HEADERS
+#include <stdlib.h>
+#include <stddef.h>
+#endif
+#include <sys/socket.h>
+#include <netinet/in.h>
+#include <arpa/inet.h>
+#include <ifaddrs.h>
+#include <netdb.h>],
+[
+struct ifaddrs *ifp = NULL;
+int ret = getifaddrs (&ifp);
+freeifaddrs(ifp);
+],
+libreplace_cv_HAVE_GETIFADDRS=yes,libreplace_cv_HAVE_GETIFADDRS=no)])
+if test x"$libreplace_cv_HAVE_GETIFADDRS" = x"yes"; then
+    AC_DEFINE(HAVE_GETIFADDRS,1,[Whether the system has getifaddrs])
+    AC_DEFINE(HAVE_FREEIFADDRS,1,[Whether the system has freeifaddrs])
+	AC_DEFINE(HAVE_STRUCT_IFADDRS,1,[Whether struct ifaddrs is available])
+fi
+
+##################
+# look for a method of finding the list of network interfaces
+iface=no;
+AC_CACHE_CHECK([for iface getifaddrs],libreplace_cv_HAVE_IFACE_GETIFADDRS,[
+AC_TRY_RUN([
+#define HAVE_IFACE_GETIFADDRS 1
+#define NO_CONFIG_H 1
+#define AUTOCONF_TEST 1
+#define SOCKET_WRAPPER_NOT_REPLACE
+#include "$libreplacedir/replace.c"
+#include "$libreplacedir/inet_ntop.c"
+#include "$libreplacedir/snprintf.c"
+#include "$libreplacedir/getifaddrs.c"
+#define getifaddrs_test main
+#include "$libreplacedir/test/getifaddrs.c"],
+           libreplace_cv_HAVE_IFACE_GETIFADDRS=yes,libreplace_cv_HAVE_IFACE_GETIFADDRS=no,libreplace_cv_HAVE_IFACE_GETIFADDRS=cross)])
+if test x"$libreplace_cv_HAVE_IFACE_GETIFADDRS" = x"yes"; then
+    iface=yes;AC_DEFINE(HAVE_IFACE_GETIFADDRS,1,[Whether iface getifaddrs is available])
+else
+	LIBREPLACE_NETWORK_OBJS="${LIBREPLACE_NETWORK_OBJS} $libreplacedir/getifaddrs.o"
+fi
+
+
+if test $iface = no; then
+AC_CACHE_CHECK([for iface AIX],libreplace_cv_HAVE_IFACE_AIX,[
+AC_TRY_RUN([
+#define HAVE_IFACE_AIX 1
+#define NO_CONFIG_H 1
+#define AUTOCONF_TEST 1
+#undef _XOPEN_SOURCE_EXTENDED
+#define SOCKET_WRAPPER_NOT_REPLACE
+#include "$libreplacedir/replace.c"
+#include "$libreplacedir/inet_ntop.c"
+#include "$libreplacedir/snprintf.c"
+#include "$libreplacedir/getifaddrs.c"
+#define getifaddrs_test main
+#include "$libreplacedir/test/getifaddrs.c"],
+           libreplace_cv_HAVE_IFACE_AIX=yes,libreplace_cv_HAVE_IFACE_AIX=no,libreplace_cv_HAVE_IFACE_AIX=cross)])
+if test x"$libreplace_cv_HAVE_IFACE_AIX" = x"yes"; then
+    iface=yes;AC_DEFINE(HAVE_IFACE_AIX,1,[Whether iface AIX is available])
+fi
+fi
+
+
+if test $iface = no; then
+AC_CACHE_CHECK([for iface ifconf],libreplace_cv_HAVE_IFACE_IFCONF,[
+AC_TRY_RUN([
+#define HAVE_IFACE_IFCONF 1
+#define NO_CONFIG_H 1
+#define AUTOCONF_TEST 1
+#define SOCKET_WRAPPER_NOT_REPLACE
+#include "$libreplacedir/replace.c"
+#include "$libreplacedir/inet_ntop.c"
+#include "$libreplacedir/snprintf.c"
+#include "$libreplacedir/getifaddrs.c"
+#define getifaddrs_test main
+#include "$libreplacedir/test/getifaddrs.c"],
+           libreplace_cv_HAVE_IFACE_IFCONF=yes,libreplace_cv_HAVE_IFACE_IFCONF=no,libreplace_cv_HAVE_IFACE_IFCONF=cross)])
+if test x"$libreplace_cv_HAVE_IFACE_IFCONF" = x"yes"; then
+    iface=yes;AC_DEFINE(HAVE_IFACE_IFCONF,1,[Whether iface ifconf is available])
+fi
+fi
+
+if test $iface = no; then
+AC_CACHE_CHECK([for iface ifreq],libreplace_cv_HAVE_IFACE_IFREQ,[
+AC_TRY_RUN([
+#define HAVE_IFACE_IFREQ 1
+#define NO_CONFIG_H 1
+#define AUTOCONF_TEST 1
+#define SOCKET_WRAPPER_NOT_REPLACE
+#include "$libreplacedir/replace.c"
+#include "$libreplacedir/inet_ntop.c"
+#include "$libreplacedir/snprintf.c"
+#include "$libreplacedir/getifaddrs.c"
+#define getifaddrs_test main
+#include "$libreplacedir/test/getifaddrs.c"],
+           libreplace_cv_HAVE_IFACE_IFREQ=yes,libreplace_cv_HAVE_IFACE_IFREQ=no,libreplace_cv_HAVE_IFACE_IFREQ=cross)])
+if test x"$libreplace_cv_HAVE_IFACE_IFREQ" = x"yes"; then
+    iface=yes;AC_DEFINE(HAVE_IFACE_IFREQ,1,[Whether iface ifreq is available])
+fi
+fi
+
+dnl Some old Linux systems have broken header files and
+dnl miss the IPV6_V6ONLY define in netinet/in.h,
+dnl but have it in linux/in6.h.
+dnl We can't include both files so we just check if the value
+dnl if defined and do the replacement in system/network.h
+AC_CACHE_CHECK([for IPV6_V6ONLY support],libreplace_cv_HAVE_IPV6_V6ONLY,[
+	AC_TRY_COMPILE([
+#include <stdlib.h> /* for NULL */
+#include <sys/socket.h>
+#include <sys/types.h>
+#include <netdb.h>
+#include <netinet/in.h>
+		],
+		[
+#ifndef IPV6_V6ONLY
+#error no IPV6_V6ONLY
+#endif
+		],[
+		libreplace_cv_HAVE_IPV6_V6ONLY=yes
+		],[
+		libreplace_cv_HAVE_IPV6_V6ONLY=no
+		])
+])
+if test x"$libreplace_cv_HAVE_IPV6_V6ONLY" != x"yes"; then
+   dnl test for IPV6_V6ONLY
+   AC_CACHE_CHECK([for IPV6_V6ONLY in linux/in6.h],libreplace_cv_HAVE_LINUX_IPV6_V6ONLY_26,[
+	AC_TRY_COMPILE([
+	#include <linux/in6.h>
+		],
+		[
+	#if (IPV6_V6ONLY != 26)
+	#error no linux IPV6_V6ONLY
+	#endif
+		],[
+		libreplace_cv_HAVE_LINUX_IPV6_V6ONLY_26=yes
+		],[
+		libreplace_cv_HAVE_LINUX_IPV6_V6ONLY_26=no
+		])
+	])
+	if test x"$libreplace_cv_HAVE_LINUX_IPV6_V6ONLY_26" = x"yes"; then
+		AC_DEFINE(HAVE_LINUX_IPV6_V6ONLY_26,1,[Whether the system has IPV6_V6ONLY in linux/in6.h])
+	fi
+fi
+
+dnl test for ipv6
+AC_CACHE_CHECK([for ipv6 support],libreplace_cv_HAVE_IPV6,[
+	AC_TRY_LINK([
+#include <stdlib.h> /* for NULL */
+#include <sys/socket.h>
+#include <sys/types.h>
+#include <netdb.h>
+#include <netinet/in.h>
+		],
+		[
+struct sockaddr_storage sa_store;
+struct addrinfo *ai = NULL;
+struct in6_addr in6addr;
+int idx = if_nametoindex("iface1");
+int s = socket(AF_INET6, SOCK_STREAM, 0);
+int ret = getaddrinfo(NULL, NULL, NULL, &ai);
+if (ret != 0) {
+	const char *es = gai_strerror(ret);
+}
+freeaddrinfo(ai);
+{
+	int val = 1;
+	#ifdef HAVE_LINUX_IPV6_V6ONLY_26
+	#define IPV6_V6ONLY 26
+	#endif
+	ret = setsockopt(s, IPPROTO_IPV6, IPV6_V6ONLY,
+			 (const void *)&val, sizeof(val));
+}
+		],[
+		libreplace_cv_HAVE_IPV6=yes
+		],[
+		libreplace_cv_HAVE_IPV6=no
+		])
+])
+if test x"$libreplace_cv_HAVE_IPV6" = x"yes"; then
+    AC_DEFINE(HAVE_IPV6,1,[Whether the system has IPv6 support])
+fi
+
+LIBS=$old_LIBS
+CPPFLAGS="$libreplace_SAVE_CPPFLAGS"
+
+AC_CACHE_CHECK([for SO_PEERCRED],libreplace_cv_HAVE_PEERCRED,[
+AC_TRY_COMPILE([#include <sys/types.h>
+#include <sys/socket.h>],
+[struct ucred cred;
+ socklen_t cred_len;
+ int ret = getsockopt(0, SOL_SOCKET, SO_PEERCRED, &cred, &cred_len);
+],
+libreplace_cv_HAVE_PEERCRED=yes,libreplace_cv_HAVE_PEERCRED=no,libreplace_cv_HAVE_PEERCRED=cross)])
+if test x"$libreplace_cv_HAVE_PEERCRED" = x"yes"; then
+    AC_DEFINE(HAVE_PEERCRED,1,[Whether we can use SO_PEERCRED to get socket credentials])
+fi
+
+AC_CACHE_CHECK([for getpeereid],libreplace_cv_HAVE_GETPEEREID,[
+AC_TRY_LINK([#include <sys/types.h>
+#include <unistd.h>],
+[uid_t uid; gid_t gid; int ret;
+ ret = getpeereid(0, &uid, &gid);
+],
+libreplace_cv_HAVE_GETPEEREID=yes,libreplace_cv_HAVE_GETPEEREID=no)])
+if test x"$libreplace_cv_HAVE_GETPEEREID" = xyes; then
+   AC_DEFINE(HAVE_GETPEEREID,1,
+	     [Whether we have getpeereid to get socket credentials])
+fi
+
+LIBREPLACEOBJ="${LIBREPLACEOBJ} ${LIBREPLACE_NETWORK_OBJS}"
+
+echo "LIBREPLACE_NETWORK_CHECKS: END"
+]) dnl end AC_LIBREPLACE_NETWORK_CHECKS
diff --git a/ctdb/lib/replace/poll.c b/ctdb/lib/replace/poll.c
new file mode 100644
index 0000000000..1105617b63
--- /dev/null
+++ b/ctdb/lib/replace/poll.c
@@ -0,0 +1,139 @@
+/*
+   Unix SMB/CIFS implementation.
+   poll.c - poll wrapper
+
+   This file is based on code from libssh (LGPLv2.1+ at the time it
+   was downloaded), thus the following copyrights:
+
+   Copyright (c) 2009-2010 by Andreas Schneider <mail@cynapses.org>
+   Copyright (c) 2003-2009 by Aris Adamantiadis
+   Copyright (c) 2009 Aleksandar Kanchev
+   Copyright (C) Volker Lendecke 2011
+
+     ** NOTE! The following LGPL license applies to the replace
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "replace.h"
+#include "system/select.h"
+#ifdef HAVE_SYS_TIME_H
+#include <sys/time.h>
+#endif
+#ifdef HAVE_SYS_IOCTL_H
+#include <sys/ioctl.h>
+#endif
+
+
+int rep_poll(struct pollfd *fds, nfds_t nfds, int timeout)
+{
+	fd_set rfds, wfds, efds;
+	struct timeval tv, *ptv;
+	int max_fd;
+	int rc;
+	nfds_t i;
+
+	if ((fds == NULL) && (nfds != 0)) {
+		errno = EFAULT;
+		return -1;
+	}
+
+	FD_ZERO(&rfds);
+	FD_ZERO(&wfds);
+	FD_ZERO(&efds);
+
+	rc = 0;
+	max_fd = 0;
+
+	/* compute fd_sets and find largest descriptor */
+	for (i = 0; i < nfds; i++) {
+		if ((fds[i].fd < 0) || (fds[i].fd >= FD_SETSIZE)) {
+			fds[i].revents = POLLNVAL;
+			continue;
+		}
+
+		if (fds[i].events & (POLLIN | POLLRDNORM)) {
+			FD_SET(fds[i].fd, &rfds);
+		}
+		if (fds[i].events & (POLLOUT | POLLWRNORM | POLLWRBAND)) {
+			FD_SET(fds[i].fd, &wfds);
+		}
+		if (fds[i].events & (POLLPRI | POLLRDBAND)) {
+			FD_SET(fds[i].fd, &efds);
+		}
+		if (fds[i].fd > max_fd &&
+		    (fds[i].events & (POLLIN | POLLOUT | POLLPRI |
+				      POLLRDNORM | POLLRDBAND |
+				      POLLWRNORM | POLLWRBAND))) {
+			max_fd = fds[i].fd;
+		}
+	}
+
+	if (timeout < 0) {
+		ptv = NULL;
+	} else {
+		ptv = &tv;
+		if (timeout == 0) {
+			tv.tv_sec = 0;
+			tv.tv_usec = 0;
+		} else {
+			tv.tv_sec = timeout / 1000;
+			tv.tv_usec = (timeout % 1000) * 1000;
+		}
+	}
+
+	rc = select(max_fd + 1, &rfds, &wfds, &efds, ptv);
+	if (rc < 0) {
+		return -1;
+	}
+
+	for (rc = 0, i = 0; i < nfds; i++) {
+		if ((fds[i].fd < 0) || (fds[i].fd >= FD_SETSIZE)) {
+			continue;
+		}
+
+		fds[i].revents = 0;
+
+		if (FD_ISSET(fds[i].fd, &rfds)) {
+			int err = errno;
+			int available = 0;
+			int ret;
+
+			/* support for POLLHUP */
+			ret = ioctl(fds[i].fd, FIONREAD, &available);
+			if ((ret == -1) || (available == 0)) {
+				fds[i].revents |= POLLHUP;
+			} else {
+				fds[i].revents |= fds[i].events
+					& (POLLIN | POLLRDNORM);
+			}
+
+			errno = err;
+		}
+		if (FD_ISSET(fds[i].fd, &wfds)) {
+			fds[i].revents |= fds[i].events
+				& (POLLOUT | POLLWRNORM | POLLWRBAND);
+		}
+		if (FD_ISSET(fds[i].fd, &efds)) {
+			fds[i].revents |= fds[i].events
+				& (POLLPRI | POLLRDBAND);
+		}
+		if (fds[i].revents & ~POLLHUP) {
+			rc++;
+		}
+	}
+	return rc;
+}
diff --git a/ctdb/lib/replace/repdir.m4 b/ctdb/lib/replace/repdir.m4
new file mode 100644
index 0000000000..682ab44c56
--- /dev/null
+++ b/ctdb/lib/replace/repdir.m4
@@ -0,0 +1,81 @@
+AC_CACHE_CHECK([for broken readdir],libreplace_cv_READDIR_NEEDED,[
+	AC_TRY_RUN([
+#define test_readdir_os2_delete main
+#include "$libreplacedir/test/os2_delete.c"],
+	[libreplace_cv_READDIR_NEEDED=no],
+	[libreplace_cv_READDIR_NEEDED=yes],
+	[libreplace_cv_READDIR_NEEDED="assuming not"])
+])
+
+AC_CHECK_FUNCS(dirfd)
+AC_HAVE_DECL(dirfd, [#include <dirent.h>])
+
+#
+# try to replace with getdirentries() if needed
+#
+if test x"$libreplace_cv_READDIR_NEEDED" = x"yes"; then
+AC_CHECK_FUNCS(getdirentries)
+AC_VERIFY_C_PROTOTYPE([long telldir(const DIR *dir)],
+	[
+	return 0;
+	],[
+	AC_DEFINE(TELLDIR_TAKES_CONST_DIR, 1, [Whether telldir takes a const pointer])
+	],[],[
+	#include <dirent.h>
+	])
+
+AC_VERIFY_C_PROTOTYPE([int seekdir(DIR *dir, long ofs)],
+	[
+	return 0;
+	],[
+	AC_DEFINE(SEEKDIR_RETURNS_INT, 1, [Whether seekdir returns an int])
+	],[],[
+	#include <dirent.h>
+	])
+AC_CACHE_CHECK([for replacing readdir using getdirentries()],libreplace_cv_READDIR_GETDIRENTRIES,[
+	AC_TRY_RUN([
+#define _LIBREPLACE_REPLACE_H
+#include "$libreplacedir/repdir_getdirentries.c"
+#define test_readdir_os2_delete main
+#include "$libreplacedir/test/os2_delete.c"],
+	[libreplace_cv_READDIR_GETDIRENTRIES=yes],
+	[libreplace_cv_READDIR_GETDIRENTRIES=no])
+])
+fi
+if test x"$libreplace_cv_READDIR_GETDIRENTRIES" = x"yes"; then
+	AC_DEFINE(REPLACE_READDIR,1,[replace readdir])
+	AC_DEFINE(REPLACE_READDIR_GETDIRENTRIES,1,[replace readdir using getdirentries()])
+	LIBREPLACEOBJ="${LIBREPLACEOBJ} $libreplacedir/repdir_getdirentries.o"
+	libreplace_cv_READDIR_NEEDED=no
+fi
+
+#
+# try to replace with getdents() if needed
+#
+if test x"$libreplace_cv_READDIR_NEEDED" = x"yes"; then
+AC_CHECK_FUNCS(getdents)
+AC_CACHE_CHECK([for replacing readdir using getdents()],libreplace_cv_READDIR_GETDENTS,[
+	AC_TRY_RUN([
+#define _LIBREPLACE_REPLACE_H
+#error _donot_use_getdents_replacement_anymore
+#include "$libreplacedir/repdir_getdents.c"
+#define test_readdir_os2_delete main
+#include "$libreplacedir/test/os2_delete.c"],
+	[libreplace_cv_READDIR_GETDENTS=yes],
+	[libreplace_cv_READDIR_GETDENTS=no])
+])
+fi
+if test x"$libreplace_cv_READDIR_GETDENTS" = x"yes"; then
+	AC_DEFINE(REPLACE_READDIR,1,[replace readdir])
+	AC_DEFINE(REPLACE_READDIR_GETDENTS,1,[replace readdir using getdents()])
+	LIBREPLACEOBJ="${LIBREPLACEOBJ} $libreplacedir/repdir_getdents.o"
+	libreplace_cv_READDIR_NEEDED=no
+fi
+
+AC_MSG_CHECKING([a usable readdir()])
+if test x"$libreplace_cv_READDIR_NEEDED" = x"yes"; then
+	AC_MSG_RESULT(no)
+	AC_MSG_WARN([the provided readdir() is broken])
+else
+	AC_MSG_RESULT(yes)
+fi
diff --git a/ctdb/lib/replace/repdir_getdents.c b/ctdb/lib/replace/repdir_getdents.c
new file mode 100644
index 0000000000..afc634a796
--- /dev/null
+++ b/ctdb/lib/replace/repdir_getdents.c
@@ -0,0 +1,166 @@
+/* 
+   Unix SMB/CIFS implementation.
+
+   Copyright (C) Andrew Tridgell 2005
+
+     ** NOTE! The following LGPL license applies to the replace
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+/*
+  a replacement for opendir/readdir/telldir/seekdir/closedir for BSD systems
+
+  This is needed because the existing directory handling in FreeBSD
+  and OpenBSD (and possibly NetBSD) doesn't correctly handle unlink()
+  on files in a directory where telldir() has been used. On a block
+  boundary it will occasionally miss a file when seekdir() is used to
+  return to a position previously recorded with telldir().
+
+  This also fixes a severe performance and memory usage problem with
+  telldir() on BSD systems. Each call to telldir() in BSD adds an
+  entry to a linked list, and those entries are cleaned up on
+  closedir(). This means with a large directory closedir() can take an
+  arbitrary amount of time, causing network timeouts as millions of
+  telldir() entries are freed
+
+  Note! This replacement code is not portable. It relies on getdents()
+  always leaving the file descriptor at a seek offset that is a
+  multiple of DIR_BUF_SIZE. If the code detects that this doesn't
+  happen then it will abort(). It also does not handle directories
+  with offsets larger than can be stored in a long,
+
+  This code is available under other free software licenses as
+  well. Contact the author.
+*/
+
+#include <stdlib.h>
+#include <sys/stat.h>
+#include <unistd.h>
+#include <sys/types.h>
+#include <errno.h>
+#include <fcntl.h>
+#include <dirent.h>
+
+#define DIR_BUF_BITS 9
+#define DIR_BUF_SIZE (1<<DIR_BUF_BITS)
+
+struct dir_buf {
+	int fd;
+	int nbytes, ofs;
+	off_t seekpos;
+	char buf[DIR_BUF_SIZE];
+};
+
+DIR *opendir(const char *dname)
+{
+	struct dir_buf *d;
+	struct stat sb;
+	d = malloc(sizeof(*d));
+	if (d == NULL) {
+		errno = ENOMEM;
+		return NULL;
+	}
+	d->fd = open(dname, O_RDONLY);
+	if (d->fd == -1) {
+		free(d);
+		return NULL;
+	}
+	if (fstat(d->fd, &sb) < 0) {
+		close(d->fd);
+		free(d);
+		return NULL;
+	}
+	if (!S_ISDIR(sb.st_mode)) {
+		close(d->fd);
+		free(d);   
+		errno = ENOTDIR;
+		return NULL;
+	}
+	d->ofs = 0;
+	d->seekpos = 0;
+	d->nbytes = 0;
+	return (DIR *)d;
+}
+
+struct dirent *readdir(DIR *dir)
+{
+	struct dir_buf *d = (struct dir_buf *)dir;
+	struct dirent *de;
+
+	if (d->ofs >= d->nbytes) {
+		d->seekpos = lseek(d->fd, 0, SEEK_CUR);
+		d->nbytes = getdents(d->fd, d->buf, DIR_BUF_SIZE);
+		d->ofs = 0;
+	}
+	if (d->ofs >= d->nbytes) {
+		return NULL;
+	}
+	de = (struct dirent *)&d->buf[d->ofs];
+	d->ofs += de->d_reclen;
+	return de;
+}
+
+long telldir(DIR *dir)
+{
+	struct dir_buf *d = (struct dir_buf *)dir;
+	if (d->ofs >= d->nbytes) {
+		d->seekpos = lseek(d->fd, 0, SEEK_CUR);
+		d->ofs = 0;
+		d->nbytes = 0;
+	}
+	/* this relies on seekpos always being a multiple of
+	   DIR_BUF_SIZE. Is that always true on BSD systems? */
+	if (d->seekpos & (DIR_BUF_SIZE-1)) {
+		abort();
+	}
+	return d->seekpos + d->ofs;
+}
+
+void seekdir(DIR *dir, long ofs)
+{
+	struct dir_buf *d = (struct dir_buf *)dir;
+	d->seekpos = lseek(d->fd, ofs & ~(DIR_BUF_SIZE-1), SEEK_SET);
+	d->nbytes = getdents(d->fd, d->buf, DIR_BUF_SIZE);
+	d->ofs = 0;
+	while (d->ofs < (ofs & (DIR_BUF_SIZE-1))) {
+		if (readdir(dir) == NULL) break;
+	}
+}
+
+void rewinddir(DIR *dir)
+{
+	seekdir(dir, 0);
+}
+
+int closedir(DIR *dir)
+{
+	struct dir_buf *d = (struct dir_buf *)dir;
+	int r = close(d->fd);
+	if (r != 0) {
+		return r;
+	}
+	free(d);
+	return 0;
+}
+
+#ifndef dirfd
+/* darn, this is a macro on some systems. */
+int dirfd(DIR *dir)
+{
+	struct dir_buf *d = (struct dir_buf *)dir;
+	return d->fd;
+}
+#endif
diff --git a/ctdb/lib/replace/repdir_getdirentries.c b/ctdb/lib/replace/repdir_getdirentries.c
new file mode 100644
index 0000000000..197e5931fc
--- /dev/null
+++ b/ctdb/lib/replace/repdir_getdirentries.c
@@ -0,0 +1,183 @@
+/* 
+   Unix SMB/CIFS implementation.
+
+   Copyright (C) Andrew Tridgell 2005
+
+     ** NOTE! The following LGPL license applies to the replace
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+/*
+  a replacement for opendir/readdir/telldir/seekdir/closedir for BSD
+  systems using getdirentries
+
+  This is needed because the existing directory handling in FreeBSD
+  and OpenBSD (and possibly NetBSD) doesn't correctly handle unlink()
+  on files in a directory where telldir() has been used. On a block
+  boundary it will occasionally miss a file when seekdir() is used to
+  return to a position previously recorded with telldir().
+
+  This also fixes a severe performance and memory usage problem with
+  telldir() on BSD systems. Each call to telldir() in BSD adds an
+  entry to a linked list, and those entries are cleaned up on
+  closedir(). This means with a large directory closedir() can take an
+  arbitrary amount of time, causing network timeouts as millions of
+  telldir() entries are freed
+
+  Note! This replacement code is not portable. It relies on
+  getdirentries() always leaving the file descriptor at a seek offset
+  that is a multiple of DIR_BUF_SIZE. If the code detects that this
+  doesn't happen then it will abort(). It also does not handle
+  directories with offsets larger than can be stored in a long,
+
+  This code is available under other free software licenses as
+  well. Contact the author.
+*/
+
+#include "replace.h"
+#include <stdlib.h>
+#include <sys/stat.h>
+#include <unistd.h>
+#include <sys/types.h>
+#include <errno.h>
+#include <fcntl.h>
+#include <dirent.h>
+
+#define DIR_BUF_BITS 9
+#define DIR_BUF_SIZE (1<<DIR_BUF_BITS)
+
+struct dir_buf {
+	int fd;
+	int nbytes, ofs;
+	off_t seekpos;
+	char buf[DIR_BUF_SIZE];
+};
+
+DIR *opendir(const char *dname)
+{
+	struct dir_buf *d;
+	struct stat sb;
+	d = malloc(sizeof(*d));
+	if (d == NULL) {
+		errno = ENOMEM;
+		return NULL;
+	}
+	d->fd = open(dname, O_RDONLY);
+	if (d->fd == -1) {
+		free(d);
+		return NULL;
+	}
+	if (fstat(d->fd, &sb) < 0) {
+		close(d->fd);
+		free(d);
+		return NULL;
+	}
+	if (!S_ISDIR(sb.st_mode)) {
+		close(d->fd);
+		free(d);   
+		errno = ENOTDIR;
+		return NULL;
+	}
+	d->ofs = 0;
+	d->seekpos = 0;
+	d->nbytes = 0;
+	return (DIR *)d;
+}
+
+struct dirent *readdir(DIR *dir)
+{
+	struct dir_buf *d = (struct dir_buf *)dir;
+	struct dirent *de;
+
+	if (d->ofs >= d->nbytes) {
+		long pos;
+		d->nbytes = getdirentries(d->fd, d->buf, DIR_BUF_SIZE, &pos);
+		d->seekpos = pos;
+		d->ofs = 0;
+	}
+	if (d->ofs >= d->nbytes) {
+		return NULL;
+	}
+	de = (struct dirent *)&d->buf[d->ofs];
+	d->ofs += de->d_reclen;
+	return de;
+}
+
+#ifdef TELLDIR_TAKES_CONST_DIR
+long telldir(const DIR *dir)
+#else
+long telldir(DIR *dir)
+#endif
+{
+	struct dir_buf *d = (struct dir_buf *)dir;
+	if (d->ofs >= d->nbytes) {
+		d->seekpos = lseek(d->fd, 0, SEEK_CUR);
+		d->ofs = 0;
+		d->nbytes = 0;
+	}
+	/* this relies on seekpos always being a multiple of
+	   DIR_BUF_SIZE. Is that always true on BSD systems? */
+	if (d->seekpos & (DIR_BUF_SIZE-1)) {
+		abort();
+	}
+	return d->seekpos + d->ofs;
+}
+
+#ifdef SEEKDIR_RETURNS_INT
+int seekdir(DIR *dir, long ofs)
+#else
+void seekdir(DIR *dir, long ofs)
+#endif
+{
+	struct dir_buf *d = (struct dir_buf *)dir;
+	long pos;
+	d->seekpos = lseek(d->fd, ofs & ~(DIR_BUF_SIZE-1), SEEK_SET);
+	d->nbytes = getdirentries(d->fd, d->buf, DIR_BUF_SIZE, &pos);
+	d->ofs = 0;
+	while (d->ofs < (ofs & (DIR_BUF_SIZE-1))) {
+		if (readdir(dir) == NULL) break;
+	}
+#ifdef SEEKDIR_RETURNS_INT
+	return -1;
+#endif
+}
+
+void rewinddir(DIR *dir)
+{
+	seekdir(dir, 0);
+}
+
+int closedir(DIR *dir)
+{
+	struct dir_buf *d = (struct dir_buf *)dir;
+	int r = close(d->fd);
+	if (r != 0) {
+		return r;
+	}
+	free(d);
+	return 0;
+}
+
+#ifndef dirfd
+/* darn, this is a macro on some systems. */
+int dirfd(DIR *dir)
+{
+	struct dir_buf *d = (struct dir_buf *)dir;
+	return d->fd;
+}
+#endif
+
+
diff --git a/ctdb/lib/replace/replace-test.h b/ctdb/lib/replace/replace-test.h
new file mode 100644
index 0000000000..ed8e75e3b1
--- /dev/null
+++ b/ctdb/lib/replace/replace-test.h
@@ -0,0 +1,9 @@
+#ifndef __LIB_REPLACE_REPLACE_TEST_H__
+#define __LIB_REPLACE_REPLACE_TEST_H__
+
+int libreplace_test_strptime(void);
+int test_readdir_os2_delete(void);
+int getifaddrs_test(void);
+
+#endif /* __LIB_REPLACE_REPLACE_TEST_H__ */
+
diff --git a/ctdb/lib/replace/replace-testsuite.h b/ctdb/lib/replace/replace-testsuite.h
new file mode 100644
index 0000000000..b28dbec067
--- /dev/null
+++ b/ctdb/lib/replace/replace-testsuite.h
@@ -0,0 +1,10 @@
+#ifndef __LIB_REPLACE_REPLACE_TESTSUITE_H__
+#define __LIB_REPLACE_REPLACE_TESTSUITE_H__
+
+#include <stdbool.h>
+struct torture_context;
+
+bool torture_local_replace(struct torture_context *ctx);
+
+#endif /* __LIB_REPLACE_REPLACE_TESTSUITE_H__ */
+
diff --git a/ctdb/lib/replace/replace.c b/ctdb/lib/replace/replace.c
new file mode 100644
index 0000000000..37edb310c5
--- /dev/null
+++ b/ctdb/lib/replace/replace.c
@@ -0,0 +1,902 @@
+/* 
+   Unix SMB/CIFS implementation.
+   replacement routines for broken systems
+   Copyright (C) Andrew Tridgell 1992-1998
+   Copyright (C) Jelmer Vernooij 2005-2008
+   Copyright (C) Matthieu Patou  2010
+
+     ** NOTE! The following LGPL license applies to the replace
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "replace.h"
+
+#include "system/filesys.h"
+#include "system/time.h"
+#include "system/network.h"
+#include "system/passwd.h"
+#include "system/syslog.h"
+#include "system/locale.h"
+#include "system/wait.h"
+
+#ifdef _WIN32
+#define mkdir(d,m) _mkdir(d)
+#endif
+
+void replace_dummy(void);
+void replace_dummy(void) {}
+
+#ifndef HAVE_FTRUNCATE
+ /*******************************************************************
+ftruncate for operating systems that don't have it
+********************************************************************/
+int rep_ftruncate(int f, off_t l)
+{
+#ifdef HAVE_CHSIZE
+      return chsize(f,l);
+#elif defined(F_FREESP)
+      struct  flock   fl;
+
+      fl.l_whence = 0;
+      fl.l_len = 0;
+      fl.l_start = l;
+      fl.l_type = F_WRLCK;
+      return fcntl(f, F_FREESP, &fl);
+#else
+#error "you must have a ftruncate function"
+#endif
+}
+#endif /* HAVE_FTRUNCATE */
+
+
+#ifndef HAVE_STRLCPY
+/* like strncpy but does not 0 fill the buffer and always null 
+   terminates. bufsize is the size of the destination buffer */
+size_t rep_strlcpy(char *d, const char *s, size_t bufsize)
+{
+	size_t len = strlen(s);
+	size_t ret = len;
+	if (bufsize <= 0) return 0;
+	if (len >= bufsize) len = bufsize-1;
+	memcpy(d, s, len);
+	d[len] = 0;
+	return ret;
+}
+#endif
+
+#ifndef HAVE_STRLCAT
+/* like strncat but does not 0 fill the buffer and always null 
+   terminates. bufsize is the length of the buffer, which should
+   be one more than the maximum resulting string length */
+size_t rep_strlcat(char *d, const char *s, size_t bufsize)
+{
+	size_t len1 = strlen(d);
+	size_t len2 = strlen(s);
+	size_t ret = len1 + len2;
+
+	if (len1+len2 >= bufsize) {
+		if (bufsize < (len1+1)) {
+			return ret;
+		}
+		len2 = bufsize - (len1+1);
+	}
+	if (len2 > 0) {
+		memcpy(d+len1, s, len2);
+		d[len1+len2] = 0;
+	}
+	return ret;
+}
+#endif
+
+#ifndef HAVE_MKTIME
+/*******************************************************************
+a mktime() replacement for those who don't have it - contributed by 
+C.A. Lademann <cal@zls.com>
+Corrections by richard.kettlewell@kewill.com
+********************************************************************/
+
+#define  MINUTE  60
+#define  HOUR    60*MINUTE
+#define  DAY             24*HOUR
+#define  YEAR    365*DAY
+time_t rep_mktime(struct tm *t)
+{
+  struct tm       *u;
+  time_t  epoch = 0;
+  int n;
+  int             mon [] = { 31, 28, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31 },
+  y, m, i;
+
+  if(t->tm_year < 70)
+    return((time_t)-1);
+
+  n = t->tm_year + 1900 - 1;
+  epoch = (t->tm_year - 70) * YEAR + 
+    ((n / 4 - n / 100 + n / 400) - (1969 / 4 - 1969 / 100 + 1969 / 400)) * DAY;
+
+  y = t->tm_year + 1900;
+  m = 0;
+
+  for(i = 0; i < t->tm_mon; i++) {
+    epoch += mon [m] * DAY;
+    if(m == 1 && y % 4 == 0 && (y % 100 != 0 || y % 400 == 0))
+      epoch += DAY;
+    
+    if(++m > 11) {
+      m = 0;
+      y++;
+    }
+  }
+
+  epoch += (t->tm_mday - 1) * DAY;
+  epoch += t->tm_hour * HOUR + t->tm_min * MINUTE + t->tm_sec;
+  
+  if((u = localtime(&epoch)) != NULL) {
+    t->tm_sec = u->tm_sec;
+    t->tm_min = u->tm_min;
+    t->tm_hour = u->tm_hour;
+    t->tm_mday = u->tm_mday;
+    t->tm_mon = u->tm_mon;
+    t->tm_year = u->tm_year;
+    t->tm_wday = u->tm_wday;
+    t->tm_yday = u->tm_yday;
+    t->tm_isdst = u->tm_isdst;
+  }
+
+  return(epoch);
+}
+#endif /* !HAVE_MKTIME */
+
+
+#ifndef HAVE_INITGROUPS
+/****************************************************************************
+ some systems don't have an initgroups call 
+****************************************************************************/
+int rep_initgroups(char *name, gid_t id)
+{
+#ifndef HAVE_SETGROUPS
+	/* yikes! no SETGROUPS or INITGROUPS? how can this work? */
+	errno = ENOSYS;
+	return -1;
+#else /* HAVE_SETGROUPS */
+
+#include <grp.h>
+
+	gid_t *grouplst = NULL;
+	int max_gr = NGROUPS_MAX;
+	int ret;
+	int    i,j;
+	struct group *g;
+	char   *gr;
+	
+	if((grouplst = malloc(sizeof(gid_t) * max_gr)) == NULL) {
+		errno = ENOMEM;
+		return -1;
+	}
+
+	grouplst[0] = id;
+	i = 1;
+	while (i < max_gr && ((g = (struct group *)getgrent()) != (struct group *)NULL)) {
+		if (g->gr_gid == id)
+			continue;
+		j = 0;
+		gr = g->gr_mem[0];
+		while (gr && (*gr != (char)NULL)) {
+			if (strcmp(name,gr) == 0) {
+				grouplst[i] = g->gr_gid;
+				i++;
+				gr = (char *)NULL;
+				break;
+			}
+			gr = g->gr_mem[++j];
+		}
+	}
+	endgrent();
+	ret = setgroups(i, grouplst);
+	free(grouplst);
+	return ret;
+#endif /* HAVE_SETGROUPS */
+}
+#endif /* HAVE_INITGROUPS */
+
+
+#ifndef HAVE_MEMMOVE
+/*******************************************************************
+safely copies memory, ensuring no overlap problems.
+this is only used if the machine does not have its own memmove().
+this is not the fastest algorithm in town, but it will do for our
+needs.
+********************************************************************/
+void *rep_memmove(void *dest,const void *src,int size)
+{
+	unsigned long d,s;
+	int i;
+	if (dest==src || !size) return(dest);
+
+	d = (unsigned long)dest;
+	s = (unsigned long)src;
+
+	if ((d >= (s+size)) || (s >= (d+size))) {
+		/* no overlap */
+		memcpy(dest,src,size);
+		return(dest);
+	}
+
+	if (d < s) {
+		/* we can forward copy */
+		if (s-d >= sizeof(int) && 
+		    !(s%sizeof(int)) && 
+		    !(d%sizeof(int)) && 
+		    !(size%sizeof(int))) {
+			/* do it all as words */
+			int *idest = (int *)dest;
+			int *isrc = (int *)src;
+			size /= sizeof(int);
+			for (i=0;i<size;i++) idest[i] = isrc[i];
+		} else {
+			/* simplest */
+			char *cdest = (char *)dest;
+			char *csrc = (char *)src;
+			for (i=0;i<size;i++) cdest[i] = csrc[i];
+		}
+	} else {
+		/* must backward copy */
+		if (d-s >= sizeof(int) && 
+		    !(s%sizeof(int)) && 
+		    !(d%sizeof(int)) && 
+		    !(size%sizeof(int))) {
+			/* do it all as words */
+			int *idest = (int *)dest;
+			int *isrc = (int *)src;
+			size /= sizeof(int);
+			for (i=size-1;i>=0;i--) idest[i] = isrc[i];
+		} else {
+			/* simplest */
+			char *cdest = (char *)dest;
+			char *csrc = (char *)src;
+			for (i=size-1;i>=0;i--) cdest[i] = csrc[i];
+		}      
+	}
+	return(dest);
+}
+#endif /* HAVE_MEMMOVE */
+
+#ifndef HAVE_STRDUP
+/****************************************************************************
+duplicate a string
+****************************************************************************/
+char *rep_strdup(const char *s)
+{
+	size_t len;
+	char *ret;
+
+	if (!s) return(NULL);
+
+	len = strlen(s)+1;
+	ret = (char *)malloc(len);
+	if (!ret) return(NULL);
+	memcpy(ret,s,len);
+	return(ret);
+}
+#endif /* HAVE_STRDUP */
+
+#ifndef HAVE_SETLINEBUF
+void rep_setlinebuf(FILE *stream)
+{
+	setvbuf(stream, (char *)NULL, _IOLBF, 0);
+}
+#endif /* HAVE_SETLINEBUF */
+
+#ifndef HAVE_VSYSLOG
+#ifdef HAVE_SYSLOG
+void rep_vsyslog (int facility_priority, const char *format, va_list arglist)
+{
+	char *msg = NULL;
+	vasprintf(&msg, format, arglist);
+	if (!msg)
+		return;
+	syslog(facility_priority, "%s", msg);
+	free(msg);
+}
+#endif /* HAVE_SYSLOG */
+#endif /* HAVE_VSYSLOG */
+
+#ifndef HAVE_STRNLEN
+/**
+ Some platforms don't have strnlen
+**/
+ size_t rep_strnlen(const char *s, size_t max)
+{
+        size_t len;
+  
+        for (len = 0; len < max; len++) {
+                if (s[len] == '\0') {
+                        break;
+                }
+        }
+        return len;  
+}
+#endif
+  
+#ifndef HAVE_STRNDUP
+/**
+ Some platforms don't have strndup.
+**/
+char *rep_strndup(const char *s, size_t n)
+{
+	char *ret;
+	
+	n = strnlen(s, n);
+	ret = malloc(n+1);
+	if (!ret)
+		return NULL;
+	memcpy(ret, s, n);
+	ret[n] = 0;
+
+	return ret;
+}
+#endif
+
+#if !defined(HAVE_WAITPID) && defined(HAVE_WAIT4)
+int rep_waitpid(pid_t pid,int *status,int options)
+{
+  return wait4(pid, status, options, NULL);
+}
+#endif
+
+#ifndef HAVE_SETEUID
+int rep_seteuid(uid_t euid)
+{
+#ifdef HAVE_SETRESUID
+	return setresuid(-1, euid, -1);
+#else
+	errno = ENOSYS;
+	return -1;
+#endif
+}
+#endif
+
+#ifndef HAVE_SETEGID
+int rep_setegid(gid_t egid)
+{
+#ifdef HAVE_SETRESGID
+	return setresgid(-1, egid, -1);
+#else
+	errno = ENOSYS;
+	return -1;
+#endif
+}
+#endif
+
+/*******************************************************************
+os/2 also doesn't have chroot
+********************************************************************/
+#ifndef HAVE_CHROOT
+int rep_chroot(const char *dname)
+{
+	errno = ENOSYS;
+	return -1;
+}
+#endif
+
+/*****************************************************************
+ Possibly replace mkstemp if it is broken.
+*****************************************************************/  
+
+#ifndef HAVE_SECURE_MKSTEMP
+int rep_mkstemp(char *template)
+{
+	/* have a reasonable go at emulating it. Hope that
+	   the system mktemp() isn't completely hopeless */
+	mktemp(template);
+	if (template[0] == 0)
+		return -1;
+	return open(template, O_CREAT|O_EXCL|O_RDWR, 0600);
+}
+#endif
+
+#ifndef HAVE_MKDTEMP
+char *rep_mkdtemp(char *template)
+{
+	char *dname;
+	
+	if ((dname = mktemp(template))) {
+		if (mkdir(dname, 0700) >= 0) {
+			return dname;
+		}
+	}
+
+	return NULL;
+}
+#endif
+
+/*****************************************************************
+ Watch out: this is not thread safe.
+*****************************************************************/
+
+#ifndef HAVE_PREAD
+ssize_t rep_pread(int __fd, void *__buf, size_t __nbytes, off_t __offset)
+{
+	if (lseek(__fd, __offset, SEEK_SET) != __offset) {
+		return -1;
+	}
+	return read(__fd, __buf, __nbytes);
+}
+#endif
+
+/*****************************************************************
+ Watch out: this is not thread safe.
+*****************************************************************/
+
+#ifndef HAVE_PWRITE
+ssize_t rep_pwrite(int __fd, const void *__buf, size_t __nbytes, off_t __offset)
+{
+	if (lseek(__fd, __offset, SEEK_SET) != __offset) {
+		return -1;
+	}
+	return write(__fd, __buf, __nbytes);
+}
+#endif
+
+#ifndef HAVE_STRCASESTR
+char *rep_strcasestr(const char *haystack, const char *needle)
+{
+	const char *s;
+	size_t nlen = strlen(needle);
+	for (s=haystack;*s;s++) {
+		if (toupper(*needle) == toupper(*s) &&
+		    strncasecmp(s, needle, nlen) == 0) {
+			return (char *)((uintptr_t)s);
+		}
+	}
+	return NULL;
+}
+#endif
+
+#ifndef HAVE_STRTOK_R
+/* based on GLIBC version, copyright Free Software Foundation */
+char *rep_strtok_r(char *s, const char *delim, char **save_ptr)
+{
+	char *token;
+
+	if (s == NULL) s = *save_ptr;
+
+	s += strspn(s, delim);
+	if (*s == '\0') {
+		*save_ptr = s;
+		return NULL;
+	}
+
+	token = s;
+	s = strpbrk(token, delim);
+	if (s == NULL) {
+		*save_ptr = token + strlen(token);
+	} else {
+		*s = '\0';
+		*save_ptr = s + 1;
+	}
+
+	return token;
+}
+#endif
+
+
+#ifndef HAVE_STRTOLL
+long long int rep_strtoll(const char *str, char **endptr, int base)
+{
+#ifdef HAVE_STRTOQ
+	return strtoq(str, endptr, base);
+#elif defined(HAVE___STRTOLL) 
+	return __strtoll(str, endptr, base);
+#elif SIZEOF_LONG == SIZEOF_LONG_LONG
+	return (long long int) strtol(str, endptr, base);
+#else
+# error "You need a strtoll function"
+#endif
+}
+#else
+#ifdef HAVE_BSD_STRTOLL
+#ifdef HAVE_STRTOQ
+long long int rep_strtoll(const char *str, char **endptr, int base)
+{
+	long long int nb = strtoq(str, endptr, base);
+	/* In linux EINVAL is only returned if base is not ok */
+	if (errno == EINVAL) {
+		if (base == 0 || (base >1 && base <37)) {
+			/* Base was ok so it's because we were not
+			 * able to make the convertion.
+			 * Let's reset errno.
+			 */
+			errno = 0;
+		}
+	}
+	return nb;
+}
+#else
+#error "You need the strtoq function"
+#endif /* HAVE_STRTOQ */
+#endif /* HAVE_BSD_STRTOLL */
+#endif /* HAVE_STRTOLL */
+
+
+#ifndef HAVE_STRTOULL
+unsigned long long int rep_strtoull(const char *str, char **endptr, int base)
+{
+#ifdef HAVE_STRTOUQ
+	return strtouq(str, endptr, base);
+#elif defined(HAVE___STRTOULL) 
+	return __strtoull(str, endptr, base);
+#elif SIZEOF_LONG == SIZEOF_LONG_LONG
+	return (unsigned long long int) strtoul(str, endptr, base);
+#else
+# error "You need a strtoull function"
+#endif
+}
+#else
+#ifdef HAVE_BSD_STRTOLL
+#ifdef HAVE_STRTOUQ
+unsigned long long int rep_strtoull(const char *str, char **endptr, int base)
+{
+	unsigned long long int nb = strtouq(str, endptr, base);
+	/* In linux EINVAL is only returned if base is not ok */
+	if (errno == EINVAL) {
+		if (base == 0 || (base >1 && base <37)) {
+			/* Base was ok so it's because we were not
+			 * able to make the convertion.
+			 * Let's reset errno.
+			 */
+			errno = 0;
+		}
+	}
+	return nb;
+}
+#else
+#error "You need the strtouq function"
+#endif /* HAVE_STRTOUQ */
+#endif /* HAVE_BSD_STRTOLL */
+#endif /* HAVE_STRTOULL */
+
+#ifndef HAVE_SETENV
+int rep_setenv(const char *name, const char *value, int overwrite) 
+{
+	char *p;
+	size_t l1, l2;
+	int ret;
+
+	if (!overwrite && getenv(name)) {
+		return 0;
+	}
+
+	l1 = strlen(name);
+	l2 = strlen(value);
+
+	p = malloc(l1+l2+2);
+	if (p == NULL) {
+		return -1;
+	}
+	memcpy(p, name, l1);
+	p[l1] = '=';
+	memcpy(p+l1+1, value, l2);
+	p[l1+l2+1] = 0;
+
+	ret = putenv(p);
+	if (ret != 0) {
+		free(p);
+	}
+
+	return ret;
+}
+#endif
+
+#ifndef HAVE_UNSETENV
+int rep_unsetenv(const char *name)
+{
+	extern char **environ;
+	size_t len = strlen(name);
+	size_t i, count;
+
+	if (environ == NULL || getenv(name) == NULL) {
+		return 0;
+	}
+
+	for (i=0;environ[i];i++) /* noop */ ;
+
+	count=i;
+	
+	for (i=0;i<count;) {
+		if (strncmp(environ[i], name, len) == 0 && environ[i][len] == '=') {
+			/* note: we do _not_ free the old variable here. It is unsafe to 
+			   do so, as the pointer may not have come from malloc */
+			memmove(&environ[i], &environ[i+1], (count-i)*sizeof(char *));
+			count--;
+		} else {
+			i++;
+		}
+	}
+
+	return 0;
+}
+#endif
+
+#ifndef HAVE_UTIME
+int rep_utime(const char *filename, const struct utimbuf *buf)
+{
+	errno = ENOSYS;
+	return -1;
+}
+#endif
+
+#ifndef HAVE_UTIMES
+int rep_utimes(const char *filename, const struct timeval tv[2])
+{
+	struct utimbuf u;
+
+	u.actime = tv[0].tv_sec;
+	if (tv[0].tv_usec > 500000) {
+		u.actime += 1;
+	}
+
+	u.modtime = tv[1].tv_sec;
+	if (tv[1].tv_usec > 500000) {
+		u.modtime += 1;
+	}
+
+	return utime(filename, &u);
+}
+#endif
+
+#ifndef HAVE_DUP2
+int rep_dup2(int oldfd, int newfd) 
+{
+	errno = ENOSYS;
+	return -1;
+}
+#endif
+
+#ifndef HAVE_CHOWN
+/**
+chown isn't used much but OS/2 doesn't have it
+**/
+int rep_chown(const char *fname, uid_t uid, gid_t gid)
+{
+	errno = ENOSYS;
+	return -1;
+}
+#endif
+
+#ifndef HAVE_LINK
+int rep_link(const char *oldpath, const char *newpath)
+{
+	errno = ENOSYS;
+	return -1;
+}
+#endif
+
+#ifndef HAVE_READLINK
+int rep_readlink(const char *path, char *buf, size_t bufsiz)
+{
+	errno = ENOSYS;
+	return -1;
+}
+#endif
+
+#ifndef HAVE_SYMLINK
+int rep_symlink(const char *oldpath, const char *newpath)
+{
+	errno = ENOSYS;
+	return -1;
+}
+#endif
+
+#ifndef HAVE_LCHOWN
+int rep_lchown(const char *fname,uid_t uid,gid_t gid)
+{
+	errno = ENOSYS;
+	return -1;
+}
+#endif
+
+#ifndef HAVE_REALPATH
+char *rep_realpath(const char *path, char *resolved_path)
+{
+	/* As realpath is not a system call we can't return ENOSYS. */
+	errno = EINVAL;
+	return NULL;
+}
+#endif
+
+
+#ifndef HAVE_MEMMEM
+void *rep_memmem(const void *haystack, size_t haystacklen,
+		 const void *needle, size_t needlelen)
+{
+	if (needlelen == 0) {
+		return discard_const(haystack);
+	}
+	while (haystacklen >= needlelen) {
+		char *p = (char *)memchr(haystack, *(const char *)needle,
+					 haystacklen-(needlelen-1));
+		if (!p) return NULL;
+		if (memcmp(p, needle, needlelen) == 0) {
+			return p;
+		}
+		haystack = p+1;
+		haystacklen -= (p - (const char *)haystack) + 1;
+	}
+	return NULL;
+}
+#endif
+
+#if !defined(HAVE_VDPRINTF) || !defined(HAVE_C99_VSNPRINTF)
+int rep_vdprintf(int fd, const char *format, va_list ap)
+{
+	char *s = NULL;
+	int ret;
+
+	vasprintf(&s, format, ap);
+	if (s == NULL) {
+		errno = ENOMEM;
+		return -1;
+	}
+	ret = write(fd, s, strlen(s));
+	free(s);
+	return ret;
+}
+#endif
+
+#if !defined(HAVE_DPRINTF) || !defined(HAVE_C99_VSNPRINTF)
+int rep_dprintf(int fd, const char *format, ...)
+{
+	int ret;
+	va_list ap;
+
+	va_start(ap, format);
+	ret = vdprintf(fd, format, ap);
+	va_end(ap);
+
+	return ret;
+}
+#endif
+
+#ifndef HAVE_GET_CURRENT_DIR_NAME
+char *rep_get_current_dir_name(void)
+{
+	char buf[PATH_MAX+1];
+	char *p;
+	p = getcwd(buf, sizeof(buf));
+	if (p == NULL) {
+		return NULL;
+	}
+	return strdup(p);
+}
+#endif
+
+#ifndef HAVE_STRERROR_R
+int rep_strerror_r(int errnum, char *buf, size_t buflen)
+{
+	char *s = strerror(errnum);
+	if (strlen(s)+1 > buflen) {
+		errno = ERANGE;
+		return -1;
+	}
+	strncpy(buf, s, buflen);
+	return 0;
+}
+#endif
+
+#ifndef HAVE_CLOCK_GETTIME
+int rep_clock_gettime(clockid_t clk_id, struct timespec *tp)
+{
+	struct timeval tval;
+	switch (clk_id) {
+		case 0: /* CLOCK_REALTIME :*/
+#ifdef HAVE_GETTIMEOFDAY_TZ
+			gettimeofday(&tval,NULL);
+#else
+			gettimeofday(&tval);
+#endif
+			tp->tv_sec = tval.tv_sec;
+			tp->tv_nsec = tval.tv_usec * 1000;
+			break;
+		default:
+			errno = EINVAL;
+			return -1;
+	}
+	return 0;
+}
+#endif
+
+#ifndef HAVE_MEMALIGN
+void *rep_memalign( size_t align, size_t size )
+{
+#if defined(HAVE_POSIX_MEMALIGN)
+	void *p = NULL;
+	int ret = posix_memalign( &p, align, size );
+	if ( ret == 0 )
+		return p;
+
+	return NULL;
+#else
+	/* On *BSD systems memaligns doesn't exist, but memory will
+	 * be aligned on allocations of > pagesize. */
+#if defined(SYSCONF_SC_PAGESIZE)
+	size_t pagesize = (size_t)sysconf(_SC_PAGESIZE);
+#elif defined(HAVE_GETPAGESIZE)
+	size_t pagesize = (size_t)getpagesize();
+#else
+	size_t pagesize = (size_t)-1;
+#endif
+	if (pagesize == (size_t)-1) {
+		errno = ENOSYS;
+		return NULL;
+	}
+	if (size < pagesize) {
+		size = pagesize;
+	}
+	return malloc(size);
+#endif
+}
+#endif
+
+#ifndef HAVE_GETPEEREID
+int rep_getpeereid(int s, uid_t *uid, gid_t *gid)
+{
+#if defined(HAVE_PEERCRED)
+	struct ucred cred;
+	socklen_t cred_len = sizeof(struct ucred);
+	int ret;
+
+#undef getsockopt
+	ret = getsockopt(s, SOL_SOCKET, SO_PEERCRED, (void *)&cred, &cred_len);
+	if (ret != 0) {
+		return -1;
+	}
+
+	if (cred_len != sizeof(struct ucred)) {
+		errno = EINVAL;
+		return -1;
+	}
+
+	*uid = cred.uid;
+	*gid = cred.gid;
+	return 0;
+#else
+	errno = ENOSYS;
+	return -1;
+#endif
+}
+#endif
+
+#ifndef HAVE_USLEEP
+int rep_usleep(useconds_t sec)
+{
+	struct timeval tval;
+	/*
+	 * Fake it with select...
+	 */
+	tval.tv_sec = 0;
+	tval.tv_usec = usecs/1000;
+	select(0,NULL,NULL,NULL,&tval);
+	return 0;
+}
+#endif /* HAVE_USLEEP */
+
+#ifndef HAVE_SETPROCTITLE
+void rep_setproctitle(const char *fmt, ...)
+{
+}
+#endif
diff --git a/ctdb/lib/replace/replace.h b/ctdb/lib/replace/replace.h
new file mode 100644
index 0000000000..c0b799763a
--- /dev/null
+++ b/ctdb/lib/replace/replace.h
@@ -0,0 +1,902 @@
+/*
+   Unix SMB/CIFS implementation.
+
+   macros to go along with the lib/replace/ portability layer code
+
+   Copyright (C) Andrew Tridgell 2005
+   Copyright (C) Jelmer Vernooij 2006-2008
+   Copyright (C) Jeremy Allison 2007.
+
+     ** NOTE! The following LGPL license applies to the replace
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#ifndef _LIBREPLACE_REPLACE_H
+#define _LIBREPLACE_REPLACE_H
+
+#ifndef NO_CONFIG_H
+#include "config.h"
+#endif
+
+#ifdef HAVE_STANDARDS_H
+#include <standards.h>
+#endif
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdarg.h>
+#include <errno.h>
+
+#if defined(_MSC_VER) || defined(__MINGW32__)
+#include "win32_replace.h"
+#endif
+
+
+#ifdef HAVE_INTTYPES_H
+#define __STDC_FORMAT_MACROS
+#include <inttypes.h>
+#elif HAVE_STDINT_H
+#include <stdint.h>
+/* force off HAVE_INTTYPES_H so that roken doesn't try to include both,
+   which causes a warning storm on irix */
+#undef HAVE_INTTYPES_H
+#endif
+
+#ifdef HAVE_MALLOC_H
+#include <malloc.h>
+#endif
+
+#ifndef __PRI64_PREFIX
+# if __WORDSIZE == 64 && ! defined __APPLE__
+#  define __PRI64_PREFIX	"l"
+# else
+#  define __PRI64_PREFIX	"ll"
+# endif
+#endif
+
+/* Decimal notation.  */
+#ifndef PRId8
+# define PRId8		"d"
+#endif
+#ifndef PRId16
+# define PRId16		"d"
+#endif
+#ifndef PRId32
+# define PRId32		"d"
+#endif
+#ifndef PRId64
+# define PRId64		__PRI64_PREFIX "d"
+#endif
+
+#ifndef PRIi8
+# define PRIi8		"i"
+#endif
+#ifndef PRIi16
+# define PRIi16		"i"
+#endif
+#ifndef PRIi32
+# define PRIi32		"i"
+#endif
+#ifndef PRIi64
+# define PRIi64		__PRI64_PREFIX "i"
+#endif
+
+#ifndef PRIu8
+# define PRIu8		"u"
+#endif
+#ifndef PRIu16
+# define PRIu16		"u"
+#endif
+#ifndef PRIu32
+# define PRIu32		"u"
+#endif
+#ifndef PRIu64
+# define PRIu64		__PRI64_PREFIX "u"
+#endif
+
+#ifndef SCNd8
+# define SCNd8		"hhd"
+#endif
+#ifndef SCNd16
+# define SCNd16		"hd"
+#endif
+#ifndef SCNd32
+# define SCNd32		"d"
+#endif
+#ifndef SCNd64
+# define SCNd64		__PRI64_PREFIX "d"
+#endif
+
+#ifndef SCNi8
+# define SCNi8		"hhi"
+#endif
+#ifndef SCNi16
+# define SCNi16		"hi"
+#endif
+#ifndef SCNi32
+# define SCNi32		"i"
+#endif
+#ifndef SCNi64
+# define SCNi64		__PRI64_PREFIX "i"
+#endif
+
+#ifndef SCNu8
+# define SCNu8		"hhu"
+#endif
+#ifndef SCNu16
+# define SCNu16		"hu"
+#endif
+#ifndef SCNu32
+# define SCNu32		"u"
+#endif
+#ifndef SCNu64
+# define SCNu64		__PRI64_PREFIX "u"
+#endif
+
+#ifdef HAVE_BSD_STRING_H
+#include <bsd/string.h>
+#endif
+
+#ifdef HAVE_BSD_UNISTD_H
+#include <bsd/unistd.h>
+#endif
+
+#ifdef HAVE_STRING_H
+#include <string.h>
+#endif
+
+#ifdef HAVE_STRINGS_H
+#include <strings.h>
+#endif
+
+#ifdef HAVE_SYS_TYPES_H
+#include <sys/types.h>
+#endif
+
+#ifdef HAVE_SETPROCTITLE_H
+#include <setproctitle.h>
+#endif
+
+#if STDC_HEADERS
+#include <stdlib.h>
+#include <stddef.h>
+#endif
+
+#ifdef HAVE_LINUX_TYPES_H
+/*
+ * This is needed as some broken header files require this to be included early
+ */
+#include <linux/types.h>
+#endif
+
+#ifndef HAVE_STRERROR
+extern char *sys_errlist[];
+#define strerror(i) sys_errlist[i]
+#endif
+
+#ifndef HAVE_ERRNO_DECL
+extern int errno;
+#endif
+
+#ifndef HAVE_STRDUP
+#define strdup rep_strdup
+char *rep_strdup(const char *s);
+#endif
+
+#ifndef HAVE_MEMMOVE
+#define memmove rep_memmove
+void *rep_memmove(void *dest,const void *src,int size);
+#endif
+
+#ifndef HAVE_MEMMEM
+#define memmem rep_memmem
+void *rep_memmem(const void *haystack, size_t haystacklen,
+		 const void *needle, size_t needlelen);
+#endif
+
+#ifndef HAVE_MEMALIGN
+#define memalign rep_memalign
+void *rep_memalign(size_t boundary, size_t size);
+#endif
+
+#ifndef HAVE_MKTIME
+#define mktime rep_mktime
+/* prototype is in "system/time.h" */
+#endif
+
+#ifndef HAVE_TIMEGM
+#define timegm rep_timegm
+/* prototype is in "system/time.h" */
+#endif
+
+#ifndef HAVE_UTIME
+#define utime rep_utime
+/* prototype is in "system/time.h" */
+#endif
+
+#ifndef HAVE_UTIMES
+#define utimes rep_utimes
+/* prototype is in "system/time.h" */
+#endif
+
+#ifndef HAVE_STRLCPY
+#define strlcpy rep_strlcpy
+size_t rep_strlcpy(char *d, const char *s, size_t bufsize);
+#endif
+
+#ifndef HAVE_STRLCAT
+#define strlcat rep_strlcat
+size_t rep_strlcat(char *d, const char *s, size_t bufsize);
+#endif
+
+#if (defined(BROKEN_STRNDUP) || !defined(HAVE_STRNDUP))
+#undef HAVE_STRNDUP
+#define strndup rep_strndup
+char *rep_strndup(const char *s, size_t n);
+#endif
+
+#if (defined(BROKEN_STRNLEN) || !defined(HAVE_STRNLEN))
+#undef HAVE_STRNLEN
+#define strnlen rep_strnlen
+size_t rep_strnlen(const char *s, size_t n);
+#endif
+
+#if !HAVE_DECL_ENVIRON
+#ifdef __APPLE__
+#include <crt_externs.h>
+#define environ (*_NSGetEnviron())
+#else
+extern char **environ;
+#endif
+#endif
+
+#ifndef HAVE_SETENV
+#define setenv rep_setenv
+int rep_setenv(const char *name, const char *value, int overwrite);
+#else
+#ifndef HAVE_SETENV_DECL
+int setenv(const char *name, const char *value, int overwrite);
+#endif
+#endif
+
+#ifndef HAVE_UNSETENV
+#define unsetenv rep_unsetenv
+int rep_unsetenv(const char *name);
+#endif
+
+#ifndef HAVE_SETEUID
+#define seteuid rep_seteuid
+int rep_seteuid(uid_t);
+#endif
+
+#ifndef HAVE_SETEGID
+#define setegid rep_setegid
+int rep_setegid(gid_t);
+#endif
+
+#if (defined(USE_SETRESUID) && !defined(HAVE_SETRESUID_DECL))
+/* stupid glibc */
+int setresuid(uid_t ruid, uid_t euid, uid_t suid);
+#endif
+#if (defined(USE_SETRESUID) && !defined(HAVE_SETRESGID_DECL))
+int setresgid(gid_t rgid, gid_t egid, gid_t sgid);
+#endif
+
+#ifndef HAVE_CHOWN
+#define chown rep_chown
+int rep_chown(const char *path, uid_t uid, gid_t gid);
+#endif
+
+#ifndef HAVE_CHROOT
+#define chroot rep_chroot
+int rep_chroot(const char *dirname);
+#endif
+
+#ifndef HAVE_LINK
+#define link rep_link
+int rep_link(const char *oldpath, const char *newpath);
+#endif
+
+#ifndef HAVE_READLINK
+#define readlink rep_readlink
+ssize_t rep_readlink(const char *path, char *buf, size_t bufsize);
+#endif
+
+#ifndef HAVE_SYMLINK
+#define symlink rep_symlink
+int rep_symlink(const char *oldpath, const char *newpath);
+#endif
+
+#ifndef HAVE_REALPATH
+#define realpath rep_realpath
+char *rep_realpath(const char *path, char *resolved_path);
+#endif
+
+#ifndef HAVE_LCHOWN
+#define lchown rep_lchown
+int rep_lchown(const char *fname,uid_t uid,gid_t gid);
+#endif
+
+#ifdef HAVE_UNIX_H
+#include <unix.h>
+#endif
+
+#ifndef HAVE_SETLINEBUF
+#define setlinebuf rep_setlinebuf
+void rep_setlinebuf(FILE *);
+#endif
+
+#ifndef HAVE_STRCASESTR
+#define strcasestr rep_strcasestr
+char *rep_strcasestr(const char *haystack, const char *needle);
+#endif
+
+#ifndef HAVE_STRTOK_R
+#define strtok_r rep_strtok_r
+char *rep_strtok_r(char *s, const char *delim, char **save_ptr);
+#endif
+
+
+
+#ifndef HAVE_STRTOLL
+#define strtoll rep_strtoll
+long long int rep_strtoll(const char *str, char **endptr, int base);
+#else
+#ifdef HAVE_BSD_STRTOLL
+#define strtoll rep_strtoll
+long long int rep_strtoll(const char *str, char **endptr, int base);
+#endif
+#endif
+
+#ifndef HAVE_STRTOULL
+#define strtoull rep_strtoull
+unsigned long long int rep_strtoull(const char *str, char **endptr, int base);
+#else
+#ifdef HAVE_BSD_STRTOLL /* yes, it's not HAVE_BSD_STRTOULL */
+#define strtoull rep_strtoull
+unsigned long long int rep_strtoull(const char *str, char **endptr, int base);
+#endif
+#endif
+
+#ifndef HAVE_FTRUNCATE
+#define ftruncate rep_ftruncate
+int rep_ftruncate(int,off_t);
+#endif
+
+#ifndef HAVE_INITGROUPS
+#define initgroups rep_initgroups
+int rep_initgroups(char *name, gid_t id);
+#endif
+
+#if !defined(HAVE_BZERO) && defined(HAVE_MEMSET)
+#define bzero(a,b) memset((a),'\0',(b))
+#endif
+
+#ifndef HAVE_DLERROR
+#define dlerror rep_dlerror
+char *rep_dlerror(void);
+#endif
+
+#ifndef HAVE_DLOPEN
+#define dlopen rep_dlopen
+#ifdef DLOPEN_TAKES_UNSIGNED_FLAGS
+void *rep_dlopen(const char *name, unsigned int flags);
+#else
+void *rep_dlopen(const char *name, int flags);
+#endif
+#endif
+
+#ifndef HAVE_DLSYM
+#define dlsym rep_dlsym
+void *rep_dlsym(void *handle, const char *symbol);
+#endif
+
+#ifndef HAVE_DLCLOSE
+#define dlclose rep_dlclose
+int rep_dlclose(void *handle);
+#endif
+
+#ifndef HAVE_SOCKETPAIR
+#define socketpair rep_socketpair
+/* prototype is in system/network.h */
+#endif
+
+#ifndef PRINTF_ATTRIBUTE
+#if (__GNUC__ >= 3) && (__GNUC_MINOR__ >= 1 )
+/** Use gcc attribute to check printf fns.  a1 is the 1-based index of
+ * the parameter containing the format, and a2 the index of the first
+ * argument. Note that some gcc 2.x versions don't handle this
+ * properly **/
+#define PRINTF_ATTRIBUTE(a1, a2) __attribute__ ((format (__printf__, a1, a2)))
+#else
+#define PRINTF_ATTRIBUTE(a1, a2)
+#endif
+#endif
+
+#ifndef _DEPRECATED_
+#if (__GNUC__ >= 3) && (__GNUC_MINOR__ >= 1 )
+#define _DEPRECATED_ __attribute__ ((deprecated))
+#else
+#define _DEPRECATED_
+#endif
+#endif
+
+#if !defined(HAVE_VDPRINTF) || !defined(HAVE_C99_VSNPRINTF)
+#define vdprintf rep_vdprintf
+int rep_vdprintf(int fd, const char *format, va_list ap) PRINTF_ATTRIBUTE(2,0);
+#endif
+
+#if !defined(HAVE_DPRINTF) || !defined(HAVE_C99_VSNPRINTF)
+#define dprintf rep_dprintf
+int rep_dprintf(int fd, const char *format, ...) PRINTF_ATTRIBUTE(2,3);
+#endif
+
+#if !defined(HAVE_VASPRINTF) || !defined(HAVE_C99_VSNPRINTF)
+#define vasprintf rep_vasprintf
+int rep_vasprintf(char **ptr, const char *format, va_list ap) PRINTF_ATTRIBUTE(2,0);
+#endif
+
+#if !defined(HAVE_SNPRINTF) || !defined(HAVE_C99_VSNPRINTF)
+#define snprintf rep_snprintf
+int rep_snprintf(char *,size_t ,const char *, ...) PRINTF_ATTRIBUTE(3,4);
+#endif
+
+#if !defined(HAVE_VSNPRINTF) || !defined(HAVE_C99_VSNPRINTF)
+#define vsnprintf rep_vsnprintf
+int rep_vsnprintf(char *,size_t ,const char *, va_list ap) PRINTF_ATTRIBUTE(3,0);
+#endif
+
+#if !defined(HAVE_ASPRINTF) || !defined(HAVE_C99_VSNPRINTF)
+#define asprintf rep_asprintf
+int rep_asprintf(char **,const char *, ...) PRINTF_ATTRIBUTE(2,3);
+#endif
+
+#if !defined(HAVE_C99_VSNPRINTF)
+#ifdef REPLACE_BROKEN_PRINTF
+/*
+ * We do not redefine printf by default
+ * as it breaks the build if system headers
+ * use __attribute__((format(printf, 3, 0)))
+ * instead of __attribute__((format(__printf__, 3, 0)))
+ */
+#define printf rep_printf
+#endif
+int rep_printf(const char *, ...) PRINTF_ATTRIBUTE(1,2);
+#endif
+
+#if !defined(HAVE_C99_VSNPRINTF)
+#define fprintf rep_fprintf
+int rep_fprintf(FILE *stream, const char *, ...) PRINTF_ATTRIBUTE(2,3);
+#endif
+
+#ifndef HAVE_VSYSLOG
+#ifdef HAVE_SYSLOG
+#define vsyslog rep_vsyslog
+void rep_vsyslog (int facility_priority, const char *format, va_list arglist) PRINTF_ATTRIBUTE(2,0);
+#endif
+#endif
+
+/* we used to use these fns, but now we have good replacements
+   for snprintf and vsnprintf */
+#define slprintf snprintf
+
+
+#ifndef HAVE_VA_COPY
+#undef va_copy
+#ifdef HAVE___VA_COPY
+#define va_copy(dest, src) __va_copy(dest, src)
+#else
+#define va_copy(dest, src) (dest) = (src)
+#endif
+#endif
+
+#ifndef HAVE_VOLATILE
+#define volatile
+#endif
+
+#ifndef HAVE_COMPARISON_FN_T
+typedef int (*comparison_fn_t)(const void *, const void *);
+#endif
+
+#ifndef HAVE_WORKING_STRPTIME
+#define strptime rep_strptime
+struct tm;
+char *rep_strptime(const char *buf, const char *format, struct tm *tm);
+#endif
+
+#ifndef HAVE_DUP2
+#define dup2 rep_dup2
+int rep_dup2(int oldfd, int newfd);
+#endif
+
+/* Load header file for dynamic linking stuff */
+#ifdef HAVE_DLFCN_H
+#include <dlfcn.h>
+#endif
+
+#ifndef RTLD_LAZY
+#define RTLD_LAZY 0
+#endif
+#ifndef RTLD_NOW
+#define RTLD_NOW 0
+#endif
+#ifndef RTLD_GLOBAL
+#define RTLD_GLOBAL 0
+#endif
+
+#ifndef HAVE_SECURE_MKSTEMP
+#define mkstemp(path) rep_mkstemp(path)
+int rep_mkstemp(char *temp);
+#endif
+
+#ifndef HAVE_MKDTEMP
+#define mkdtemp rep_mkdtemp
+char *rep_mkdtemp(char *template);
+#endif
+
+#ifndef HAVE_PREAD
+#define pread rep_pread
+ssize_t rep_pread(int __fd, void *__buf, size_t __nbytes, off_t __offset);
+#define LIBREPLACE_PREAD_REPLACED 1
+#else
+#define LIBREPLACE_PREAD_NOT_REPLACED 1
+#endif
+
+#ifndef HAVE_PWRITE
+#define pwrite rep_pwrite
+ssize_t rep_pwrite(int __fd, const void *__buf, size_t __nbytes, off_t __offset);
+#define LIBREPLACE_PWRITE_REPLACED 1
+#else
+#define LIBREPLACE_PWRITE_NOT_REPLACED 1
+#endif
+
+#if !defined(HAVE_INET_NTOA) || defined(REPLACE_INET_NTOA)
+#define inet_ntoa rep_inet_ntoa
+/* prototype is in "system/network.h" */
+#endif
+
+#ifndef HAVE_INET_PTON
+#define inet_pton rep_inet_pton
+/* prototype is in "system/network.h" */
+#endif
+
+#ifndef HAVE_INET_NTOP
+#define inet_ntop rep_inet_ntop
+/* prototype is in "system/network.h" */
+#endif
+
+#ifndef HAVE_INET_ATON
+#define inet_aton rep_inet_aton
+/* prototype is in "system/network.h" */
+#endif
+
+#ifndef HAVE_CONNECT
+#define connect rep_connect
+/* prototype is in "system/network.h" */
+#endif
+
+#ifndef HAVE_GETHOSTBYNAME
+#define gethostbyname rep_gethostbyname
+/* prototype is in "system/network.h" */
+#endif
+
+#ifndef HAVE_GETIFADDRS
+#define getifaddrs rep_getifaddrs
+/* prototype is in "system/network.h" */
+#endif
+
+#ifndef HAVE_FREEIFADDRS
+#define freeifaddrs rep_freeifaddrs
+/* prototype is in "system/network.h" */
+#endif
+
+#ifndef HAVE_GET_CURRENT_DIR_NAME
+#define get_current_dir_name rep_get_current_dir_name
+char *rep_get_current_dir_name(void);
+#endif
+
+#ifndef HAVE_STRERROR_R
+#define strerror_r rep_strerror_r
+int rep_strerror_r(int errnum, char *buf, size_t buflen);
+#endif
+
+#if !defined(HAVE_CLOCK_GETTIME)
+#define clock_gettime rep_clock_gettime
+#endif
+
+#ifdef HAVE_LIMITS_H
+#include <limits.h>
+#endif
+
+#ifdef HAVE_SYS_PARAM_H
+#include <sys/param.h>
+#endif
+
+/* The extra casts work around common compiler bugs.  */
+#define _TYPE_SIGNED(t) (! ((t) 0 < (t) -1))
+/* The outer cast is needed to work around a bug in Cray C 5.0.3.0.
+   It is necessary at least when t == time_t.  */
+#define _TYPE_MINIMUM(t) ((t) (_TYPE_SIGNED (t) \
+  			      ? ~ (t) 0 << (sizeof (t) * CHAR_BIT - 1) : (t) 0))
+#define _TYPE_MAXIMUM(t) ((t) (~ (t) 0 - _TYPE_MINIMUM (t)))
+
+#ifndef UINT16_MAX
+#define UINT16_MAX 65535
+#endif
+
+#ifndef UINT32_MAX
+#define UINT32_MAX (4294967295U)
+#endif
+
+#ifndef UINT64_MAX
+#define UINT64_MAX ((uint64_t)-1)
+#endif
+
+#ifndef INT64_MAX
+#define INT64_MAX 9223372036854775807LL
+#endif
+
+#ifndef CHAR_BIT
+#define CHAR_BIT 8
+#endif
+
+#ifndef INT32_MAX
+#define INT32_MAX _TYPE_MAXIMUM(int32_t)
+#endif
+
+#ifdef HAVE_STDBOOL_H
+#include <stdbool.h>
+#endif
+
+#if !defined(HAVE_BOOL)
+#ifdef HAVE__Bool
+#define bool _Bool
+#else
+typedef int bool;
+#endif
+#endif
+
+#if !defined(HAVE_INTPTR_T)
+typedef long long intptr_t ;
+#endif
+
+#if !defined(HAVE_UINTPTR_T)
+typedef unsigned long long uintptr_t ;
+#endif
+
+#if !defined(HAVE_PTRDIFF_T)
+typedef unsigned long long ptrdiff_t ;
+#endif
+
+/*
+ * to prevent <rpcsvc/yp_prot.h> from doing a redefine of 'bool'
+ *
+ * IRIX, HPUX, MacOS 10 and Solaris need BOOL_DEFINED
+ * Tru64 needs _BOOL_EXISTS
+ * AIX needs _BOOL,_TRUE,_FALSE
+ */
+#ifndef BOOL_DEFINED
+#define BOOL_DEFINED
+#endif
+#ifndef _BOOL_EXISTS
+#define _BOOL_EXISTS
+#endif
+#ifndef _BOOL
+#define _BOOL
+#endif
+
+#ifndef __bool_true_false_are_defined
+#define __bool_true_false_are_defined
+#endif
+
+#ifndef true
+#define true (1)
+#endif
+#ifndef false
+#define false (0)
+#endif
+
+#ifndef _TRUE
+#define _TRUE true
+#endif
+#ifndef _FALSE
+#define _FALSE false
+#endif
+
+#ifndef HAVE_FUNCTION_MACRO
+#ifdef HAVE_func_MACRO
+#define __FUNCTION__ __func__
+#else
+#define __FUNCTION__ ("")
+#endif
+#endif
+
+
+#ifndef MIN
+#define MIN(a,b) ((a)<(b)?(a):(b))
+#endif
+
+#ifndef MAX
+#define MAX(a,b) ((a)>(b)?(a):(b))
+#endif
+
+#if !defined(HAVE_VOLATILE)
+#define volatile
+#endif
+
+/**
+  this is a warning hack. The idea is to use this everywhere that we
+  get the "discarding const" warning from gcc. That doesn't actually
+  fix the problem of course, but it means that when we do get to
+  cleaning them up we can do it by searching the code for
+  discard_const.
+
+  It also means that other error types aren't as swamped by the noise
+  of hundreds of const warnings, so we are more likely to notice when
+  we get new errors.
+
+  Please only add more uses of this macro when you find it
+  _really_ hard to fix const warnings. Our aim is to eventually use
+  this function in only a very few places.
+
+  Also, please call this via the discard_const_p() macro interface, as that
+  makes the return type safe.
+*/
+#define discard_const(ptr) ((void *)((uintptr_t)(ptr)))
+
+/** Type-safe version of discard_const */
+#define discard_const_p(type, ptr) ((type *)discard_const(ptr))
+
+#ifndef __STRING
+#define __STRING(x)    #x
+#endif
+
+#ifndef __STRINGSTRING
+#define __STRINGSTRING(x) __STRING(x)
+#endif
+
+#ifndef __LINESTR__
+#define __LINESTR__ __STRINGSTRING(__LINE__)
+#endif
+
+#ifndef __location__
+#define __location__ __FILE__ ":" __LINESTR__
+#endif
+
+/** 
+ * zero a structure 
+ */
+#define ZERO_STRUCT(x) memset((char *)&(x), 0, sizeof(x))
+
+/** 
+ * zero a structure given a pointer to the structure 
+ */
+#define ZERO_STRUCTP(x) do { if ((x) != NULL) memset((char *)(x), 0, sizeof(*(x))); } while(0)
+
+/** 
+ * zero a structure given a pointer to the structure - no zero check 
+ */
+#define ZERO_STRUCTPN(x) memset((char *)(x), 0, sizeof(*(x)))
+
+/* zero an array - note that sizeof(array) must work - ie. it must not be a
+   pointer */
+#define ZERO_ARRAY(x) memset((char *)(x), 0, sizeof(x))
+
+/**
+ * work out how many elements there are in a static array 
+ */
+#define ARRAY_SIZE(a) (sizeof(a)/sizeof(a[0]))
+
+/** 
+ * pointer difference macro 
+ */
+#define PTR_DIFF(p1,p2) ((ptrdiff_t)(((const char *)(p1)) - (const char *)(p2)))
+
+#if MMAP_BLACKLIST
+#undef HAVE_MMAP
+#endif
+
+#ifdef __COMPAR_FN_T
+#define QSORT_CAST (__compar_fn_t)
+#endif
+
+#ifndef QSORT_CAST
+#define QSORT_CAST (int (*)(const void *, const void *))
+#endif
+
+#ifndef PATH_MAX
+#define PATH_MAX 1024
+#endif
+
+#ifndef MAX_DNS_NAME_LENGTH
+#define MAX_DNS_NAME_LENGTH 256 /* Actually 255 but +1 for terminating null. */
+#endif
+
+#ifndef HAVE_CRYPT
+char *ufc_crypt(const char *key, const char *salt);
+#define crypt ufc_crypt
+#else
+#ifdef HAVE_CRYPT_H
+#include <crypt.h>
+#endif
+#endif
+
+/* these macros gain us a few percent of speed on gcc */
+#if (__GNUC__ >= 3)
+/* the strange !! is to ensure that __builtin_expect() takes either 0 or 1
+   as its first argument */
+#ifndef likely
+#define likely(x)   __builtin_expect(!!(x), 1)
+#endif
+#ifndef unlikely
+#define unlikely(x) __builtin_expect(!!(x), 0)
+#endif
+#else
+#ifndef likely
+#define likely(x) (x)
+#endif
+#ifndef unlikely
+#define unlikely(x) (x)
+#endif
+#endif
+
+#ifndef HAVE_FDATASYNC
+#define fdatasync(fd) fsync(fd)
+#elif !defined(HAVE_DECL_FDATASYNC)
+int fdatasync(int );
+#endif
+
+/* these are used to mark symbols as local to a shared lib, or
+ * publicly available via the shared lib API */
+#ifndef _PUBLIC_
+#ifdef HAVE_VISIBILITY_ATTR
+#define _PUBLIC_ __attribute__((visibility("default")))
+#else
+#define _PUBLIC_
+#endif
+#endif
+
+#ifndef _PRIVATE_
+#ifdef HAVE_VISIBILITY_ATTR
+#  define _PRIVATE_ __attribute__((visibility("hidden")))
+#else
+#  define _PRIVATE_
+#endif
+#endif
+
+#ifndef HAVE_POLL
+#define poll rep_poll
+/* prototype is in "system/network.h" */
+#endif
+
+#ifndef HAVE_GETPEEREID
+#define getpeereid rep_getpeereid
+int rep_getpeereid(int s, uid_t *uid, gid_t *gid);
+#endif
+
+#ifndef HAVE_USLEEP
+#define usleep rep_usleep
+typedef long useconds_t;
+int usleep(useconds_t);
+#endif
+
+#ifndef HAVE_SETPROCTITLE
+#define setproctitle rep_setproctitle
+void rep_setproctitle(const char *fmt, ...) PRINTF_ATTRIBUTE(1, 2);
+#endif
+
+#endif /* _LIBREPLACE_REPLACE_H */
diff --git a/ctdb/lib/replace/snprintf.c b/ctdb/lib/replace/snprintf.c
new file mode 100644
index 0000000000..6b4a711045
--- /dev/null
+++ b/ctdb/lib/replace/snprintf.c
@@ -0,0 +1,1530 @@
+/*
+ * NOTE: If you change this file, please merge it into rsync, samba, etc.
+ */
+
+/*
+ * Copyright Patrick Powell 1995
+ * This code is based on code written by Patrick Powell (papowell@astart.com)
+ * It may be used for any purpose as long as this notice remains intact
+ * on all source code distributions
+ */
+
+/**************************************************************
+ * Original:
+ * Patrick Powell Tue Apr 11 09:48:21 PDT 1995
+ * A bombproof version of doprnt (dopr) included.
+ * Sigh.  This sort of thing is always nasty do deal with.  Note that
+ * the version here does not include floating point...
+ *
+ * snprintf() is used instead of sprintf() as it does limit checks
+ * for string length.  This covers a nasty loophole.
+ *
+ * The other functions are there to prevent NULL pointers from
+ * causing nast effects.
+ *
+ * More Recently:
+ *  Brandon Long <blong@fiction.net> 9/15/96 for mutt 0.43
+ *  This was ugly.  It is still ugly.  I opted out of floating point
+ *  numbers, but the formatter understands just about everything
+ *  from the normal C string format, at least as far as I can tell from
+ *  the Solaris 2.5 printf(3S) man page.
+ *
+ *  Brandon Long <blong@fiction.net> 10/22/97 for mutt 0.87.1
+ *    Ok, added some minimal floating point support, which means this
+ *    probably requires libm on most operating systems.  Don't yet
+ *    support the exponent (e,E) and sigfig (g,G).  Also, fmtint()
+ *    was pretty badly broken, it just wasn't being exercised in ways
+ *    which showed it, so that's been fixed.  Also, formated the code
+ *    to mutt conventions, and removed dead code left over from the
+ *    original.  Also, there is now a builtin-test, just compile with:
+ *           gcc -DTEST_SNPRINTF -o snprintf snprintf.c -lm
+ *    and run snprintf for results.
+ * 
+ *  Thomas Roessler <roessler@guug.de> 01/27/98 for mutt 0.89i
+ *    The PGP code was using unsigned hexadecimal formats. 
+ *    Unfortunately, unsigned formats simply didn't work.
+ *
+ *  Michael Elkins <me@cs.hmc.edu> 03/05/98 for mutt 0.90.8
+ *    The original code assumed that both snprintf() and vsnprintf() were
+ *    missing.  Some systems only have snprintf() but not vsnprintf(), so
+ *    the code is now broken down under HAVE_SNPRINTF and HAVE_VSNPRINTF.
+ *
+ *  Andrew Tridgell (tridge@samba.org) Oct 1998
+ *    fixed handling of %.0f
+ *    added test for HAVE_LONG_DOUBLE
+ *
+ * tridge@samba.org, idra@samba.org, April 2001
+ *    got rid of fcvt code (twas buggy and made testing harder)
+ *    added C99 semantics
+ *
+ * date: 2002/12/19 19:56:31;  author: herb;  state: Exp;  lines: +2 -0
+ * actually print args for %g and %e
+ * 
+ * date: 2002/06/03 13:37:52;  author: jmcd;  state: Exp;  lines: +8 -0
+ * Since includes.h isn't included here, VA_COPY has to be defined here.  I don't
+ * see any include file that is guaranteed to be here, so I'm defining it
+ * locally.  Fixes AIX and Solaris builds.
+ * 
+ * date: 2002/06/03 03:07:24;  author: tridge;  state: Exp;  lines: +5 -13
+ * put the ifdef for HAVE_VA_COPY in one place rather than in lots of
+ * functions
+ * 
+ * date: 2002/05/17 14:51:22;  author: jmcd;  state: Exp;  lines: +21 -4
+ * Fix usage of va_list passed as an arg.  Use __va_copy before using it
+ * when it exists.
+ * 
+ * date: 2002/04/16 22:38:04;  author: idra;  state: Exp;  lines: +20 -14
+ * Fix incorrect zpadlen handling in fmtfp.
+ * Thanks to Ollie Oldham <ollie.oldham@metro-optix.com> for spotting it.
+ * few mods to make it easier to compile the tests.
+ * addedd the "Ollie" test to the floating point ones.
+ *
+ * Martin Pool (mbp@samba.org) April 2003
+ *    Remove NO_CONFIG_H so that the test case can be built within a source
+ *    tree with less trouble.
+ *    Remove unnecessary SAFE_FREE() definition.
+ *
+ * Martin Pool (mbp@samba.org) May 2003
+ *    Put in a prototype for dummy_snprintf() to quiet compiler warnings.
+ *
+ *    Move #endif to make sure VA_COPY, LDOUBLE, etc are defined even
+ *    if the C library has some snprintf functions already.
+ *
+ * Darren Tucker (dtucker@zip.com.au) 2005
+ *    Fix bug allowing read overruns of the source string with "%.*s"
+ *    Usually harmless unless the read runs outside the process' allocation
+ *    (eg if your malloc does guard pages) in which case it will segfault.
+ *    From OpenSSH.  Also added test for same.
+ *
+ * Simo Sorce (idra@samba.org) Jan 2006
+ * 
+ *    Add support for position independent parameters 
+ *    fix fmtstr now it conforms to sprintf wrt min.max
+ *
+ **************************************************************/
+
+#include "replace.h"
+#include "system/locale.h"
+
+#ifdef TEST_SNPRINTF /* need math library headers for testing */
+
+/* In test mode, we pretend that this system doesn't have any snprintf
+ * functions, regardless of what config.h says. */
+#  undef HAVE_SNPRINTF
+#  undef HAVE_VSNPRINTF
+#  undef HAVE_C99_VSNPRINTF
+#  undef HAVE_ASPRINTF
+#  undef HAVE_VASPRINTF
+#  include <math.h>
+#endif /* TEST_SNPRINTF */
+
+#if defined(HAVE_SNPRINTF) && defined(HAVE_VSNPRINTF) && defined(HAVE_C99_VSNPRINTF)
+/* only include stdio.h if we are not re-defining snprintf or vsnprintf */
+#include <stdio.h>
+ /* make the compiler happy with an empty file */
+ void dummy_snprintf(void);
+ void dummy_snprintf(void) {} 
+#endif /* HAVE_SNPRINTF, etc */
+
+/* yes this really must be a ||. Don't muck with this (tridge) */
+#if !defined(HAVE_VSNPRINTF) || !defined(HAVE_C99_VSNPRINTF)
+
+#ifdef HAVE_LONG_DOUBLE
+#define LDOUBLE long double
+#else
+#define LDOUBLE double
+#endif
+
+#ifdef HAVE_LONG_LONG
+#define LLONG long long
+#else
+#define LLONG long
+#endif
+
+#ifndef VA_COPY
+#ifdef HAVE_VA_COPY
+#define VA_COPY(dest, src) va_copy(dest, src)
+#else
+#ifdef HAVE___VA_COPY
+#define VA_COPY(dest, src) __va_copy(dest, src)
+#else
+#define VA_COPY(dest, src) (dest) = (src)
+#endif
+#endif
+
+/*
+ * dopr(): poor man's version of doprintf
+ */
+
+/* format read states */
+#define DP_S_DEFAULT 0
+#define DP_S_FLAGS   1
+#define DP_S_MIN     2
+#define DP_S_DOT     3
+#define DP_S_MAX     4
+#define DP_S_MOD     5
+#define DP_S_CONV    6
+#define DP_S_DONE    7
+
+/* format flags - Bits */
+#define DP_F_MINUS 	(1 << 0)
+#define DP_F_PLUS  	(1 << 1)
+#define DP_F_SPACE 	(1 << 2)
+#define DP_F_NUM   	(1 << 3)
+#define DP_F_ZERO  	(1 << 4)
+#define DP_F_UP    	(1 << 5)
+#define DP_F_UNSIGNED 	(1 << 6)
+
+/* Conversion Flags */
+#define DP_C_CHAR    1
+#define DP_C_SHORT   2
+#define DP_C_LONG    3
+#define DP_C_LDOUBLE 4
+#define DP_C_LLONG   5
+#define DP_C_SIZET   6
+
+/* Chunk types */
+#define CNK_FMT_STR 0
+#define CNK_INT     1
+#define CNK_OCTAL   2
+#define CNK_UINT    3
+#define CNK_HEX     4
+#define CNK_FLOAT   5
+#define CNK_CHAR    6
+#define CNK_STRING  7
+#define CNK_PTR     8
+#define CNK_NUM     9
+#define CNK_PRCNT   10
+
+#define char_to_int(p) ((p)- '0')
+#ifndef MAX
+#define MAX(p,q) (((p) >= (q)) ? (p) : (q))
+#endif
+
+struct pr_chunk {
+	int type; /* chunk type */
+	int num; /* parameter number */
+	int min; 
+	int max;
+	int flags;
+	int cflags;
+	int start;
+	int len;
+	LLONG value;
+	LDOUBLE fvalue;
+	char *strvalue;
+	void *pnum;
+	struct pr_chunk *min_star;
+	struct pr_chunk *max_star;
+	struct pr_chunk *next;
+};
+
+struct pr_chunk_x {
+	struct pr_chunk **chunks;
+	int num;
+};
+
+static int dopr(char *buffer, size_t maxlen, const char *format, 
+		   va_list args_in);
+static void fmtstr(char *buffer, size_t *currlen, size_t maxlen,
+		    char *value, int flags, int min, int max);
+static void fmtint(char *buffer, size_t *currlen, size_t maxlen,
+		    LLONG value, int base, int min, int max, int flags);
+static void fmtfp(char *buffer, size_t *currlen, size_t maxlen,
+		   LDOUBLE fvalue, int min, int max, int flags);
+static void dopr_outch(char *buffer, size_t *currlen, size_t maxlen, char c);
+static struct pr_chunk *new_chunk(void);
+static int add_cnk_list_entry(struct pr_chunk_x **list,
+				int max_num, struct pr_chunk *chunk);
+
+static int dopr(char *buffer, size_t maxlen, const char *format, va_list args_in)
+{
+	char ch;
+	int state;
+	int pflag;
+	int pnum;
+	int pfirst;
+	size_t currlen;
+	va_list args;
+	const char *base;
+	struct pr_chunk *chunks = NULL;
+	struct pr_chunk *cnk = NULL;
+	struct pr_chunk_x *clist = NULL;
+	int max_pos;
+	int ret = -1;
+
+	VA_COPY(args, args_in);
+
+	state = DP_S_DEFAULT;
+	pfirst = 1;
+	pflag = 0;
+	pnum = 0;
+
+	max_pos = 0;
+	base = format;
+	ch = *format++;
+	
+	/* retrieve the string structure as chunks */
+	while (state != DP_S_DONE) {
+		if (ch == '\0') 
+			state = DP_S_DONE;
+
+		switch(state) {
+		case DP_S_DEFAULT:
+			
+			if (cnk) {
+				cnk->next = new_chunk();
+				cnk = cnk->next;
+			} else {
+				cnk = new_chunk();
+			}
+			if (!cnk) goto done;
+			if (!chunks) chunks = cnk;
+			
+			if (ch == '%') {
+				state = DP_S_FLAGS;
+				ch = *format++;
+			} else {
+				cnk->type = CNK_FMT_STR;
+				cnk->start = format - base -1;
+				while ((ch != '\0') && (ch != '%')) ch = *format++;
+				cnk->len = format - base - cnk->start -1;
+			}
+			break;
+		case DP_S_FLAGS:
+			switch (ch) {
+			case '-':
+				cnk->flags |= DP_F_MINUS;
+				ch = *format++;
+				break;
+			case '+':
+				cnk->flags |= DP_F_PLUS;
+				ch = *format++;
+				break;
+			case ' ':
+				cnk->flags |= DP_F_SPACE;
+				ch = *format++;
+				break;
+			case '#':
+				cnk->flags |= DP_F_NUM;
+				ch = *format++;
+				break;
+			case '0':
+				cnk->flags |= DP_F_ZERO;
+				ch = *format++;
+				break;
+			case 'I':
+				/* internationalization not supported yet */
+				ch = *format++;
+				break;
+			default:
+				state = DP_S_MIN;
+				break;
+			}
+			break;
+		case DP_S_MIN:
+			if (isdigit((unsigned char)ch)) {
+				cnk->min = 10 * cnk->min + char_to_int (ch);
+				ch = *format++;
+			} else if (ch == '$') {
+				if (!pfirst && !pflag) {
+					/* parameters must be all positioned or none */
+					goto done;
+				}
+				if (pfirst) {
+					pfirst = 0;
+					pflag = 1;
+				}
+				if (cnk->min == 0) /* what ?? */
+					goto done;
+				cnk->num = cnk->min;
+				cnk->min = 0;
+				ch = *format++;
+			} else if (ch == '*') {
+				if (pfirst) pfirst = 0;
+				cnk->min_star = new_chunk();
+				if (!cnk->min_star) /* out of memory :-( */
+					goto done;
+				cnk->min_star->type = CNK_INT;
+				if (pflag) {
+					int num;
+					ch = *format++;
+					if (!isdigit((unsigned char)ch)) {
+						/* parameters must be all positioned or none */
+						goto done;
+					}
+					for (num = 0; isdigit((unsigned char)ch); ch = *format++) {
+						num = 10 * num + char_to_int(ch);
+					}
+					cnk->min_star->num = num;
+					if (ch != '$') /* what ?? */
+						goto done;
+				} else {
+					cnk->min_star->num = ++pnum;
+				}
+				max_pos = add_cnk_list_entry(&clist, max_pos, cnk->min_star);
+				if (max_pos == 0) /* out of memory :-( */
+					goto done;
+				ch = *format++;
+				state = DP_S_DOT;
+			} else {
+				if (pfirst) pfirst = 0;
+				state = DP_S_DOT;
+			}
+			break;
+		case DP_S_DOT:
+			if (ch == '.') {
+				state = DP_S_MAX;
+				ch = *format++;
+			} else { 
+				state = DP_S_MOD;
+			}
+			break;
+		case DP_S_MAX:
+			if (isdigit((unsigned char)ch)) {
+				if (cnk->max < 0)
+					cnk->max = 0;
+				cnk->max = 10 * cnk->max + char_to_int (ch);
+				ch = *format++;
+			} else if (ch == '$') {
+				if (!pfirst && !pflag) {
+					/* parameters must be all positioned or none */
+					goto done;
+				}
+				if (cnk->max <= 0) /* what ?? */
+					goto done;
+				cnk->num = cnk->max;
+				cnk->max = -1;
+				ch = *format++;
+			} else if (ch == '*') {
+				cnk->max_star = new_chunk();
+				if (!cnk->max_star) /* out of memory :-( */
+					goto done;
+				cnk->max_star->type = CNK_INT;
+				if (pflag) {
+					int num;
+					ch = *format++;
+					if (!isdigit((unsigned char)ch)) {
+						/* parameters must be all positioned or none */
+						goto done;
+					}
+					for (num = 0; isdigit((unsigned char)ch); ch = *format++) {
+						num = 10 * num + char_to_int(ch);
+					}
+					cnk->max_star->num = num;
+					if (ch != '$') /* what ?? */
+						goto done;
+				} else {
+					cnk->max_star->num = ++pnum;
+				}
+				max_pos = add_cnk_list_entry(&clist, max_pos, cnk->max_star);
+				if (max_pos == 0) /* out of memory :-( */
+					goto done;
+
+				ch = *format++;
+				state = DP_S_MOD;
+			} else {
+				state = DP_S_MOD;
+			}
+			break;
+		case DP_S_MOD:
+			switch (ch) {
+			case 'h':
+				cnk->cflags = DP_C_SHORT;
+				ch = *format++;
+				if (ch == 'h') {
+					cnk->cflags = DP_C_CHAR;
+					ch = *format++;
+				}
+				break;
+			case 'l':
+				cnk->cflags = DP_C_LONG;
+				ch = *format++;
+				if (ch == 'l') {	/* It's a long long */
+					cnk->cflags = DP_C_LLONG;
+					ch = *format++;
+				}
+				break;
+			case 'L':
+				cnk->cflags = DP_C_LDOUBLE;
+				ch = *format++;
+				break;
+			case 'z':
+				cnk->cflags = DP_C_SIZET;
+				ch = *format++;
+				break;
+			default:
+				break;
+			}
+			state = DP_S_CONV;
+			break;
+		case DP_S_CONV:
+			if (cnk->num == 0) cnk->num = ++pnum;
+			max_pos = add_cnk_list_entry(&clist, max_pos, cnk);
+			if (max_pos == 0) /* out of memory :-( */
+				goto done;
+			
+			switch (ch) {
+			case 'd':
+			case 'i':
+				cnk->type = CNK_INT;
+				break;
+			case 'o':
+				cnk->type = CNK_OCTAL;
+				cnk->flags |= DP_F_UNSIGNED;
+				break;
+			case 'u':
+				cnk->type = CNK_UINT;
+				cnk->flags |= DP_F_UNSIGNED;
+				break;
+			case 'X':
+				cnk->flags |= DP_F_UP;
+			case 'x':
+				cnk->type = CNK_HEX;
+				cnk->flags |= DP_F_UNSIGNED;
+				break;
+			case 'A':
+				/* hex float not supported yet */
+			case 'E':
+			case 'G':
+			case 'F':
+				cnk->flags |= DP_F_UP;
+			case 'a':
+				/* hex float not supported yet */
+			case 'e':
+			case 'f':
+			case 'g':
+				cnk->type = CNK_FLOAT;
+				break;
+			case 'c':
+				cnk->type = CNK_CHAR;
+				break;
+			case 's':
+				cnk->type = CNK_STRING;
+				break;
+			case 'p':
+				cnk->type = CNK_PTR;
+				cnk->flags |= DP_F_UNSIGNED;
+				break;
+			case 'n':
+				cnk->type = CNK_NUM;
+				break;
+			case '%':
+				cnk->type = CNK_PRCNT;
+				break;
+			default:
+				/* Unknown, bail out*/
+				goto done;
+			}
+			ch = *format++;
+			state = DP_S_DEFAULT;
+			break;
+		case DP_S_DONE:
+			break;
+		default:
+			/* hmm? */
+			break; /* some picky compilers need this */
+		}
+	}
+
+	/* retrieve the format arguments */
+	for (pnum = 0; pnum < max_pos; pnum++) {
+		int i;
+
+		if (clist[pnum].num == 0) {
+			/* ignoring a parameter should not be permitted
+			 * all parameters must be matched at least once
+			 * BUT seem some system ignore this rule ...
+			 * at least my glibc based system does --SSS
+			 */
+#ifdef DEBUG_SNPRINTF
+			printf("parameter at position %d not used\n", pnum+1);
+#endif
+			/* eat the parameter */
+			va_arg (args, int);
+			continue;
+		}
+		for (i = 1; i < clist[pnum].num; i++) {
+			if (clist[pnum].chunks[0]->type != clist[pnum].chunks[i]->type) {
+				/* nooo noo no!
+				 * all the references to a parameter
+				 * must be of the same type
+				 */
+				goto done;
+			}
+		}
+		cnk = clist[pnum].chunks[0];
+		switch (cnk->type) {
+		case CNK_INT:
+			if (cnk->cflags == DP_C_SHORT) 
+				cnk->value = va_arg (args, int);
+			else if (cnk->cflags == DP_C_LONG)
+				cnk->value = va_arg (args, long int);
+			else if (cnk->cflags == DP_C_LLONG)
+				cnk->value = va_arg (args, LLONG);
+			else if (cnk->cflags == DP_C_SIZET)
+				cnk->value = va_arg (args, ssize_t);
+			else
+				cnk->value = va_arg (args, int);
+
+			for (i = 1; i < clist[pnum].num; i++) {
+				clist[pnum].chunks[i]->value = cnk->value;
+			}
+			break;
+
+		case CNK_OCTAL:
+		case CNK_UINT:
+		case CNK_HEX:
+			if (cnk->cflags == DP_C_SHORT)
+				cnk->value = va_arg (args, unsigned int);
+			else if (cnk->cflags == DP_C_LONG)
+				cnk->value = (unsigned long int)va_arg (args, unsigned long int);
+			else if (cnk->cflags == DP_C_LLONG)
+				cnk->value = (LLONG)va_arg (args, unsigned LLONG);
+			else if (cnk->cflags == DP_C_SIZET)
+				cnk->value = (size_t)va_arg (args, size_t);
+			else
+				cnk->value = (unsigned int)va_arg (args, unsigned int);
+
+			for (i = 1; i < clist[pnum].num; i++) {
+				clist[pnum].chunks[i]->value = cnk->value;
+			}
+			break;
+
+		case CNK_FLOAT:
+			if (cnk->cflags == DP_C_LDOUBLE)
+				cnk->fvalue = va_arg (args, LDOUBLE);
+			else
+				cnk->fvalue = va_arg (args, double);
+
+			for (i = 1; i < clist[pnum].num; i++) {
+				clist[pnum].chunks[i]->fvalue = cnk->fvalue;
+			}
+			break;
+
+		case CNK_CHAR:
+			cnk->value = va_arg (args, int);
+
+			for (i = 1; i < clist[pnum].num; i++) {
+				clist[pnum].chunks[i]->value = cnk->value;
+			}
+			break;
+
+		case CNK_STRING:
+			cnk->strvalue = va_arg (args, char *);
+			if (!cnk->strvalue) cnk->strvalue = "(NULL)";
+
+			for (i = 1; i < clist[pnum].num; i++) {
+				clist[pnum].chunks[i]->strvalue = cnk->strvalue;
+			}
+			break;
+
+		case CNK_PTR:
+			cnk->strvalue = va_arg (args, void *);
+			for (i = 1; i < clist[pnum].num; i++) {
+				clist[pnum].chunks[i]->strvalue = cnk->strvalue;
+			}
+			break;
+
+		case CNK_NUM:
+			if (cnk->cflags == DP_C_CHAR)
+				cnk->pnum = va_arg (args, char *);
+			else if (cnk->cflags == DP_C_SHORT)
+				cnk->pnum = va_arg (args, short int *);
+			else if (cnk->cflags == DP_C_LONG)
+				cnk->pnum = va_arg (args, long int *);
+			else if (cnk->cflags == DP_C_LLONG)
+				cnk->pnum = va_arg (args, LLONG *);
+			else if (cnk->cflags == DP_C_SIZET)
+				cnk->pnum = va_arg (args, ssize_t *);
+			else
+				cnk->pnum = va_arg (args, int *);
+
+			for (i = 1; i < clist[pnum].num; i++) {
+				clist[pnum].chunks[i]->pnum = cnk->pnum;
+			}
+			break;
+
+		case CNK_PRCNT:
+			break;
+
+		default:
+			/* what ?? */
+			goto done;
+		}
+	}
+	/* print out the actual string from chunks */
+	currlen = 0;
+	cnk = chunks;
+	while (cnk) {
+		int len, min, max;
+
+		if (cnk->min_star) min = cnk->min_star->value;
+		else min = cnk->min;
+		if (cnk->max_star) max = cnk->max_star->value;
+		else max = cnk->max;
+
+		switch (cnk->type) {
+
+		case CNK_FMT_STR:
+			if (maxlen != 0 && maxlen > currlen) {
+				if (maxlen > (currlen + cnk->len)) len = cnk->len;
+				else len = maxlen - currlen;
+
+				memcpy(&(buffer[currlen]), &(base[cnk->start]), len);
+			}
+			currlen += cnk->len;
+				
+			break;
+
+		case CNK_INT:
+		case CNK_UINT:
+			fmtint (buffer, &currlen, maxlen, cnk->value, 10, min, max, cnk->flags);
+			break;
+
+		case CNK_OCTAL:
+			fmtint (buffer, &currlen, maxlen, cnk->value, 8, min, max, cnk->flags);
+			break;
+
+		case CNK_HEX:
+			fmtint (buffer, &currlen, maxlen, cnk->value, 16, min, max, cnk->flags);
+			break;
+
+		case CNK_FLOAT:
+			fmtfp (buffer, &currlen, maxlen, cnk->fvalue, min, max, cnk->flags);
+			break;
+
+		case CNK_CHAR:
+			dopr_outch (buffer, &currlen, maxlen, cnk->value);
+			break;
+
+		case CNK_STRING:
+			if (max == -1) {
+				max = strlen(cnk->strvalue);
+			}
+			fmtstr (buffer, &currlen, maxlen, cnk->strvalue, cnk->flags, min, max);
+			break;
+
+		case CNK_PTR:
+			fmtint (buffer, &currlen, maxlen, (long)(cnk->strvalue), 16, min, max, cnk->flags);
+			break;
+
+		case CNK_NUM:
+			if (cnk->cflags == DP_C_CHAR)
+				*((char *)(cnk->pnum)) = (char)currlen;
+			else if (cnk->cflags == DP_C_SHORT)
+				*((short int *)(cnk->pnum)) = (short int)currlen;
+			else if (cnk->cflags == DP_C_LONG)
+				*((long int *)(cnk->pnum)) = (long int)currlen;
+			else if (cnk->cflags == DP_C_LLONG)
+				*((LLONG *)(cnk->pnum)) = (LLONG)currlen;
+			else if (cnk->cflags == DP_C_SIZET)
+				*((ssize_t *)(cnk->pnum)) = (ssize_t)currlen;
+			else
+				*((int *)(cnk->pnum)) = (int)currlen;
+			break;
+
+		case CNK_PRCNT:
+			dopr_outch (buffer, &currlen, maxlen, '%');
+			break;
+
+		default:
+			/* what ?? */
+			goto done;
+		}
+		cnk = cnk->next;
+	}
+	if (maxlen != 0) {
+		if (currlen < maxlen - 1) 
+			buffer[currlen] = '\0';
+		else if (maxlen > 0) 
+			buffer[maxlen - 1] = '\0';
+	}
+	ret = currlen;
+
+done:
+	va_end(args);
+
+	while (chunks) {
+		cnk = chunks->next;
+		free(chunks);
+		chunks = cnk;
+	}
+	if (clist) {
+		for (pnum = 0; pnum < max_pos; pnum++) {
+			if (clist[pnum].chunks) free(clist[pnum].chunks);
+		}
+		free(clist);
+	}
+	return ret;
+}
+
+static void fmtstr(char *buffer, size_t *currlen, size_t maxlen,
+		    char *value, int flags, int min, int max)
+{
+	int padlen, strln;     /* amount to pad */
+	int cnt = 0;
+
+#ifdef DEBUG_SNPRINTF
+	printf("fmtstr min=%d max=%d s=[%s]\n", min, max, value);
+#endif
+	if (value == 0) {
+		value = "<NULL>";
+	}
+
+	for (strln = 0; strln < max && value[strln]; ++strln); /* strlen */
+	padlen = min - strln;
+	if (padlen < 0) 
+		padlen = 0;
+	if (flags & DP_F_MINUS) 
+		padlen = -padlen; /* Left Justify */
+	
+	while (padlen > 0) {
+		dopr_outch (buffer, currlen, maxlen, ' ');
+		--padlen;
+	}
+	while (*value && (cnt < max)) {
+		dopr_outch (buffer, currlen, maxlen, *value++);
+		++cnt;
+	}
+	while (padlen < 0) {
+		dopr_outch (buffer, currlen, maxlen, ' ');
+		++padlen;
+	}
+}
+
+/* Have to handle DP_F_NUM (ie 0x and 0 alternates) */
+
+static void fmtint(char *buffer, size_t *currlen, size_t maxlen,
+		    LLONG value, int base, int min, int max, int flags)
+{
+	int signvalue = 0;
+	unsigned LLONG uvalue;
+	char convert[20];
+	int place = 0;
+	int spadlen = 0; /* amount to space pad */
+	int zpadlen = 0; /* amount to zero pad */
+	int caps = 0;
+	
+	if (max < 0)
+		max = 0;
+	
+	uvalue = value;
+	
+	if(!(flags & DP_F_UNSIGNED)) {
+		if( value < 0 ) {
+			signvalue = '-';
+			uvalue = -value;
+		} else {
+			if (flags & DP_F_PLUS)  /* Do a sign (+/i) */
+				signvalue = '+';
+			else if (flags & DP_F_SPACE)
+				signvalue = ' ';
+		}
+	}
+  
+	if (flags & DP_F_UP) caps = 1; /* Should characters be upper case? */
+
+	do {
+		convert[place++] =
+			(caps? "0123456789ABCDEF":"0123456789abcdef")
+			[uvalue % (unsigned)base  ];
+		uvalue = (uvalue / (unsigned)base );
+	} while(uvalue && (place < 20));
+	if (place == 20) place--;
+	convert[place] = 0;
+
+	zpadlen = max - place;
+	spadlen = min - MAX (max, place) - (signvalue ? 1 : 0);
+	if (zpadlen < 0) zpadlen = 0;
+	if (spadlen < 0) spadlen = 0;
+	if (flags & DP_F_ZERO) {
+		zpadlen = MAX(zpadlen, spadlen);
+		spadlen = 0;
+	}
+	if (flags & DP_F_MINUS) 
+		spadlen = -spadlen; /* Left Justifty */
+
+#ifdef DEBUG_SNPRINTF
+	printf("zpad: %d, spad: %d, min: %d, max: %d, place: %d\n",
+	       zpadlen, spadlen, min, max, place);
+#endif
+
+	/* Spaces */
+	while (spadlen > 0) {
+		dopr_outch (buffer, currlen, maxlen, ' ');
+		--spadlen;
+	}
+
+	/* Sign */
+	if (signvalue) 
+		dopr_outch (buffer, currlen, maxlen, signvalue);
+
+	/* Zeros */
+	if (zpadlen > 0) {
+		while (zpadlen > 0) {
+			dopr_outch (buffer, currlen, maxlen, '0');
+			--zpadlen;
+		}
+	}
+
+	/* Digits */
+	while (place > 0) 
+		dopr_outch (buffer, currlen, maxlen, convert[--place]);
+  
+	/* Left Justified spaces */
+	while (spadlen < 0) {
+		dopr_outch (buffer, currlen, maxlen, ' ');
+		++spadlen;
+	}
+}
+
+static LDOUBLE abs_val(LDOUBLE value)
+{
+	LDOUBLE result = value;
+
+	if (value < 0)
+		result = -value;
+	
+	return result;
+}
+
+static LDOUBLE POW10(int exp)
+{
+	LDOUBLE result = 1;
+	
+	while (exp) {
+		result *= 10;
+		exp--;
+	}
+  
+	return result;
+}
+
+static LLONG ROUND(LDOUBLE value)
+{
+	LLONG intpart;
+
+	intpart = (LLONG)value;
+	value = value - intpart;
+	if (value >= 0.5) intpart++;
+	
+	return intpart;
+}
+
+/* a replacement for modf that doesn't need the math library. Should
+   be portable, but slow */
+static double my_modf(double x0, double *iptr)
+{
+	int i;
+	LLONG l=0;
+	double x = x0;
+	double f = 1.0;
+
+	for (i=0;i<100;i++) {
+		l = (long)x;
+		if (l <= (x+1) && l >= (x-1)) break;
+		x *= 0.1;
+		f *= 10.0;
+	}
+
+	if (i == 100) {
+		/* yikes! the number is beyond what we can handle. What do we do? */
+		(*iptr) = 0;
+		return 0;
+	}
+
+	if (i != 0) {
+		double i2;
+		double ret;
+
+		ret = my_modf(x0-l*f, &i2);
+		(*iptr) = l*f + i2;
+		return ret;
+	} 
+
+	(*iptr) = l;
+	return x - (*iptr);
+}
+
+
+static void fmtfp (char *buffer, size_t *currlen, size_t maxlen,
+		   LDOUBLE fvalue, int min, int max, int flags)
+{
+	int signvalue = 0;
+	double ufvalue;
+	char iconvert[311];
+	char fconvert[311];
+	int iplace = 0;
+	int fplace = 0;
+	int padlen = 0; /* amount to pad */
+	int zpadlen = 0; 
+	int caps = 0;
+	int idx;
+	double intpart;
+	double fracpart;
+	double temp;
+  
+	/* 
+	 * AIX manpage says the default is 0, but Solaris says the default
+	 * is 6, and sprintf on AIX defaults to 6
+	 */
+	if (max < 0)
+		max = 6;
+
+	ufvalue = abs_val (fvalue);
+
+	if (fvalue < 0) {
+		signvalue = '-';
+	} else {
+		if (flags & DP_F_PLUS) { /* Do a sign (+/i) */
+			signvalue = '+';
+		} else {
+			if (flags & DP_F_SPACE)
+				signvalue = ' ';
+		}
+	}
+
+#if 0
+	if (flags & DP_F_UP) caps = 1; /* Should characters be upper case? */
+#endif
+
+#if 0
+	 if (max == 0) ufvalue += 0.5; /* if max = 0 we must round */
+#endif
+
+	/* 
+	 * Sorry, we only support 9 digits past the decimal because of our 
+	 * conversion method
+	 */
+	if (max > 9)
+		max = 9;
+
+	/* We "cheat" by converting the fractional part to integer by
+	 * multiplying by a factor of 10
+	 */
+
+	temp = ufvalue;
+	my_modf(temp, &intpart);
+
+	fracpart = ROUND((POW10(max)) * (ufvalue - intpart));
+	
+	if (fracpart >= POW10(max)) {
+		intpart++;
+		fracpart -= POW10(max);
+	}
+
+
+	/* Convert integer part */
+	do {
+		temp = intpart*0.1;
+		my_modf(temp, &intpart);
+		idx = (int) ((temp -intpart +0.05)* 10.0);
+		/* idx = (int) (((double)(temp*0.1) -intpart +0.05) *10.0); */
+		/* printf ("%llf, %f, %x\n", temp, intpart, idx); */
+		iconvert[iplace++] =
+			(caps? "0123456789ABCDEF":"0123456789abcdef")[idx];
+	} while (intpart && (iplace < 311));
+	if (iplace == 311) iplace--;
+	iconvert[iplace] = 0;
+
+	/* Convert fractional part */
+	if (fracpart)
+	{
+		do {
+			temp = fracpart*0.1;
+			my_modf(temp, &fracpart);
+			idx = (int) ((temp -fracpart +0.05)* 10.0);
+			/* idx = (int) ((((temp/10) -fracpart) +0.05) *10); */
+			/* printf ("%lf, %lf, %ld\n", temp, fracpart, idx ); */
+			fconvert[fplace++] =
+			(caps? "0123456789ABCDEF":"0123456789abcdef")[idx];
+		} while(fracpart && (fplace < 311));
+		if (fplace == 311) fplace--;
+	}
+	fconvert[fplace] = 0;
+  
+	/* -1 for decimal point, another -1 if we are printing a sign */
+	padlen = min - iplace - max - 1 - ((signvalue) ? 1 : 0); 
+	zpadlen = max - fplace;
+	if (zpadlen < 0) zpadlen = 0;
+	if (padlen < 0) 
+		padlen = 0;
+	if (flags & DP_F_MINUS) 
+		padlen = -padlen; /* Left Justifty */
+	
+	if ((flags & DP_F_ZERO) && (padlen > 0)) {
+		if (signvalue) {
+			dopr_outch (buffer, currlen, maxlen, signvalue);
+			--padlen;
+			signvalue = 0;
+		}
+		while (padlen > 0) {
+			dopr_outch (buffer, currlen, maxlen, '0');
+			--padlen;
+		}
+	}
+	while (padlen > 0) {
+		dopr_outch (buffer, currlen, maxlen, ' ');
+		--padlen;
+	}
+	if (signvalue) 
+		dopr_outch (buffer, currlen, maxlen, signvalue);
+	
+	while (iplace > 0) 
+		dopr_outch (buffer, currlen, maxlen, iconvert[--iplace]);
+
+#ifdef DEBUG_SNPRINTF
+	printf("fmtfp: fplace=%d zpadlen=%d\n", fplace, zpadlen);
+#endif
+
+	/*
+	 * Decimal point.  This should probably use locale to find the correct
+	 * char to print out.
+	 */
+	if (max > 0) {
+		dopr_outch (buffer, currlen, maxlen, '.');
+		
+		while (zpadlen > 0) {
+			dopr_outch (buffer, currlen, maxlen, '0');
+			--zpadlen;
+		}
+
+		while (fplace > 0) 
+			dopr_outch (buffer, currlen, maxlen, fconvert[--fplace]);
+	}
+
+	while (padlen < 0) {
+		dopr_outch (buffer, currlen, maxlen, ' ');
+		++padlen;
+	}
+}
+
+static void dopr_outch(char *buffer, size_t *currlen, size_t maxlen, char c)
+{
+	if (*currlen < maxlen) {
+		buffer[(*currlen)] = c;
+	}
+	(*currlen)++;
+}
+
+static struct pr_chunk *new_chunk(void) {
+	struct pr_chunk *new_c = (struct pr_chunk *)malloc(sizeof(struct pr_chunk));
+
+	if (!new_c)
+		return NULL;
+
+	new_c->type = 0;
+	new_c->num = 0;
+	new_c->min = 0;
+	new_c->min_star = NULL;
+	new_c->max = -1;
+	new_c->max_star = NULL;
+	new_c->flags = 0;
+	new_c->cflags = 0;
+	new_c->start = 0;
+	new_c->len = 0;
+	new_c->value = 0;
+	new_c->fvalue = 0;
+	new_c->strvalue = NULL;
+	new_c->pnum = NULL;
+	new_c->next = NULL;
+
+	return new_c;
+}
+
+static int add_cnk_list_entry(struct pr_chunk_x **list,
+				int max_num, struct pr_chunk *chunk) {
+	struct pr_chunk_x *l;
+	struct pr_chunk **c;
+	int max;
+	int cnum;
+	int i, pos;
+
+	if (chunk->num > max_num) {
+		max = chunk->num;
+	
+		if (*list == NULL) {
+			l = (struct pr_chunk_x *)malloc(sizeof(struct pr_chunk_x) * max);
+			pos = 0;
+		} else {
+			l = (struct pr_chunk_x *)realloc(*list, sizeof(struct pr_chunk_x) * max);
+			pos = max_num;
+		}
+		if (l == NULL) {
+			for (i = 0; i < max; i++) {
+				if ((*list)[i].chunks) free((*list)[i].chunks);
+			}
+			return 0;
+		}
+		for (i = pos; i < max; i++) {
+			l[i].chunks = NULL;
+			l[i].num = 0;
+		}
+	} else {
+		l = *list;
+		max = max_num;
+	}
+
+	i = chunk->num - 1;
+	cnum = l[i].num + 1;
+	if (l[i].chunks == NULL) {
+		c = (struct pr_chunk **)malloc(sizeof(struct pr_chunk *) * cnum); 
+	} else {
+		c = (struct pr_chunk **)realloc(l[i].chunks, sizeof(struct pr_chunk *) * cnum);
+	}
+	if (c == NULL) {
+		for (i = 0; i < max; i++) {
+			if (l[i].chunks) free(l[i].chunks);
+		}
+		return 0;
+	}
+	c[l[i].num] = chunk;
+	l[i].chunks = c;
+	l[i].num = cnum;
+
+	*list = l;
+	return max;
+}
+
+ int rep_vsnprintf (char *str, size_t count, const char *fmt, va_list args)
+{
+	return dopr(str, count, fmt, args);
+}
+#endif
+
+/* yes this really must be a ||. Don't muck with this (tridge)
+ *
+ * The logic for these two is that we need our own definition if the
+ * OS *either* has no definition of *sprintf, or if it does have one
+ * that doesn't work properly according to the autoconf test.
+ */
+#if !defined(HAVE_SNPRINTF) || !defined(HAVE_C99_VSNPRINTF)
+ int rep_snprintf(char *str,size_t count,const char *fmt,...)
+{
+	size_t ret;
+	va_list ap;
+    
+	va_start(ap, fmt);
+	ret = vsnprintf(str, count, fmt, ap);
+	va_end(ap);
+	return ret;
+}
+#endif
+
+#ifndef HAVE_C99_VSNPRINTF
+ int rep_printf(const char *fmt, ...)
+{
+	va_list ap;
+	int ret;
+	char *s;
+
+	s = NULL;
+	va_start(ap, fmt);
+	ret = vasprintf(&s, fmt, ap);
+	va_end(ap);
+
+	if (s) {
+		fwrite(s, 1, strlen(s), stdout);
+	}
+	free(s);
+
+	return ret;
+}
+#endif
+
+#ifndef HAVE_C99_VSNPRINTF
+ int rep_fprintf(FILE *stream, const char *fmt, ...)
+{
+	va_list ap;
+	int ret;
+	char *s;
+
+	s = NULL;
+	va_start(ap, fmt);
+	ret = vasprintf(&s, fmt, ap);
+	va_end(ap);
+
+	if (s) {
+		fwrite(s, 1, strlen(s), stream);
+	}
+	free(s);
+
+	return ret;
+}
+#endif
+
+#endif 
+
+#if !defined(HAVE_VASPRINTF) || !defined(HAVE_C99_VSNPRINTF)
+ int rep_vasprintf(char **ptr, const char *format, va_list ap)
+{
+	int ret;
+	va_list ap2;
+
+	VA_COPY(ap2, ap);
+	ret = vsnprintf(NULL, 0, format, ap2);
+	va_end(ap2);
+	if (ret < 0) return ret;
+
+	(*ptr) = (char *)malloc(ret+1);
+	if (!*ptr) return -1;
+
+	VA_COPY(ap2, ap);
+	ret = vsnprintf(*ptr, ret+1, format, ap2);
+	va_end(ap2);
+
+	return ret;
+}
+#endif
+
+#if !defined(HAVE_ASPRINTF) || !defined(HAVE_C99_VSNPRINTF)
+ int rep_asprintf(char **ptr, const char *format, ...)
+{
+	va_list ap;
+	int ret;
+	
+	*ptr = NULL;
+	va_start(ap, format);
+	ret = vasprintf(ptr, format, ap);
+	va_end(ap);
+
+	return ret;
+}
+#endif
+
+#ifdef TEST_SNPRINTF
+
+ int sprintf(char *str,const char *fmt,...);
+ int printf(const char *fmt,...);
+
+ int main (void)
+{
+	char buf1[1024];
+	char buf2[1024];
+	char *buf3;
+	char *fp_fmt[] = {
+		"%1.1f",
+		"%-1.5f",
+		"%1.5f",
+		"%123.9f",
+		"%10.5f",
+		"% 10.5f",
+		"%+22.9f",
+		"%+4.9f",
+		"%01.3f",
+		"%4f",
+		"%3.1f",
+		"%3.2f",
+		"%.0f",
+		"%f",
+		"%-8.8f",
+		"%-9.9f",
+		NULL
+	};
+	double fp_nums[] = { 6442452944.1234, -1.5, 134.21, 91340.2, 341.1234, 203.9, 0.96, 0.996, 
+			     0.9996, 1.996, 4.136, 5.030201, 0.00205,
+			     /* END LIST */ 0};
+	char *int_fmt[] = {
+		"%-1.5d",
+		"%1.5d",
+		"%123.9d",
+		"%5.5d",
+		"%10.5d",
+		"% 10.5d",
+		"%+22.33d",
+		"%01.3d",
+		"%4d",
+		"%d",
+		NULL
+	};
+	long int_nums[] = { -1, 134, 91340, 341, 0203, 1234567890, 0};
+	char *str_fmt[] = {
+		"%10.5s",
+		"%-10.5s",
+		"%5.10s",
+		"%-5.10s",
+		"%10.1s",
+		"%0.10s",
+		"%10.0s",
+		"%1.10s",
+		"%s",
+		"%.1s",
+		"%.10s",
+		"%10s",
+		NULL
+	};
+	char *str_vals[] = {"hello", "a", "", "a longer string", NULL};
+#ifdef HAVE_LONG_LONG
+	char *ll_fmt[] = {
+		"%llu",
+		NULL
+	};
+	LLONG ll_nums[] = { 134, 91340, 341, 0203, 1234567890, 128006186140000000LL, 0};
+#endif
+	int x, y;
+	int fail = 0;
+	int num = 0;
+	int l1, l2;
+	char *ss_fmt[] = {
+		"%zd",
+		"%zu",
+		NULL
+	};
+	size_t ss_nums[] = {134, 91340, 123456789, 0203, 1234567890, 0};
+
+	printf ("Testing snprintf format codes against system sprintf...\n");
+
+	for (x = 0; fp_fmt[x] ; x++) {
+		for (y = 0; fp_nums[y] != 0 ; y++) {
+			buf1[0] = buf2[0] = '\0';
+			l1 = snprintf(buf1, sizeof(buf1), fp_fmt[x], fp_nums[y]);
+			l2 = sprintf (buf2, fp_fmt[x], fp_nums[y]);
+			buf1[1023] = buf2[1023] = '\0';
+			if (strcmp (buf1, buf2) || (l1 != l2)) {
+				printf("snprintf doesn't match Format: %s\n\tsnprintf(%d) = [%s]\n\t sprintf(%d) = [%s]\n", 
+				       fp_fmt[x], l1, buf1, l2, buf2);
+				fail++;
+			}
+			num++;
+		}
+	}
+
+	for (x = 0; int_fmt[x] ; x++) {
+		for (y = 0; int_nums[y] != 0 ; y++) {
+			buf1[0] = buf2[0] = '\0';
+			l1 = snprintf(buf1, sizeof(buf1), int_fmt[x], int_nums[y]);
+			l2 = sprintf (buf2, int_fmt[x], int_nums[y]);
+			buf1[1023] = buf2[1023] = '\0';
+			if (strcmp (buf1, buf2) || (l1 != l2)) {
+				printf("snprintf doesn't match Format: %s\n\tsnprintf(%d) = [%s]\n\t sprintf(%d) = [%s]\n", 
+				       int_fmt[x], l1, buf1, l2, buf2);
+				fail++;
+			}
+			num++;
+		}
+	}
+
+	for (x = 0; str_fmt[x] ; x++) {
+		for (y = 0; str_vals[y] != 0 ; y++) {
+			buf1[0] = buf2[0] = '\0';
+			l1 = snprintf(buf1, sizeof(buf1), str_fmt[x], str_vals[y]);
+			l2 = sprintf (buf2, str_fmt[x], str_vals[y]);
+			buf1[1023] = buf2[1023] = '\0';
+			if (strcmp (buf1, buf2) || (l1 != l2)) {
+				printf("snprintf doesn't match Format: %s\n\tsnprintf(%d) = [%s]\n\t sprintf(%d) = [%s]\n", 
+				       str_fmt[x], l1, buf1, l2, buf2);
+				fail++;
+			}
+			num++;
+		}
+	}
+
+#ifdef HAVE_LONG_LONG
+	for (x = 0; ll_fmt[x] ; x++) {
+		for (y = 0; ll_nums[y] != 0 ; y++) {
+			buf1[0] = buf2[0] = '\0';
+			l1 = snprintf(buf1, sizeof(buf1), ll_fmt[x], ll_nums[y]);
+			l2 = sprintf (buf2, ll_fmt[x], ll_nums[y]);
+			buf1[1023] = buf2[1023] = '\0';
+			if (strcmp (buf1, buf2) || (l1 != l2)) {
+				printf("snprintf doesn't match Format: %s\n\tsnprintf(%d) = [%s]\n\t sprintf(%d) = [%s]\n", 
+				       ll_fmt[x], l1, buf1, l2, buf2);
+				fail++;
+			}
+			num++;
+		}
+	}
+#endif
+
+#define BUFSZ 2048
+
+	buf1[0] = buf2[0] = '\0';
+	if ((buf3 = malloc(BUFSZ)) == NULL) {
+		fail++;
+	} else {
+		num++;
+		memset(buf3, 'a', BUFSZ);
+		snprintf(buf1, sizeof(buf1), "%.*s", 1, buf3);
+		buf1[1023] = '\0';
+		if (strcmp(buf1, "a") != 0) {
+			printf("length limit buf1 '%s' expected 'a'\n", buf1);
+			fail++;
+		}
+        }
+
+	buf1[0] = buf2[0] = '\0';
+	l1 = snprintf(buf1, sizeof(buf1), "%4$*1$d %2$s %3$*1$.*1$f", 3, "pos test", 12.3456, 9);
+	l2 = sprintf(buf2, "%4$*1$d %2$s %3$*1$.*1$f", 3, "pos test", 12.3456, 9);
+	buf1[1023] = buf2[1023] = '\0';
+	if (strcmp(buf1, buf2) || (l1 != l2)) {
+		printf("snprintf doesn't match Format: %s\n\tsnprintf(%d) = [%s]\n\t sprintf(%d) = [%s]\n",
+				"%4$*1$d %2$s %3$*1$.*1$f", l1, buf1, l2, buf2);
+		fail++;
+	}
+
+	buf1[0] = buf2[0] = '\0';
+	l1 = snprintf(buf1, sizeof(buf1), "%4$*4$d %2$s %3$*4$.*4$f", 3, "pos test", 12.3456, 9);
+	l2 = sprintf(buf2, "%4$*4$d %2$s %3$*4$.*4$f", 3, "pos test", 12.3456, 9);
+	buf1[1023] = buf2[1023] = '\0';
+	if (strcmp(buf1, buf2)) {
+		printf("snprintf doesn't match Format: %s\n\tsnprintf(%d) = [%s]\n\t sprintf(%d) = [%s]\n",
+				"%4$*1$d %2$s %3$*1$.*1$f", l1, buf1, l2, buf2);
+		fail++;
+	}
+
+	for (x = 0; ss_fmt[x] ; x++) {
+		for (y = 0; ss_nums[y] != 0 ; y++) {
+			buf1[0] = buf2[0] = '\0';
+			l1 = snprintf(buf1, sizeof(buf1), ss_fmt[x], ss_nums[y]);
+			l2 = sprintf (buf2, ss_fmt[x], ss_nums[y]);
+			buf1[1023] = buf2[1023] = '\0';
+			if (strcmp (buf1, buf2) || (l1 != l2)) {
+				printf("snprintf doesn't match Format: %s\n\tsnprintf(%d) = [%s]\n\t sprintf(%d) = [%s]\n", 
+				       ss_fmt[x], l1, buf1, l2, buf2);
+				fail++;
+			}
+			num++;
+		}
+	}
+#if 0
+	buf1[0] = buf2[0] = '\0';
+	l1 = snprintf(buf1, sizeof(buf1), "%lld", (LLONG)1234567890);
+	l2 = sprintf(buf2, "%lld", (LLONG)1234567890);
+	buf1[1023] = buf2[1023] = '\0';
+	if (strcmp(buf1, buf2)) {
+		printf("snprintf doesn't match Format: %s\n\tsnprintf(%d) = [%s]\n\t sprintf(%d) = [%s]\n",
+				"%lld", l1, buf1, l2, buf2);
+		fail++;
+	}
+
+	buf1[0] = buf2[0] = '\0';
+	l1 = snprintf(buf1, sizeof(buf1), "%Lf", (LDOUBLE)890.1234567890123);
+	l2 = sprintf(buf2, "%Lf", (LDOUBLE)890.1234567890123);
+	buf1[1023] = buf2[1023] = '\0';
+	if (strcmp(buf1, buf2)) {
+		printf("snprintf doesn't match Format: %s\n\tsnprintf(%d) = [%s]\n\t sprintf(%d) = [%s]\n",
+				"%Lf", l1, buf1, l2, buf2);
+		fail++;
+	}
+#endif
+	printf ("%d tests failed out of %d.\n", fail, num);
+
+	printf("seeing how many digits we support\n");
+	{
+		double v0 = 0.12345678901234567890123456789012345678901;
+		for (x=0; x<100; x++) {
+			double p = pow(10, x); 
+			double r = v0*p;
+			snprintf(buf1, sizeof(buf1), "%1.1f", r);
+			sprintf(buf2,                "%1.1f", r);
+			if (strcmp(buf1, buf2)) {
+				printf("we seem to support %d digits\n", x-1);
+				break;
+			}
+		}
+	}
+
+	return 0;
+}
+#endif /* TEST_SNPRINTF */
diff --git a/ctdb/lib/replace/socket.c b/ctdb/lib/replace/socket.c
new file mode 100644
index 0000000000..35e975fce7
--- /dev/null
+++ b/ctdb/lib/replace/socket.c
@@ -0,0 +1,35 @@
+/*
+ * Unix SMB/CIFS implementation.
+ *
+ * Dummy replacements for socket functions.
+ *
+ * Copyright (C) Michael Adam <obnox@samba.org> 2008
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 3 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "replace.h"
+#include "system/network.h"
+
+int rep_connect(int sockfd, const struct sockaddr *serv_addr, socklen_t addrlen)
+{
+	errno = ENOSYS;
+	return -1;
+}
+
+struct hostent *rep_gethostbyname(const char *name)
+{
+	errno = ENOSYS;
+	return NULL;
+}
diff --git a/ctdb/lib/replace/socketpair.c b/ctdb/lib/replace/socketpair.c
new file mode 100644
index 0000000000..c775730952
--- /dev/null
+++ b/ctdb/lib/replace/socketpair.c
@@ -0,0 +1,46 @@
+/*
+ * Unix SMB/CIFS implementation.
+ * replacement routines for broken systems
+ * Copyright (C) Jelmer Vernooij <jelmer@samba.org> 2006
+ * Copyright (C) Michael Adam <obnox@samba.org> 2008
+ *
+ *  ** NOTE! The following LGPL license applies to the replace
+ *  ** library. This does NOT imply that all of Samba is released
+ *  ** under the LGPL
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 3 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "replace.h"
+#include "system/network.h"
+
+int rep_socketpair(int d, int type, int protocol, int sv[2])
+{
+	if (d != AF_UNIX) {
+		errno = EAFNOSUPPORT;
+		return -1;
+	}
+
+	if (protocol != 0) {
+		errno = EPROTONOSUPPORT;
+		return -1;
+	}
+
+	if (type != SOCK_STREAM) {
+		errno = EOPNOTSUPP;
+		return -1;
+	}
+
+	return pipe(sv);
+}
diff --git a/ctdb/lib/replace/strptime.c b/ctdb/lib/replace/strptime.c
new file mode 100644
index 0000000000..20e5d8c39b
--- /dev/null
+++ b/ctdb/lib/replace/strptime.c
@@ -0,0 +1,993 @@
+/* Convert a string representation of time to a time value.
+   Copyright (C) 1996, 1997, 1998, 1999, 2000 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Contributed by Ulrich Drepper <drepper@cygnus.com>, 1996.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public License as
+   published by the Free Software Foundation; either version 3 of the
+   License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Library General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; see the file COPYING.LIB.  If not, 
+   see <http://www.gnu.org/licenses/>.  */
+
+/* XXX This version of the implementation is not really complete.
+   Some of the fields cannot add information alone.  But if seeing
+   some of them in the same format (such as year, week and weekday)
+   this is enough information for determining the date.  */
+
+#include "replace.h"
+#include "system/locale.h"
+#include "system/time.h"
+
+#ifndef __P
+# if defined (__GNUC__) || (defined (__STDC__) && __STDC__)
+#  define __P(args) args
+# else
+#  define __P(args) ()
+# endif  /* GCC.  */
+#endif  /* Not __P.  */
+
+#if ! HAVE_LOCALTIME_R && ! defined localtime_r
+# ifdef _LIBC
+#  define localtime_r __localtime_r
+# else
+/* Approximate localtime_r as best we can in its absence.  */
+#  define localtime_r my_localtime_r
+static struct tm *localtime_r __P ((const time_t *, struct tm *));
+static struct tm *
+localtime_r (t, tp)
+     const time_t *t;
+     struct tm *tp;
+{
+  struct tm *l = localtime (t);
+  if (! l)
+    return 0;
+  *tp = *l;
+  return tp;
+}
+# endif /* ! _LIBC */
+#endif /* ! HAVE_LOCALTIME_R && ! defined (localtime_r) */
+
+
+#define match_char(ch1, ch2) if (ch1 != ch2) return NULL
+#if defined __GNUC__ && __GNUC__ >= 2
+# define match_string(cs1, s2) \
+  ({ size_t len = strlen (cs1);						      \
+     int result = strncasecmp ((cs1), (s2), len) == 0;			      \
+     if (result) (s2) += len;						      \
+     result; })
+#else
+/* Oh come on.  Get a reasonable compiler.  */
+# define match_string(cs1, s2) \
+  (strncasecmp ((cs1), (s2), strlen (cs1)) ? 0 : ((s2) += strlen (cs1), 1))
+#endif
+/* We intentionally do not use isdigit() for testing because this will
+   lead to problems with the wide character version.  */
+#define get_number(from, to, n) \
+  do {									      \
+    int __n = n;							      \
+    val = 0;								      \
+    while (*rp == ' ')							      \
+      ++rp;								      \
+    if (*rp < '0' || *rp > '9')						      \
+      return NULL;							      \
+    do {								      \
+      val *= 10;							      \
+      val += *rp++ - '0';						      \
+    } while (--__n > 0 && val * 10 <= to && *rp >= '0' && *rp <= '9');	      \
+    if (val < from || val > to)						      \
+      return NULL;							      \
+  } while (0)
+#ifdef _NL_CURRENT
+# define get_alt_number(from, to, n) \
+  ({									      \
+    __label__ do_normal;						      \
+    if (*decided != raw)						      \
+      {									      \
+	const char *alts = _NL_CURRENT (LC_TIME, ALT_DIGITS);		      \
+	int __n = n;							      \
+	int any = 0;							      \
+	while (*rp == ' ')						      \
+	  ++rp;								      \
+	val = 0;							      \
+	do {								      \
+	  val *= 10;							      \
+	  while (*alts != '\0')						      \
+	    {								      \
+	      size_t len = strlen (alts);				      \
+	      if (strncasecmp (alts, rp, len) == 0)			      \
+	        break;							      \
+	      alts += len + 1;						      \
+	      ++val;							      \
+	    }								      \
+	  if (*alts == '\0')						      \
+	    {								      \
+	      if (*decided == not && ! any)				      \
+		goto do_normal;						      \
+	      /* If we haven't read anything it's an error.  */		      \
+	      if (! any)						      \
+		return NULL;						      \
+	      /* Correct the premature multiplication.  */		      \
+	      val /= 10;						      \
+	      break;							      \
+	    }								      \
+	  else								      \
+	    *decided = loc;						      \
+	} while (--__n > 0 && val * 10 <= to);				      \
+	if (val < from || val > to)					      \
+	  return NULL;							      \
+      }									      \
+    else								      \
+      {									      \
+       do_normal:							      \
+        get_number (from, to, n);					      \
+      }									      \
+    0;									      \
+  })
+#else
+# define get_alt_number(from, to, n) \
+  /* We don't have the alternate representation.  */			      \
+  get_number(from, to, n)
+#endif
+#define recursive(new_fmt) \
+  (*(new_fmt) != '\0'							      \
+   && (rp = strptime_internal (rp, (new_fmt), tm, decided, era_cnt)) != NULL)
+
+
+#ifdef _LIBC
+/* This is defined in locale/C-time.c in the GNU libc.  */
+extern const struct locale_data _nl_C_LC_TIME;
+extern const unsigned short int __mon_yday[2][13];
+
+# define weekday_name (&_nl_C_LC_TIME.values[_NL_ITEM_INDEX (DAY_1)].string)
+# define ab_weekday_name \
+  (&_nl_C_LC_TIME.values[_NL_ITEM_INDEX (ABDAY_1)].string)
+# define month_name (&_nl_C_LC_TIME.values[_NL_ITEM_INDEX (MON_1)].string)
+# define ab_month_name (&_nl_C_LC_TIME.values[_NL_ITEM_INDEX (ABMON_1)].string)
+# define HERE_D_T_FMT (_nl_C_LC_TIME.values[_NL_ITEM_INDEX (D_T_FMT)].string)
+# define HERE_D_FMT (_nl_C_LC_TIME.values[_NL_ITEM_INDEX (D_FMT)].string)
+# define HERE_AM_STR (_nl_C_LC_TIME.values[_NL_ITEM_INDEX (AM_STR)].string)
+# define HERE_PM_STR (_nl_C_LC_TIME.values[_NL_ITEM_INDEX (PM_STR)].string)
+# define HERE_T_FMT_AMPM \
+  (_nl_C_LC_TIME.values[_NL_ITEM_INDEX (T_FMT_AMPM)].string)
+# define HERE_T_FMT (_nl_C_LC_TIME.values[_NL_ITEM_INDEX (T_FMT)].string)
+
+# define strncasecmp(s1, s2, n) __strncasecmp (s1, s2, n)
+#else
+static char const weekday_name[][10] =
+  {
+    "Sunday", "Monday", "Tuesday", "Wednesday",
+    "Thursday", "Friday", "Saturday"
+  };
+static char const ab_weekday_name[][4] =
+  {
+    "Sun", "Mon", "Tue", "Wed", "Thu", "Fri", "Sat"
+  };
+static char const month_name[][10] =
+  {
+    "January", "February", "March", "April", "May", "June",
+    "July", "August", "September", "October", "November", "December"
+  };
+static char const ab_month_name[][4] =
+  {
+    "Jan", "Feb", "Mar", "Apr", "May", "Jun",
+    "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
+  };
+# define HERE_D_T_FMT "%a %b %e %H:%M:%S %Y"
+# define HERE_D_FMT "%m/%d/%y"
+# define HERE_AM_STR "AM"
+# define HERE_PM_STR "PM"
+# define HERE_T_FMT_AMPM "%I:%M:%S %p"
+# define HERE_T_FMT "%H:%M:%S"
+
+static const unsigned short int __mon_yday[2][13] =
+  {
+    /* Normal years.  */
+    { 0, 31, 59, 90, 120, 151, 181, 212, 243, 273, 304, 334, 365 },
+    /* Leap years.  */
+    { 0, 31, 60, 91, 121, 152, 182, 213, 244, 274, 305, 335, 366 }
+  };
+#endif
+
+/* Status of lookup: do we use the locale data or the raw data?  */
+enum locale_status { not, loc, raw };
+
+
+#ifndef __isleap
+/* Nonzero if YEAR is a leap year (every 4 years,
+   except every 100th isn't, and every 400th is).  */
+# define __isleap(year)	\
+  ((year) % 4 == 0 && ((year) % 100 != 0 || (year) % 400 == 0))
+#endif
+
+/* Compute the day of the week.  */
+static void
+day_of_the_week (struct tm *tm)
+{
+  /* We know that January 1st 1970 was a Thursday (= 4).  Compute the
+     the difference between this data in the one on TM and so determine
+     the weekday.  */
+  int corr_year = 1900 + tm->tm_year - (tm->tm_mon < 2);
+  int wday = (-473
+	      + (365 * (tm->tm_year - 70))
+	      + (corr_year / 4)
+	      - ((corr_year / 4) / 25) + ((corr_year / 4) % 25 < 0)
+	      + (((corr_year / 4) / 25) / 4)
+	      + __mon_yday[0][tm->tm_mon]
+	      + tm->tm_mday - 1);
+  tm->tm_wday = ((wday % 7) + 7) % 7;
+}
+
+/* Compute the day of the year.  */
+static void
+day_of_the_year (struct tm *tm)
+{
+  tm->tm_yday = (__mon_yday[__isleap (1900 + tm->tm_year)][tm->tm_mon]
+		 + (tm->tm_mday - 1));
+}
+
+static char *
+#ifdef _LIBC
+internal_function
+#endif
+strptime_internal __P ((const char *rp, const char *fmt, struct tm *tm,
+			enum locale_status *decided, int era_cnt));
+
+static char *
+#ifdef _LIBC
+internal_function
+#endif
+strptime_internal (rp, fmt, tm, decided, era_cnt)
+     const char *rp;
+     const char *fmt;
+     struct tm *tm;
+     enum locale_status *decided;
+     int era_cnt;
+{
+  int cnt;
+  size_t val;
+  int have_I, is_pm;
+  int century, want_century;
+  int want_era;
+  int have_wday, want_xday;
+  int have_yday;
+  int have_mon, have_mday;
+#ifdef _NL_CURRENT
+  const char *rp_backup;
+  size_t num_eras;
+  struct era_entry *era;
+
+  era = NULL;
+#endif
+
+  have_I = is_pm = 0;
+  century = -1;
+  want_century = 0;
+  want_era = 0;
+
+  have_wday = want_xday = have_yday = have_mon = have_mday = 0;
+
+  while (*fmt != '\0')
+    {
+      /* A white space in the format string matches 0 more or white
+	 space in the input string.  */
+      if (isspace (*fmt))
+	{
+	  while (isspace (*rp))
+	    ++rp;
+	  ++fmt;
+	  continue;
+	}
+
+      /* Any character but `%' must be matched by the same character
+	 in the iput string.  */
+      if (*fmt != '%')
+	{
+	  match_char (*fmt++, *rp++);
+	  continue;
+	}
+
+      ++fmt;
+#ifndef _NL_CURRENT
+      /* We need this for handling the `E' modifier.  */
+    start_over:
+#endif
+
+#ifdef _NL_CURRENT
+      /* Make back up of current processing pointer.  */
+      rp_backup = rp;
+#endif
+
+      switch (*fmt++)
+	{
+	case '%':
+	  /* Match the `%' character itself.  */
+	  match_char ('%', *rp++);
+	  break;
+	case 'a':
+	case 'A':
+	  /* Match day of week.  */
+	  for (cnt = 0; cnt < 7; ++cnt)
+	    {
+#ifdef _NL_CURRENT
+	      if (*decided !=raw)
+		{
+		  if (match_string (_NL_CURRENT (LC_TIME, DAY_1 + cnt), rp))
+		    {
+		      if (*decided == not
+			  && strcmp (_NL_CURRENT (LC_TIME, DAY_1 + cnt),
+				     weekday_name[cnt]))
+			*decided = loc;
+		      break;
+		    }
+		  if (match_string (_NL_CURRENT (LC_TIME, ABDAY_1 + cnt), rp))
+		    {
+		      if (*decided == not
+			  && strcmp (_NL_CURRENT (LC_TIME, ABDAY_1 + cnt),
+				     ab_weekday_name[cnt]))
+			*decided = loc;
+		      break;
+		    }
+		}
+#endif
+	      if (*decided != loc
+		  && (match_string (weekday_name[cnt], rp)
+		      || match_string (ab_weekday_name[cnt], rp)))
+		{
+		  *decided = raw;
+		  break;
+		}
+	    }
+	  if (cnt == 7)
+	    /* Does not match a weekday name.  */
+	    return NULL;
+	  tm->tm_wday = cnt;
+	  have_wday = 1;
+	  break;
+	case 'b':
+	case 'B':
+	case 'h':
+	  /* Match month name.  */
+	  for (cnt = 0; cnt < 12; ++cnt)
+	    {
+#ifdef _NL_CURRENT
+	      if (*decided !=raw)
+		{
+		  if (match_string (_NL_CURRENT (LC_TIME, MON_1 + cnt), rp))
+		    {
+		      if (*decided == not
+			  && strcmp (_NL_CURRENT (LC_TIME, MON_1 + cnt),
+				     month_name[cnt]))
+			*decided = loc;
+		      break;
+		    }
+		  if (match_string (_NL_CURRENT (LC_TIME, ABMON_1 + cnt), rp))
+		    {
+		      if (*decided == not
+			  && strcmp (_NL_CURRENT (LC_TIME, ABMON_1 + cnt),
+				     ab_month_name[cnt]))
+			*decided = loc;
+		      break;
+		    }
+		}
+#endif
+	      if (match_string (month_name[cnt], rp)
+		  || match_string (ab_month_name[cnt], rp))
+		{
+		  *decided = raw;
+		  break;
+		}
+	    }
+	  if (cnt == 12)
+	    /* Does not match a month name.  */
+	    return NULL;
+	  tm->tm_mon = cnt;
+	  want_xday = 1;
+	  break;
+	case 'c':
+	  /* Match locale's date and time format.  */
+#ifdef _NL_CURRENT
+	  if (*decided != raw)
+	    {
+	      if (!recursive (_NL_CURRENT (LC_TIME, D_T_FMT)))
+		{
+		  if (*decided == loc)
+		    return NULL;
+		  else
+		    rp = rp_backup;
+		}
+	      else
+		{
+		  if (*decided == not &&
+		      strcmp (_NL_CURRENT (LC_TIME, D_T_FMT), HERE_D_T_FMT))
+		    *decided = loc;
+		  want_xday = 1;
+		  break;
+		}
+	      *decided = raw;
+	    }
+#endif
+	  if (!recursive (HERE_D_T_FMT))
+	    return NULL;
+	  want_xday = 1;
+	  break;
+	case 'C':
+	  /* Match century number.  */
+#ifdef _NL_CURRENT
+	match_century:
+#endif
+	  get_number (0, 99, 2);
+	  century = val;
+	  want_xday = 1;
+	  break;
+	case 'd':
+	case 'e':
+	  /* Match day of month.  */
+	  get_number (1, 31, 2);
+	  tm->tm_mday = val;
+	  have_mday = 1;
+	  want_xday = 1;
+	  break;
+	case 'F':
+	  if (!recursive ("%Y-%m-%d"))
+	    return NULL;
+	  want_xday = 1;
+	  break;
+	case 'x':
+#ifdef _NL_CURRENT
+	  if (*decided != raw)
+	    {
+	      if (!recursive (_NL_CURRENT (LC_TIME, D_FMT)))
+		{
+		  if (*decided == loc)
+		    return NULL;
+		  else
+		    rp = rp_backup;
+		}
+	      else
+		{
+		  if (*decided == not
+		      && strcmp (_NL_CURRENT (LC_TIME, D_FMT), HERE_D_FMT))
+		    *decided = loc;
+		  want_xday = 1;
+		  break;
+		}
+	      *decided = raw;
+	    }
+#endif
+	  /* Fall through.  */
+	case 'D':
+	  /* Match standard day format.  */
+	  if (!recursive (HERE_D_FMT))
+	    return NULL;
+	  want_xday = 1;
+	  break;
+	case 'k':
+	case 'H':
+	  /* Match hour in 24-hour clock.  */
+	  get_number (0, 23, 2);
+	  tm->tm_hour = val;
+	  have_I = 0;
+	  break;
+	case 'I':
+	  /* Match hour in 12-hour clock.  */
+	  get_number (1, 12, 2);
+	  tm->tm_hour = val % 12;
+	  have_I = 1;
+	  break;
+	case 'j':
+	  /* Match day number of year.  */
+	  get_number (1, 366, 3);
+	  tm->tm_yday = val - 1;
+	  have_yday = 1;
+	  break;
+	case 'm':
+	  /* Match number of month.  */
+	  get_number (1, 12, 2);
+	  tm->tm_mon = val - 1;
+	  have_mon = 1;
+	  want_xday = 1;
+	  break;
+	case 'M':
+	  /* Match minute.  */
+	  get_number (0, 59, 2);
+	  tm->tm_min = val;
+	  break;
+	case 'n':
+	case 't':
+	  /* Match any white space.  */
+	  while (isspace (*rp))
+	    ++rp;
+	  break;
+	case 'p':
+	  /* Match locale's equivalent of AM/PM.  */
+#ifdef _NL_CURRENT
+	  if (*decided != raw)
+	    {
+	      if (match_string (_NL_CURRENT (LC_TIME, AM_STR), rp))
+		{
+		  if (strcmp (_NL_CURRENT (LC_TIME, AM_STR), HERE_AM_STR))
+		    *decided = loc;
+		  break;
+		}
+	      if (match_string (_NL_CURRENT (LC_TIME, PM_STR), rp))
+		{
+		  if (strcmp (_NL_CURRENT (LC_TIME, PM_STR), HERE_PM_STR))
+		    *decided = loc;
+		  is_pm = 1;
+		  break;
+		}
+	      *decided = raw;
+	    }
+#endif
+	  if (!match_string (HERE_AM_STR, rp)) {
+	    if (match_string (HERE_PM_STR, rp)) {
+	      is_pm = 1;
+	    } else {
+	      return NULL;
+	    }
+	  }
+	  break;
+	case 'r':
+#ifdef _NL_CURRENT
+	  if (*decided != raw)
+	    {
+	      if (!recursive (_NL_CURRENT (LC_TIME, T_FMT_AMPM)))
+		{
+		  if (*decided == loc)
+		    return NULL;
+		  else
+		    rp = rp_backup;
+		}
+	      else
+		{
+		  if (*decided == not &&
+		      strcmp (_NL_CURRENT (LC_TIME, T_FMT_AMPM),
+			      HERE_T_FMT_AMPM))
+		    *decided = loc;
+		  break;
+		}
+	      *decided = raw;
+	    }
+#endif
+	  if (!recursive (HERE_T_FMT_AMPM))
+	    return NULL;
+	  break;
+	case 'R':
+	  if (!recursive ("%H:%M"))
+	    return NULL;
+	  break;
+	case 's':
+	  {
+	    /* The number of seconds may be very high so we cannot use
+	       the `get_number' macro.  Instead read the number
+	       character for character and construct the result while
+	       doing this.  */
+	    time_t secs = 0;
+	    if (*rp < '0' || *rp > '9')
+	      /* We need at least one digit.  */
+	      return NULL;
+
+	    do
+	      {
+		secs *= 10;
+		secs += *rp++ - '0';
+	      }
+	    while (*rp >= '0' && *rp <= '9');
+
+	    if (localtime_r (&secs, tm) == NULL)
+	      /* Error in function.  */
+	      return NULL;
+	  }
+	  break;
+	case 'S':
+	  get_number (0, 61, 2);
+	  tm->tm_sec = val;
+	  break;
+	case 'X':
+#ifdef _NL_CURRENT
+	  if (*decided != raw)
+	    {
+	      if (!recursive (_NL_CURRENT (LC_TIME, T_FMT)))
+		{
+		  if (*decided == loc)
+		    return NULL;
+		  else
+		    rp = rp_backup;
+		}
+	      else
+		{
+		  if (strcmp (_NL_CURRENT (LC_TIME, T_FMT), HERE_T_FMT))
+		    *decided = loc;
+		  break;
+		}
+	      *decided = raw;
+	    }
+#endif
+	  /* Fall through.  */
+	case 'T':
+	  if (!recursive (HERE_T_FMT))
+	    return NULL;
+	  break;
+	case 'u':
+	  get_number (1, 7, 1);
+	  tm->tm_wday = val % 7;
+	  have_wday = 1;
+	  break;
+	case 'g':
+	  get_number (0, 99, 2);
+	  /* XXX This cannot determine any field in TM.  */
+	  break;
+	case 'G':
+	  if (*rp < '0' || *rp > '9')
+	    return NULL;
+	  /* XXX Ignore the number since we would need some more
+	     information to compute a real date.  */
+	  do
+	    ++rp;
+	  while (*rp >= '0' && *rp <= '9');
+	  break;
+	case 'U':
+	case 'V':
+	case 'W':
+	  get_number (0, 53, 2);
+	  /* XXX This cannot determine any field in TM without some
+	     information.  */
+	  break;
+	case 'w':
+	  /* Match number of weekday.  */
+	  get_number (0, 6, 1);
+	  tm->tm_wday = val;
+	  have_wday = 1;
+	  break;
+	case 'y':
+#ifdef _NL_CURRENT
+	match_year_in_century:
+#endif
+	  /* Match year within century.  */
+	  get_number (0, 99, 2);
+	  /* The "Year 2000: The Millennium Rollover" paper suggests that
+	     values in the range 69-99 refer to the twentieth century.  */
+	  tm->tm_year = val >= 69 ? val : val + 100;
+	  /* Indicate that we want to use the century, if specified.  */
+	  want_century = 1;
+	  want_xday = 1;
+	  break;
+	case 'Y':
+	  /* Match year including century number.  */
+	  get_number (0, 9999, 4);
+	  tm->tm_year = val - 1900;
+	  want_century = 0;
+	  want_xday = 1;
+	  break;
+	case 'Z':
+	  /* XXX How to handle this?  */
+	  break;
+	case 'E':
+#ifdef _NL_CURRENT
+	  switch (*fmt++)
+	    {
+	    case 'c':
+	      /* Match locale's alternate date and time format.  */
+	      if (*decided != raw)
+		{
+		  const char *fmt = _NL_CURRENT (LC_TIME, ERA_D_T_FMT);
+
+		  if (*fmt == '\0')
+		    fmt = _NL_CURRENT (LC_TIME, D_T_FMT);
+
+		  if (!recursive (fmt))
+		    {
+		      if (*decided == loc)
+			return NULL;
+		      else
+			rp = rp_backup;
+		    }
+		  else
+		    {
+		      if (strcmp (fmt, HERE_D_T_FMT))
+			*decided = loc;
+		      want_xday = 1;
+		      break;
+		    }
+		  *decided = raw;
+		}
+	      /* The C locale has no era information, so use the
+		 normal representation.  */
+	      if (!recursive (HERE_D_T_FMT))
+		return NULL;
+	      want_xday = 1;
+	      break;
+	    case 'C':
+	      if (*decided != raw)
+		{
+		  if (era_cnt >= 0)
+		    {
+		      era = _nl_select_era_entry (era_cnt);
+		      if (match_string (era->era_name, rp))
+			{
+			  *decided = loc;
+			  break;
+			}
+		      else
+			return NULL;
+		    }
+		  else
+		    {
+		      num_eras = _NL_CURRENT_WORD (LC_TIME,
+						   _NL_TIME_ERA_NUM_ENTRIES);
+		      for (era_cnt = 0; era_cnt < (int) num_eras;
+			   ++era_cnt, rp = rp_backup)
+			{
+			  era = _nl_select_era_entry (era_cnt);
+			  if (match_string (era->era_name, rp))
+			    {
+			      *decided = loc;
+			      break;
+			    }
+			}
+		      if (era_cnt == (int) num_eras)
+			{
+			  era_cnt = -1;
+			  if (*decided == loc)
+			    return NULL;
+			}
+		      else
+			break;
+		    }
+
+		  *decided = raw;
+		}
+	      /* The C locale has no era information, so use the
+		 normal representation.  */
+	      goto match_century;
+ 	    case 'y':
+	      if (*decided == raw)
+		goto match_year_in_century;
+
+	      get_number(0, 9999, 4);
+	      tm->tm_year = val;
+	      want_era = 1;
+	      want_xday = 1;
+	      break;
+	    case 'Y':
+	      if (*decided != raw)
+		{
+		  num_eras = _NL_CURRENT_WORD (LC_TIME,
+					       _NL_TIME_ERA_NUM_ENTRIES);
+		  for (era_cnt = 0; era_cnt < (int) num_eras;
+		       ++era_cnt, rp = rp_backup)
+		    {
+		      era = _nl_select_era_entry (era_cnt);
+		      if (recursive (era->era_format))
+			break;
+		    }
+		  if (era_cnt == (int) num_eras)
+		    {
+		      era_cnt = -1;
+		      if (*decided == loc)
+			return NULL;
+		      else
+			rp = rp_backup;
+		    }
+		  else
+		    {
+		      *decided = loc;
+		      era_cnt = -1;
+		      break;
+		    }
+
+		  *decided = raw;
+		}
+	      get_number (0, 9999, 4);
+	      tm->tm_year = val - 1900;
+	      want_century = 0;
+	      want_xday = 1;
+	      break;
+	    case 'x':
+	      if (*decided != raw)
+		{
+		  const char *fmt = _NL_CURRENT (LC_TIME, ERA_D_FMT);
+
+		  if (*fmt == '\0')
+		    fmt = _NL_CURRENT (LC_TIME, D_FMT);
+
+		  if (!recursive (fmt))
+		    {
+		      if (*decided == loc)
+			return NULL;
+		      else
+			rp = rp_backup;
+		    }
+		  else
+		    {
+		      if (strcmp (fmt, HERE_D_FMT))
+			*decided = loc;
+		      break;
+		    }
+		  *decided = raw;
+		}
+	      if (!recursive (HERE_D_FMT))
+		return NULL;
+	      break;
+	    case 'X':
+	      if (*decided != raw)
+		{
+		  const char *fmt = _NL_CURRENT (LC_TIME, ERA_T_FMT);
+
+		  if (*fmt == '\0')
+		    fmt = _NL_CURRENT (LC_TIME, T_FMT);
+
+		  if (!recursive (fmt))
+		    {
+		      if (*decided == loc)
+			return NULL;
+		      else
+			rp = rp_backup;
+		    }
+		  else
+		    {
+		      if (strcmp (fmt, HERE_T_FMT))
+			*decided = loc;
+		      break;
+		    }
+		  *decided = raw;
+		}
+	      if (!recursive (HERE_T_FMT))
+		return NULL;
+	      break;
+	    default:
+	      return NULL;
+	    }
+	  break;
+#else
+	  /* We have no information about the era format.  Just use
+	     the normal format.  */
+	  if (*fmt != 'c' && *fmt != 'C' && *fmt != 'y' && *fmt != 'Y'
+	      && *fmt != 'x' && *fmt != 'X')
+	    /* This is an illegal format.  */
+	    return NULL;
+
+	  goto start_over;
+#endif
+	case 'O':
+	  switch (*fmt++)
+	    {
+	    case 'd':
+	    case 'e':
+	      /* Match day of month using alternate numeric symbols.  */
+	      get_alt_number (1, 31, 2);
+	      tm->tm_mday = val;
+	      have_mday = 1;
+	      want_xday = 1;
+	      break;
+	    case 'H':
+	      /* Match hour in 24-hour clock using alternate numeric
+		 symbols.  */
+	      get_alt_number (0, 23, 2);
+	      tm->tm_hour = val;
+	      have_I = 0;
+	      break;
+	    case 'I':
+	      /* Match hour in 12-hour clock using alternate numeric
+		 symbols.  */
+	      get_alt_number (1, 12, 2);
+	      tm->tm_hour = val - 1;
+	      have_I = 1;
+	      break;
+	    case 'm':
+	      /* Match month using alternate numeric symbols.  */
+	      get_alt_number (1, 12, 2);
+	      tm->tm_mon = val - 1;
+	      have_mon = 1;
+	      want_xday = 1;
+	      break;
+	    case 'M':
+	      /* Match minutes using alternate numeric symbols.  */
+	      get_alt_number (0, 59, 2);
+	      tm->tm_min = val;
+	      break;
+	    case 'S':
+	      /* Match seconds using alternate numeric symbols.  */
+	      get_alt_number (0, 61, 2);
+	      tm->tm_sec = val;
+	      break;
+	    case 'U':
+	    case 'V':
+	    case 'W':
+	      get_alt_number (0, 53, 2);
+	      /* XXX This cannot determine any field in TM without
+		 further information.  */
+	      break;
+	    case 'w':
+	      /* Match number of weekday using alternate numeric symbols.  */
+	      get_alt_number (0, 6, 1);
+	      tm->tm_wday = val;
+	      have_wday = 1;
+	      break;
+	    case 'y':
+	      /* Match year within century using alternate numeric symbols.  */
+	      get_alt_number (0, 99, 2);
+	      tm->tm_year = val >= 69 ? val : val + 100;
+	      want_xday = 1;
+	      break;
+	    default:
+	      return NULL;
+	    }
+	  break;
+	default:
+	  return NULL;
+	}
+    }
+
+  if (have_I && is_pm)
+    tm->tm_hour += 12;
+
+  if (century != -1)
+    {
+      if (want_century)
+	tm->tm_year = tm->tm_year % 100 + (century - 19) * 100;
+      else
+	/* Only the century, but not the year.  Strange, but so be it.  */
+	tm->tm_year = (century - 19) * 100;
+    }
+
+#ifdef _NL_CURRENT
+  if (era_cnt != -1)
+    {
+      era = _nl_select_era_entry(era_cnt);
+      if (want_era)
+	tm->tm_year = (era->start_date[0]
+		       + ((tm->tm_year - era->offset)
+			  * era->absolute_direction));
+      else
+	/* Era start year assumed.  */
+	tm->tm_year = era->start_date[0];
+    }
+  else
+#endif
+    if (want_era)
+      return NULL;
+
+  if (want_xday && !have_wday)
+    {
+      if ( !(have_mon && have_mday) && have_yday)
+	{
+	  /* We don't have tm_mon and/or tm_mday, compute them.  */
+	  int t_mon = 0;
+	  while (__mon_yday[__isleap(1900 + tm->tm_year)][t_mon] <= tm->tm_yday)
+	      t_mon++;
+	  if (!have_mon)
+	      tm->tm_mon = t_mon - 1;
+	  if (!have_mday)
+	      tm->tm_mday =
+		(tm->tm_yday
+		 - __mon_yday[__isleap(1900 + tm->tm_year)][t_mon - 1] + 1);
+	}
+      day_of_the_week (tm);
+    }
+  if (want_xday && !have_yday)
+    day_of_the_year (tm);
+
+  return discard_const_p(char, rp);
+}
+
+
+char *rep_strptime(const char *buf, const char *format, struct tm *tm)
+{
+  enum locale_status decided;
+
+#ifdef _NL_CURRENT
+  decided = not;
+#else
+  decided = raw;
+#endif
+  return strptime_internal (buf, format, tm, &decided, -1);
+}
diff --git a/ctdb/lib/replace/strptime.m4 b/ctdb/lib/replace/strptime.m4
new file mode 100644
index 0000000000..8ac22f6342
--- /dev/null
+++ b/ctdb/lib/replace/strptime.m4
@@ -0,0 +1,16 @@
+AC_CHECK_FUNCS(strptime)
+AC_CHECK_DECLS(strptime, [], [], [#include <time.h>])
+AC_CACHE_CHECK([whether strptime is available and works],libreplace_cv_STRPTIME_OK,[
+	AC_TRY_RUN([
+		#define LIBREPLACE_CONFIGURE_TEST_STRPTIME
+		#include "$libreplacedir/test/strptime.c"
+		],
+		[libreplace_cv_STRPTIME_OK=yes],
+		[libreplace_cv_STRPTIME_OK=no],
+		[libreplace_cv_STRPTIME_OK="assuming not"])
+])
+if test x"$libreplace_cv_STRPTIME_OK" != x"yes"; then
+        LIBREPLACEOBJ="${LIBREPLACEOBJ} $libreplacedir/strptime.o"
+else
+        AC_DEFINE(HAVE_WORKING_STRPTIME,1,[Whether strptime is working correct])
+fi
diff --git a/ctdb/lib/replace/system/README b/ctdb/lib/replace/system/README
new file mode 100644
index 0000000000..69a2b80b56
--- /dev/null
+++ b/ctdb/lib/replace/system/README
@@ -0,0 +1,4 @@
+This directory contains wrappers around logical groups of system
+include files. The idea is to avoid #ifdef blocks in the main code,
+and instead put all the necessary conditional includes in subsystem
+specific header files in this directory.
diff --git a/ctdb/lib/replace/system/aio.h b/ctdb/lib/replace/system/aio.h
new file mode 100644
index 0000000000..784d77fa28
--- /dev/null
+++ b/ctdb/lib/replace/system/aio.h
@@ -0,0 +1,32 @@
+#ifndef _system_aio_h
+#define _system_aio_h
+/* 
+   Unix SMB/CIFS implementation.
+
+   AIO system include wrappers
+
+   Copyright (C) Andrew Tridgell 2006
+   
+     ** NOTE! The following LGPL license applies to the replace
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#ifdef HAVE_LIBAIO_H
+#include <libaio.h>
+#endif
+
+#endif
diff --git a/ctdb/lib/replace/system/capability.h b/ctdb/lib/replace/system/capability.h
new file mode 100644
index 0000000000..a7b78f0275
--- /dev/null
+++ b/ctdb/lib/replace/system/capability.h
@@ -0,0 +1,55 @@
+#ifndef _system_capability_h
+#define _system_capability_h
+/* 
+   Unix SMB/CIFS implementation.
+
+   capability system include wrappers
+
+   Copyright (C) Andrew Tridgell 2004
+   
+     ** NOTE! The following LGPL license applies to the replace
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#ifdef HAVE_SYS_CAPABILITY_H
+
+#if defined(BROKEN_REDHAT_7_SYSTEM_HEADERS) && !defined(_I386_STATFS_H) && !defined(_PPC_STATFS_H)
+#define _I386_STATFS_H
+#define _PPC_STATFS_H
+#define BROKEN_REDHAT_7_STATFS_WORKAROUND
+#endif
+
+#if defined(BROKEN_RHEL5_SYS_CAP_HEADER) && !defined(_LINUX_TYPES_H)
+#define BROKEN_RHEL5_SYS_CAP_HEADER_WORKAROUND
+#endif
+
+#include <sys/capability.h>
+
+#ifdef BROKEN_RHEL5_SYS_CAP_HEADER_WORKAROUND
+#undef _LINUX_TYPES_H
+#undef BROKEN_RHEL5_SYS_CAP_HEADER_WORKAROUND
+#endif
+
+#ifdef BROKEN_REDHAT_7_STATFS_WORKAROUND
+#undef _PPC_STATFS_H
+#undef _I386_STATFS_H
+#undef BROKEN_REDHAT_7_STATFS_WORKAROUND
+#endif
+
+#endif
+
+#endif
diff --git a/ctdb/lib/replace/system/config.m4 b/ctdb/lib/replace/system/config.m4
new file mode 100644
index 0000000000..b7cdf1414a
--- /dev/null
+++ b/ctdb/lib/replace/system/config.m4
@@ -0,0 +1,145 @@
+# filesys
+AC_HEADER_DIRENT 
+AC_CHECK_HEADERS(fcntl.h sys/fcntl.h sys/resource.h sys/ioctl.h sys/mode.h sys/filio.h sys/fs/s5param.h sys/filsys.h)
+AC_CHECK_HEADERS(sys/acl.h acl/libacl.h sys/file.h)
+
+# select
+AC_CHECK_HEADERS(sys/select.h)
+
+# poll
+AC_CHECK_HEADERS(poll.h)
+AC_CHECK_FUNCS(poll,[],[LIBREPLACEOBJ="${LIBREPLACEOBJ} $libreplacedir/poll.o"])
+
+# time
+AC_CHECK_HEADERS(sys/time.h utime.h)
+AC_HEADER_TIME
+AC_CHECK_FUNCS(utime utimes)
+
+AC_CACHE_CHECK([if gettimeofday takes TZ argument],libreplace_cv_HAVE_GETTIMEOFDAY_TZ,[
+AC_TRY_RUN([
+#include <sys/time.h>
+#include <unistd.h>
+main() { struct timeval tv; exit(gettimeofday(&tv, NULL));}],
+           libreplace_cv_HAVE_GETTIMEOFDAY_TZ=yes,libreplace_cv_HAVE_GETTIMEOFDAY_TZ=no,libreplace_cv_HAVE_GETTIMEOFDAY_TZ=yes)])
+if test x"$libreplace_cv_HAVE_GETTIMEOFDAY_TZ" = x"yes"; then
+    AC_DEFINE(HAVE_GETTIMEOFDAY_TZ,1,[Whether gettimeofday() is available])
+fi
+
+# wait
+AC_HEADER_SYS_WAIT
+
+# capability
+AC_CHECK_HEADERS(sys/capability.h)
+
+case "$host_os" in
+*linux*)
+AC_CACHE_CHECK([for broken RedHat 7.2 system header files],libreplace_cv_BROKEN_REDHAT_7_SYSTEM_HEADERS,[
+AC_TRY_COMPILE([
+	#ifdef HAVE_SYS_VFS_H
+	#include <sys/vfs.h>
+	#endif
+	#ifdef HAVE_SYS_CAPABILITY_H
+	#include <sys/capability.h>
+	#endif
+	],[
+	int i;
+	],
+	libreplace_cv_BROKEN_REDHAT_7_SYSTEM_HEADERS=no,
+	libreplace_cv_BROKEN_REDHAT_7_SYSTEM_HEADERS=yes
+)])
+if test x"$libreplace_cv_BROKEN_REDHAT_7_SYSTEM_HEADERS" = x"yes"; then
+	AC_DEFINE(BROKEN_REDHAT_7_SYSTEM_HEADERS,1,[Broken RedHat 7.2 system header files])
+fi
+
+AC_CACHE_CHECK([for broken RHEL5 sys/capability.h],libreplace_cv_BROKEN_RHEL5_SYS_CAP_HEADER,[
+AC_TRY_COMPILE([
+	#ifdef HAVE_SYS_CAPABILITY_H
+	#include <sys/capability.h>
+	#endif
+	#include <linux/types.h>
+	],[
+	__s8 i;
+	],
+	libreplace_cv_BROKEN_RHEL5_SYS_CAP_HEADER=no,
+	libreplace_cv_BROKEN_RHEL5_SYS_CAP_HEADER=yes
+)])
+if test x"$libreplace_cv_BROKEN_RHEL5_SYS_CAP_HEADER" = x"yes"; then
+	AC_DEFINE(BROKEN_RHEL5_SYS_CAP_HEADER,1,[Broken RHEL5 sys/capability.h])
+fi
+;;
+esac
+
+# passwd
+AC_CHECK_HEADERS(grp.h sys/id.h compat.h shadow.h sys/priv.h pwd.h sys/security.h)
+AC_CHECK_FUNCS(getpwnam_r getpwuid_r getpwent_r)
+AC_HAVE_DECL(getpwent_r, [
+	#include <unistd.h>
+	#include <pwd.h>
+	])
+AC_VERIFY_C_PROTOTYPE([struct passwd *getpwent_r(struct passwd *src, char *buf, int buflen)],
+	[
+	#ifndef HAVE_GETPWENT_R_DECL
+	#error missing getpwent_r prototype
+	#endif
+	return NULL;
+	],[
+	AC_DEFINE(SOLARIS_GETPWENT_R, 1, [getpwent_r solaris function prototype])
+	],[],[
+	#include <unistd.h>
+	#include <pwd.h>
+	])
+AC_VERIFY_C_PROTOTYPE([struct passwd *getpwent_r(struct passwd *src, char *buf, size_t buflen)],
+	[
+	#ifndef HAVE_GETPWENT_R_DECL
+	#error missing getpwent_r prototype
+	#endif
+	return NULL;
+	],[
+	AC_DEFINE(SOLARIS_GETPWENT_R, 1, [getpwent_r irix (similar to solaris) function prototype])
+	],[],[
+	#include <unistd.h>
+	#include <pwd.h>
+	])
+AC_CHECK_FUNCS(getgrnam_r getgrgid_r getgrent_r)
+AC_HAVE_DECL(getgrent_r, [
+	#include <unistd.h>
+	#include <grp.h>
+	])
+AC_VERIFY_C_PROTOTYPE([struct group *getgrent_r(struct group *src, char *buf, int buflen)],
+	[
+	#ifndef HAVE_GETGRENT_R_DECL
+	#error missing getgrent_r prototype
+	#endif
+	return NULL;
+	],[
+	AC_DEFINE(SOLARIS_GETGRENT_R, 1, [getgrent_r solaris function prototype])
+	],[],[
+	#include <unistd.h>
+	#include <grp.h>
+	])
+
+AC_VERIFY_C_PROTOTYPE([struct group *getgrent_r(struct group *src, char *buf, size_t buflen)],
+	[
+	#ifndef HAVE_GETGRENT_R_DECL
+	#error missing getgrent_r prototype
+	#endif
+	return NULL;
+	],[
+	AC_DEFINE(SOLARIS_GETGRENT_R, 1, [getgrent_r irix (similar to solaris)  function prototype])
+	],[],[
+	#include <unistd.h>
+	#include <grp.h>
+	])
+AC_CHECK_FUNCS(getgrouplist)
+
+# locale
+AC_CHECK_HEADERS(ctype.h locale.h langinfo.h)
+
+# glob
+AC_CHECK_HEADERS(fnmatch.h)
+
+# shmem
+AC_CHECK_HEADERS(sys/ipc.h sys/mman.h sys/shm.h )
+
+# terminal
+AC_CHECK_HEADERS(termios.h termio.h sys/termio.h )
diff --git a/ctdb/lib/replace/system/dir.h b/ctdb/lib/replace/system/dir.h
new file mode 100644
index 0000000000..dec2d54649
--- /dev/null
+++ b/ctdb/lib/replace/system/dir.h
@@ -0,0 +1,67 @@
+#ifndef _system_dir_h
+#define _system_dir_h
+/* 
+   Unix SMB/CIFS implementation.
+
+   directory system include wrappers
+
+   Copyright (C) Andrew Tridgell 2004
+   
+     ** NOTE! The following LGPL license applies to the replace
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#if HAVE_DIRENT_H
+# include <dirent.h>
+# define NAMLEN(dirent) strlen((dirent)->d_name)
+#else
+# define dirent direct
+# define NAMLEN(dirent) (dirent)->d_namlen
+# if HAVE_SYS_NDIR_H
+#  include <sys/ndir.h>
+# endif
+# if HAVE_SYS_DIR_H
+#  include <sys/dir.h>
+# endif
+# if HAVE_NDIR_H
+#  include <ndir.h>
+# endif
+#endif
+
+#ifndef HAVE_MKDIR_MODE
+#define mkdir(dir, mode) mkdir(dir)
+#endif
+
+/* Test whether a file name is the "." or ".." directory entries.
+ * These really should be inline functions.
+ */
+#ifndef ISDOT
+#define ISDOT(path) ( \
+			*((const char *)(path)) == '.' && \
+			*(((const char *)(path)) + 1) == '\0' \
+		    )
+#endif
+
+#ifndef ISDOTDOT
+#define ISDOTDOT(path)	( \
+			    *((const char *)(path)) == '.' && \
+			    *(((const char *)(path)) + 1) == '.' && \
+			    *(((const char *)(path)) + 2) == '\0' \
+			)
+#endif
+
+#endif
diff --git a/ctdb/lib/replace/system/filesys.h b/ctdb/lib/replace/system/filesys.h
new file mode 100644
index 0000000000..c8ac2b4f74
--- /dev/null
+++ b/ctdb/lib/replace/system/filesys.h
@@ -0,0 +1,277 @@
+#ifndef _system_filesys_h
+#define _system_filesys_h
+/*
+   Unix SMB/CIFS implementation.
+
+   filesystem system include wrappers
+
+   Copyright (C) Andrew Tridgell 2004
+
+     ** NOTE! The following LGPL license applies to the replace
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+
+*/
+
+#include <unistd.h>
+#include <sys/stat.h>
+
+#ifdef HAVE_SYS_PARAM_H
+#include <sys/param.h>
+#endif
+
+#ifdef HAVE_SYS_MOUNT_H
+#include <sys/mount.h>
+#endif
+
+#ifdef HAVE_MNTENT_H
+#include <mntent.h>
+#endif
+
+#ifdef HAVE_SYS_VFS_H
+#include <sys/vfs.h>
+#endif
+
+#ifdef HAVE_SYS_ACL_H
+#include <sys/acl.h>
+#endif
+
+#ifdef HAVE_ACL_LIBACL_H
+#include <acl/libacl.h>
+#endif
+
+#ifdef HAVE_SYS_FS_S5PARAM_H
+#include <sys/fs/s5param.h>
+#endif
+
+#if defined (HAVE_SYS_FILSYS_H) && !defined (_CRAY)
+#include <sys/filsys.h>
+#endif
+
+#ifdef HAVE_SYS_STATFS_H
+# include <sys/statfs.h>
+#endif
+
+#ifdef HAVE_DUSTAT_H
+#include <sys/dustat.h>
+#endif
+
+#ifdef HAVE_SYS_STATVFS_H
+#include <sys/statvfs.h>
+#endif
+
+#ifdef HAVE_SYS_FILIO_H
+#include <sys/filio.h>
+#endif
+
+#ifdef HAVE_SYS_FILE_H
+#include <sys/file.h>
+#endif
+
+#ifdef HAVE_FCNTL_H
+#include <fcntl.h>
+#else
+#ifdef HAVE_SYS_FCNTL_H
+#include <sys/fcntl.h>
+#endif
+#endif
+
+#ifdef HAVE_SYS_MODE_H
+/* apparently AIX needs this for S_ISLNK */
+#ifndef S_ISLNK
+#include <sys/mode.h>
+#endif
+#endif
+
+#ifdef HAVE_SYS_IOCTL_H
+#include <sys/ioctl.h>
+#endif
+
+#ifdef HAVE_SYS_UIO_H
+#include <sys/uio.h>
+#endif
+
+/*
+ * Veritas File System.  Often in addition to native.
+ * Quotas different.
+ */
+#if defined(HAVE_SYS_FS_VX_QUOTA_H)
+#define VXFS_QUOTA
+#endif
+
+#if HAVE_SYS_ATTRIBUTES_H
+#include <sys/attributes.h>
+#elif HAVE_ATTR_ATTRIBUTES_H
+#include <attr/attributes.h>
+#endif
+
+/* mutually exclusive (SuSE 8.2) */
+#if HAVE_ATTR_XATTR_H
+#include <attr/xattr.h>
+#elif HAVE_SYS_XATTR_H
+#include <sys/xattr.h>
+#endif
+
+#ifdef HAVE_SYS_EA_H
+#include <sys/ea.h>
+#endif
+
+#ifdef HAVE_SYS_EXTATTR_H
+#include <sys/extattr.h>
+#endif
+
+#ifdef HAVE_SYS_RESOURCE_H
+#include <sys/resource.h>
+#endif
+
+#ifndef XATTR_CREATE
+#define XATTR_CREATE  0x1       /* set value, fail if attr already exists */
+#endif
+
+#ifndef XATTR_REPLACE
+#define XATTR_REPLACE 0x2       /* set value, fail if attr does not exist */
+#endif
+
+/* Some POSIX definitions for those without */
+
+#ifndef S_IFDIR
+#define S_IFDIR         0x4000
+#endif
+#ifndef S_ISDIR
+#define S_ISDIR(mode)   ((mode & 0xF000) == S_IFDIR)
+#endif
+#ifndef S_IRWXU
+#define S_IRWXU 00700           /* read, write, execute: owner */
+#endif
+#ifndef S_IRUSR
+#define S_IRUSR 00400           /* read permission: owner */
+#endif
+#ifndef S_IWUSR
+#define S_IWUSR 00200           /* write permission: owner */
+#endif
+#ifndef S_IXUSR
+#define S_IXUSR 00100           /* execute permission: owner */
+#endif
+#ifndef S_IRWXG
+#define S_IRWXG 00070           /* read, write, execute: group */
+#endif
+#ifndef S_IRGRP
+#define S_IRGRP 00040           /* read permission: group */
+#endif
+#ifndef S_IWGRP
+#define S_IWGRP 00020           /* write permission: group */
+#endif
+#ifndef S_IXGRP
+#define S_IXGRP 00010           /* execute permission: group */
+#endif
+#ifndef S_IRWXO
+#define S_IRWXO 00007           /* read, write, execute: other */
+#endif
+#ifndef S_IROTH
+#define S_IROTH 00004           /* read permission: other */
+#endif
+#ifndef S_IWOTH
+#define S_IWOTH 00002           /* write permission: other */
+#endif
+#ifndef S_IXOTH
+#define S_IXOTH 00001           /* execute permission: other */
+#endif
+
+#ifndef O_ACCMODE
+#define O_ACCMODE (O_RDONLY | O_WRONLY | O_RDWR)
+#endif
+
+#ifndef MAXPATHLEN
+#define MAXPATHLEN 256
+#endif
+
+#ifndef SEEK_SET
+#define SEEK_SET 0
+#endif
+
+#ifdef _WIN32
+#define mkdir(d,m) _mkdir(d)
+#endif
+
+#ifdef UID_WRAPPER
+# ifndef UID_WRAPPER_DISABLE
+#  ifndef UID_WRAPPER_NOT_REPLACE
+#   define UID_WRAPPER_REPLACE
+#  endif /* UID_WRAPPER_NOT_REPLACE */
+#  include "../uid_wrapper/uid_wrapper.h"
+# endif /* UID_WRAPPER_DISABLE */
+#else /* UID_WRAPPER */
+# define uwrap_enabled() 0
+#endif /* UID_WRAPPER */
+
+/*
+   this allows us to use a uniform error handling for our xattr
+   wrappers
+*/
+#ifndef ENOATTR
+#define ENOATTR ENODATA
+#endif
+
+
+#if !defined(HAVE_GETXATTR) || defined(XATTR_ADDITIONAL_OPTIONS)
+ssize_t rep_getxattr (const char *path, const char *name, void *value, size_t size);
+#define getxattr(path, name, value, size) rep_getxattr(path, name, value, size)
+/* define is in "replace.h" */
+#endif
+
+#if !defined(HAVE_FGETXATTR) || defined(XATTR_ADDITIONAL_OPTIONS)
+ssize_t rep_fgetxattr (int filedes, const char *name, void *value, size_t size);
+#define fgetxattr(filedes, name, value, size) rep_fgetxattr(filedes, name, value, size)
+/* define is in "replace.h" */
+#endif
+
+#if !defined(HAVE_LISTXATTR) || defined(XATTR_ADDITIONAL_OPTIONS)
+ssize_t rep_listxattr (const char *path, char *list, size_t size);
+#define listxattr(path, list, size) rep_listxattr(path, list, size)
+/* define is in "replace.h" */
+#endif
+
+#if !defined(HAVE_FLISTXATTR) || defined(XATTR_ADDITIONAL_OPTIONS)
+ssize_t rep_flistxattr (int filedes, char *list, size_t size);
+#define flistxattr(filedes, value, size) rep_flistxattr(filedes, value, size)
+/* define is in "replace.h" */
+#endif
+
+#if !defined(HAVE_REMOVEXATTR) || defined(XATTR_ADDITIONAL_OPTIONS)
+int rep_removexattr (const char *path, const char *name);
+#define removexattr(path, name) rep_removexattr(path, name)
+/* define is in "replace.h" */
+#endif
+
+#if !defined(HAVE_FREMOVEXATTR) || defined(XATTR_ADDITIONAL_OPTIONS)
+int rep_fremovexattr (int filedes, const char *name);
+#define fremovexattr(filedes, name) rep_fremovexattr(filedes, name)
+/* define is in "replace.h" */
+#endif
+
+#if !defined(HAVE_SETXATTR) || defined(XATTR_ADDITIONAL_OPTIONS)
+int rep_setxattr (const char *path, const char *name, const void *value, size_t size, int flags);
+#define setxattr(path, name, value, size, flags) rep_setxattr(path, name, value, size, flags)
+/* define is in "replace.h" */
+#endif
+
+#if !defined(HAVE_FSETXATTR) || defined(XATTR_ADDITIONAL_OPTIONS)
+int rep_fsetxattr (int filedes, const char *name, const void *value, size_t size, int flags);
+#define fsetxattr(filedes, name, value, size, flags) rep_fsetxattr(filedes, name, value, size, flags)
+/* define is in "replace.h" */
+#endif
+
+#endif
diff --git a/ctdb/lib/replace/system/glob.h b/ctdb/lib/replace/system/glob.h
new file mode 100644
index 0000000000..3e23db6828
--- /dev/null
+++ b/ctdb/lib/replace/system/glob.h
@@ -0,0 +1,37 @@
+#ifndef _system_glob_h
+#define _system_glob_h
+/* 
+   Unix SMB/CIFS implementation.
+
+   glob system include wrappers
+
+   Copyright (C) Andrew Tridgell 2004
+   
+     ** NOTE! The following LGPL license applies to the replace
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+
+*/
+
+#ifdef HAVE_GLOB_H
+#include <glob.h>
+#endif
+
+#ifdef HAVE_FNMATCH_H
+#include <fnmatch.h>
+#endif
+
+#endif
diff --git a/ctdb/lib/replace/system/gssapi.h b/ctdb/lib/replace/system/gssapi.h
new file mode 100644
index 0000000000..6386c7b21a
--- /dev/null
+++ b/ctdb/lib/replace/system/gssapi.h
@@ -0,0 +1,53 @@
+#ifndef _system_gssapi_h
+#define _system_gssapi_h
+
+/*
+   Unix SMB/CIFS implementation.
+
+   GSSAPI system include wrappers
+
+   Copyright (C) Andrew Tridgell 2004
+
+     ** NOTE! The following LGPL license applies to the replace
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+
+*/
+
+#ifdef HAVE_GSSAPI
+
+#ifdef HAVE_GSSAPI_GSSAPI_EXT_H
+#include <gssapi/gssapi_ext.h>
+#elif HAVE_GSSAPI_GSSAPI_H
+#include <gssapi/gssapi.h>
+#elif HAVE_GSSAPI_GSSAPI_GENERIC_H
+#include <gssapi/gssapi_generic.h>
+#elif HAVE_GSSAPI_H
+#include <gssapi.h>
+#endif
+
+#if HAVE_GSSAPI_GSSAPI_KRB5_H
+#include <gssapi/gssapi_krb5.h>
+#endif
+
+#if HAVE_GSSAPI_GSSAPI_SPNEGO_H
+#include <gssapi/gssapi_spnego.h>
+#elif HAVE_GSSAPI_SPNEGO_H
+#include <gssapi_spnego.h>
+#endif
+
+#endif
+#endif
diff --git a/ctdb/lib/replace/system/iconv.h b/ctdb/lib/replace/system/iconv.h
new file mode 100644
index 0000000000..3c8a71f2f7
--- /dev/null
+++ b/ctdb/lib/replace/system/iconv.h
@@ -0,0 +1,57 @@
+#ifndef _system_iconv_h
+#define _system_iconv_h
+/* 
+   Unix SMB/CIFS implementation.
+
+   iconv memory system include wrappers
+
+   Copyright (C) Andrew Tridgell 2004
+   
+     ** NOTE! The following LGPL license applies to the replace
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+
+*/
+
+#if !defined(HAVE_ICONV) && defined(HAVE_ICONV_H)
+#define HAVE_ICONV
+#endif
+
+#if !defined(HAVE_GICONV) && defined(HAVE_GICONV_H)
+#define HAVE_GICONV
+#endif
+
+#if !defined(HAVE_BICONV) && defined(HAVE_BICONV_H)
+#define HAVE_BICONV
+#endif
+
+#ifdef HAVE_NATIVE_ICONV
+#if defined(HAVE_ICONV)
+#include <iconv.h>
+#elif defined(HAVE_GICONV)
+#include <giconv.h>
+#elif defined(HAVE_BICONV)
+#include <biconv.h>
+#endif
+#endif /* HAVE_NATIVE_ICONV */
+
+/* needed for some systems without iconv. Doesn't really matter
+   what error code we use */
+#ifndef EILSEQ
+#define EILSEQ EIO
+#endif
+
+#endif
diff --git a/ctdb/lib/replace/system/kerberos.h b/ctdb/lib/replace/system/kerberos.h
new file mode 100644
index 0000000000..636ce0f2e0
--- /dev/null
+++ b/ctdb/lib/replace/system/kerberos.h
@@ -0,0 +1,41 @@
+#ifndef _system_kerberos_h
+#define _system_kerberos_h
+
+/* 
+   Unix SMB/CIFS implementation.
+
+   kerberos system include wrappers
+
+   Copyright (C) Andrew Tridgell 2004
+
+     ** NOTE! The following LGPL license applies to the replace
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+
+*/
+
+#ifdef HAVE_KRB5
+
+#if HAVE_KRB5_H
+#include <krb5.h>
+#endif
+
+#if HAVE_COM_ERR_H
+#include <com_err.h>
+#endif
+
+#endif
+#endif
diff --git a/ctdb/lib/replace/system/locale.h b/ctdb/lib/replace/system/locale.h
new file mode 100644
index 0000000000..504a3bb470
--- /dev/null
+++ b/ctdb/lib/replace/system/locale.h
@@ -0,0 +1,42 @@
+#ifndef _system_locale_h
+#define _system_locale_h
+
+/* 
+   Unix SMB/CIFS implementation.
+
+   locale include wrappers
+
+   Copyright (C) Andrew Tridgell 2004
+   
+     ** NOTE! The following LGPL license applies to the replace
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+
+*/
+
+#ifdef HAVE_CTYPE_H
+#include <ctype.h>
+#endif
+
+#ifdef HAVE_LOCALE_H
+#include <locale.h>
+#endif
+
+#ifdef HAVE_LANGINFO_H
+#include <langinfo.h>
+#endif
+
+#endif
diff --git a/ctdb/lib/replace/system/network.h b/ctdb/lib/replace/system/network.h
new file mode 100644
index 0000000000..7cb8d7becf
--- /dev/null
+++ b/ctdb/lib/replace/system/network.h
@@ -0,0 +1,390 @@
+#ifndef _system_network_h
+#define _system_network_h
+/* 
+   Unix SMB/CIFS implementation.
+
+   networking system include wrappers
+
+   Copyright (C) Andrew Tridgell 2004
+   Copyright (C) Jelmer Vernooij 2007
+   
+     ** NOTE! The following LGPL license applies to the replace
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+
+*/
+
+#ifndef LIBREPLACE_NETWORK_CHECKS
+#error "AC_LIBREPLACE_NETWORK_CHECKS missing in configure"
+#endif
+
+#include <unistd.h>
+
+#ifdef HAVE_SYS_SOCKET_H
+#include <sys/socket.h>
+#endif
+
+#ifdef HAVE_UNIXSOCKET
+#include <sys/un.h>
+#endif
+
+#ifdef HAVE_NETINET_IN_H
+#include <netinet/in.h>
+#endif
+#ifdef HAVE_ARPA_INET_H
+#include <arpa/inet.h>
+#endif
+
+#ifdef HAVE_NETDB_H
+#include <netdb.h>
+#endif
+
+#ifdef HAVE_NETINET_TCP_H
+#include <netinet/tcp.h>
+#endif
+
+/*
+ * The next three defines are needed to access the IPTOS_* options
+ * on some systems.
+ */
+
+#ifdef HAVE_NETINET_IN_SYSTM_H
+#include <netinet/in_systm.h>
+#endif
+
+#ifdef HAVE_NETINET_IN_IP_H
+#include <netinet/in_ip.h>
+#endif
+
+#ifdef HAVE_NETINET_IP_H
+#include <netinet/ip.h>
+#endif
+
+#ifdef HAVE_NET_IF_H
+#include <net/if.h>
+#endif
+
+#ifdef HAVE_UNISTD_H
+#include <unistd.h>
+#endif
+
+#ifdef HAVE_SYS_IOCTL_H
+#include <sys/ioctl.h>
+#endif
+
+#ifdef HAVE_SYS_UIO_H
+#include <sys/uio.h>
+#endif
+
+#ifdef HAVE_STROPTS_H
+#include <stropts.h>
+#endif
+
+#ifndef HAVE_SOCKLEN_T
+#define HAVE_SOCKLEN_T
+typedef int socklen_t;
+#endif
+
+#if !defined (HAVE_INET_NTOA) || defined(REPLACE_INET_NTOA)
+/* define is in "replace.h" */
+char *rep_inet_ntoa(struct in_addr ip);
+#endif
+
+#ifndef HAVE_INET_PTON
+/* define is in "replace.h" */
+int rep_inet_pton(int af, const char *src, void *dst);
+#endif
+
+#ifndef HAVE_INET_NTOP
+/* define is in "replace.h" */
+const char *rep_inet_ntop(int af, const void *src, char *dst, socklen_t size);
+#endif
+
+#ifndef HAVE_INET_ATON
+/* define is in "replace.h" */
+int rep_inet_aton(const char *src, struct in_addr *dst);
+#endif
+
+#ifndef HAVE_CONNECT
+/* define is in "replace.h" */
+int rep_connect(int sockfd, const struct sockaddr *addr, socklen_t addrlen);
+#endif
+
+#ifndef HAVE_GETHOSTBYNAME
+/* define is in "replace.h" */
+struct hostent *rep_gethostbyname(const char *name);
+#endif
+
+#ifdef HAVE_IFADDRS_H
+#include <ifaddrs.h>
+#endif
+
+#ifndef HAVE_STRUCT_IFADDRS
+struct ifaddrs {
+	struct ifaddrs   *ifa_next;         /* Pointer to next struct */
+	char             *ifa_name;         /* Interface name */
+	unsigned int     ifa_flags;         /* Interface flags */
+	struct sockaddr  *ifa_addr;         /* Interface address */
+	struct sockaddr  *ifa_netmask;      /* Interface netmask */
+#undef ifa_dstaddr
+	struct sockaddr  *ifa_dstaddr;      /* P2P interface destination */
+	void             *ifa_data;         /* Address specific data */
+};
+#endif
+
+#ifndef HAVE_GETIFADDRS
+int rep_getifaddrs(struct ifaddrs **);
+#endif
+
+#ifndef HAVE_FREEIFADDRS
+void rep_freeifaddrs(struct ifaddrs *);
+#endif
+
+#ifndef HAVE_SOCKETPAIR
+/* define is in "replace.h" */
+int rep_socketpair(int d, int type, int protocol, int sv[2]);
+#endif
+
+/*
+ * Some systems have getaddrinfo but not the
+ * defines needed to use it.
+ */
+
+/* Various macros that ought to be in <netdb.h>, but might not be */
+
+#ifndef EAI_FAIL
+#define EAI_BADFLAGS	(-1)
+#define EAI_NONAME	(-2)
+#define EAI_AGAIN	(-3)
+#define EAI_FAIL	(-4)
+#define EAI_FAMILY	(-6)
+#define EAI_SOCKTYPE	(-7)
+#define EAI_SERVICE	(-8)
+#define EAI_MEMORY	(-10)
+#define EAI_SYSTEM	(-11)
+#endif   /* !EAI_FAIL */
+
+#ifndef AI_PASSIVE
+#define AI_PASSIVE	0x0001
+#endif
+
+#ifndef AI_CANONNAME
+#define AI_CANONNAME	0x0002
+#endif
+
+#ifndef AI_NUMERICHOST
+/*
+ * some platforms don't support AI_NUMERICHOST; define as zero if using
+ * the system version of getaddrinfo...
+ */
+#if defined(HAVE_STRUCT_ADDRINFO) && defined(HAVE_GETADDRINFO)
+#define AI_NUMERICHOST	0
+#else
+#define AI_NUMERICHOST	0x0004
+#endif
+#endif
+
+/*
+ * Some of the functions in source3/lib/util_sock.c use AI_ADDRCONFIG. On QNX
+ * 6.3.0, this macro is defined but, if it's used, getaddrinfo will fail. This
+ * prevents smbd from opening any sockets.
+ *
+ * If I undefine AI_ADDRCONFIG on such systems and define it to be 0,
+ * this works around the issue.
+ */
+#ifdef __QNX__
+#include <sys/neutrino.h>
+#if _NTO_VERSION == 630
+#undef AI_ADDRCONFIG
+#endif
+#endif
+#ifndef AI_ADDRCONFIG
+/*
+ * logic copied from AI_NUMERICHOST
+ */
+#if defined(HAVE_STRUCT_ADDRINFO) && defined(HAVE_GETADDRINFO)
+#define AI_ADDRCONFIG	0
+#else
+#define AI_ADDRCONFIG	0x0020
+#endif
+#endif
+
+#ifndef AI_NUMERICSERV
+/*
+ * logic copied from AI_NUMERICHOST
+ */
+#if defined(HAVE_STRUCT_ADDRINFO) && defined(HAVE_GETADDRINFO)
+#define AI_NUMERICSERV	0
+#else
+#define AI_NUMERICSERV	0x0400
+#endif
+#endif
+
+#ifndef NI_NUMERICHOST
+#define NI_NUMERICHOST	1
+#endif
+
+#ifndef NI_NUMERICSERV
+#define NI_NUMERICSERV	2
+#endif
+
+#ifndef NI_NOFQDN
+#define NI_NOFQDN	4
+#endif
+
+#ifndef NI_NAMEREQD
+#define NI_NAMEREQD 	8
+#endif
+
+#ifndef NI_DGRAM
+#define NI_DGRAM	16
+#endif
+
+
+#ifndef NI_MAXHOST
+#define NI_MAXHOST	1025
+#endif
+
+#ifndef NI_MAXSERV
+#define NI_MAXSERV	32
+#endif
+
+/*
+ * glibc on linux doesn't seem to have MSG_WAITALL
+ * defined. I think the kernel has it though..
+ */
+#ifndef MSG_WAITALL
+#define MSG_WAITALL 0
+#endif
+
+#ifndef INADDR_LOOPBACK
+#define INADDR_LOOPBACK 0x7f000001
+#endif
+
+#ifndef INADDR_NONE
+#define INADDR_NONE 0xffffffff
+#endif
+
+#ifndef EAFNOSUPPORT
+#define EAFNOSUPPORT EINVAL
+#endif
+
+#ifndef INET_ADDRSTRLEN
+#define INET_ADDRSTRLEN 16
+#endif
+
+#ifndef INET6_ADDRSTRLEN
+#define INET6_ADDRSTRLEN 46
+#endif
+
+#ifndef HOST_NAME_MAX
+#define HOST_NAME_MAX 255
+#endif
+
+#ifndef MAXHOSTNAMELEN
+#define MAXHOSTNAMELEN HOST_NAME_MAX
+#endif
+
+#ifndef HAVE_SA_FAMILY_T
+#define HAVE_SA_FAMILY_T
+typedef unsigned short int sa_family_t;
+#endif
+
+#ifndef HAVE_STRUCT_SOCKADDR_STORAGE
+#define HAVE_STRUCT_SOCKADDR_STORAGE
+#ifdef HAVE_STRUCT_SOCKADDR_IN6
+#define sockaddr_storage sockaddr_in6
+#define ss_family sin6_family
+#define HAVE_SS_FAMILY 1
+#else /*HAVE_STRUCT_SOCKADDR_IN6*/
+#define sockaddr_storage sockaddr_in
+#define ss_family sin_family
+#define HAVE_SS_FAMILY 1
+#endif /*HAVE_STRUCT_SOCKADDR_IN6*/
+#endif /*HAVE_STRUCT_SOCKADDR_STORAGE*/
+
+#ifndef HAVE_SS_FAMILY
+#ifdef HAVE___SS_FAMILY
+#define ss_family __ss_family
+#define HAVE_SS_FAMILY 1
+#endif
+#endif
+
+#ifndef IOV_MAX
+# ifdef UIO_MAXIOV
+#  define IOV_MAX UIO_MAXIOV
+# else
+#  ifdef __sgi
+    /*
+     * IRIX 6.5 has sysconf(_SC_IOV_MAX)
+     * which might return 512 or bigger
+     */
+#   define IOV_MAX 512
+#  endif
+# endif
+#endif
+
+#ifndef HAVE_STRUCT_ADDRINFO
+#define HAVE_STRUCT_ADDRINFO
+struct addrinfo {
+	int			ai_flags;
+	int			ai_family;
+	int			ai_socktype;
+	int			ai_protocol;
+	socklen_t		ai_addrlen;
+	struct sockaddr 	*ai_addr;
+	char			*ai_canonname;
+	struct addrinfo		*ai_next;
+};
+#endif   /* HAVE_STRUCT_ADDRINFO */
+
+#if !defined(HAVE_GETADDRINFO)
+#include "getaddrinfo.h"
+#endif
+
+/* Needed for some systems that don't define it (Solaris). */
+#ifndef ifr_netmask
+#define ifr_netmask ifr_addr
+#endif
+
+/* Some old Linux systems have broken header files */
+#ifdef HAVE_IPV6
+#ifdef HAVE_LINUX_IPV6_V6ONLY_26
+#define IPV6_V6ONLY 26
+#endif /* HAVE_LINUX_IPV6_V6ONLY_26 */
+#endif /* HAVE_IPV6 */
+
+#ifdef SOCKET_WRAPPER
+#ifndef SOCKET_WRAPPER_DISABLE
+#ifndef SOCKET_WRAPPER_NOT_REPLACE
+#define SOCKET_WRAPPER_REPLACE
+#endif /* SOCKET_WRAPPER_NOT_REPLACE */
+#include "../socket_wrapper/socket_wrapper.h"
+#endif /* SOCKET_WRAPPER_DISABLE */
+#endif /* SOCKET_WRAPPER */
+
+#ifdef UID_WRAPPER
+# ifndef UID_WRAPPER_DISABLE
+#  ifndef UID_WRAPPER_NOT_REPLACE
+#   define UID_WRAPPER_REPLACE
+#  endif /* UID_WRAPPER_NOT_REPLACE */
+#  include "../uid_wrapper/uid_wrapper.h"
+# endif /* UID_WRAPPER_DISABLE */
+#else /* UID_WRAPPER */
+# define uwrap_enabled() 0
+#endif /* UID_WRAPPER */
+
+#endif
diff --git a/ctdb/lib/replace/system/passwd.h b/ctdb/lib/replace/system/passwd.h
new file mode 100644
index 0000000000..8257e066ac
--- /dev/null
+++ b/ctdb/lib/replace/system/passwd.h
@@ -0,0 +1,114 @@
+#ifndef _system_passwd_h
+#define _system_passwd_h
+
+/*
+   Unix SMB/CIFS implementation.
+
+   passwd system include wrappers
+
+   Copyright (C) Andrew Tridgell 2004
+
+     ** NOTE! The following LGPL license applies to the replace
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+
+*/
+
+/* this needs to be included before nss_wrapper.h on some systems */
+#include <unistd.h>
+
+#ifdef HAVE_PWD_H
+#include <pwd.h>
+#endif
+#ifdef HAVE_GRP_H
+#include <grp.h>
+#endif
+#ifdef HAVE_SYS_PRIV_H
+#include <sys/priv.h>
+#endif
+#ifdef HAVE_SYS_ID_H
+#include <sys/id.h>
+#endif
+
+#ifdef HAVE_CRYPT_H
+#include <crypt.h>
+#endif
+
+#ifdef HAVE_SHADOW_H
+#include <shadow.h>
+#endif
+
+#ifdef HAVE_SYS_SECURITY_H
+#include <sys/security.h>
+#include <prot.h>
+#define PASSWORD_LENGTH 16
+#endif  /* HAVE_SYS_SECURITY_H */
+
+#ifdef HAVE_GETPWANAM
+#include <sys/label.h>
+#include <sys/audit.h>
+#include <pwdadj.h>
+#endif
+
+#ifdef HAVE_COMPAT_H
+#include <compat.h>
+#endif
+
+#ifndef NGROUPS_MAX
+#define NGROUPS_MAX 32 /* Guess... */
+#endif
+
+/* what is the longest significant password available on your system?
+ Knowing this speeds up password searches a lot */
+#ifndef PASSWORD_LENGTH
+#define PASSWORD_LENGTH 8
+#endif
+
+#if defined(HAVE_PUTPRPWNAM) && defined(AUTH_CLEARTEXT_SEG_CHARS)
+#define OSF1_ENH_SEC 1
+#endif
+
+#ifndef ALLOW_CHANGE_PASSWORD
+#if (defined(HAVE_TERMIOS_H) && defined(HAVE_DUP2) && defined(HAVE_SETSID))
+#define ALLOW_CHANGE_PASSWORD 1
+#endif
+#endif
+
+#if defined(HAVE_CRYPT16) && defined(HAVE_GETAUTHUID)
+#define ULTRIX_AUTH 1
+#endif
+
+#ifdef NSS_WRAPPER
+#ifndef NSS_WRAPPER_DISABLE
+#ifndef NSS_WRAPPER_NOT_REPLACE
+#define NSS_WRAPPER_REPLACE
+#endif /* NSS_WRAPPER_NOT_REPLACE */
+#include "../nss_wrapper/nss_wrapper.h"
+#endif /* NSS_WRAPPER_DISABLE */
+#endif /* NSS_WRAPPER */
+
+#ifdef UID_WRAPPER
+# ifndef UID_WRAPPER_DISABLE
+#  ifndef UID_WRAPPER_NOT_REPLACE
+#   define UID_WRAPPER_REPLACE
+#  endif /* UID_WRAPPER_NOT_REPLACE */
+#  include "../uid_wrapper/uid_wrapper.h"
+# endif /* UID_WRAPPER_DISABLE */
+#else /* UID_WRAPPER */
+# define uwrap_enabled() 0
+#endif /* UID_WRAPPER */
+
+#endif
diff --git a/ctdb/lib/replace/system/readline.h b/ctdb/lib/replace/system/readline.h
new file mode 100644
index 0000000000..e6b8fb9129
--- /dev/null
+++ b/ctdb/lib/replace/system/readline.h
@@ -0,0 +1,58 @@
+#ifndef _system_readline_h
+#define _system_readline_h
+/* 
+   Unix SMB/CIFS implementation.
+
+   Readline wrappers
+   
+     ** NOTE! The following LGPL license applies to the replace
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+
+*/
+
+#ifdef HAVE_LIBREADLINE
+#  ifdef HAVE_READLINE_READLINE_H
+#    include <readline/readline.h>
+#    ifdef HAVE_READLINE_HISTORY_H
+#      include <readline/history.h>
+#    endif
+#  else
+#    ifdef HAVE_READLINE_H
+#      include <readline.h>
+#      ifdef HAVE_HISTORY_H
+#        include <history.h>
+#      endif
+#    else
+#      undef HAVE_LIBREADLINE
+#    endif
+#  endif
+#endif
+
+#ifdef HAVE_NEW_LIBREADLINE
+#ifdef HAVE_CPPFUNCTION
+#  define RL_COMPLETION_CAST (CPPFunction *)
+#elif HAVE_RL_COMPLETION_T
+#  define RL_COMPLETION_CAST (rl_completion_t *)
+#else
+#  define RL_COMPLETION_CAST
+#endif
+#else
+/* This type is missing from libreadline<4.0  (approximately) */
+#  define RL_COMPLETION_CAST
+#endif /* HAVE_NEW_LIBREADLINE */
+
+#endif
diff --git a/ctdb/lib/replace/system/select.h b/ctdb/lib/replace/system/select.h
new file mode 100644
index 0000000000..11c5390d90
--- /dev/null
+++ b/ctdb/lib/replace/system/select.h
@@ -0,0 +1,77 @@
+#ifndef _system_select_h
+#define _system_select_h
+/* 
+   Unix SMB/CIFS implementation.
+
+   select system include wrappers
+
+   Copyright (C) Andrew Tridgell 2004
+   
+     ** NOTE! The following LGPL license applies to the replace
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+
+*/
+
+#ifdef HAVE_SYS_SELECT_H
+#include <sys/select.h>
+#endif
+
+#ifdef HAVE_SYS_EPOLL_H
+#include <sys/epoll.h>
+#endif
+
+#ifndef SELECT_CAST
+#define SELECT_CAST
+#endif
+
+#ifdef HAVE_POLL
+
+#include <poll.h>
+
+#else
+
+/* Type used for the number of file descriptors.  */
+typedef unsigned long int nfds_t;
+
+/* Data structure describing a polling request.  */
+struct pollfd
+{
+	int fd;		   /* File descriptor to poll.  */
+	short int events;  /* Types of events poller cares about.  */
+	short int revents; /* Types of events that actually occurred.  */
+};
+
+/* Event types that can be polled for.  These bits may be set in `events'
+   to indicate the interesting event types; they will appear in `revents'
+   to indicate the status of the file descriptor.  */
+#define POLLIN		0x001		/* There is data to read.  */
+#define POLLPRI		0x002		/* There is urgent data to read.  */
+#define POLLOUT		0x004		/* Writing now will not block.  */
+#define POLLRDNORM	0x040		/* Normal data may be read.  */
+#define POLLRDBAND	0x080		/* Priority data may be read.  */
+#define POLLWRNORM	0x100		/* Writing now will not block.  */
+#define POLLWRBAND	0x200		/* Priority data may be written.  */
+#define POLLERR		0x008		/* Error condition.  */
+#define POLLHUP		0x010		/* Hung up.  */
+#define POLLNVAL	0x020		/* Invalid polling request.  */
+
+/* define is in "replace.h" */
+int rep_poll(struct pollfd *fds, nfds_t nfds, int timeout);
+
+#endif
+
+#endif
diff --git a/ctdb/lib/replace/system/shmem.h b/ctdb/lib/replace/system/shmem.h
new file mode 100644
index 0000000000..64fe39b6cb
--- /dev/null
+++ b/ctdb/lib/replace/system/shmem.h
@@ -0,0 +1,59 @@
+#ifndef _system_shmem_h
+#define _system_shmem_h
+/* 
+   Unix SMB/CIFS implementation.
+
+   shared memory system include wrappers
+
+   Copyright (C) Andrew Tridgell 2004
+   
+     ** NOTE! The following LGPL license applies to the replace
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+
+*/
+
+#if defined(HAVE_SYS_IPC_H)
+#include <sys/ipc.h>
+#endif /* HAVE_SYS_IPC_H */
+
+#if defined(HAVE_SYS_SHM_H)
+#include <sys/shm.h>
+#endif /* HAVE_SYS_SHM_H */
+
+#ifdef HAVE_SYS_MMAN_H
+#include <sys/mman.h>
+#endif
+
+/* NetBSD doesn't have these */
+#ifndef SHM_R
+#define SHM_R 0400
+#endif
+
+#ifndef SHM_W
+#define SHM_W 0200
+#endif
+
+
+#ifndef MAP_FILE
+#define MAP_FILE 0
+#endif
+
+#ifndef MAP_FAILED
+#define MAP_FAILED ((void *)-1)
+#endif
+
+#endif
diff --git a/ctdb/lib/replace/system/syslog.h b/ctdb/lib/replace/system/syslog.h
new file mode 100644
index 0000000000..104be1df84
--- /dev/null
+++ b/ctdb/lib/replace/system/syslog.h
@@ -0,0 +1,70 @@
+#ifndef _system_syslog_h
+#define _system_syslog_h
+/* 
+   Unix SMB/CIFS implementation.
+
+   syslog system include wrappers
+
+   Copyright (C) Andrew Tridgell 2004
+   
+     ** NOTE! The following LGPL license applies to the replace
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+
+*/
+
+#ifdef HAVE_SYSLOG_H
+#include <syslog.h>
+#else
+#ifdef HAVE_SYS_SYSLOG_H
+#include <sys/syslog.h>
+#endif
+#endif
+
+/* For sys_adminlog(). */
+#ifndef LOG_EMERG
+#define LOG_EMERG       0       /* system is unusable */
+#endif
+
+#ifndef LOG_ALERT
+#define LOG_ALERT       1       /* action must be taken immediately */
+#endif
+
+#ifndef LOG_CRIT
+#define LOG_CRIT        2       /* critical conditions */
+#endif
+
+#ifndef LOG_ERR
+#define LOG_ERR         3       /* error conditions */
+#endif
+
+#ifndef LOG_WARNING
+#define LOG_WARNING     4       /* warning conditions */
+#endif
+
+#ifndef LOG_NOTICE
+#define LOG_NOTICE      5       /* normal but significant condition */
+#endif
+
+#ifndef LOG_INFO
+#define LOG_INFO        6       /* informational */
+#endif
+
+#ifndef LOG_DEBUG
+#define LOG_DEBUG       7       /* debug-level messages */
+#endif
+
+#endif
diff --git a/ctdb/lib/replace/system/terminal.h b/ctdb/lib/replace/system/terminal.h
new file mode 100644
index 0000000000..9ad601ace0
--- /dev/null
+++ b/ctdb/lib/replace/system/terminal.h
@@ -0,0 +1,46 @@
+#ifndef _system_terminal_h
+#define _system_terminal_h
+/* 
+   Unix SMB/CIFS implementation.
+
+   terminal system include wrappers
+
+   Copyright (C) Andrew Tridgell 2004
+   
+     ** NOTE! The following LGPL license applies to the replace
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+
+*/
+
+#ifdef SUNOS4
+/* on SUNOS4 termios.h conflicts with sys/ioctl.h */
+#undef HAVE_TERMIOS_H
+#endif
+
+
+#if defined(HAVE_TERMIOS_H)
+/* POSIX terminal handling. */
+#include <termios.h>
+#elif defined(HAVE_TERMIO_H)
+/* Older SYSV terminal handling - don't use if we can avoid it. */
+#include <termio.h>
+#elif defined(HAVE_SYS_TERMIO_H)
+/* Older SYSV terminal handling - don't use if we can avoid it. */
+#include <sys/termio.h>
+#endif
+
+#endif
diff --git a/ctdb/lib/replace/system/time.h b/ctdb/lib/replace/system/time.h
new file mode 100644
index 0000000000..b6d2609289
--- /dev/null
+++ b/ctdb/lib/replace/system/time.h
@@ -0,0 +1,91 @@
+#ifndef _system_time_h
+#define _system_time_h
+/* 
+   Unix SMB/CIFS implementation.
+
+   time system include wrappers
+
+   Copyright (C) Andrew Tridgell 2004
+
+     ** NOTE! The following LGPL license applies to the replace
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+
+*/
+
+#ifdef TIME_WITH_SYS_TIME
+#include <sys/time.h>
+#include <time.h>
+#else
+#ifdef HAVE_SYS_TIME_H
+#include <sys/time.h>
+#else
+#include <time.h>
+#endif
+#endif
+
+#ifdef HAVE_UTIME_H
+#include <utime.h>
+#else
+struct utimbuf {
+	time_t actime;       /* access time */
+	time_t modtime;      /* modification time */
+};
+#endif
+
+#ifndef HAVE_STRUCT_TIMESPEC
+struct timespec {
+	time_t tv_sec;            /* Seconds.  */
+	long tv_nsec;           /* Nanoseconds.  */
+};
+#endif
+
+#ifndef HAVE_MKTIME
+/* define is in "replace.h" */
+time_t rep_mktime(struct tm *t);
+#endif
+
+#ifndef HAVE_TIMEGM
+/* define is in "replace.h" */
+time_t rep_timegm(struct tm *tm);
+#endif
+
+#ifndef HAVE_UTIME
+/* define is in "replace.h" */
+int rep_utime(const char *filename, const struct utimbuf *buf);
+#endif
+
+#ifndef HAVE_UTIMES
+/* define is in "replace.h" */
+int rep_utimes(const char *filename, const struct timeval tv[2]);
+#endif
+
+#ifndef HAVE_CLOCK_GETTIME
+/* CLOCK_REALTIME is required by POSIX */
+#define CLOCK_REALTIME 0
+typedef int clockid_t;
+int rep_clock_gettime(clockid_t clk_id, struct timespec *tp);
+#endif
+/* make sure we have a best effort CUSTOM_CLOCK_MONOTONIC we can rely on */
+#if defined(CLOCK_MONOTONIC)
+#define CUSTOM_CLOCK_MONOTONIC CLOCK_MONOTONIC
+#elif defined(CLOCK_HIGHRES)
+#define CUSTOM_CLOCK_MONOTONIC CLOCK_HIGHRES
+#else
+#define CUSTOM_CLOCK_MONOTONIC CLOCK_REALTIME
+#endif
+
+#endif
diff --git a/ctdb/lib/replace/system/wait.h b/ctdb/lib/replace/system/wait.h
new file mode 100644
index 0000000000..146c61a094
--- /dev/null
+++ b/ctdb/lib/replace/system/wait.h
@@ -0,0 +1,55 @@
+#ifndef _system_wait_h
+#define _system_wait_h
+/* 
+   Unix SMB/CIFS implementation.
+
+   waitpid system include wrappers
+
+   Copyright (C) Andrew Tridgell 2004
+
+     ** NOTE! The following LGPL license applies to the replace
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+
+*/
+
+#ifdef HAVE_SYS_WAIT_H
+#include <sys/wait.h>
+#endif
+
+#include <signal.h>
+
+#ifndef SIGCLD
+#define SIGCLD SIGCHLD
+#endif
+
+#ifdef HAVE_SETJMP_H
+#include <setjmp.h>
+#endif
+
+#ifdef HAVE_SYS_UCONTEXT_H
+#include <sys/ucontext.h>
+#endif
+
+#if !defined(HAVE_SIG_ATOMIC_T_TYPE)
+typedef int sig_atomic_t;
+#endif
+
+#if !defined(HAVE_WAITPID) && defined(HAVE_WAIT4)
+int rep_waitpid(pid_t pid,int *status,int options)
+#endif
+
+#endif
diff --git a/ctdb/lib/replace/system/wscript_configure b/ctdb/lib/replace/system/wscript_configure
new file mode 100644
index 0000000000..2035474b32
--- /dev/null
+++ b/ctdb/lib/replace/system/wscript_configure
@@ -0,0 +1,26 @@
+#!/usr/bin/env python
+
+conf.CHECK_HEADERS('sys/capability.h')
+conf.CHECK_FUNCS('getpwnam_r getpwuid_r getpwent_r')
+
+# solaris varients of getXXent_r
+conf.CHECK_C_PROTOTYPE('getpwent_r',
+                       'struct passwd *getpwent_r(struct passwd *src, char *buf, int buflen)',
+                       define='SOLARIS_GETPWENT_R', headers='pwd.h')
+conf.CHECK_C_PROTOTYPE('getgrent_r',
+                       'struct group *getgrent_r(struct group *src, char *buf, int buflen)',
+                       define='SOLARIS_GETGRENT_R', headers='grp.h')
+
+# the irix varients
+conf.CHECK_C_PROTOTYPE('getpwent_r',
+                       'struct passwd *getpwent_r(struct passwd *src, char *buf, size_t buflen)',
+                       define='SOLARIS_GETPWENT_R', headers='pwd.h')
+conf.CHECK_C_PROTOTYPE('getgrent_r',
+                       'struct group *getgrent_r(struct group *src, char *buf, size_t buflen)',
+                       define='SOLARIS_GETGRENT_R', headers='grp.h')
+
+conf.CHECK_FUNCS('getgrouplist')
+conf.CHECK_HEADERS('ctype.h locale.h langinfo.h')
+conf.CHECK_HEADERS('fnmatch.h locale.h langinfo.h')
+conf.CHECK_HEADERS('sys/ipc.h sys/mman.h sys/shm.h')
+conf.CHECK_HEADERS('termios.h termio.h sys/termio.h')
diff --git a/ctdb/lib/replace/test/getifaddrs.c b/ctdb/lib/replace/test/getifaddrs.c
new file mode 100644
index 0000000000..d325d8bb1b
--- /dev/null
+++ b/ctdb/lib/replace/test/getifaddrs.c
@@ -0,0 +1,101 @@
+/*
+ * Unix SMB/CIFS implementation.
+ *
+ * libreplace getifaddrs test
+ *
+ * Copyright (C) Michael Adam <obnox@samba.org> 2008
+ *
+ *  This program is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; either version 3 of the License, or
+ *  (at your option) any later version.
+ *
+ *  This program is distributed in the hope that it will be useful,
+ *  but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *  GNU General Public License for more details.
+ *
+ *  You should have received a copy of the GNU General Public License
+ *  along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef AUTOCONF_TEST
+#include "replace.h"
+#include "system/network.h"
+#include "replace-test.h"
+#endif
+
+#ifdef HAVE_INET_NTOP
+#define rep_inet_ntop inet_ntop
+#endif
+
+static const char *format_sockaddr(struct sockaddr *addr,
+				   char *addrstring,
+				   socklen_t addrlen)
+{
+	const char *result = NULL;
+
+	if (addr->sa_family == AF_INET) {
+		result = rep_inet_ntop(AF_INET,
+				       &((struct sockaddr_in *)addr)->sin_addr,
+				       addrstring,
+				       addrlen);
+#ifdef HAVE_STRUCT_SOCKADDR_IN6
+	} else if (addr->sa_family == AF_INET6) {
+		result = rep_inet_ntop(AF_INET6,
+				       &((struct sockaddr_in6 *)addr)->sin6_addr,
+				       addrstring,
+				       addrlen);
+#endif
+	}
+	return result;
+}
+
+int getifaddrs_test(void)
+{
+	struct ifaddrs *ifs = NULL;
+	struct ifaddrs *ifs_head = NULL;
+	int ret;
+
+	ret = getifaddrs(&ifs);
+	ifs_head = ifs;
+	if (ret != 0) {
+		fprintf(stderr, "getifaddrs() failed: %s\n", strerror(errno));
+		return 1;
+	}
+
+	while (ifs) {
+		printf("%-10s ", ifs->ifa_name);
+		if (ifs->ifa_addr != NULL) {
+			char addrstring[INET6_ADDRSTRLEN];
+			const char *result;
+
+			result = format_sockaddr(ifs->ifa_addr,
+						 addrstring,
+						 sizeof(addrstring));
+			if (result != NULL) {
+				printf("IP=%s ", addrstring);
+			}
+
+			if (ifs->ifa_netmask != NULL) {
+				result = format_sockaddr(ifs->ifa_netmask,
+							 addrstring,
+							 sizeof(addrstring));
+				if (result != NULL) {
+					printf("NETMASK=%s", addrstring);
+				}
+			} else {
+				printf("AF=%d ", ifs->ifa_addr->sa_family);
+			}
+		} else {
+			printf("<no address>");
+		}
+
+		printf("\n");
+		ifs = ifs->ifa_next;
+	}
+
+	freeifaddrs(ifs_head);
+
+	return 0;
+}
diff --git a/ctdb/lib/replace/test/incoherent_mmap.c b/ctdb/lib/replace/test/incoherent_mmap.c
new file mode 100644
index 0000000000..ee288fd806
--- /dev/null
+++ b/ctdb/lib/replace/test/incoherent_mmap.c
@@ -0,0 +1,83 @@
+/* In OpenBSD, if you write to a file, another process doesn't see it
+ * in its mmap.  Returns with exit status 0 if that is the case, 1 if
+ * it's coherent, and other if there's a problem. */
+#include <err.h>
+#include <sys/mman.h>
+#include <unistd.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include <stdlib.h>
+#include <string.h>
+#include <errno.h>
+
+#define DATA "coherent.mmap"
+
+int main(int argc, char *argv[])
+{
+	int tochild[2], toparent[2];
+	int fd;
+	volatile unsigned char *map;
+	unsigned char *page;
+        const char *fname = argv[1];
+	char c = 0;
+
+	if (pipe(tochild) != 0 || pipe(toparent) != 0)
+		err(2, "Creating pipe");
+
+	if (!fname)
+		fname = DATA;
+
+	fd = open(fname, O_RDWR|O_CREAT|O_TRUNC, 0600);
+	if (fd < 0)
+		err(2, "opening %s", fname);
+	unlink(fname);
+
+	switch (fork()) {
+	case -1:
+		err(2, "Fork");
+	case 0:
+		close(tochild[1]);
+		close(toparent[0]);
+
+		/* Wait for parent to create file. */
+		if (read(tochild[0], &c, 1) != 1)
+			err(2, "reading from parent");
+
+		/* Alter first byte. */
+		pwrite(fd, &c, 1, 0);
+
+		if (write(toparent[1], &c, 1) != 1)
+			err(2, "writing to parent");
+		exit(0);
+
+	default:
+		close(tochild[0]);
+		close(toparent[1]);
+
+		/* Create a file and mmap it. */
+		page = malloc(getpagesize());
+		memset(page, 0x42, getpagesize());
+		if (write(fd, page, getpagesize()) != getpagesize())
+			err(2, "writing first page");
+		map = mmap(NULL, getpagesize(), PROT_READ|PROT_WRITE,
+			   MAP_SHARED, fd, 0);
+		if (map == MAP_FAILED)
+			err(2, "mapping file");
+
+		if (*map != 0x42)
+			errx(2, "first byte isn't 0x42!");
+
+		/* Tell child to alter file. */
+		if (write(tochild[1], &c, 1) != 1)
+			err(2, "writing to child");
+
+		if (read(toparent[0], &c, 1) != 1)
+			err(2, "reading from child");
+
+		if (*map)
+			errx(0, "mmap incoherent: first byte isn't 0.");
+
+		exit(1);
+	}
+}
diff --git a/ctdb/lib/replace/test/main.c b/ctdb/lib/replace/test/main.c
new file mode 100644
index 0000000000..94264d7367
--- /dev/null
+++ b/ctdb/lib/replace/test/main.c
@@ -0,0 +1,35 @@
+/*
+   Unix SMB/CIFS implementation.
+
+   libreplace tests
+
+   Copyright (C) Jelmer Vernooij 2006
+
+     ** NOTE! The following LGPL license applies to the talloc
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "replace.h"
+#include "replace-testsuite.h"
+
+int main(void)
+{
+	bool ret = torture_local_replace(NULL);
+	if (ret)
+		return 0;
+	return -1;
+}
diff --git a/ctdb/lib/replace/test/os2_delete.c b/ctdb/lib/replace/test/os2_delete.c
new file mode 100644
index 0000000000..a11ed3b370
--- /dev/null
+++ b/ctdb/lib/replace/test/os2_delete.c
@@ -0,0 +1,133 @@
+/*
+  test readdir/unlink pattern that OS/2 uses
+  tridge@samba.org July 2005
+*/
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <sys/stat.h>
+#include <unistd.h>
+#include <sys/types.h>
+#include <dirent.h>
+#include <errno.h>
+#include <string.h>
+#include <fcntl.h>
+#include "replace-test.h"
+
+#define NUM_FILES 700
+#define READDIR_SIZE 100
+#define DELETE_SIZE 4
+
+#define TESTDIR "test.dir"
+
+static int test_readdir_os2_delete_ret;
+
+#define FAILED(d) (printf("failure: readdir [\nFailed for %s - %d = %s\n]\n", d, errno, strerror(errno)), test_readdir_os2_delete_ret = 1, 1)
+
+#ifndef MIN
+#define MIN(a,b) ((a)<(b)?(a):(b))
+#endif
+
+#ifdef _WIN32
+#define mkdir(d,m) _mkdir(d)
+#endif
+
+static void cleanup(void)
+{
+	/* I'm a lazy bastard */
+	if (system("rm -rf " TESTDIR)) {
+		FAILED("system");
+	}
+	mkdir(TESTDIR, 0700) == 0 || FAILED("mkdir");
+}
+
+static void create_files(void)
+{
+	int i;
+	for (i=0;i<NUM_FILES;i++) {
+		char fname[40];
+		int fd;
+		snprintf(fname, sizeof(fname), TESTDIR "/test%u.txt", i);
+		fd = open(fname, O_CREAT|O_RDWR, 0600);
+		if (fd < 0) {
+			FAILED("open");
+		}
+		if (close(fd) != 0) {
+			FAILED("close");
+		}
+	}
+}
+
+static int os2_delete(DIR *d)
+{
+	off_t offsets[READDIR_SIZE];
+	int i, j;
+	struct dirent *de;
+	char names[READDIR_SIZE][30];
+
+	/* scan, remembering offsets */
+	for (i=0, de=readdir(d); 
+	     de && i < READDIR_SIZE; 
+	     de=readdir(d), i++) {
+		offsets[i] = telldir(d);
+		strcpy(names[i], de->d_name);
+	}
+
+	if (i == 0) {
+		return 0;
+	}
+
+	/* delete the first few */
+	for (j=0; j<MIN(i, DELETE_SIZE); j++) {
+		char fname[40];
+		snprintf(fname, sizeof(fname), TESTDIR "/%s", names[j]);
+		unlink(fname) == 0 || FAILED("unlink");
+	}
+
+	/* seek to just after the deletion */
+	seekdir(d, offsets[j-1]);
+
+	/* return number deleted */
+	return j;
+}
+
+int test_readdir_os2_delete(void)
+{
+	int total_deleted = 0;
+	DIR *d;
+	struct dirent *de;
+
+	test_readdir_os2_delete_ret = 0;
+
+	cleanup();
+	create_files();
+
+	d = opendir(TESTDIR "/test0.txt");
+	if (d != NULL) FAILED("opendir() on file succeed");
+	if (errno != ENOTDIR) FAILED("opendir() on file didn't give ENOTDIR");
+
+	d = opendir(TESTDIR);
+
+	/* skip past . and .. */
+	de = readdir(d);
+	strcmp(de->d_name, ".") == 0 || FAILED("match .");
+	de = readdir(d);
+	strcmp(de->d_name, "..") == 0 || FAILED("match ..");
+
+	while (1) {
+		int n = os2_delete(d);
+		if (n == 0) break;
+		total_deleted += n;
+	}
+	closedir(d);
+
+	fprintf(stderr, "Deleted %d files of %d\n", total_deleted, NUM_FILES);
+
+	rmdir(TESTDIR) == 0 || FAILED("rmdir");
+
+	if (system("rm -rf " TESTDIR) == -1) {
+		FAILED("system");
+	}
+
+	return test_readdir_os2_delete_ret;
+}
diff --git a/ctdb/lib/replace/test/shared_mmap.c b/ctdb/lib/replace/test/shared_mmap.c
new file mode 100644
index 0000000000..50dad8d696
--- /dev/null
+++ b/ctdb/lib/replace/test/shared_mmap.c
@@ -0,0 +1,68 @@
+/* this tests whether we can use a shared writeable mmap on a file -
+   as needed for the mmap variant of FAST_SHARE_MODES */
+
+#if defined(HAVE_UNISTD_H)
+#include <unistd.h>
+#endif
+#include <sys/mman.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+
+#define DATA "conftest.mmap"
+
+#ifndef MAP_FILE
+#define MAP_FILE 0
+#endif
+
+main()
+{
+	int *buf;
+	int i; 
+	int fd = open(DATA,O_RDWR|O_CREAT|O_TRUNC,0666);
+	int count=7;
+
+	if (fd == -1) exit(1);
+
+	for (i=0;i<10000;i++) {
+		write(fd,&i,sizeof(i));
+	}
+
+	close(fd);
+
+	if (fork() == 0) {
+		fd = open(DATA,O_RDWR);
+		if (fd == -1) exit(1);
+
+		buf = (int *)mmap(NULL, 10000*sizeof(int), 
+				   (PROT_READ | PROT_WRITE), 
+				   MAP_FILE | MAP_SHARED, 
+				   fd, 0);
+
+		while (count-- && buf[9124] != 55732) sleep(1);
+
+		if (count <= 0) exit(1);
+
+		buf[1763] = 7268;
+		exit(0);
+	}
+
+	fd = open(DATA,O_RDWR);
+	if (fd == -1) exit(1);
+
+	buf = (int *)mmap(NULL, 10000*sizeof(int), 
+			   (PROT_READ | PROT_WRITE), 
+			   MAP_FILE | MAP_SHARED, 
+			   fd, 0);
+
+	if (buf == (int *)-1) exit(1);
+
+	buf[9124] = 55732;
+
+	while (count-- && buf[1763] != 7268) sleep(1);
+
+	unlink(DATA);
+		
+	if (count > 0) exit(0);
+	exit(1);
+}
diff --git a/ctdb/lib/replace/test/shared_mremap.c b/ctdb/lib/replace/test/shared_mremap.c
new file mode 100644
index 0000000000..05032ad12e
--- /dev/null
+++ b/ctdb/lib/replace/test/shared_mremap.c
@@ -0,0 +1,48 @@
+/* this tests whether we can use mremap */
+
+#if defined(HAVE_UNISTD_H)
+#include <unistd.h>
+#endif
+#include <sys/mman.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+
+#define DATA "conftest.mmap"
+
+#ifndef MAP_FILE
+#define MAP_FILE 0
+#endif
+
+#ifndef MAP_FAILED
+#define MAP_FAILED (int *)-1
+#endif
+
+main()
+{
+	int *buf;
+	int fd;
+	int err = 1;
+
+	fd = open(DATA, O_RDWR|O_CREAT|O_TRUNC, 0666);
+	if (fd == -1) {
+		exit(1);
+	}
+
+	buf = (int *)mmap(NULL, 0x1000, PROT_READ | PROT_WRITE,
+			  MAP_FILE | MAP_SHARED, fd, 0);
+	if (buf == MAP_FAILED) {
+		goto done;
+	}
+
+	buf = mremap(buf, 0x1000, 0x2000, MREMAP_MAYMOVE);
+	if (buf == MAP_FAILED) {
+		goto done;
+	}
+
+	err = 0;
+done:
+	close(fd);
+	unlink(DATA);
+	exit(err);
+}
diff --git a/ctdb/lib/replace/test/snprintf.c b/ctdb/lib/replace/test/snprintf.c
new file mode 100644
index 0000000000..d06630bcc9
--- /dev/null
+++ b/ctdb/lib/replace/test/snprintf.c
@@ -0,0 +1,29 @@
+void foo(const char *format, ...)
+{
+	va_list ap;
+	int len;
+	char buf[20];
+	long long l = 1234567890;
+	l *= 100;
+
+	va_start(ap, format);
+	len = vsnprintf(buf, 0, format, ap);
+	va_end(ap);
+	if (len != 5) exit(1);
+
+	va_start(ap, format);
+	len = vsnprintf(0, 0, format, ap);
+	va_end(ap);
+	if (len != 5) exit(2);
+
+	if (snprintf(buf, 3, "hello") != 5 || strcmp(buf, "he") != 0) exit(3);
+
+	if (snprintf(buf, 20, "%lld", l) != 12 || strcmp(buf, "123456789000") != 0) exit(4);
+	if (snprintf(buf, 20, "%zu", 123456789) != 9 || strcmp(buf, "123456789") != 0) exit(5);
+	if (snprintf(buf, 20, "%2\$d %1\$d", 3, 4) != 3 || strcmp(buf, "4 3") != 0) exit(6);
+	if (snprintf(buf, 20, "%s", 0) < 3) exit(7);
+
+	printf("1");
+	exit(0);
+}
+main() { foo("hello"); }
diff --git a/ctdb/lib/replace/test/strptime.c b/ctdb/lib/replace/test/strptime.c
new file mode 100644
index 0000000000..5bf03f5b35
--- /dev/null
+++ b/ctdb/lib/replace/test/strptime.c
@@ -0,0 +1,173 @@
+
+#ifdef LIBREPLACE_CONFIGURE_TEST_STRPTIME
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <time.h>
+
+#define true 1
+#define false 0
+
+#ifndef __STRING
+#define __STRING(x)    #x
+#endif
+
+/* make printf a no-op */
+#define printf if(0) printf
+
+#else /* LIBREPLACE_CONFIGURE_TEST_STRPTIME */
+
+#include "replace.h"
+#include "system/time.h"
+#include "replace-test.h"
+
+#endif /* LIBREPLACE_CONFIGURE_TEST_STRPTIME */
+
+int libreplace_test_strptime(void)
+{
+	const char *s = "20070414101546Z";
+	char *ret;
+	struct tm t, t2;
+
+	memset(&t, 0, sizeof(t));
+	memset(&t2, 0, sizeof(t2));
+
+	printf("test: strptime\n");
+
+	ret = strptime(s, "%Y%m%d%H%M%S", &t);
+	if ( ret == NULL ) {
+		printf("failure: strptime [\n"
+		       "returned NULL\n"
+		       "]\n");
+		return false;
+	}
+
+	if ( *ret != 'Z' ) {
+		printf("failure: strptime [\n"
+		       "ret doesn't point to 'Z'\n"
+		       "]\n");
+		return false;
+	}
+
+	ret = strptime(s, "%Y%m%d%H%M%SZ", &t2);
+	if ( ret == NULL ) {
+		printf("failure: strptime [\n"
+		       "returned NULL with Z\n"
+		       "]\n");
+		return false;
+	}
+
+	if ( *ret != '\0' ) {
+		printf("failure: strptime [\n"
+		       "ret doesn't point to '\\0'\n"
+		       "]\n");
+		return false;
+	}
+
+#define CMP_TM_ELEMENT(t1,t2,elem) \
+	if (t1.elem != t2.elem) { \
+		printf("failure: strptime [\n" \
+		       "result differs if the format string has a 'Z' at the end\n" \
+		       "element: %s %d != %d\n" \
+		       "]\n", \
+		       __STRING(elen), t1.elem, t2.elem); \
+		return false; \
+	}
+
+	CMP_TM_ELEMENT(t,t2,tm_sec);
+	CMP_TM_ELEMENT(t,t2,tm_min);
+	CMP_TM_ELEMENT(t,t2,tm_hour);
+	CMP_TM_ELEMENT(t,t2,tm_mday);
+	CMP_TM_ELEMENT(t,t2,tm_mon);
+	CMP_TM_ELEMENT(t,t2,tm_year);
+	CMP_TM_ELEMENT(t,t2,tm_wday);
+	CMP_TM_ELEMENT(t,t2,tm_yday);
+	CMP_TM_ELEMENT(t,t2,tm_isdst);
+
+	if (t.tm_sec != 46) {
+		printf("failure: strptime [\n"
+		       "tm_sec: expected: 46, got: %d\n"
+		       "]\n",
+		       t.tm_sec);
+		return false;
+	}
+
+	if (t.tm_min != 15) {
+		printf("failure: strptime [\n"
+		       "tm_min: expected: 15, got: %d\n"
+		       "]\n",
+		       t.tm_min);
+		return false;
+	}
+
+	if (t.tm_hour != 10) {
+		printf("failure: strptime [\n"
+		       "tm_hour: expected: 10, got: %d\n"
+		       "]\n",
+		       t.tm_hour);
+		return false;
+	}
+
+	if (t.tm_mday != 14) {
+		printf("failure: strptime [\n"
+		       "tm_mday: expected: 14, got: %d\n"
+		       "]\n",
+		       t.tm_mday);
+		return false;
+	}
+
+	if (t.tm_mon != 3) {
+		printf("failure: strptime [\n"
+		       "tm_mon: expected: 3, got: %d\n"
+		       "]\n",
+		       t.tm_mon);
+		return false;
+	}
+
+	if (t.tm_year != 107) {
+		printf("failure: strptime [\n"
+		       "tm_year: expected: 107, got: %d\n"
+		       "]\n",
+		       t.tm_year);
+		return false;
+	}
+
+	if (t.tm_wday != 6) { /* saturday */
+		printf("failure: strptime [\n"
+		       "tm_wday: expected: 6, got: %d\n"
+		       "]\n",
+		       t.tm_wday);
+		return false;
+	}
+
+	if (t.tm_yday != 103) {
+		printf("failure: strptime [\n"
+		       "tm_yday: expected: 103, got: %d\n"
+		       "]\n",
+		       t.tm_yday);
+		return false;
+	}
+
+	/* we don't test this as it depends on the host configuration
+	if (t.tm_isdst != 0) {
+		printf("failure: strptime [\n"
+		       "tm_isdst: expected: 0, got: %d\n"
+		       "]\n",
+		       t.tm_isdst);
+		return false;
+	}*/
+
+	printf("success: strptime\n");
+
+	return true;
+}
+
+#ifdef LIBREPLACE_CONFIGURE_TEST_STRPTIME
+int main (void)
+{
+	int ret;
+	ret = libreplace_test_strptime();
+	if (ret == false) return 1;
+	return 0;
+}
+#endif
diff --git a/ctdb/lib/replace/test/testsuite.c b/ctdb/lib/replace/test/testsuite.c
new file mode 100644
index 0000000000..52629ec82e
--- /dev/null
+++ b/ctdb/lib/replace/test/testsuite.c
@@ -0,0 +1,1101 @@
+/* 
+   Unix SMB/CIFS implementation.
+
+   libreplace tests
+
+   Copyright (C) Jelmer Vernooij 2006
+
+     ** NOTE! The following LGPL license applies to the talloc
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "replace.h"
+#include "replace-test.h"
+#include "replace-testsuite.h"
+
+/*
+  we include all the system/ include files here so that libreplace tests
+  them in the build farm
+*/
+#include "system/capability.h"
+#include "system/dir.h"
+#include "system/filesys.h"
+#include "system/glob.h"
+#include "system/iconv.h"
+#include "system/locale.h"
+#include "system/network.h"
+#include "system/passwd.h"
+#include "system/readline.h"
+#include "system/select.h"
+#include "system/shmem.h"
+#include "system/syslog.h"
+#include "system/terminal.h"
+#include "system/time.h"
+#include "system/wait.h"
+#include "system/aio.h"
+
+#define TESTFILE "testfile.dat"
+
+
+/*
+  test ftruncate() function
+ */
+static int test_ftruncate(void)
+{
+	struct stat st;
+	int fd;
+	const int size = 1234;
+	printf("test: ftruncate\n");
+	unlink(TESTFILE);
+	fd = open(TESTFILE, O_RDWR|O_CREAT, 0600);
+	if (fd == -1) {
+		printf("failure: ftruncate [\n"
+			   "creating '%s' failed - %s\n]\n", TESTFILE, strerror(errno));
+		return false;
+	}
+	if (ftruncate(fd, size) != 0) {
+		printf("failure: ftruncate [\n%s\n]\n", strerror(errno));
+		return false;
+	}
+	if (fstat(fd, &st) != 0) {
+		printf("failure: ftruncate [\nfstat failed - %s\n]\n", strerror(errno));
+		return false;
+	}
+	if (st.st_size != size) {
+		printf("failure: ftruncate [\ngave wrong size %d - expected %d\n]\n",
+		       (int)st.st_size, size);
+		return false;
+	}
+	unlink(TESTFILE);
+	printf("success: ftruncate\n");
+	return true;
+}
+
+/*
+  test strlcpy() function.
+  see http://www.gratisoft.us/todd/papers/strlcpy.html
+ */
+static int test_strlcpy(void)
+{
+	char buf[4];
+	const struct {
+		const char *src;
+		size_t result;
+	} tests[] = {
+		{ "abc", 3 },
+		{ "abcdef", 6 },
+		{ "abcd", 4 },
+		{ "", 0 },
+		{ NULL, 0 }
+	};
+	int i;
+	printf("test: strlcpy\n");
+	for (i=0;tests[i].src;i++) {
+		if (strlcpy(buf, tests[i].src, sizeof(buf)) != tests[i].result) {
+			printf("failure: strlcpy [\ntest %d failed\n]\n", i);
+			return false;
+		}
+	}
+	printf("success: strlcpy\n");
+	return true;
+}
+
+static int test_strlcat(void)
+{
+	char tmp[10];
+	printf("test: strlcat\n");
+	strlcpy(tmp, "", sizeof(tmp));
+	if (strlcat(tmp, "bla", 3) != 3) {
+		printf("failure: strlcat [\ninvalid return code\n]\n");
+		return false;
+	}
+	if (strcmp(tmp, "bl") != 0) {
+		printf("failure: strlcat [\nexpected \"bl\", got \"%s\"\n]\n", 
+			   tmp);
+		return false;
+	}
+
+	strlcpy(tmp, "da", sizeof(tmp));
+	if (strlcat(tmp, "me", 4) != 4) {
+		printf("failure: strlcat [\nexpected \"dam\", got \"%s\"\n]\n",
+			   tmp);
+		return false;
+	}
+
+	printf("success: strlcat\n");
+	return true;
+}
+
+static int test_mktime(void)
+{
+	/* FIXME */
+	return true;
+}
+
+static int test_initgroups(void)
+{
+	/* FIXME */
+	return true;
+}
+
+static int test_memmove(void)
+{
+	/* FIXME */
+	return true;
+}
+
+static int test_strdup(void)
+{
+	char *x;
+	printf("test: strdup\n");
+	x = strdup("bla");
+	if (strcmp("bla", x) != 0) {
+		printf("failure: strdup [\nfailed: expected \"bla\", got \"%s\"\n]\n",
+			   x);
+		return false;
+	}
+	free(x);
+	printf("success: strdup\n");
+	return true;
+}	
+
+static int test_setlinebuf(void)
+{
+	printf("test: setlinebuf\n");
+	setlinebuf(stdout);
+	printf("success: setlinebuf\n");
+	return true;
+}
+
+static int test_vsyslog(void)
+{
+	/* FIXME */
+	return true;
+}
+
+static int test_timegm(void)
+{
+	/* FIXME */
+	return true;
+}
+
+static int test_setenv(void)
+{
+#define TEST_SETENV(key, value, overwrite, result) do { \
+	int _ret; \
+	char *_v; \
+	_ret = setenv(key, value, overwrite); \
+	if (_ret != 0) { \
+		printf("failure: setenv [\n" \
+			"setenv(%s, %s, %d) failed\n" \
+			"]\n", \
+			key, value, overwrite); \
+		return false; \
+	} \
+	_v=getenv(key); \
+	if (!_v) { \
+		printf("failure: setenv [\n" \
+			"getenv(%s) returned NULL\n" \
+			"]\n", \
+			key); \
+		return false; \
+	} \
+	if (strcmp(result, _v) != 0) { \
+		printf("failure: setenv [\n" \
+			"getenv(%s): '%s' != '%s'\n" \
+			"]\n", \
+			key, result, _v); \
+		return false; \
+	} \
+} while(0)
+
+#define TEST_UNSETENV(key) do { \
+	char *_v; \
+	unsetenv(key); \
+	_v=getenv(key); \
+	if (_v) { \
+		printf("failure: setenv [\n" \
+			"getenv(%s): NULL != '%s'\n" \
+			"]\n", \
+			SETENVTEST_KEY, _v); \
+		return false; \
+	} \
+} while (0)
+
+#define SETENVTEST_KEY "SETENVTESTKEY"
+#define SETENVTEST_VAL "SETENVTESTVAL"
+
+	printf("test: setenv\n");
+	TEST_SETENV(SETENVTEST_KEY, SETENVTEST_VAL"1", 0, SETENVTEST_VAL"1");
+	TEST_SETENV(SETENVTEST_KEY, SETENVTEST_VAL"2", 0, SETENVTEST_VAL"1");
+	TEST_SETENV(SETENVTEST_KEY, SETENVTEST_VAL"3", 1, SETENVTEST_VAL"3");
+	TEST_SETENV(SETENVTEST_KEY, SETENVTEST_VAL"4", 1, SETENVTEST_VAL"4");
+	TEST_UNSETENV(SETENVTEST_KEY);
+	TEST_UNSETENV(SETENVTEST_KEY);
+	TEST_SETENV(SETENVTEST_KEY, SETENVTEST_VAL"5", 0, SETENVTEST_VAL"5");
+	TEST_UNSETENV(SETENVTEST_KEY);
+	TEST_UNSETENV(SETENVTEST_KEY);
+	printf("success: setenv\n");
+	return true;
+}
+
+static int test_strndup(void)
+{
+	char *x;
+	printf("test: strndup\n");
+	x = strndup("bla", 0);
+	if (strcmp(x, "") != 0) {
+		printf("failure: strndup [\ninvalid\n]\n");
+		return false;
+	}
+	free(x);
+	x = strndup("bla", 2);
+	if (strcmp(x, "bl") != 0) {
+		printf("failure: strndup [\ninvalid\n]\n");
+		return false;
+	}
+	free(x);
+	x = strndup("bla", 10);
+	if (strcmp(x, "bla") != 0) {
+		printf("failure: strndup [\ninvalid\n]\n");
+		return false;
+	}
+	free(x);
+	printf("success: strndup\n");
+	return true;
+}
+
+static int test_strnlen(void)
+{
+	printf("test: strnlen\n");
+	if (strnlen("bla", 2) != 2) {
+		printf("failure: strnlen [\nunexpected length\n]\n");
+		return false;
+	}
+
+	if (strnlen("some text\n", 0) != 0) {
+		printf("failure: strnlen [\nunexpected length\n]\n");
+		return false;
+	}
+
+	if (strnlen("some text", 20) != 9) {
+		printf("failure: strnlen [\nunexpected length\n]\n");
+		return false;
+	}
+
+	printf("success: strnlen\n");
+	return true;
+}
+
+static int test_waitpid(void)
+{
+	/* FIXME */
+	return true;
+}
+
+static int test_seteuid(void)
+{
+	/* FIXME */
+	return true;
+}
+
+static int test_setegid(void)
+{
+	/* FIXME */
+	return true;
+}
+
+static int test_asprintf(void)
+{
+	char *x;
+	printf("test: asprintf\n");
+	if (asprintf(&x, "%d", 9) != 1) {
+		printf("failure: asprintf [\ngenerate asprintf\n]\n");
+		return false;
+	}
+	if (strcmp(x, "9") != 0) {
+		printf("failure: asprintf [\ngenerate asprintf\n]\n");
+		return false;
+	}
+	if (asprintf(&x, "dat%s", "a") != 4) {
+		printf("failure: asprintf [\ngenerate asprintf\n]\n");
+		return false;
+	}
+	if (strcmp(x, "data") != 0) {
+		printf("failure: asprintf [\ngenerate asprintf\n]\n");
+		return false;
+	}
+	printf("success: asprintf\n");
+	return true;
+}
+
+static int test_snprintf(void)
+{
+	char tmp[10];
+	printf("test: snprintf\n");
+	if (snprintf(tmp, 3, "foo%d", 9) != 4) {
+		printf("failure: snprintf [\nsnprintf return code failed\n]\n");
+		return false;
+	}
+
+	if (strcmp(tmp, "fo") != 0) {
+		printf("failure: snprintf [\nsnprintf failed\n]\n");
+		return false;
+	}
+
+	printf("success: snprintf\n");
+	return true;
+}
+
+static int test_vasprintf(void)
+{
+	/* FIXME */
+	return true;
+}
+
+static int test_vsnprintf(void)
+{
+	/* FIXME */
+	return true;
+}
+
+static int test_opendir(void)
+{
+	/* FIXME */
+	return true;
+}
+
+static int test_readdir(void)
+{
+	printf("test: readdir\n");
+	if (test_readdir_os2_delete() != 0) {
+		return false;
+	}
+	printf("success: readdir\n");
+	return true;
+}
+
+static int test_telldir(void)
+{
+	/* FIXME */
+	return true;
+}
+
+static int test_seekdir(void)
+{
+	/* FIXME */
+	return true;
+}
+
+static int test_dlopen(void)
+{
+	/* FIXME: test dlopen, dlsym, dlclose, dlerror */
+	return true;
+}
+
+
+static int test_chroot(void)
+{
+	/* FIXME: chroot() */
+	return true;
+}
+
+static int test_bzero(void)
+{
+	/* FIXME: bzero */
+	return true;
+}
+
+static int test_strerror(void)
+{
+	/* FIXME */
+	return true;
+}
+
+static int test_errno(void)
+{
+	printf("test: errno\n");
+	errno = 3;
+	if (errno != 3) {
+		printf("failure: errno [\nerrno failed\n]\n");
+		return false;
+	}
+
+	printf("success: errno\n");
+	return true;
+}
+
+static int test_mkdtemp(void)
+{
+	/* FIXME */
+	return true;
+}
+
+static int test_mkstemp(void)
+{
+	/* FIXME */
+	return true;
+}
+
+static int test_pread(void)
+{
+	/* FIXME */
+	return true;
+}
+
+static int test_pwrite(void)
+{
+	/* FIXME */
+	return true;
+}
+
+static int test_inet_ntoa(void)
+{
+	/* FIXME */
+	return true;
+}
+
+#define TEST_STRTO_X(type,fmt,func,str,base,res,diff,rrnoo) do {\
+	type _v; \
+	char _s[64]; \
+	char *_p = NULL;\
+	char *_ep = NULL; \
+	strlcpy(_s, str, sizeof(_s));\
+	if (diff >= 0) { \
+		_ep = &_s[diff]; \
+	} \
+	errno = 0; \
+	_v = func(_s, &_p, base); \
+	if (errno != rrnoo) { \
+		printf("failure: %s [\n" \
+		       "\t%s\n" \
+		       "\t%s(\"%s\",%d,%d): " fmt " (=/!)= " fmt "\n" \
+		       "\terrno: %d != %d\n" \
+		       "]\n", \
+		        __STRING(func), __location__, __STRING(func), \
+		       str, diff, base, res, _v, rrnoo, errno); \
+		return false; \
+	} else if (_v != res) { \
+		printf("failure: %s [\n" \
+		       "\t%s\n" \
+		       "\t%s(\"%s\",%d,%d): " fmt " != " fmt "\n" \
+		       "]\n", \
+		       __STRING(func), __location__, __STRING(func), \
+		       str, diff, base, res, _v); \
+		return false; \
+	} else if (_p != _ep) { \
+		printf("failure: %s [\n" \
+		       "\t%s\n" \
+		       "\t%s(\"%s\",%d,%d): " fmt " (=/!)= " fmt "\n" \
+		       "\tptr: %p - %p = %d != %d\n" \
+		       "]\n", \
+		       __STRING(func), __location__, __STRING(func), \
+		       str, diff, base, res, _v, _ep, _p, (int)(diff - (_ep - _p)), diff); \
+		return false; \
+	} \
+} while (0)
+
+static int test_strtoll(void)
+{
+	printf("test: strtoll\n");
+
+#define TEST_STRTOLL(str,base,res,diff,errnoo) TEST_STRTO_X(long long int, "%lld", strtoll,str,base,res,diff,errnoo)
+
+	TEST_STRTOLL("15",	10,	15LL,	2, 0);
+	TEST_STRTOLL("  15",	10,	15LL,	4, 0);
+	TEST_STRTOLL("15",	0,	15LL,	2, 0);
+	TEST_STRTOLL(" 15 ",	0,	15LL,	3, 0);
+	TEST_STRTOLL("+15",	10,	15LL,	3, 0);
+	TEST_STRTOLL("  +15",	10,	15LL,	5, 0);
+	TEST_STRTOLL("+15",	0,	15LL,	3, 0);
+	TEST_STRTOLL(" +15 ",	0,	15LL,	4, 0);
+	TEST_STRTOLL("-15",	10,	-15LL,	3, 0);
+	TEST_STRTOLL("  -15",	10,	-15LL,	5, 0);
+	TEST_STRTOLL("-15",	0,	-15LL,	3, 0);
+	TEST_STRTOLL(" -15 ",	0,	-15LL,	4, 0);
+	TEST_STRTOLL("015",	10,	15LL,	3, 0);
+	TEST_STRTOLL("  015",	10,	15LL,	5, 0);
+	TEST_STRTOLL("015",	0,	13LL,	3, 0);
+	TEST_STRTOLL("  015",	0,	13LL,	5, 0);
+	TEST_STRTOLL("0x15",	10,	0LL,	1, 0);
+	TEST_STRTOLL("  0x15",	10,	0LL,	3, 0);
+	TEST_STRTOLL("0x15",	0,	21LL,	4, 0);
+	TEST_STRTOLL("  0x15",	0,	21LL,	6, 0);
+
+	TEST_STRTOLL("10",	16,	16LL,	2, 0);
+	TEST_STRTOLL("  10 ",	16,	16LL,	4, 0);
+	TEST_STRTOLL("0x10",	16,	16LL,	4, 0);
+	TEST_STRTOLL("0x10",	0,	16LL,	4, 0);
+	TEST_STRTOLL(" 0x10 ",	0,	16LL,	5, 0);
+	TEST_STRTOLL("+10",	16,	16LL,	3, 0);
+	TEST_STRTOLL("  +10 ",	16,	16LL,	5, 0);
+	TEST_STRTOLL("+0x10",	16,	16LL,	5, 0);
+	TEST_STRTOLL("+0x10",	0,	16LL,	5, 0);
+	TEST_STRTOLL(" +0x10 ",	0,	16LL,	6, 0);
+	TEST_STRTOLL("-10",	16,	-16LL,	3, 0);
+	TEST_STRTOLL("  -10 ",	16,	-16LL,	5, 0);
+	TEST_STRTOLL("-0x10",	16,	-16LL,	5, 0);
+	TEST_STRTOLL("-0x10",	0,	-16LL,	5, 0);
+	TEST_STRTOLL(" -0x10 ",	0,	-16LL,	6, 0);
+	TEST_STRTOLL("010",	16,	16LL,	3, 0);
+	TEST_STRTOLL("  010 ",	16,	16LL,	5, 0);
+	TEST_STRTOLL("-010",	16,	-16LL,	4, 0);
+
+	TEST_STRTOLL("11",	8,	9LL,	2, 0);
+	TEST_STRTOLL("011",	8,	9LL,	3, 0);
+	TEST_STRTOLL("011",	0,	9LL,	3, 0);
+	TEST_STRTOLL("-11",	8,	-9LL,	3, 0);
+	TEST_STRTOLL("-011",	8,	-9LL,	4, 0);
+	TEST_STRTOLL("-011",	0,	-9LL,	4, 0);
+
+	TEST_STRTOLL("011",	8,	9LL,	3, 0);
+	TEST_STRTOLL("011",	0,	9LL,	3, 0);
+	TEST_STRTOLL("-11",	8,	-9LL,	3, 0);
+	TEST_STRTOLL("-011",	8,	-9LL,	4, 0);
+	TEST_STRTOLL("-011",	0,	-9LL,	4, 0);
+
+	TEST_STRTOLL("Text",	0,	0LL,	0, 0);
+
+	TEST_STRTOLL("9223372036854775807",	10,	9223372036854775807LL,	19, 0);
+	TEST_STRTOLL("9223372036854775807",	0,	9223372036854775807LL,	19, 0);
+	TEST_STRTOLL("9223372036854775808",	0,	9223372036854775807LL,	19, ERANGE);
+	TEST_STRTOLL("9223372036854775808",	10,	9223372036854775807LL,	19, ERANGE);
+	TEST_STRTOLL("0x7FFFFFFFFFFFFFFF",	0,	9223372036854775807LL,	18, 0);
+	TEST_STRTOLL("0x7FFFFFFFFFFFFFFF",	16,	9223372036854775807LL,	18, 0);
+	TEST_STRTOLL("7FFFFFFFFFFFFFFF",	16,	9223372036854775807LL,	16, 0);
+	TEST_STRTOLL("0x8000000000000000",	0,	9223372036854775807LL,	18, ERANGE);
+	TEST_STRTOLL("0x8000000000000000",	16,	9223372036854775807LL,	18, ERANGE);
+	TEST_STRTOLL("80000000000000000",	16,	9223372036854775807LL,	17, ERANGE);
+	TEST_STRTOLL("0777777777777777777777",	0,	9223372036854775807LL,	22, 0);
+	TEST_STRTOLL("0777777777777777777777",	8,	9223372036854775807LL,	22, 0);
+	TEST_STRTOLL("777777777777777777777",	8,	9223372036854775807LL,	21, 0);
+	TEST_STRTOLL("01000000000000000000000",	0,	9223372036854775807LL,	23, ERANGE);
+	TEST_STRTOLL("01000000000000000000000",	8,	9223372036854775807LL,	23, ERANGE);
+	TEST_STRTOLL("1000000000000000000000",	8,	9223372036854775807LL,	22, ERANGE);
+
+	TEST_STRTOLL("-9223372036854775808",	10,	-9223372036854775807LL -1,	20, 0);
+	TEST_STRTOLL("-9223372036854775808",	0,	-9223372036854775807LL -1,	20, 0);
+	TEST_STRTOLL("-9223372036854775809",	0,	-9223372036854775807LL -1,	20, ERANGE);
+	TEST_STRTOLL("-9223372036854775809",	10,	-9223372036854775807LL -1,	20, ERANGE);
+	TEST_STRTOLL("-0x8000000000000000",	0,	-9223372036854775807LL -1,	19, 0);
+	TEST_STRTOLL("-0x8000000000000000",	16,	-9223372036854775807LL -1,	19, 0);
+	TEST_STRTOLL("-8000000000000000",	16,	-9223372036854775807LL -1,	17, 0);
+	TEST_STRTOLL("-0x8000000000000001",	0,	-9223372036854775807LL -1,	19, ERANGE);
+	TEST_STRTOLL("-0x8000000000000001",	16,	-9223372036854775807LL -1,	19, ERANGE);
+	TEST_STRTOLL("-80000000000000001",	16,	-9223372036854775807LL -1,	18, ERANGE);
+	TEST_STRTOLL("-01000000000000000000000",0,	-9223372036854775807LL -1,	24, 0);
+	TEST_STRTOLL("-01000000000000000000000",8,	-9223372036854775807LL -1,	24, 0);
+	TEST_STRTOLL("-1000000000000000000000",	8,	-9223372036854775807LL -1,	23, 0);
+	TEST_STRTOLL("-01000000000000000000001",0,	-9223372036854775807LL -1,	24, ERANGE);
+	TEST_STRTOLL("-01000000000000000000001",8,	-9223372036854775807LL -1,	24, ERANGE);
+	TEST_STRTOLL("-1000000000000000000001",	8,	-9223372036854775807LL -1,	23, ERANGE);
+
+	printf("success: strtoll\n");
+	return true;
+}
+
+static int test_strtoull(void)
+{
+	printf("test: strtoull\n");
+
+#define TEST_STRTOULL(str,base,res,diff,errnoo) TEST_STRTO_X(long long unsigned int,"%llu",strtoull,str,base,res,diff,errnoo)
+
+	TEST_STRTOULL("15",	10,	15LLU,	2, 0);
+	TEST_STRTOULL("  15",	10,	15LLU,	4, 0);
+	TEST_STRTOULL("15",	0,	15LLU,	2, 0);
+	TEST_STRTOULL(" 15 ",	0,	15LLU,	3, 0);
+	TEST_STRTOULL("+15",	10,	15LLU,	3, 0);
+	TEST_STRTOULL("  +15",	10,	15LLU,	5, 0);
+	TEST_STRTOULL("+15",	0,	15LLU,	3, 0);
+	TEST_STRTOULL(" +15 ",	0,	15LLU,	4, 0);
+	TEST_STRTOULL("-15",	10,	18446744073709551601LLU,	3, 0);
+	TEST_STRTOULL("  -15",	10,	18446744073709551601LLU,	5, 0);
+	TEST_STRTOULL("-15",	0,	18446744073709551601LLU,	3, 0);
+	TEST_STRTOULL(" -15 ",	0,	18446744073709551601LLU,	4, 0);
+	TEST_STRTOULL("015",	10,	15LLU,	3, 0);
+	TEST_STRTOULL("  015",	10,	15LLU,	5, 0);
+	TEST_STRTOULL("015",	0,	13LLU,	3, 0);
+	TEST_STRTOULL("  015",	0,	13LLU,	5, 0);
+	TEST_STRTOULL("0x15",	10,	0LLU,	1, 0);
+	TEST_STRTOULL("  0x15",	10,	0LLU,	3, 0);
+	TEST_STRTOULL("0x15",	0,	21LLU,	4, 0);
+	TEST_STRTOULL("  0x15",	0,	21LLU,	6, 0);
+
+	TEST_STRTOULL("10",	16,	16LLU,	2, 0);
+	TEST_STRTOULL("  10 ",	16,	16LLU,	4, 0);
+	TEST_STRTOULL("0x10",	16,	16LLU,	4, 0);
+	TEST_STRTOULL("0x10",	0,	16LLU,	4, 0);
+	TEST_STRTOULL(" 0x10 ",	0,	16LLU,	5, 0);
+	TEST_STRTOULL("+10",	16,	16LLU,	3, 0);
+	TEST_STRTOULL("  +10 ",	16,	16LLU,	5, 0);
+	TEST_STRTOULL("+0x10",	16,	16LLU,	5, 0);
+	TEST_STRTOULL("+0x10",	0,	16LLU,	5, 0);
+	TEST_STRTOULL(" +0x10 ",	0,	16LLU,	6, 0);
+	TEST_STRTOULL("-10",	16,	-16LLU,	3, 0);
+	TEST_STRTOULL("  -10 ",	16,	-16LLU,	5, 0);
+	TEST_STRTOULL("-0x10",	16,	-16LLU,	5, 0);
+	TEST_STRTOULL("-0x10",	0,	-16LLU,	5, 0);
+	TEST_STRTOULL(" -0x10 ",	0,	-16LLU,	6, 0);
+	TEST_STRTOULL("010",	16,	16LLU,	3, 0);
+	TEST_STRTOULL("  010 ",	16,	16LLU,	5, 0);
+	TEST_STRTOULL("-010",	16,	-16LLU,	4, 0);
+
+	TEST_STRTOULL("11",	8,	9LLU,	2, 0);
+	TEST_STRTOULL("011",	8,	9LLU,	3, 0);
+	TEST_STRTOULL("011",	0,	9LLU,	3, 0);
+	TEST_STRTOULL("-11",	8,	-9LLU,	3, 0);
+	TEST_STRTOULL("-011",	8,	-9LLU,	4, 0);
+	TEST_STRTOULL("-011",	0,	-9LLU,	4, 0);
+
+	TEST_STRTOULL("011",	8,	9LLU,	3, 0);
+	TEST_STRTOULL("011",	0,	9LLU,	3, 0);
+	TEST_STRTOULL("-11",	8,	-9LLU,	3, 0);
+	TEST_STRTOULL("-011",	8,	-9LLU,	4, 0);
+	TEST_STRTOULL("-011",	0,	-9LLU,	4, 0);
+
+	TEST_STRTOULL("Text",	0,	0LLU,	0, 0);
+
+	TEST_STRTOULL("9223372036854775807",	10,	9223372036854775807LLU,	19, 0);
+	TEST_STRTOULL("9223372036854775807",	0,	9223372036854775807LLU,	19, 0);
+	TEST_STRTOULL("9223372036854775808",	0,	9223372036854775808LLU,	19, 0);
+	TEST_STRTOULL("9223372036854775808",	10,	9223372036854775808LLU,	19, 0);
+	TEST_STRTOULL("0x7FFFFFFFFFFFFFFF",	0,	9223372036854775807LLU,	18, 0);
+	TEST_STRTOULL("0x7FFFFFFFFFFFFFFF",	16,	9223372036854775807LLU,	18, 0);
+	TEST_STRTOULL("7FFFFFFFFFFFFFFF",	16,	9223372036854775807LLU,	16, 0);
+	TEST_STRTOULL("0x8000000000000000",	0,	9223372036854775808LLU,	18, 0);
+	TEST_STRTOULL("0x8000000000000000",	16,	9223372036854775808LLU,	18, 0);
+	TEST_STRTOULL("8000000000000000",	16,	9223372036854775808LLU,	16, 0);
+	TEST_STRTOULL("0777777777777777777777",	0,	9223372036854775807LLU,	22, 0);
+	TEST_STRTOULL("0777777777777777777777",	8,	9223372036854775807LLU,	22, 0);
+	TEST_STRTOULL("777777777777777777777",	8,	9223372036854775807LLU,	21, 0);
+	TEST_STRTOULL("01000000000000000000000",0,	9223372036854775808LLU,	23, 0);
+	TEST_STRTOULL("01000000000000000000000",8,	9223372036854775808LLU,	23, 0);
+	TEST_STRTOULL("1000000000000000000000",	8,	9223372036854775808LLU,	22, 0);
+
+	TEST_STRTOULL("-9223372036854775808",	10,	9223372036854775808LLU,	20, 0);
+	TEST_STRTOULL("-9223372036854775808",	0,	9223372036854775808LLU,	20, 0);
+	TEST_STRTOULL("-9223372036854775809",	0,	9223372036854775807LLU,	20, 0);
+	TEST_STRTOULL("-9223372036854775809",	10,	9223372036854775807LLU,	20, 0);
+	TEST_STRTOULL("-0x8000000000000000",	0,	9223372036854775808LLU,	19, 0);
+	TEST_STRTOULL("-0x8000000000000000",	16,	9223372036854775808LLU,	19, 0);
+	TEST_STRTOULL("-8000000000000000",	16,	9223372036854775808LLU,	17, 0);
+	TEST_STRTOULL("-0x8000000000000001",	0,	9223372036854775807LLU,	19, 0);
+	TEST_STRTOULL("-0x8000000000000001",	16,	9223372036854775807LLU,	19, 0);
+	TEST_STRTOULL("-8000000000000001",	16,	9223372036854775807LLU,	17, 0);
+	TEST_STRTOULL("-01000000000000000000000",0,	9223372036854775808LLU,	24, 0);
+	TEST_STRTOULL("-01000000000000000000000",8,	9223372036854775808LLU,	24, 0);
+	TEST_STRTOULL("-1000000000000000000000",8,	9223372036854775808LLU,	23, 0);
+	TEST_STRTOULL("-01000000000000000000001",0,	9223372036854775807LLU,	24, 0);
+	TEST_STRTOULL("-01000000000000000000001",8,	9223372036854775807LLU,	24, 0);
+	TEST_STRTOULL("-1000000000000000000001",8,	9223372036854775807LLU,	23, 0);
+
+	TEST_STRTOULL("18446744073709551615",	0,	18446744073709551615LLU,	20, 0);
+	TEST_STRTOULL("18446744073709551615",	10,	18446744073709551615LLU,	20, 0);
+	TEST_STRTOULL("18446744073709551616",	0,	18446744073709551615LLU,	20, ERANGE);
+	TEST_STRTOULL("18446744073709551616",	10,	18446744073709551615LLU,	20, ERANGE);
+	TEST_STRTOULL("0xFFFFFFFFFFFFFFFF",	0,	18446744073709551615LLU,	18, 0);
+	TEST_STRTOULL("0xFFFFFFFFFFFFFFFF",	16,	18446744073709551615LLU,	18, 0);
+	TEST_STRTOULL("FFFFFFFFFFFFFFFF",	16,	18446744073709551615LLU,	16, 0);
+	TEST_STRTOULL("0x10000000000000000",	0,	18446744073709551615LLU,	19, ERANGE);
+	TEST_STRTOULL("0x10000000000000000",	16,	18446744073709551615LLU,	19, ERANGE);
+	TEST_STRTOULL("10000000000000000",	16,	18446744073709551615LLU,	17, ERANGE);
+	TEST_STRTOULL("01777777777777777777777",0,	18446744073709551615LLU,	23, 0);
+	TEST_STRTOULL("01777777777777777777777",8,	18446744073709551615LLU,	23, 0);
+	TEST_STRTOULL("1777777777777777777777",	8,	18446744073709551615LLU,	22, 0);
+	TEST_STRTOULL("02000000000000000000000",0,	18446744073709551615LLU,	23, ERANGE);
+	TEST_STRTOULL("02000000000000000000000",8,	18446744073709551615LLU,	23, ERANGE);
+	TEST_STRTOULL("2000000000000000000000",	8,	18446744073709551615LLU,	22, ERANGE);
+
+	TEST_STRTOULL("-18446744073709551615",	0,	1LLU,				21, 0);
+	TEST_STRTOULL("-18446744073709551615",	10,	1LLU,				21, 0);
+	TEST_STRTOULL("-18446744073709551616",	0,	18446744073709551615LLU,	21, ERANGE);
+	TEST_STRTOULL("-18446744073709551616",	10,	18446744073709551615LLU,	21, ERANGE);
+	TEST_STRTOULL("-0xFFFFFFFFFFFFFFFF",	0,	1LLU,				19, 0);
+	TEST_STRTOULL("-0xFFFFFFFFFFFFFFFF",	16,	1LLU,				19, 0);
+	TEST_STRTOULL("-FFFFFFFFFFFFFFFF",	16,	1LLU,				17, 0);
+	TEST_STRTOULL("-0x10000000000000000",	0,	18446744073709551615LLU,	20, ERANGE);
+	TEST_STRTOULL("-0x10000000000000000",	16,	18446744073709551615LLU,	20, ERANGE);
+	TEST_STRTOULL("-10000000000000000",	16,	18446744073709551615LLU,	18, ERANGE);
+	TEST_STRTOULL("-01777777777777777777777",0,	1LLU,				24, 0);
+	TEST_STRTOULL("-01777777777777777777777",8,	1LLU,				24, 0);
+	TEST_STRTOULL("-1777777777777777777777",8,	1LLU,				23, 0);
+	TEST_STRTOULL("-02000000000000000000000",0,	18446744073709551615LLU,	24, ERANGE);
+	TEST_STRTOULL("-02000000000000000000000",8,	18446744073709551615LLU,	24, ERANGE);
+	TEST_STRTOULL("-2000000000000000000000",8,	18446744073709551615LLU,	23, ERANGE);
+
+	printf("success: strtoull\n");
+	return true;
+}
+
+/* 
+FIXME:
+Types:
+bool
+socklen_t
+uint{8,16,32,64}_t
+int{8,16,32,64}_t
+intptr_t
+
+Constants:
+PATH_NAME_MAX
+UINT{16,32,64}_MAX
+INT32_MAX
+*/
+
+static int test_va_copy(void)
+{
+	/* FIXME */
+	return true;
+}
+
+static int test_FUNCTION(void)
+{
+	printf("test: FUNCTION\n");
+	if (strcmp(__FUNCTION__, "test_FUNCTION") != 0) {
+		printf("failure: FUNCTION [\nFUNCTION invalid\n]\n");
+		return false;
+	}
+	printf("success: FUNCTION\n");
+	return true;
+}
+
+static int test_MIN(void)
+{
+	printf("test: MIN\n");
+	if (MIN(20, 1) != 1) {
+		printf("failure: MIN [\nMIN invalid\n]\n");
+		return false;
+	}
+	if (MIN(1, 20) != 1) {
+		printf("failure: MIN [\nMIN invalid\n]\n");
+		return false;
+	}
+	printf("success: MIN\n");
+	return true;
+}
+
+static int test_MAX(void)
+{
+	printf("test: MAX\n");
+	if (MAX(20, 1) != 20) {
+		printf("failure: MAX [\nMAX invalid\n]\n");
+		return false;
+	}
+	if (MAX(1, 20) != 20) {
+		printf("failure: MAX [\nMAX invalid\n]\n");
+		return false;
+	}
+	printf("success: MAX\n");
+	return true;
+}
+
+static int test_socketpair(void)
+{
+	int sock[2];
+	char buf[20];
+
+	printf("test: socketpair\n");
+
+	if (socketpair(AF_UNIX, SOCK_STREAM, 0, sock) == -1) {
+		printf("failure: socketpair [\n"
+			   "socketpair() failed\n"
+			   "]\n");
+		return false;
+	}
+
+	if (write(sock[1], "automatisch", 12) == -1) {
+		printf("failure: socketpair [\n"
+			   "write() failed: %s\n"
+			   "]\n", strerror(errno));
+		return false;
+	}
+
+	if (read(sock[0], buf, 12) == -1) {
+		printf("failure: socketpair [\n"
+			   "read() failed: %s\n"
+			   "]\n", strerror(errno));
+		return false;
+	}
+
+	if (strcmp(buf, "automatisch") != 0) {
+		printf("failure: socketpair [\n"
+			   "expected: automatisch, got: %s\n"
+			   "]\n", buf);
+		return false;
+	}
+
+	printf("success: socketpair\n");
+
+	return true;
+}
+
+extern int libreplace_test_strptime(void);
+
+static int test_strptime(void)
+{
+	return libreplace_test_strptime();
+}
+
+extern int getifaddrs_test(void);
+
+static int test_getifaddrs(void)
+{
+
+	printf("test: getifaddrs\n");
+
+	if (getifaddrs_test() != 0) {
+		printf("failure: getifaddrs\n");
+		return false;
+	}
+
+	printf("success: getifaddrs\n");
+	return true;
+}
+
+static int test_utime(void)
+{
+	struct utimbuf u;
+	struct stat st1, st2, st3;
+	int fd;
+
+	printf("test: utime\n");
+	unlink(TESTFILE);
+
+	fd = open(TESTFILE, O_RDWR|O_CREAT, 0600);
+	if (fd == -1) {
+		printf("failure: utime [\n"
+		       "creating '%s' failed - %s\n]\n",
+		       TESTFILE, strerror(errno));
+		return false;
+	}
+
+	if (fstat(fd, &st1) != 0) {
+		printf("failure: utime [\n"
+		       "fstat (1) failed - %s\n]\n",
+		       strerror(errno));
+		return false;
+	}
+
+	u.actime = st1.st_atime + 300;
+	u.modtime = st1.st_mtime - 300;
+	if (utime(TESTFILE, &u) != 0) {
+		printf("failure: utime [\n"
+		       "utime(&u) failed - %s\n]\n",
+		       strerror(errno));
+		return false;
+	}
+
+	if (fstat(fd, &st2) != 0) {
+		printf("failure: utime [\n"
+		       "fstat (2) failed - %s\n]\n",
+		       strerror(errno));
+		return false;
+	}
+
+	if (utime(TESTFILE, NULL) != 0) {
+		printf("failure: utime [\n"
+		       "utime(NULL) failed - %s\n]\n",
+		       strerror(errno));
+		return false;
+	}
+
+	if (fstat(fd, &st3) != 0) {
+		printf("failure: utime [\n"
+		       "fstat (3) failed - %s\n]\n",
+		       strerror(errno));
+		return false;
+	}
+
+#define CMP_VAL(a,c,b) do { \
+	if (a c b) { \
+		printf("failure: utime [\n" \
+		       "%s: %s(%d) %s %s(%d)\n]\n", \
+		       __location__, \
+		       #a, (int)a, #c, #b, (int)b); \
+		return false; \
+	} \
+} while(0)
+#define EQUAL_VAL(a,b) CMP_VAL(a,!=,b)
+#define GREATER_VAL(a,b) CMP_VAL(a,<=,b)
+#define LESSER_VAL(a,b) CMP_VAL(a,>=,b)
+
+	EQUAL_VAL(st2.st_atime, st1.st_atime + 300);
+	EQUAL_VAL(st2.st_mtime, st1.st_mtime - 300);
+	LESSER_VAL(st3.st_atime, st2.st_atime);
+	GREATER_VAL(st3.st_mtime, st2.st_mtime);
+
+#undef CMP_VAL
+#undef EQUAL_VAL
+#undef GREATER_VAL
+#undef LESSER_VAL
+
+	unlink(TESTFILE);
+	printf("success: utime\n");
+	return true;
+}
+
+static int test_utimes(void)
+{
+	struct timeval tv[2];
+	struct stat st1, st2;
+	int fd;
+
+	printf("test: utimes\n");
+	unlink(TESTFILE);
+
+	fd = open(TESTFILE, O_RDWR|O_CREAT, 0600);
+	if (fd == -1) {
+		printf("failure: utimes [\n"
+		       "creating '%s' failed - %s\n]\n",
+		       TESTFILE, strerror(errno));
+		return false;
+	}
+
+	if (fstat(fd, &st1) != 0) {
+		printf("failure: utimes [\n"
+		       "fstat (1) failed - %s\n]\n",
+		       strerror(errno));
+		return false;
+	}
+
+	ZERO_STRUCT(tv);
+	tv[0].tv_sec = st1.st_atime + 300;
+	tv[1].tv_sec = st1.st_mtime - 300;
+	if (utimes(TESTFILE, tv) != 0) {
+		printf("failure: utimes [\n"
+		       "utimes(tv) failed - %s\n]\n",
+		       strerror(errno));
+		return false;
+	}
+
+	if (fstat(fd, &st2) != 0) {
+		printf("failure: utimes [\n"
+		       "fstat (2) failed - %s\n]\n",
+		       strerror(errno));
+		return false;
+	}
+
+#define EQUAL_VAL(a,b) do { \
+	if (a != b) { \
+		printf("failure: utimes [\n" \
+		       "%s: %s(%d) != %s(%d)\n]\n", \
+		       __location__, \
+		       #a, (int)a, #b, (int)b); \
+		return false; \
+	} \
+} while(0)
+
+	EQUAL_VAL(st2.st_atime, st1.st_atime + 300);
+	EQUAL_VAL(st2.st_mtime, st1.st_mtime - 300);
+
+#undef EQUAL_VAL
+
+	unlink(TESTFILE);
+	printf("success: utimes\n");
+	return true;
+}
+
+static int test_memmem(void)
+{
+	char *s;
+
+	printf("test: memmem\n");
+
+	s = (char *)memmem("foo", 3, "fo", 2);
+	if (strcmp(s, "foo") != 0) {
+		printf(__location__ ": Failed memmem\n");
+		return false;
+	}
+
+	s = (char *)memmem("foo", 3, "", 0);
+	/* it is allowable for this to return NULL (as happens on
+	   FreeBSD) */
+	if (s && strcmp(s, "foo") != 0) {
+		printf(__location__ ": Failed memmem\n");
+		return false;
+	}
+
+	s = (char *)memmem("foo", 4, "o", 1);
+	if (strcmp(s, "oo") != 0) {
+		printf(__location__ ": Failed memmem\n");
+		return false;
+	}
+
+	s = (char *)memmem("foobarfodx", 11, "fod", 3);
+	if (strcmp(s, "fodx") != 0) {
+		printf(__location__ ": Failed memmem\n");
+		return false;
+	}
+
+	printf("success: memmem\n");
+
+	return true;
+}
+
+
+bool torture_local_replace(struct torture_context *ctx)
+{
+	bool ret = true;
+	ret &= test_ftruncate();
+	ret &= test_strlcpy();
+	ret &= test_strlcat();
+	ret &= test_mktime();
+	ret &= test_initgroups();
+	ret &= test_memmove();
+	ret &= test_strdup();
+	ret &= test_setlinebuf();
+	ret &= test_vsyslog();
+	ret &= test_timegm();
+	ret &= test_setenv();
+	ret &= test_strndup();
+	ret &= test_strnlen();
+	ret &= test_waitpid();
+	ret &= test_seteuid();
+	ret &= test_setegid();
+	ret &= test_asprintf();
+	ret &= test_snprintf();
+	ret &= test_vasprintf();
+	ret &= test_vsnprintf();
+	ret &= test_opendir();
+	ret &= test_readdir();
+	ret &= test_telldir();
+	ret &= test_seekdir();
+	ret &= test_dlopen();
+	ret &= test_chroot();
+	ret &= test_bzero();
+	ret &= test_strerror();
+	ret &= test_errno();
+	ret &= test_mkdtemp();
+	ret &= test_mkstemp();
+	ret &= test_pread();
+	ret &= test_pwrite();
+	ret &= test_inet_ntoa();
+	ret &= test_strtoll();
+	ret &= test_strtoull();
+	ret &= test_va_copy();
+	ret &= test_FUNCTION();
+	ret &= test_MIN();
+	ret &= test_MAX();
+	ret &= test_socketpair();
+	ret &= test_strptime();
+	ret &= test_getifaddrs();
+	ret &= test_utime();
+	ret &= test_utimes();
+	ret &= test_memmem();
+
+	return ret;
+}
diff --git a/ctdb/lib/replace/timegm.c b/ctdb/lib/replace/timegm.c
new file mode 100644
index 0000000000..395c684e11
--- /dev/null
+++ b/ctdb/lib/replace/timegm.c
@@ -0,0 +1,78 @@
+/*
+ * Copyright (c) 1997 Kungliga Tekniska Högskolan
+ * (Royal Institute of Technology, Stockholm, Sweden). 
+ * All rights reserved. 
+ *
+ * Redistribution and use in source and binary forms, with or without 
+ * modification, are permitted provided that the following conditions 
+ * are met: 
+ *
+ * 1. Redistributions of source code must retain the above copyright 
+ *    notice, this list of conditions and the following disclaimer. 
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright 
+ *    notice, this list of conditions and the following disclaimer in the 
+ *    documentation and/or other materials provided with the distribution. 
+ *
+ * 3. Neither the name of the Institute nor the names of its contributors 
+ *    may be used to endorse or promote products derived from this software 
+ *    without specific prior written permission. 
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE INSTITUTE AND CONTRIBUTORS ``AS IS'' AND 
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE INSTITUTE OR CONTRIBUTORS BE LIABLE 
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 
+ * SUCH DAMAGE. 
+ */
+
+/*
+  adapted for Samba4 by Andrew Tridgell
+*/
+
+#include "replace.h"
+#include "system/time.h"
+
+static int is_leap(unsigned y)
+{
+	y += 1900;
+	return (y % 4) == 0 && ((y % 100) != 0 || (y % 400) == 0);
+}
+
+time_t rep_timegm(struct tm *tm)
+{
+	static const unsigned ndays[2][12] ={
+		{31, 28, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31},
+		{31, 29, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31}};
+	time_t res = 0;
+	unsigned i;
+
+	if (tm->tm_mon > 12 ||
+	    tm->tm_mon < 0 ||
+	    tm->tm_mday > 31 ||
+	    tm->tm_min > 60 ||
+	    tm->tm_sec > 60 ||
+	    tm->tm_hour > 24) {
+		/* invalid tm structure */
+		return 0;
+	}
+	
+	for (i = 70; i < tm->tm_year; ++i)
+		res += is_leap(i) ? 366 : 365;
+	
+	for (i = 0; i < tm->tm_mon; ++i)
+		res += ndays[is_leap(tm->tm_year)][i];
+	res += tm->tm_mday - 1;
+	res *= 24;
+	res += tm->tm_hour;
+	res *= 60;
+	res += tm->tm_min;
+	res *= 60;
+	res += tm->tm_sec;
+	return res;
+}
diff --git a/ctdb/lib/replace/timegm.m4 b/ctdb/lib/replace/timegm.m4
new file mode 100644
index 0000000000..9b76d0c727
--- /dev/null
+++ b/ctdb/lib/replace/timegm.m4
@@ -0,0 +1 @@
+AC_CHECK_FUNCS(timegm,[],[LIBREPLACEOBJ="${LIBREPLACEOBJ} $libreplacedir/timegm.o"])
diff --git a/ctdb/lib/replace/win32.m4 b/ctdb/lib/replace/win32.m4
new file mode 100644
index 0000000000..eb364e2cb9
--- /dev/null
+++ b/ctdb/lib/replace/win32.m4
@@ -0,0 +1,20 @@
+AC_CHECK_HEADERS(direct.h windows.h winsock2.h ws2tcpip.h)
+
+#######################################
+# Check for mkdir mode
+AC_CACHE_CHECK( [whether mkdir supports mode], libreplace_cv_mkdir_has_mode,
+	AC_TRY_COMPILE([
+		#include <stdio.h>
+		#ifdef HAVE_DIRECT_H
+		#include <direct.h>
+		#endif],[
+			mkdir("foo",0777);
+			return 0;
+	],
+    libreplace_cv_mkdir_has_mode="yes",
+    libreplace_cv_mkdir_has_mode="no") )
+
+if test "$libreplace_cv_mkdir_has_mode" = "yes"
+then
+    AC_DEFINE(HAVE_MKDIR_MODE, 1, [Define if target mkdir supports mode option])
+fi
diff --git a/ctdb/lib/replace/win32_replace.h b/ctdb/lib/replace/win32_replace.h
new file mode 100644
index 0000000000..9901e72f6e
--- /dev/null
+++ b/ctdb/lib/replace/win32_replace.h
@@ -0,0 +1,159 @@
+#ifndef _WIN32_REPLACE_H
+#define _WIN32_REPLACE_H
+
+#ifdef HAVE_WINSOCK2_H
+#include <winsock2.h>
+#endif
+
+#ifdef HAVE_WS2TCPIP_H
+#include <ws2tcpip.h>
+#endif
+
+#ifdef HAVE_WINDOWS_H
+#include <windows.h>
+#endif
+
+/* Map BSD Socket errorcodes to the WSA errorcodes (if possible) */ 
+
+#define EAFNOSUPPORT	WSAEAFNOSUPPORT
+#define ECONNREFUSED    WSAECONNREFUSED 
+#define EINPROGRESS	WSAEINPROGRESS
+#define EMSGSIZE	WSAEMSGSIZE 
+#define ENOBUFS         WSAENOBUFS
+#define ENOTSOCK	WSAENOTSOCK
+#define ENETUNREACH	WSAENETUNREACH
+#define ENOPROTOOPT	WSAENOPROTOOPT
+#define ENOTCONN	WSAENOTCONN 
+#define ENOTSUP		134 
+
+/* We undefine the following constants due to conflicts with the w32api headers
+ * and the Windows Platform SDK/DDK.
+ */
+
+#undef interface
+
+#undef ERROR_INVALID_PARAMETER
+#undef ERROR_INSUFFICIENT_BUFFER
+#undef ERROR_INVALID_DATATYPE
+
+#undef FILE_GENERIC_READ
+#undef FILE_GENERIC_WRITE
+#undef FILE_GENERIC_EXECUTE
+#undef FILE_ATTRIBUTE_READONLY
+#undef FILE_ATTRIBUTE_HIDDEN
+#undef FILE_ATTRIBUTE_SYSTEM
+#undef FILE_ATTRIBUTE_DIRECTORY
+#undef FILE_ATTRIBUTE_ARCHIVE
+#undef FILE_ATTRIBUTE_DEVICE
+#undef FILE_ATTRIBUTE_NORMAL
+#undef FILE_ATTRIBUTE_TEMPORARY
+#undef FILE_ATTRIBUTE_REPARSE_POINT
+#undef FILE_ATTRIBUTE_COMPRESSED
+#undef FILE_ATTRIBUTE_OFFLINE
+#undef FILE_ATTRIBUTE_ENCRYPTED
+#undef FILE_FLAG_WRITE_THROUGH
+#undef FILE_FLAG_NO_BUFFERING
+#undef FILE_FLAG_RANDOM_ACCESS
+#undef FILE_FLAG_SEQUENTIAL_SCAN
+#undef FILE_FLAG_DELETE_ON_CLOSE
+#undef FILE_FLAG_BACKUP_SEMANTICS
+#undef FILE_FLAG_POSIX_SEMANTICS
+#undef FILE_TYPE_DISK
+#undef FILE_TYPE_UNKNOWN
+#undef FILE_CASE_SENSITIVE_SEARCH
+#undef FILE_CASE_PRESERVED_NAMES
+#undef FILE_UNICODE_ON_DISK
+#undef FILE_PERSISTENT_ACLS
+#undef FILE_FILE_COMPRESSION
+#undef FILE_VOLUME_QUOTAS
+#undef FILE_VOLUME_IS_COMPRESSED
+#undef FILE_NOTIFY_CHANGE_FILE_NAME
+#undef FILE_NOTIFY_CHANGE_DIR_NAME
+#undef FILE_NOTIFY_CHANGE_ATTRIBUTES
+#undef FILE_NOTIFY_CHANGE_SIZE
+#undef FILE_NOTIFY_CHANGE_LAST_WRITE
+#undef FILE_NOTIFY_CHANGE_LAST_ACCESS
+#undef FILE_NOTIFY_CHANGE_CREATION
+#undef FILE_NOTIFY_CHANGE_EA
+#undef FILE_NOTIFY_CHANGE_SECURITY
+#undef FILE_NOTIFY_CHANGE_STREAM_NAME
+#undef FILE_NOTIFY_CHANGE_STREAM_SIZE
+#undef FILE_NOTIFY_CHANGE_STREAM_WRITE
+#undef FILE_NOTIFY_CHANGE_NAME
+
+#undef PRINTER_ATTRIBUTE_QUEUED
+#undef PRINTER_ATTRIBUTE_DIRECT
+#undef PRINTER_ATTRIBUTE_DEFAULT
+#undef PRINTER_ATTRIBUTE_SHARED
+#undef PRINTER_ATTRIBUTE_NETWORK
+#undef PRINTER_ATTRIBUTE_HIDDEN
+#undef PRINTER_ATTRIBUTE_LOCAL
+#undef PRINTER_ATTRIBUTE_ENABLE_DEVQ
+#undef PRINTER_ATTRIBUTE_KEEPPRINTEDJOBS
+#undef PRINTER_ATTRIBUTE_DO_COMPLETE_FIRST
+#undef PRINTER_ATTRIBUTE_WORK_OFFLINE
+#undef PRINTER_ATTRIBUTE_ENABLE_BIDI
+#undef PRINTER_ATTRIBUTE_RAW_ONLY
+#undef PRINTER_ATTRIBUTE_PUBLISHED
+#undef PRINTER_ENUM_DEFAULT
+#undef PRINTER_ENUM_LOCAL
+#undef PRINTER_ENUM_CONNECTIONS
+#undef PRINTER_ENUM_FAVORITE
+#undef PRINTER_ENUM_NAME
+#undef PRINTER_ENUM_REMOTE
+#undef PRINTER_ENUM_SHARED
+#undef PRINTER_ENUM_NETWORK
+#undef PRINTER_ENUM_EXPAND
+#undef PRINTER_ENUM_CONTAINER
+#undef PRINTER_ENUM_ICON1
+#undef PRINTER_ENUM_ICON2
+#undef PRINTER_ENUM_ICON3
+#undef PRINTER_ENUM_ICON4
+#undef PRINTER_ENUM_ICON5
+#undef PRINTER_ENUM_ICON6
+#undef PRINTER_ENUM_ICON7
+#undef PRINTER_ENUM_ICON8
+#undef PRINTER_STATUS_PAUSED
+#undef PRINTER_STATUS_ERROR
+#undef PRINTER_STATUS_PENDING_DELETION
+#undef PRINTER_STATUS_PAPER_JAM
+#undef PRINTER_STATUS_PAPER_OUT
+#undef PRINTER_STATUS_MANUAL_FEED
+#undef PRINTER_STATUS_PAPER_PROBLEM
+#undef PRINTER_STATUS_OFFLINE
+#undef PRINTER_STATUS_IO_ACTIVE
+#undef PRINTER_STATUS_BUSY
+#undef PRINTER_STATUS_PRINTING
+#undef PRINTER_STATUS_OUTPUT_BIN_FULL
+#undef PRINTER_STATUS_NOT_AVAILABLE
+#undef PRINTER_STATUS_WAITING
+#undef PRINTER_STATUS_PROCESSING
+#undef PRINTER_STATUS_INITIALIZING
+#undef PRINTER_STATUS_WARMING_UP
+#undef PRINTER_STATUS_TONER_LOW
+#undef PRINTER_STATUS_NO_TONER
+#undef PRINTER_STATUS_PAGE_PUNT
+#undef PRINTER_STATUS_USER_INTERVENTION
+#undef PRINTER_STATUS_OUT_OF_MEMORY
+#undef PRINTER_STATUS_DOOR_OPEN
+#undef PRINTER_STATUS_SERVER_UNKNOWN
+#undef PRINTER_STATUS_POWER_SAVE
+
+#undef DWORD
+#undef HKEY_CLASSES_ROOT
+#undef HKEY_CURRENT_USER
+#undef HKEY_LOCAL_MACHINE
+#undef HKEY_USERS
+#undef HKEY_PERFORMANCE_DATA
+#undef HKEY_CURRENT_CONFIG
+#undef HKEY_DYN_DATA
+#undef REG_DWORD
+#undef REG_QWORD
+
+#undef SERVICE_STATE_ALL
+
+#undef SE_GROUP_MANDATORY
+#undef SE_GROUP_ENABLED_BY_DEFAULT
+#undef SE_GROUP_ENABLED
+
+#endif /* _WIN32_REPLACE_H */
diff --git a/ctdb/lib/replace/wscript b/ctdb/lib/replace/wscript
new file mode 100644
index 0000000000..2117f56e22
--- /dev/null
+++ b/ctdb/lib/replace/wscript
@@ -0,0 +1,632 @@
+#!/usr/bin/env python
+
+APPNAME = 'libreplace'
+VERSION = '1.2.1'
+
+blddir = 'bin'
+
+import sys, os, Utils
+
+# find the buildtools directory
+srcdir = '.'
+while not os.path.exists(srcdir+'/buildtools') and len(srcdir.split('/')) < 5:
+    srcdir = '../' + srcdir
+sys.path.insert(0, srcdir + '/buildtools/wafsamba')
+
+import wafsamba, samba_dist
+import Options, os, preproc
+
+samba_dist.DIST_DIRS('lib/replace buildtools:buildtools')
+
+def set_options(opt):
+    opt.BUILTIN_DEFAULT('NONE')
+    opt.PRIVATE_EXTENSION_DEFAULT('')
+    opt.RECURSE('buildtools/wafsamba')
+
+@wafsamba.runonce
+def configure(conf):
+    conf.RECURSE('buildtools/wafsamba')
+
+    conf.env.standalone_replace = conf.IN_LAUNCH_DIR()
+
+    conf.DEFINE('HAVE_LIBREPLACE', 1)
+    conf.DEFINE('LIBREPLACE_NETWORK_CHECKS', 1)
+
+    # on Tru64 certain features are only available with _OSF_SOURCE set to 1
+    # and _XOPEN_SOURCE set to 600
+    if conf.env['SYSTEM_UNAME_SYSNAME'] == 'OSF1':
+        conf.DEFINE('_OSF_SOURCE', 1, add_to_cflags=True)
+        conf.DEFINE('_XOPEN_SOURCE', 600, add_to_cflags=True)
+
+    # SCM_RIGHTS is only avail if _XOPEN_SOURCE iѕ defined on IRIX
+    if conf.env['SYSTEM_UNAME_SYSNAME'] == 'IRIX':
+        conf.DEFINE('_XOPEN_SOURCE', 600, add_to_cflags=True)
+        conf.DEFINE('_BSD_TYPES', 1, add_to_cflags=True)
+
+    # Try to find the right extra flags for C99 initialisers
+    for f in ["", "-AC99", "-qlanglvl=extc99", "-qlanglvl=stdc99", "-c99"]:
+        if conf.CHECK_CFLAGS([f], '''
+struct foo {int x;char y;};
+struct foo bar = { .y = 'X', .x = 1 };
+'''):
+            if f != "":
+                conf.ADD_CFLAGS(f)
+            break
+
+    if conf.CHECK_CFLAGS(['-fstack-protector']) and conf.CHECK_LDFLAGS(['-fstack-protector']):
+        conf.ADD_CFLAGS('-fstack-protector')
+        conf.ADD_LDFLAGS('-fstack-protector')
+
+    # Try to find the right extra flags for -Werror behaviour
+    for f in ["-Werror",       # GCC
+	      "-errwarn=%all", # Sun Studio
+	      "-qhalt=w",     # IBM xlc
+	      "-w2",           # Tru64
+	      ]:
+        if conf.CHECK_CFLAGS([f], '''
+'''):
+            if not 'WERROR_CFLAGS' in conf.env:
+                conf.env['WERROR_CFLAGS'] = []
+            conf.env['WERROR_CFLAGS'].extend([f])
+            break
+
+    conf.CHECK_HEADERS('linux/types.h crypt.h locale.h acl/libacl.h compat.h')
+    conf.CHECK_HEADERS('acl/libacl.h attr/xattr.h compat.h ctype.h dustat.h')
+    conf.CHECK_HEADERS('fcntl.h fnmatch.h glob.h history.h krb5.h langinfo.h')
+    conf.CHECK_HEADERS('libaio.h locale.h ndir.h pwd.h')
+    conf.CHECK_HEADERS('shadow.h sys/acl.h')
+    conf.CHECK_HEADERS('sys/attributes.h attr/attributes.h sys/capability.h sys/dir.h sys/epoll.h')
+    conf.CHECK_HEADERS('sys/fcntl.h sys/filio.h sys/filsys.h sys/fs/s5param.h sys/fs/vx/quota.h')
+    conf.CHECK_HEADERS('sys/id.h sys/ioctl.h sys/ipc.h sys/mman.h sys/mode.h sys/ndir.h sys/priv.h')
+    conf.CHECK_HEADERS('sys/resource.h sys/security.h sys/shm.h sys/statfs.h sys/statvfs.h sys/termio.h')
+    conf.CHECK_HEADERS('sys/vfs.h sys/xattr.h termio.h termios.h sys/file.h')
+    conf.CHECK_HEADERS('sys/ucontext.h sys/wait.h sys/stat.h malloc.h grp.h')
+    conf.CHECK_HEADERS('sys/select.h setjmp.h utime.h sys/syslog.h syslog.h')
+    conf.CHECK_HEADERS('stdarg.h vararg.h sys/mount.h mntent.h')
+    conf.CHECK_HEADERS('stropts.h unix.h string.h strings.h sys/param.h limits.h')
+    conf.CHECK_HEADERS('''sys/socket.h netinet/in.h netdb.h arpa/inet.h netinet/in_systm.h
+                          netinet/ip.h netinet/tcp.h netinet/in_ip.h
+                          sys/sockio.h sys/un.h''', together=True)
+    conf.CHECK_HEADERS('sys/uio.h ifaddrs.h direct.h dirent.h')
+    conf.CHECK_HEADERS('windows.h winsock2.h ws2tcpip.h')
+    conf.CHECK_HEADERS('libintl.h errno.h')
+    conf.CHECK_HEADERS('gcrypt.h getopt.h iconv.h')
+    conf.CHECK_HEADERS('sys/inotify.h memory.h nss.h sasl/sasl.h')
+    conf.CHECK_HEADERS('security/pam_appl.h zlib.h asm/unistd.h')
+    conf.CHECK_HEADERS('aio.h sys/unistd.h rpc/rpc.h rpc/nettype.h alloca.h float.h')
+
+    conf.CHECK_HEADERS('rpcsvc/nis.h rpcsvc/ypclnt.h sys/sysctl.h')
+    conf.CHECK_HEADERS('sys/fileio.h sys/filesys.h sys/dustat.h sys/sysmacros.h')
+    conf.CHECK_HEADERS('xfs/libxfs.h netgroup.h')
+
+    conf.CHECK_CODE('', headers='rpc/rpc.h rpcsvc/yp_prot.h', define='HAVE_RPCSVC_YP_PROT_H')
+
+    conf.CHECK_HEADERS('valgrind.h valgrind/valgrind.h valgrind/memcheck.h')
+    conf.CHECK_HEADERS('nss_common.h nsswitch.h ns_api.h')
+    conf.CHECK_HEADERS('sys/extattr.h sys/ea.h sys/proplist.h sys/cdefs.h')
+    conf.CHECK_HEADERS('utmp.h utmpx.h lastlog.h malloc.h')
+    conf.CHECK_HEADERS('syscall.h sys/syscall.h inttypes.h')
+
+    # Check for process set name support
+    conf.CHECK_CODE('''
+                    #include <sys/prctl.h>
+                    int main(void) {
+                        prctl(0);
+                        return 0;
+                    }
+                    ''',
+                    'HAVE_PRCTL',
+                    headers='sys/prctl.h',
+                    msg='Checking for prctl syscall')
+
+    conf.CHECK_CODE('''
+                    #include <unistd.h>
+                    #ifdef HAVE_FCNTL_H
+                    #include <fcntl.h>
+                    #endif
+                    int main(void) { int fd = open("/dev/null", O_DIRECT); }
+                    ''',
+                    define='HAVE_OPEN_O_DIRECT',
+                    addmain=False,
+                    msg='Checking for O_DIRECT flag to open(2)')
+
+    conf.CHECK_TYPES('"long long" intptr_t uintptr_t ptrdiff_t comparison_fn_t')
+    conf.CHECK_TYPE('_Bool', define='HAVE__Bool')
+    conf.CHECK_TYPE('bool', define='HAVE_BOOL')
+
+    conf.CHECK_TYPE('int8_t', 'char')
+    conf.CHECK_TYPE('uint8_t', 'unsigned char')
+    conf.CHECK_TYPE('int16_t', 'short')
+    conf.CHECK_TYPE('uint16_t', 'unsigned short')
+    conf.CHECK_TYPE('int32_t', 'int')
+    conf.CHECK_TYPE('uint32_t', 'unsigned')
+    conf.CHECK_TYPE('int64_t', 'long long')
+    conf.CHECK_TYPE('uint64_t', 'unsigned long long')
+    conf.CHECK_TYPE('size_t', 'unsigned int')
+    conf.CHECK_TYPE('ssize_t', 'int')
+    conf.CHECK_TYPE('ino_t', 'unsigned')
+    conf.CHECK_TYPE('loff_t', 'off_t')
+    conf.CHECK_TYPE('offset_t', 'loff_t')
+    conf.CHECK_TYPE('volatile int', define='HAVE_VOLATILE')
+    conf.CHECK_TYPE('uint_t', 'unsigned int')
+    conf.CHECK_TYPE('blksize_t', 'long', headers='sys/types.h sys/stat.h unistd.h')
+    conf.CHECK_TYPE('blkcnt_t', 'long', headers='sys/types.h sys/stat.h unistd.h')
+
+    conf.CHECK_SIZEOF('bool char int "long long" long short size_t ssize_t')
+    conf.CHECK_SIZEOF('int8_t uint8_t int16_t uint16_t int32_t uint32_t int64_t uint64_t')
+    conf.CHECK_SIZEOF('void*', define='SIZEOF_VOID_P')
+    conf.CHECK_SIZEOF('off_t dev_t ino_t time_t')
+
+    conf.CHECK_TYPES('socklen_t', headers='sys/socket.h')
+    conf.CHECK_TYPE_IN('struct ifaddrs', 'ifaddrs.h')
+    conf.CHECK_TYPE_IN('struct addrinfo', 'netdb.h')
+    conf.CHECK_TYPE_IN('struct sockaddr', 'sys/socket.h')
+    conf.CHECK_CODE('struct sockaddr_in6 x', define='HAVE_STRUCT_SOCKADDR_IN6',
+                    headers='sys/socket.h netdb.h netinet/in.h')
+    conf.CHECK_TYPE_IN('struct sockaddr_storage', 'sys/socket.h')
+    conf.CHECK_TYPE_IN('sa_family_t', 'sys/socket.h')
+
+    conf.CHECK_TYPE_IN('sig_atomic_t', 'signal.h', define='HAVE_SIG_ATOMIC_T_TYPE')
+
+    conf.CHECK_FUNCS_IN('''inet_ntoa inet_aton inet_ntop inet_pton connect gethostbyname
+                           getaddrinfo getnameinfo freeaddrinfo gai_strerror socketpair''',
+                        'socket nsl', checklibc=True,
+                        headers='sys/socket.h netinet/in.h arpa/inet.h netdb.h')
+
+    # Some old Linux systems have broken header files and
+    # miss the IPV6_V6ONLY define in netinet/in.h,
+    # but have it in linux/in6.h.
+    # We can't include both files so we just check if the value
+    # if defined and do the replacement in system/network.h
+    if not conf.CHECK_VARIABLE('IPV6_V6ONLY',
+                               headers='sys/socket.h netdb.h netinet/in.h'):
+        conf.CHECK_CODE('''
+                        #include <linux/in6.h>
+                        #if (IPV6_V6ONLY != 26)
+                        #error no IPV6_V6ONLY support on linux
+                        #endif
+                        int main(void) { return IPV6_V6ONLY; }
+                        ''',
+                        define='HAVE_LINUX_IPV6_V6ONLY_26',
+                        addmain=False,
+                        msg='Checking for IPV6_V6ONLY in linux/in6.h',
+                        local_include=False)
+
+    conf.CHECK_CODE('''
+                       struct sockaddr_storage sa_store;
+                       struct addrinfo *ai = NULL;
+                       struct in6_addr in6addr;
+                       int idx = if_nametoindex("iface1");
+                       int s = socket(AF_INET6, SOCK_STREAM, 0);
+                       int ret = getaddrinfo(NULL, NULL, NULL, &ai);
+                       if (ret != 0) {
+                           const char *es = gai_strerror(ret);
+                       }
+                       freeaddrinfo(ai);
+                       {
+                          int val = 1;
+                          #ifdef HAVE_LINUX_IPV6_V6ONLY_26
+                          #define IPV6_V6ONLY 26
+                          #endif
+                          ret = setsockopt(s, IPPROTO_IPV6, IPV6_V6ONLY,
+                                           (const void *)&val, sizeof(val));
+                       }
+                       ''',
+                    define='HAVE_IPV6',
+                    lib='nsl socket',
+                    headers='sys/socket.h netdb.h netinet/in.h')
+
+    if conf.CONFIG_SET('HAVE_SYS_UCONTEXT_H') and conf.CONFIG_SET('HAVE_SIGNAL_H'):
+        conf.CHECK_CODE('''
+                       ucontext_t uc;
+                       sigaddset(&uc.uc_sigmask, SIGUSR1);
+                       ''',
+                       'HAVE_UCONTEXT_T',
+                       msg="Checking whether we have ucontext_t",
+                       headers='signal.h sys/ucontext.h')
+
+    # these may be builtins, so we need the link=False strategy
+    conf.CHECK_FUNCS('strdup memmem printf memset memcpy memmove strcpy strncpy bzero', link=False)
+
+    # See https://bugzilla.samba.org/show_bug.cgi?id=1097
+    #
+    # Ported in from autoconf where it was added with this commit:
+    # commit 804cfb20a067b4b687089dc72a8271b3abf20f31
+    # Author: Simo Sorce <idra@samba.org>
+    # Date:   Wed Aug 25 14:24:16 2004 +0000
+    #     r2070: Let's try to overload srnlen and strndup for AIX where they are natly broken.
+
+    host_os = sys.platform
+    if host_os.rfind('aix') > -1:
+        conf.DEFINE('BROKEN_STRNLEN', 1)
+        conf.DEFINE('BROKEN_STRNDUP', 1)
+
+    conf.CHECK_FUNCS('shl_load shl_unload shl_findsym')
+    conf.CHECK_FUNCS('pipe strftime srandom random srand rand usleep setbuffer')
+    conf.CHECK_FUNCS('lstat getpgrp utime utimes setuid seteuid setreuid setresuid setgid setegid')
+    conf.CHECK_FUNCS('setregid setresgid chroot strerror vsyslog setlinebuf mktime')
+    conf.CHECK_FUNCS('ftruncate chsize rename waitpid wait4')
+    conf.CHECK_FUNCS('initgroups pread pwrite strndup strcasestr')
+    conf.CHECK_FUNCS('strtok_r mkdtemp dup2 dprintf vdprintf isatty chown lchown')
+    conf.CHECK_FUNCS('link readlink symlink realpath snprintf vsnprintf')
+    conf.CHECK_FUNCS('asprintf vasprintf setenv unsetenv strnlen strtoull __strtoull')
+    conf.CHECK_FUNCS('strtouq strtoll __strtoll strtoq memalign posix_memalign')
+    conf.CHECK_FUNCS('prctl')
+
+    # libbsd on some platforms provides strlcpy and strlcat
+    if not conf.CHECK_FUNCS('strlcpy strlcat'):
+        conf.CHECK_FUNCS_IN('strlcpy strlcat', 'bsd', headers='bsd/string.h',
+                checklibc=True)
+    if not conf.CHECK_FUNCS('getpeereid'):
+        conf.CHECK_FUNCS_IN('getpeereid', 'bsd', headers='sys/types.h bsd/unistd.h')
+    if not conf.CHECK_FUNCS_IN('setproctitle', 'bsd', headers='sys/types.h bsd/unistd.h'):
+        conf.CHECK_FUNCS_IN('setproctitle', 'setproctitle', headers='setproctitle.h')
+
+    conf.CHECK_CODE('''
+                struct ucred cred;
+                socklen_t cred_len;
+                int ret = getsockopt(0, SOL_SOCKET, SO_PEERCRED, &cred, &cred_len);''',
+                'HAVE_PEERCRED',
+                msg="Checking whether we can use SO_PEERCRED to get socket credentials",
+                headers='sys/types.h sys/socket.h')
+
+    #Some OS (ie. freebsd) return EINVAL if the convertion could not be done, it's not what we expect
+    #Let's detect those cases
+    if conf.CONFIG_SET('HAVE_STRTOLL'):
+        conf.CHECK_CODE('''
+                        long long nb = strtoll("Text", NULL, 0);
+                        if (errno == EINVAL) {
+                            return 0;
+                        } else {
+                            return 1;
+                        }
+                        ''',
+                        msg="Checking correct behavior of strtoll",
+                        headers = 'errno.h',
+                        execute = True,
+                        define = 'HAVE_BSD_STRTOLL',
+                        )
+    conf.CHECK_FUNCS('if_nametoindex strerror_r')
+    conf.CHECK_FUNCS('getdirentries getdents syslog')
+    conf.CHECK_FUNCS('gai_strerror get_current_dir_name')
+    conf.CHECK_FUNCS('timegm getifaddrs freeifaddrs mmap setgroups syscall setsid')
+    conf.CHECK_FUNCS('getgrent_r getgrgid_r getgrnam_r getgrouplist getpagesize')
+    conf.CHECK_FUNCS('getpwent_r getpwnam_r getpwuid_r epoll_create')
+
+    conf.SET_TARGET_TYPE('attr', 'EMPTY')
+
+    xattr_headers='sys/attributes.h attr/xattr.h sys/xattr.h'
+
+    conf.CHECK_FUNCS_IN('''
+fgetxattr flistea flistxattr
+fremovexattr fsetxattr getxattr
+listxattr removexattr setxattr
+''', 'attr', checklibc=True, headers=xattr_headers)
+
+    # We need to check for linux xattrs first, as we do not wish to link to -lattr
+    # (the XFS compat API) on Linux systems with the native xattr API
+    if not conf.CONFIG_SET('HAVE_GETXATTR'):
+        conf.CHECK_FUNCS_IN('''
+attr_get attr_getf attr_list attr_listf attropen attr_remove
+attr_removef attr_set attr_setf extattr_delete_fd extattr_delete_file
+extattr_get_fd extattr_get_file extattr_list_fd extattr_list_file
+extattr_set_fd extattr_set_file fgetea
+fremoveea fsetea getea listea
+removeea setea
+''', 'attr', checklibc=True, headers=xattr_headers)
+
+    if (conf.CONFIG_SET('HAVE_ATTR_LISTF') or
+        conf.CONFIG_SET('HAVE_EXTATTR_LIST_FD') or
+        conf.CONFIG_SET('HAVE_FLISTEA') or
+        conf.CONFIG_SET('HAVE_FLISTXATTR')):
+            conf.DEFINE('HAVE_XATTR_SUPPORT', 1)
+
+    # Darwin has extra options to xattr-family functions
+    conf.CHECK_CODE('getxattr(NULL, NULL, NULL, 0, 0, 0)',
+                    headers=xattr_headers, local_include=False,
+                    define='XATTR_ADDITIONAL_OPTIONS',
+                    msg="Checking whether xattr interface takes additional options")
+
+    conf.CHECK_FUNCS_IN('dlopen dlsym dlerror dlclose', 'dl',
+                        checklibc=True, headers='dlfcn.h dl.h')
+
+    conf.CHECK_C_PROTOTYPE('dlopen', 'void *dlopen(const char* filename, unsigned int flags)',
+                           define='DLOPEN_TAKES_UNSIGNED_FLAGS', headers='dlfcn.h dl.h')
+
+    if conf.CHECK_FUNCS_IN('fdatasync', 'rt', checklibc=True):
+        # some systems are missing the declaration
+        conf.CHECK_DECLS('fdatasync')
+
+    if conf.CHECK_FUNCS_IN('clock_gettime', 'rt', checklibc=True):
+        for c in ['CLOCK_MONOTONIC', 'CLOCK_PROCESS_CPUTIME_ID', 'CLOCK_REALTIME']:
+            conf.CHECK_CODE('''
+                #if TIME_WITH_SYS_TIME
+                # include <sys/time.h>
+                # include <time.h>
+                #else
+                # if HAVE_SYS_TIME_H
+                #  include <sys/time.h>
+                # else
+                #  include <time.h>
+                # endif
+                #endif
+                clockid_t clk = %s''' % c,
+                'HAVE_%s' % c,
+                msg='Checking whether the clock_gettime clock ID %s is available' % c)
+
+    conf.CHECK_TYPE('struct timespec', headers='sys/time.h time.h')
+
+    # these headers need to be tested as a group on freebsd
+    conf.CHECK_HEADERS(headers='sys/socket.h net/if.h', together=True)
+    conf.CHECK_HEADERS(headers='netinet/in.h arpa/nameser.h resolv.h', together=True)
+    conf.CHECK_FUNCS_IN('res_search', 'resolv', checklibc=True,
+                        headers='netinet/in.h arpa/nameser.h resolv.h')
+
+
+    if not conf.CHECK_FUNCS_IN('gettext', 'intl', checklibc=True, headers='libintl.h'):
+    # Some hosts need lib iconv for linking with lib intl
+    # So we try with flags just in case it helps.
+        oldflags = conf.env['LDFLAGS_INTL']
+        conf.env['LDFLAGS_INTL'] = "-liconv"
+        if not conf.CHECK_LIB('intl'):
+            conf.env['LDFLAGS_INTL'] = oldflags
+        else:
+            conf.CHECK_FUNCS_IN('gettext', 'intl', checklibc=True, headers='libintl.h')
+
+    conf.CHECK_FUNCS_IN('dgettext gettext', 'intl', headers='libintl.h')
+    conf.CHECK_FUNCS_IN('pthread_create', 'pthread', checklibc=True, headers='pthread.h')
+
+    conf.CHECK_FUNCS_IN('crypt', 'crypt', checklibc=True)
+
+    conf.CHECK_VARIABLE('rl_event_hook', define='HAVE_DECL_RL_EVENT_HOOK', always=True,
+                        headers='readline.h readline/readline.h readline/history.h')
+
+    conf.CHECK_DECLS('snprintf vsnprintf asprintf vasprintf')
+
+    conf.CHECK_DECLS('errno', headers='errno.h', reverse=True)
+    conf.CHECK_DECLS('environ getgrent_r getpwent_r', reverse=True, headers='pwd.h grp.h')
+    conf.CHECK_DECLS('pread pwrite setenv setresgid setresuid', reverse=True)
+
+    if conf.CONFIG_SET('HAVE_EPOLL_CREATE') and conf.CONFIG_SET('HAVE_SYS_EPOLL_H'):
+        conf.DEFINE('HAVE_EPOLL', 1)
+
+    conf.CHECK_HEADERS('poll.h')
+    conf.CHECK_FUNCS('poll')
+
+    conf.CHECK_FUNCS('strptime')
+    conf.CHECK_DECLS('strptime', headers='time.h')
+    conf.CHECK_CODE('''#define LIBREPLACE_CONFIGURE_TEST_STRPTIME
+                       #include "test/strptime.c"''',
+                       define='HAVE_WORKING_STRPTIME',
+                       execute=True,
+                       addmain=False,
+                       msg='Checking for working strptime')
+
+    conf.CHECK_CODE('gettimeofday(NULL, NULL)', 'HAVE_GETTIMEOFDAY_TZ', execute=False)
+
+    conf.CHECK_CODE('#include "test/snprintf.c"',
+                    define="HAVE_C99_VSNPRINTF",
+                    execute=True,
+                    addmain=False,
+                    msg="Checking for C99 vsnprintf")
+
+    conf.CHECK_CODE('#include "test/shared_mmap.c"',
+                    addmain=False, add_headers=False, execute=True,
+                    define='HAVE_SHARED_MMAP',
+                    msg="Checking for HAVE_SHARED_MMAP")
+
+    conf.CHECK_CODE('#include "test/shared_mremap.c"',
+                    addmain=False, add_headers=False, execute=True,
+                    define='HAVE_MREMAP',
+                    msg="Checking for HAVE_MREMAP")
+
+    # OpenBSD (and I've heard HPUX) doesn't sync between mmap and write.
+    # FIXME: Anything other than a 0 or 1 exit code should abort configure!
+    conf.CHECK_CODE('#include "test/incoherent_mmap.c"',
+                    addmain=False, add_headers=False, execute=True,
+                    define='HAVE_INCOHERENT_MMAP',
+                    msg="Checking for HAVE_INCOHERENT_MMAP")
+
+    conf.SAMBA_BUILD_ENV()
+
+    conf.CHECK_CODE('''
+                    typedef struct {unsigned x;} FOOBAR;
+                    #define X_FOOBAR(x) ((FOOBAR) { x })
+                    #define FOO_ONE X_FOOBAR(1)
+                    FOOBAR f = FOO_ONE;
+                    static const struct {
+                        FOOBAR y;
+                    } f2[] = {
+                        {FOO_ONE}
+                    };
+                    static const FOOBAR f3[] = {FOO_ONE};
+                    ''',
+                    define='HAVE_IMMEDIATE_STRUCTURES')
+
+    conf.CHECK_CODE('mkdir("foo",0777)', define='HAVE_MKDIR_MODE', headers='sys/stat.h')
+
+    conf.CHECK_STRUCTURE_MEMBER('struct stat', 'st_mtim.tv_nsec', define='HAVE_STAT_TV_NSEC',
+                                headers='sys/stat.h')
+    # we need the st_rdev test under two names
+    conf.CHECK_STRUCTURE_MEMBER('struct stat', 'st_rdev',
+                                define='HAVE_STRUCT_STAT_ST_RDEV',
+                                headers='sys/stat.h')
+    conf.CHECK_STRUCTURE_MEMBER('struct stat', 'st_rdev', define='HAVE_ST_RDEV',
+                                headers='sys/stat.h')
+    conf.CHECK_STRUCTURE_MEMBER('struct sockaddr_storage', 'ss_family',
+                                headers='sys/socket.h netinet/in.h')
+    conf.CHECK_STRUCTURE_MEMBER('struct sockaddr_storage', '__ss_family',
+                                headers='sys/socket.h netinet/in.h')
+
+
+    if conf.CHECK_STRUCTURE_MEMBER('struct sockaddr', 'sa_len',
+                                   headers='sys/socket.h netinet/in.h',
+                                   define='HAVE_SOCKADDR_SA_LEN'):
+        # the old build system produced both defines
+        conf.DEFINE('HAVE_STRUCT_SOCKADDR_SA_LEN', 1)
+
+    conf.CHECK_STRUCTURE_MEMBER('struct sockaddr_in', 'sin_len',
+                                headers='sys/socket.h netinet/in.h',
+                                define='HAVE_SOCK_SIN_LEN')
+
+    conf.CHECK_CODE('struct sockaddr_un sunaddr; sunaddr.sun_family = AF_UNIX;',
+                    define='HAVE_UNIXSOCKET', headers='sys/socket.h sys/un.h')
+
+
+    conf.CHECK_CODE('''
+                    struct stat st;
+                    char tpl[20]="/tmp/test.XXXXXX";
+                    char tpl2[20]="/tmp/test.XXXXXX";
+                    int fd = mkstemp(tpl);
+                    int fd2 = mkstemp(tpl2);
+                    if (fd == -1) {
+                          if (fd2 != -1) {
+                                  unlink(tpl2);
+                          }
+                          exit(1);
+                    }
+                    if (fd2 == -1) exit(1);
+                    unlink(tpl);
+                    unlink(tpl2);
+                    if (fstat(fd, &st) != 0) exit(1);
+                    if ((st.st_mode & 0777) != 0600) exit(1);
+                    if (strcmp(tpl, "/tmp/test.XXXXXX") == 0) {
+                          exit(1);
+                    }
+                    if (strcmp(tpl, tpl2) == 0) {
+                          exit(1);
+                    }
+                    exit(0);
+                    ''',
+                    define='HAVE_SECURE_MKSTEMP',
+                    execute=True,
+                    mandatory=True) # lets see if we get a mandatory failure for this one
+
+    if conf.CHECK_CFLAGS('-fvisibility=hidden'):
+        conf.env.VISIBILITY_CFLAGS = '-fvisibility=hidden'
+        conf.CHECK_CODE('''void vis_foo1(void) {}
+                           __attribute__((visibility("default"))) void vis_foo2(void) {}''',
+                        cflags=conf.env.VISIBILITY_CFLAGS,
+                        define='HAVE_VISIBILITY_ATTR')
+
+    # look for a method of finding the list of network interfaces
+    for method in ['HAVE_IFACE_GETIFADDRS', 'HAVE_IFACE_AIX', 'HAVE_IFACE_IFCONF', 'HAVE_IFACE_IFREQ']:
+        if conf.CHECK_CODE('''
+                           #define %s 1
+                           #define NO_CONFIG_H 1
+                           #define AUTOCONF_TEST 1
+                           #define SOCKET_WRAPPER_NOT_REPLACE
+                           #include "replace.c"
+                           #include "inet_ntop.c"
+                           #include "snprintf.c"
+                           #include "getifaddrs.c"
+                           #define getifaddrs_test main
+                           #include "test/getifaddrs.c"
+                           ''' % method,
+                           method,
+                           lib='nsl socket',
+                           addmain=False,
+                           execute=True):
+            break
+
+    conf.RECURSE('system')
+    conf.SAMBA_CONFIG_H()
+
+
+REPLACEMENT_FUNCTIONS = {
+    'replace.c': ['ftruncate', 'strlcpy', 'strlcat', 'mktime', 'initgroups',
+                  'memmove', 'strdup', 'setlinebuf', 'vsyslog', 'strnlen',
+                  'strndup', 'waitpid', 'seteuid', 'setegid', 'chroot',
+                  'mkstemp', 'mkdtemp', 'pread', 'pwrite', 'strcasestr',
+                  'strtok_r', 'strtoll', 'strtoull', 'setenv', 'unsetenv',
+                  'utime', 'utimes', 'dup2', 'chown', 'link', 'readlink',
+                  'symlink', 'lchown', 'realpath', 'memmem', 'vdprintf',
+                  'dprintf', 'get_current_dir_name',
+                  'strerror_r', 'clock_gettime'],
+    'timegm.c': ['timegm'],
+    # Note: C99_VSNPRINTF is not a function, but a special condition
+    # for replacement
+    'snprintf.c': ['C99_VSNPRINTF', 'snprintf', 'vsnprintf', 'asprintf', 'vasprintf'],
+    # Note: WORKING_STRPTIME is not a function, but a special condition
+    # for replacement
+    'strptime.c': ['WORKING_STRPTIME', 'strptime'],
+    }
+
+
+def build(bld):
+    bld.RECURSE('buildtools/wafsamba')
+
+    REPLACE_HOSTCC_SOURCE = ''
+
+    for filename, functions in REPLACEMENT_FUNCTIONS.iteritems():
+        for function in functions:
+            if not bld.CONFIG_SET('HAVE_%s' % function.upper()):
+                REPLACE_HOSTCC_SOURCE += ' %s' % filename
+                break
+
+    extra_libs = ''
+    if bld.CONFIG_SET('HAVE_LIBBSD'): extra_libs += ' bsd'
+
+    bld.SAMBA_SUBSYSTEM('LIBREPLACE_HOSTCC',
+        REPLACE_HOSTCC_SOURCE,
+        use_hostcc=True,
+        use_global_deps=False,
+        cflags='-DSOCKET_WRAPPER_DISABLE=1 -DNSS_WRAPPER_DISABLE=1 -DUID_WRAPPER_DISABLE=1 -D_SAMBA_HOSTCC_',
+        group='compiler_libraries',
+        deps = extra_libs
+    )
+
+    REPLACE_SOURCE = REPLACE_HOSTCC_SOURCE
+
+    if not bld.CONFIG_SET('HAVE_CRYPT'):         REPLACE_SOURCE += ' crypt.c'
+    if not bld.CONFIG_SET('HAVE_DLOPEN'):        REPLACE_SOURCE += ' dlfcn.c'
+    if not bld.CONFIG_SET('HAVE_POLL'):          REPLACE_SOURCE += ' poll.c'
+
+    if not bld.CONFIG_SET('HAVE_SOCKETPAIR'):    REPLACE_SOURCE += ' socketpair.c'
+    if not bld.CONFIG_SET('HAVE_CONNECT'):       REPLACE_SOURCE += ' socket.c'
+    if not bld.CONFIG_SET('HAVE_GETIFADDRS'):    REPLACE_SOURCE += ' getifaddrs.c'
+    if not bld.CONFIG_SET('HAVE_GETADDRINFO'):   REPLACE_SOURCE += ' getaddrinfo.c'
+    if not bld.CONFIG_SET('HAVE_INET_NTOA'):     REPLACE_SOURCE += ' inet_ntoa.c'
+    if not bld.CONFIG_SET('HAVE_INET_ATON'):     REPLACE_SOURCE += ' inet_aton.c'
+    if not bld.CONFIG_SET('HAVE_INET_NTOP'):     REPLACE_SOURCE += ' inet_ntop.c'
+    if not bld.CONFIG_SET('HAVE_INET_PTON'):     REPLACE_SOURCE += ' inet_pton.c'
+    if not bld.CONFIG_SET('HAVE_GETXATTR') or bld.CONFIG_SET('XATTR_ADDITIONAL_OPTIONS'):
+                                                 REPLACE_SOURCE += ' xattr.c'
+
+    bld.SAMBA_LIBRARY('replace',
+                      source=REPLACE_SOURCE,
+                      group='base_libraries',
+                      # FIXME: Ideally symbols should be hidden here so they 
+                      # don't appear in the global namespace when Samba 
+                      # libraries are loaded, but this doesn't appear to work 
+                      # at the moment:
+                      # hide_symbols=bld.BUILTIN_LIBRARY('replace'),
+                      private_library=True,
+                      deps='crypt dl nsl socket rt attr' + extra_libs)
+
+    bld.SAMBA_SUBSYSTEM('replace-test',
+                      source='''test/testsuite.c test/strptime.c
+                      test/os2_delete.c test/getifaddrs.c''',
+                      deps='replace')
+
+    if bld.env.standalone_replace:
+        bld.SAMBA_BINARY('replace_testsuite',
+                         source='test/main.c',
+                         deps='replace replace-test',
+                         install=False)
+
+    # build replacements for stdint.h and stdbool.h if needed
+    bld.SAMBA_GENERATOR('replace_stdint_h',
+                        rule='cp ${SRC} ${TGT}',
+                        source='hdr_replace.h',
+                        target='stdint.h',
+                        enabled = not bld.CONFIG_SET('HAVE_STDINT_H'))
+    bld.SAMBA_GENERATOR('replace_stdbool_h',
+                        rule='cp ${SRC} ${TGT}',
+                        source='hdr_replace.h',
+                        target='stdbool.h',
+                        enabled = not bld.CONFIG_SET('HAVE_STDBOOL_H'))
+
+def dist():
+    '''makes a tarball for distribution'''
+    samba_dist.dist()
diff --git a/ctdb/lib/replace/xattr.c b/ctdb/lib/replace/xattr.c
new file mode 100644
index 0000000000..a26ff674a1
--- /dev/null
+++ b/ctdb/lib/replace/xattr.c
@@ -0,0 +1,729 @@
+/* 
+   Unix SMB/CIFS implementation.
+   replacement routines for xattr implementations
+   Copyright (C) Jeremy Allison  1998-2005
+   Copyright (C) Timur Bakeyev        2005
+   Copyright (C) Bjoern Jacke    2006-2007
+   Copyright (C) Herb Lewis           2003
+   Copyright (C) Andrew Bartlett      2012
+
+     ** NOTE! The following LGPL license applies to the replace
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "replace.h"
+#include "system/filesys.h"
+#include "system/dir.h"
+
+/******** Solaris EA helper function prototypes ********/
+#ifdef HAVE_ATTROPEN
+#define SOLARIS_ATTRMODE S_IRUSR|S_IWUSR|S_IRGRP|S_IWGRP
+static int solaris_write_xattr(int attrfd, const char *value, size_t size);
+static ssize_t solaris_read_xattr(int attrfd, void *value, size_t size);
+static ssize_t solaris_list_xattr(int attrdirfd, char *list, size_t size);
+static int solaris_unlinkat(int attrdirfd, const char *name);
+static int solaris_attropen(const char *path, const char *attrpath, int oflag, mode_t mode);
+static int solaris_openat(int fildes, const char *path, int oflag, mode_t mode);
+#endif
+
+/**************************************************************************
+ Wrappers for extented attribute calls. Based on the Linux package with
+ support for IRIX and (Net|Free)BSD also. Expand as other systems have them.
+****************************************************************************/
+
+ssize_t rep_getxattr (const char *path, const char *name, void *value, size_t size)
+{
+#if defined(HAVE_GETXATTR)
+#ifndef XATTR_ADDITIONAL_OPTIONS
+	return getxattr(path, name, value, size);
+#else
+
+/* So that we do not recursivly call this function */
+#undef getxattr
+	int options = 0;
+	return getxattr(path, name, value, size, 0, options);
+#endif
+#elif defined(HAVE_GETEA)
+	return getea(path, name, value, size);
+#elif defined(HAVE_EXTATTR_GET_FILE)
+	char *s;
+	ssize_t retval;
+	int attrnamespace = (strncmp(name, "system", 6) == 0) ? 
+		EXTATTR_NAMESPACE_SYSTEM : EXTATTR_NAMESPACE_USER;
+	const char *attrname = ((s=strchr(name, '.')) == NULL) ? name : s + 1;
+	/*
+	 * The BSD implementation has a nasty habit of silently truncating
+	 * the returned value to the size of the buffer, so we have to check
+	 * that the buffer is large enough to fit the returned value.
+	 */
+	if((retval=extattr_get_file(path, attrnamespace, attrname, NULL, 0)) >= 0) {
+		if (size == 0) {
+			return retval;
+		} else if (retval > size) {
+			errno = ERANGE;
+			return -1;
+		}
+		if((retval=extattr_get_file(path, attrnamespace, attrname, value, size)) >= 0)
+			return retval;
+	}
+
+	return -1;
+#elif defined(HAVE_ATTR_GET)
+	int retval, flags = 0;
+	int valuelength = (int)size;
+	char *attrname = strchr(name,'.') + 1;
+
+	if (strncmp(name, "system", 6) == 0) flags |= ATTR_ROOT;
+
+	retval = attr_get(path, attrname, (char *)value, &valuelength, flags);
+	if (size == 0 && retval == -1 && errno == E2BIG) {
+		return valuelength;
+	}
+
+	return retval ? retval : valuelength;
+#elif defined(HAVE_ATTROPEN)
+	ssize_t ret = -1;
+	int attrfd = solaris_attropen(path, name, O_RDONLY, 0);
+	if (attrfd >= 0) {
+		ret = solaris_read_xattr(attrfd, value, size);
+		close(attrfd);
+	}
+	return ret;
+#else
+	errno = ENOSYS;
+	return -1;
+#endif
+}
+
+ssize_t rep_fgetxattr (int filedes, const char *name, void *value, size_t size)
+{
+#if defined(HAVE_FGETXATTR)
+#ifndef XATTR_ADDITIONAL_OPTIONS
+	return fgetxattr(filedes, name, value, size);
+#else
+
+/* So that we do not recursivly call this function */
+#undef fgetxattr
+	int options = 0;
+	return fgetxattr(filedes, name, value, size, 0, options);
+#endif
+#elif defined(HAVE_FGETEA)
+	return fgetea(filedes, name, value, size);
+#elif defined(HAVE_EXTATTR_GET_FD)
+	char *s;
+	ssize_t retval;
+	int attrnamespace = (strncmp(name, "system", 6) == 0) ? 
+		EXTATTR_NAMESPACE_SYSTEM : EXTATTR_NAMESPACE_USER;
+	const char *attrname = ((s=strchr(name, '.')) == NULL) ? name : s + 1;
+
+	if((retval=extattr_get_fd(filedes, attrnamespace, attrname, NULL, 0)) >= 0) {
+		if (size == 0) {
+			return retval;
+		} else if (retval > size) {
+			errno = ERANGE;
+			return -1;
+		}
+		if((retval=extattr_get_fd(filedes, attrnamespace, attrname, value, size)) >= 0)
+			return retval;
+	}
+
+	return -1;
+#elif defined(HAVE_ATTR_GETF)
+	int retval, flags = 0;
+	int valuelength = (int)size;
+	char *attrname = strchr(name,'.') + 1;
+
+	if (strncmp(name, "system", 6) == 0) flags |= ATTR_ROOT;
+
+	retval = attr_getf(filedes, attrname, (char *)value, &valuelength, flags);
+	if (size == 0 && retval == -1 && errno == E2BIG) {
+		return valuelength;
+	}
+	return retval ? retval : valuelength;
+#elif defined(HAVE_ATTROPEN)
+	ssize_t ret = -1;
+	int attrfd = solaris_openat(filedes, name, O_RDONLY|O_XATTR, 0);
+	if (attrfd >= 0) {
+		ret = solaris_read_xattr(attrfd, value, size);
+		close(attrfd);
+	}
+	return ret;
+#else
+	errno = ENOSYS;
+	return -1;
+#endif
+}
+
+#if defined(HAVE_EXTATTR_LIST_FILE)
+
+#define EXTATTR_PREFIX(s)	(s), (sizeof((s))-1)
+
+static struct {
+        int space;
+	const char *name;
+	size_t len;
+} 
+extattr[] = {
+	{ EXTATTR_NAMESPACE_SYSTEM, EXTATTR_PREFIX("system.") },
+        { EXTATTR_NAMESPACE_USER, EXTATTR_PREFIX("user.") },
+};
+
+typedef union {
+	const char *path;
+	int filedes;
+} extattr_arg;
+
+static ssize_t bsd_attr_list (int type, extattr_arg arg, char *list, size_t size)
+{
+	ssize_t list_size, total_size = 0;
+	int i, t, len;
+	char *buf;
+	/* Iterate through extattr(2) namespaces */
+	for(t = 0; t < ARRAY_SIZE(extattr); t++) {
+		switch(type) {
+#if defined(HAVE_EXTATTR_LIST_FILE)
+			case 0:
+				list_size = extattr_list_file(arg.path, extattr[t].space, list, size);
+				break;
+#endif
+#if defined(HAVE_EXTATTR_LIST_LINK)
+			case 1:
+				list_size = extattr_list_link(arg.path, extattr[t].space, list, size);
+				break;
+#endif
+#if defined(HAVE_EXTATTR_LIST_FD)
+			case 2:
+				list_size = extattr_list_fd(arg.filedes, extattr[t].space, list, size);
+				break;
+#endif
+			default:
+				errno = ENOSYS;
+				return -1;
+		}
+		/* Some error happend. Errno should be set by the previous call */
+		if(list_size < 0)
+			return -1;
+		/* No attributes */
+		if(list_size == 0)
+			continue;
+		/* XXX: Call with an empty buffer may be used to calculate
+		   necessary buffer size. Unfortunately, we can't say, how
+		   many attributes were returned, so here is the potential
+		   problem with the emulation.
+		*/
+		if(list == NULL) {
+			/* Take the worse case of one char attribute names - 
+			   two bytes per name plus one more for sanity.
+			*/
+			total_size += list_size + (list_size/2 + 1)*extattr[t].len;
+			continue;
+		}
+		/* Count necessary offset to fit namespace prefixes */
+		len = 0;
+		for(i = 0; i < list_size; i += list[i] + 1)
+			len += extattr[t].len;
+
+		total_size += list_size + len;
+		/* Buffer is too small to fit the results */
+		if(total_size > size) {
+			errno = ERANGE;
+			return -1;
+		}
+		/* Shift results back, so we can prepend prefixes */
+		buf = (char *)memmove(list + len, list, list_size);
+
+		for(i = 0; i < list_size; i += len + 1) {
+			len = buf[i];
+			strncpy(list, extattr[t].name, extattr[t].len + 1);
+			list += extattr[t].len;
+			strncpy(list, buf + i + 1, len);
+			list[len] = '\0';
+			list += len + 1;
+		}
+		size -= total_size;
+	}
+	return total_size;
+}
+
+#endif
+
+#if defined(HAVE_ATTR_LIST) && (defined(HAVE_SYS_ATTRIBUTES_H) || defined(HAVE_ATTR_ATTRIBUTES_H))
+static char attr_buffer[ATTR_MAX_VALUELEN];
+
+static ssize_t irix_attr_list(const char *path, int filedes, char *list, size_t size, int flags)
+{
+	int retval = 0, index;
+	attrlist_cursor_t *cursor = 0;
+	int total_size = 0;
+	attrlist_t * al = (attrlist_t *)attr_buffer;
+	attrlist_ent_t *ae;
+	size_t ent_size, left = size;
+	char *bp = list;
+
+	while (true) {
+	    if (filedes)
+		retval = attr_listf(filedes, attr_buffer, ATTR_MAX_VALUELEN, flags, cursor);
+	    else
+		retval = attr_list(path, attr_buffer, ATTR_MAX_VALUELEN, flags, cursor);
+	    if (retval) break;
+	    for (index = 0; index < al->al_count; index++) {
+		ae = ATTR_ENTRY(attr_buffer, index);
+		ent_size = strlen(ae->a_name) + sizeof("user.");
+		if (left >= ent_size) {
+		    strncpy(bp, "user.", sizeof("user."));
+		    strncat(bp, ae->a_name, ent_size - sizeof("user."));
+		    bp += ent_size;
+		    left -= ent_size;
+		} else if (size) {
+		    errno = ERANGE;
+		    retval = -1;
+		    break;
+		}
+		total_size += ent_size;
+	    }
+	    if (al->al_more == 0) break;
+	}
+	if (retval == 0) {
+	    flags |= ATTR_ROOT;
+	    cursor = 0;
+	    while (true) {
+		if (filedes)
+		    retval = attr_listf(filedes, attr_buffer, ATTR_MAX_VALUELEN, flags, cursor);
+		else
+		    retval = attr_list(path, attr_buffer, ATTR_MAX_VALUELEN, flags, cursor);
+		if (retval) break;
+		for (index = 0; index < al->al_count; index++) {
+		    ae = ATTR_ENTRY(attr_buffer, index);
+		    ent_size = strlen(ae->a_name) + sizeof("system.");
+		    if (left >= ent_size) {
+			strncpy(bp, "system.", sizeof("system."));
+			strncat(bp, ae->a_name, ent_size - sizeof("system."));
+			bp += ent_size;
+			left -= ent_size;
+		    } else if (size) {
+			errno = ERANGE;
+			retval = -1;
+			break;
+		    }
+		    total_size += ent_size;
+		}
+		if (al->al_more == 0) break;
+	    }
+	}
+	return (ssize_t)(retval ? retval : total_size);
+}
+
+#endif
+
+ssize_t rep_listxattr (const char *path, char *list, size_t size)
+{
+#if defined(HAVE_LISTXATTR)
+#ifndef XATTR_ADDITIONAL_OPTIONS
+	return listxattr(path, list, size);
+#else
+/* So that we do not recursivly call this function */
+#undef listxattr
+	int options = 0;
+	return listxattr(path, list, size, options);
+#endif
+#elif defined(HAVE_LISTEA)
+	return listea(path, list, size);
+#elif defined(HAVE_EXTATTR_LIST_FILE)
+	extattr_arg arg;
+	arg.path = path;
+	return bsd_attr_list(0, arg, list, size);
+#elif defined(HAVE_ATTR_LIST) && defined(HAVE_SYS_ATTRIBUTES_H)
+	return irix_attr_list(path, 0, list, size, 0);
+#elif defined(HAVE_ATTROPEN)
+	ssize_t ret = -1;
+	int attrdirfd = solaris_attropen(path, ".", O_RDONLY, 0);
+	if (attrdirfd >= 0) {
+		ret = solaris_list_xattr(attrdirfd, list, size);
+		close(attrdirfd);
+	}
+	return ret;
+#else
+	errno = ENOSYS;
+	return -1;
+#endif
+}
+
+ssize_t rep_flistxattr (int filedes, char *list, size_t size)
+{
+#if defined(HAVE_FLISTXATTR)
+#ifndef XATTR_ADDITIONAL_OPTIONS
+	return flistxattr(filedes, list, size);
+#else
+/* So that we do not recursivly call this function */
+#undef flistxattr
+	int options = 0;
+	return flistxattr(filedes, list, size, options);
+#endif
+#elif defined(HAVE_FLISTEA)
+	return flistea(filedes, list, size);
+#elif defined(HAVE_EXTATTR_LIST_FD)
+	extattr_arg arg;
+	arg.filedes = filedes;
+	return bsd_attr_list(2, arg, list, size);
+#elif defined(HAVE_ATTR_LISTF)
+	return irix_attr_list(NULL, filedes, list, size, 0);
+#elif defined(HAVE_ATTROPEN)
+	ssize_t ret = -1;
+	int attrdirfd = solaris_openat(filedes, ".", O_RDONLY|O_XATTR, 0);
+	if (attrdirfd >= 0) {
+		ret = solaris_list_xattr(attrdirfd, list, size);
+		close(attrdirfd);
+	}
+	return ret;
+#else
+	errno = ENOSYS;
+	return -1;
+#endif
+}
+
+int rep_removexattr (const char *path, const char *name)
+{
+#if defined(HAVE_REMOVEXATTR)
+#ifndef XATTR_ADDITIONAL_OPTIONS
+	return removexattr(path, name);
+#else
+/* So that we do not recursivly call this function */
+#undef removexattr
+	int options = 0;
+	return removexattr(path, name, options);
+#endif
+#elif defined(HAVE_REMOVEEA)
+	return removeea(path, name);
+#elif defined(HAVE_EXTATTR_DELETE_FILE)
+	char *s;
+	int attrnamespace = (strncmp(name, "system", 6) == 0) ? 
+		EXTATTR_NAMESPACE_SYSTEM : EXTATTR_NAMESPACE_USER;
+	const char *attrname = ((s=strchr(name, '.')) == NULL) ? name : s + 1;
+
+	return extattr_delete_file(path, attrnamespace, attrname);
+#elif defined(HAVE_ATTR_REMOVE)
+	int flags = 0;
+	char *attrname = strchr(name,'.') + 1;
+
+	if (strncmp(name, "system", 6) == 0) flags |= ATTR_ROOT;
+
+	return attr_remove(path, attrname, flags);
+#elif defined(HAVE_ATTROPEN)
+	int ret = -1;
+	int attrdirfd = solaris_attropen(path, ".", O_RDONLY, 0);
+	if (attrdirfd >= 0) {
+		ret = solaris_unlinkat(attrdirfd, name);
+		close(attrdirfd);
+	}
+	return ret;
+#else
+	errno = ENOSYS;
+	return -1;
+#endif
+}
+
+int rep_fremovexattr (int filedes, const char *name)
+{
+#if defined(HAVE_FREMOVEXATTR)
+#ifndef XATTR_ADDITIONAL_OPTIONS
+	return fremovexattr(filedes, name);
+#else
+/* So that we do not recursivly call this function */
+#undef fremovexattr
+	int options = 0;
+	return fremovexattr(filedes, name, options);
+#endif
+#elif defined(HAVE_FREMOVEEA)
+	return fremoveea(filedes, name);
+#elif defined(HAVE_EXTATTR_DELETE_FD)
+	char *s;
+	int attrnamespace = (strncmp(name, "system", 6) == 0) ? 
+		EXTATTR_NAMESPACE_SYSTEM : EXTATTR_NAMESPACE_USER;
+	const char *attrname = ((s=strchr(name, '.')) == NULL) ? name : s + 1;
+
+	return extattr_delete_fd(filedes, attrnamespace, attrname);
+#elif defined(HAVE_ATTR_REMOVEF)
+	int flags = 0;
+	char *attrname = strchr(name,'.') + 1;
+
+	if (strncmp(name, "system", 6) == 0) flags |= ATTR_ROOT;
+
+	return attr_removef(filedes, attrname, flags);
+#elif defined(HAVE_ATTROPEN)
+	int ret = -1;
+	int attrdirfd = solaris_openat(filedes, ".", O_RDONLY|O_XATTR, 0);
+	if (attrdirfd >= 0) {
+		ret = solaris_unlinkat(attrdirfd, name);
+		close(attrdirfd);
+	}
+	return ret;
+#else
+	errno = ENOSYS;
+	return -1;
+#endif
+}
+
+int rep_setxattr (const char *path, const char *name, const void *value, size_t size, int flags)
+{
+#if defined(HAVE_SETXATTR)
+#ifndef XATTR_ADDITIONAL_OPTIONS
+	return setxattr(path, name, value, size, flags);
+#else
+/* So that we do not recursivly call this function */
+#undef setxattr
+	int options = 0;
+	return setxattr(path, name, value, size, 0, options);
+#endif
+#elif defined(HAVE_SETEA)
+	return setea(path, name, value, size, flags);
+#elif defined(HAVE_EXTATTR_SET_FILE)
+	char *s;
+	int retval = 0;
+	int attrnamespace = (strncmp(name, "system", 6) == 0) ? 
+		EXTATTR_NAMESPACE_SYSTEM : EXTATTR_NAMESPACE_USER;
+	const char *attrname = ((s=strchr(name, '.')) == NULL) ? name : s + 1;
+	if (flags) {
+		/* Check attribute existence */
+		retval = extattr_get_file(path, attrnamespace, attrname, NULL, 0);
+		if (retval < 0) {
+			/* REPLACE attribute, that doesn't exist */
+			if (flags & XATTR_REPLACE && errno == ENOATTR) {
+				errno = ENOATTR;
+				return -1;
+			}
+			/* Ignore other errors */
+		}
+		else {
+			/* CREATE attribute, that already exists */
+			if (flags & XATTR_CREATE) {
+				errno = EEXIST;
+				return -1;
+			}
+		}
+	}
+	retval = extattr_set_file(path, attrnamespace, attrname, value, size);
+	return (retval < 0) ? -1 : 0;
+#elif defined(HAVE_ATTR_SET)
+	int myflags = 0;
+	char *attrname = strchr(name,'.') + 1;
+
+	if (strncmp(name, "system", 6) == 0) myflags |= ATTR_ROOT;
+	if (flags & XATTR_CREATE) myflags |= ATTR_CREATE;
+	if (flags & XATTR_REPLACE) myflags |= ATTR_REPLACE;
+
+	return attr_set(path, attrname, (const char *)value, size, myflags);
+#elif defined(HAVE_ATTROPEN)
+	int ret = -1;
+	int myflags = O_RDWR;
+	int attrfd;
+	if (flags & XATTR_CREATE) myflags |= O_EXCL;
+	if (!(flags & XATTR_REPLACE)) myflags |= O_CREAT;
+	attrfd = solaris_attropen(path, name, myflags, (mode_t) SOLARIS_ATTRMODE);
+	if (attrfd >= 0) {
+		ret = solaris_write_xattr(attrfd, value, size);
+		close(attrfd);
+	}
+	return ret;
+#else
+	errno = ENOSYS;
+	return -1;
+#endif
+}
+
+int rep_fsetxattr (int filedes, const char *name, const void *value, size_t size, int flags)
+{
+#if defined(HAVE_FSETXATTR)
+#ifndef XATTR_ADDITIONAL_OPTIONS
+	return fsetxattr(filedes, name, value, size, flags);
+#else
+/* So that we do not recursivly call this function */
+#undef fsetxattr
+	int options = 0;
+	return fsetxattr(filedes, name, value, size, 0, options);
+#endif
+#elif defined(HAVE_FSETEA)
+	return fsetea(filedes, name, value, size, flags);
+#elif defined(HAVE_EXTATTR_SET_FD)
+	char *s;
+	int retval = 0;
+	int attrnamespace = (strncmp(name, "system", 6) == 0) ? 
+		EXTATTR_NAMESPACE_SYSTEM : EXTATTR_NAMESPACE_USER;
+	const char *attrname = ((s=strchr(name, '.')) == NULL) ? name : s + 1;
+	if (flags) {
+		/* Check attribute existence */
+		retval = extattr_get_fd(filedes, attrnamespace, attrname, NULL, 0);
+		if (retval < 0) {
+			/* REPLACE attribute, that doesn't exist */
+			if (flags & XATTR_REPLACE && errno == ENOATTR) {
+				errno = ENOATTR;
+				return -1;
+			}
+			/* Ignore other errors */
+		}
+		else {
+			/* CREATE attribute, that already exists */
+			if (flags & XATTR_CREATE) {
+				errno = EEXIST;
+				return -1;
+			}
+		}
+	}
+	retval = extattr_set_fd(filedes, attrnamespace, attrname, value, size);
+	return (retval < 0) ? -1 : 0;
+#elif defined(HAVE_ATTR_SETF)
+	int myflags = 0;
+	char *attrname = strchr(name,'.') + 1;
+
+	if (strncmp(name, "system", 6) == 0) myflags |= ATTR_ROOT;
+	if (flags & XATTR_CREATE) myflags |= ATTR_CREATE;
+	if (flags & XATTR_REPLACE) myflags |= ATTR_REPLACE;
+
+	return attr_setf(filedes, attrname, (const char *)value, size, myflags);
+#elif defined(HAVE_ATTROPEN)
+	int ret = -1;
+	int myflags = O_RDWR | O_XATTR;
+	int attrfd;
+	if (flags & XATTR_CREATE) myflags |= O_EXCL;
+	if (!(flags & XATTR_REPLACE)) myflags |= O_CREAT;
+	attrfd = solaris_openat(filedes, name, myflags, (mode_t) SOLARIS_ATTRMODE);
+	if (attrfd >= 0) {
+		ret = solaris_write_xattr(attrfd, value, size);
+		close(attrfd);
+	}
+	return ret;
+#else
+	errno = ENOSYS;
+	return -1;
+#endif
+}
+
+/**************************************************************************
+ helper functions for Solaris' EA support
+****************************************************************************/
+#ifdef HAVE_ATTROPEN
+static ssize_t solaris_read_xattr(int attrfd, void *value, size_t size)
+{
+	struct stat sbuf;
+
+	if (fstat(attrfd, &sbuf) == -1) {
+		errno = ENOATTR;
+		return -1;
+	}
+
+	/* This is to return the current size of the named extended attribute */
+	if (size == 0) {
+		return sbuf.st_size;
+	}
+
+	/* check size and read xattr */
+	if (sbuf.st_size > size) {
+		errno = ERANGE;
+		return -1;
+	}
+
+	return read(attrfd, value, sbuf.st_size);
+}
+
+static ssize_t solaris_list_xattr(int attrdirfd, char *list, size_t size)
+{
+	ssize_t len = 0;
+	DIR *dirp;
+	struct dirent *de;
+	int newfd = dup(attrdirfd);
+	/* CAUTION: The originating file descriptor should not be
+	            used again following the call to fdopendir().
+	            For that reason we dup() the file descriptor
+		    here to make things more clear. */
+	dirp = fdopendir(newfd);
+
+	while ((de = readdir(dirp))) {
+		size_t listlen = strlen(de->d_name) + 1;
+		if (!strcmp(de->d_name, ".") || !strcmp(de->d_name, "..")) {
+			/* we don't want "." and ".." here: */
+			continue;
+		}
+
+		if (size == 0) {
+			/* return the current size of the list of extended attribute names*/
+			len += listlen;
+		} else {
+			/* check size and copy entrieѕ + nul into list. */
+			if ((len + listlen) > size) {
+				errno = ERANGE;
+				len = -1;
+				break;
+			} else {
+				strlcpy(list + len, de->d_name, listlen);
+				len += listlen;
+			}
+		}
+	}
+
+	if (closedir(dirp) == -1) {
+		return -1;
+	}
+	return len;
+}
+
+static int solaris_unlinkat(int attrdirfd, const char *name)
+{
+	if (unlinkat(attrdirfd, name, 0) == -1) {
+		if (errno == ENOENT) {
+			errno = ENOATTR;
+		}
+		return -1;
+	}
+	return 0;
+}
+
+static int solaris_attropen(const char *path, const char *attrpath, int oflag, mode_t mode)
+{
+	int filedes = attropen(path, attrpath, oflag, mode);
+	if (filedes == -1) {
+		if (errno == EINVAL) {
+			errno = ENOTSUP;
+		} else {
+			errno = ENOATTR;
+		}
+	}
+	return filedes;
+}
+
+static int solaris_openat(int fildes, const char *path, int oflag, mode_t mode)
+{
+	int filedes = openat(fildes, path, oflag, mode);
+	if (filedes == -1) {
+		if (errno == EINVAL) {
+			errno = ENOTSUP;
+		} else {
+			errno = ENOATTR;
+		}
+	}
+	return filedes;
+}
+
+static int solaris_write_xattr(int attrfd, const char *value, size_t size)
+{
+	if ((ftruncate(attrfd, 0) == 0) && (write(attrfd, value, size) == size)) {
+		return 0;
+	} else {
+		return -1;
+	}
+}
+#endif /*HAVE_ATTROPEN*/
+
+
diff --git a/ctdb/lib/socket_wrapper/config.m4 b/ctdb/lib/socket_wrapper/config.m4
new file mode 100644
index 0000000000..6db392fce4
--- /dev/null
+++ b/ctdb/lib/socket_wrapper/config.m4
@@ -0,0 +1,21 @@
+AC_ARG_ENABLE(socket-wrapper, 
+AS_HELP_STRING([--enable-socket-wrapper], [Turn on socket wrapper library (default=no)]))
+
+DEFAULT_TEST_OPTIONS=
+HAVE_SOCKET_WRAPPER=no
+
+if eval "test x$developer = xyes"; then
+	enable_socket_wrapper=yes
+fi
+    
+if eval "test x$enable_socket_wrapper = xyes"; then
+        AC_DEFINE(SOCKET_WRAPPER,1,[Use socket wrapper library])
+	DEFAULT_TEST_OPTIONS=--socket-wrapper
+	HAVE_SOCKET_WRAPPER=yes
+
+	SOCKET_WRAPPER_OBJS="lib/socket_wrapper/socket_wrapper.o"
+fi
+
+AC_SUBST(DEFAULT_TEST_OPTIONS)
+AC_SUBST(HAVE_SOCKET_WRAPPER)
+AC_SUBST(SOCKET_WRAPPER_OBJS)
diff --git a/ctdb/lib/socket_wrapper/socket_wrapper.c b/ctdb/lib/socket_wrapper/socket_wrapper.c
new file mode 100644
index 0000000000..2c24ab7924
--- /dev/null
+++ b/ctdb/lib/socket_wrapper/socket_wrapper.c
@@ -0,0 +1,2654 @@
+/*
+ * Copyright (C) Jelmer Vernooij 2005,2008 <jelmer@samba.org>
+ * Copyright (C) Stefan Metzmacher 2006-2009 <metze@samba.org>
+ *
+ * All rights reserved.
+ * 
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ * 
+ * 3. Neither the name of the author nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ * 
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ *
+ */
+
+/*
+   Socket wrapper library. Passes all socket communication over
+   unix domain sockets if the environment variable SOCKET_WRAPPER_DIR
+   is set.
+*/
+
+#include "config.h"
+
+#ifdef HAVE_LIBREPLACE
+
+#define SOCKET_WRAPPER_NOT_REPLACE
+#include "replace.h"
+#include "system/network.h"
+#include "system/filesys.h"
+#include "system/time.h"
+
+#else /* HAVE_LIBREPLACE */
+
+#include <sys/types.h>
+#include <sys/time.h>
+#include <sys/stat.h>
+#include <sys/socket.h>
+#include <sys/ioctl.h>
+#include <sys/filio.h>
+#include <errno.h>
+#include <sys/un.h>
+#include <netinet/in.h>
+#include <netinet/tcp.h>
+#include <fcntl.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <string.h>
+#include <stdio.h>
+#include <stdint.h>
+
+#endif /* HAVE_LIBREPLACE */
+
+#ifndef _PUBLIC_
+#define _PUBLIC_
+#endif
+
+#define SWRAP_DLIST_ADD(list,item) do { \
+	if (!(list)) { \
+		(item)->prev	= NULL; \
+		(item)->next	= NULL; \
+		(list)		= (item); \
+	} else { \
+		(item)->prev	= NULL; \
+		(item)->next	= (list); \
+		(list)->prev	= (item); \
+		(list)		= (item); \
+	} \
+} while (0)
+
+#define SWRAP_DLIST_REMOVE(list,item) do { \
+	if ((list) == (item)) { \
+		(list)		= (item)->next; \
+		if (list) { \
+			(list)->prev	= NULL; \
+		} \
+	} else { \
+		if ((item)->prev) { \
+			(item)->prev->next	= (item)->next; \
+		} \
+		if ((item)->next) { \
+			(item)->next->prev	= (item)->prev; \
+		} \
+	} \
+	(item)->prev	= NULL; \
+	(item)->next	= NULL; \
+} while (0)
+
+/* LD_PRELOAD doesn't work yet, so REWRITE_CALLS is all we support
+ * for now */
+#define REWRITE_CALLS 
+
+#ifdef REWRITE_CALLS
+#define real_accept accept
+#define real_connect connect
+#define real_bind bind
+#define real_listen listen
+#define real_getpeername getpeername
+#define real_getsockname getsockname
+#define real_getsockopt getsockopt
+#define real_setsockopt setsockopt
+#define real_recvfrom recvfrom
+#define real_sendto sendto
+#define real_sendmsg sendmsg
+#define real_ioctl ioctl
+#define real_recv recv
+#define real_read read
+#define real_send send
+#define real_readv readv
+#define real_writev writev
+#define real_socket socket
+#define real_close close
+#define real_dup dup
+#define real_dup2 dup2
+#endif
+
+#ifdef HAVE_GETTIMEOFDAY_TZ
+#define swrapGetTimeOfDay(tval) gettimeofday(tval,NULL)
+#else
+#define swrapGetTimeOfDay(tval)	gettimeofday(tval)
+#endif
+
+/* we need to use a very terse format here as IRIX 6.4 silently
+   truncates names to 16 chars, so if we use a longer name then we
+   can't tell which port a packet came from with recvfrom() 
+   
+   with this format we have 8 chars left for the directory name
+*/
+#define SOCKET_FORMAT "%c%02X%04X"
+#define SOCKET_TYPE_CHAR_TCP		'T'
+#define SOCKET_TYPE_CHAR_UDP		'U'
+#define SOCKET_TYPE_CHAR_TCP_V6		'X'
+#define SOCKET_TYPE_CHAR_UDP_V6		'Y'
+
+/* This limit is to avoid broadcast sendto() needing to stat too many
+ * files.  It may be raised (with a performance cost) to up to 254
+ * without changing the format above */
+#define MAX_WRAPPED_INTERFACES 32
+
+#ifdef HAVE_IPV6
+/*
+ * FD00::5357:5FXX
+ */
+static const struct in6_addr *swrap_ipv6(void)
+{
+	static struct in6_addr v;
+	static int initialized;
+	int ret;
+
+	if (initialized) {
+		return &v;
+	}
+	initialized = 1;
+
+	ret = inet_pton(AF_INET6, "FD00::5357:5F00", &v);
+	if (ret <= 0) {
+		abort();
+	}
+
+	return &v;
+}
+#endif
+
+static struct sockaddr *sockaddr_dup(const void *data, socklen_t len)
+{
+	struct sockaddr *ret = (struct sockaddr *)malloc(len);
+	memcpy(ret, data, len);
+	return ret;
+}
+
+static void set_port(int family, int prt, struct sockaddr *addr)
+{
+	switch (family) {
+	case AF_INET:
+		((struct sockaddr_in *)addr)->sin_port = htons(prt);
+		break;
+#ifdef HAVE_IPV6
+	case AF_INET6:
+		((struct sockaddr_in6 *)addr)->sin6_port = htons(prt);
+		break;
+#endif
+	}
+}
+
+static size_t socket_length(int family)
+{
+	switch (family) {
+	case AF_INET:
+		return sizeof(struct sockaddr_in);
+#ifdef HAVE_IPV6
+	case AF_INET6:
+		return sizeof(struct sockaddr_in6);
+#endif
+	}
+	return 0;
+}
+
+struct socket_info_fd {
+	struct socket_info_fd *prev, *next;
+	int fd;
+};
+
+struct socket_info
+{
+	struct socket_info_fd *fds;
+
+	int family;
+	int type;
+	int protocol;
+	int bound;
+	int bcast;
+	int is_server;
+	int connected;
+	int defer_connect;
+
+	char *tmp_path;
+
+	struct sockaddr *myname;
+	socklen_t myname_len;
+
+	struct sockaddr *peername;
+	socklen_t peername_len;
+
+	struct {
+		unsigned long pck_snd;
+		unsigned long pck_rcv;
+	} io;
+
+	struct socket_info *prev, *next;
+};
+
+static struct socket_info *sockets;
+
+const char *socket_wrapper_dir(void)
+{
+	const char *s = getenv("SOCKET_WRAPPER_DIR");
+	if (s == NULL) {
+		return NULL;
+	}
+	if (strncmp(s, "./", 2) == 0) {
+		s += 2;
+	}
+	return s;
+}
+
+unsigned int socket_wrapper_default_iface(void)
+{
+	const char *s = getenv("SOCKET_WRAPPER_DEFAULT_IFACE");
+	if (s) {
+		unsigned int iface;
+		if (sscanf(s, "%u", &iface) == 1) {
+			if (iface >= 1 && iface <= MAX_WRAPPED_INTERFACES) {
+				return iface;
+			}
+		}
+	}
+
+	return 1;/* 127.0.0.1 */
+}
+
+static int convert_un_in(const struct sockaddr_un *un, struct sockaddr *in, socklen_t *len)
+{
+	unsigned int iface;
+	unsigned int prt;
+	const char *p;
+	char type;
+
+	p = strrchr(un->sun_path, '/');
+	if (p) p++; else p = un->sun_path;
+
+	if (sscanf(p, SOCKET_FORMAT, &type, &iface, &prt) != 3) {
+		errno = EINVAL;
+		return -1;
+	}
+
+	if (iface == 0 || iface > MAX_WRAPPED_INTERFACES) {
+		errno = EINVAL;
+		return -1;
+	}
+
+	if (prt > 0xFFFF) {
+		errno = EINVAL;
+		return -1;
+	}
+
+	switch(type) {
+	case SOCKET_TYPE_CHAR_TCP:
+	case SOCKET_TYPE_CHAR_UDP: {
+		struct sockaddr_in *in2 = (struct sockaddr_in *)(void *)in;
+
+		if ((*len) < sizeof(*in2)) {
+		    errno = EINVAL;
+		    return -1;
+		}
+
+		memset(in2, 0, sizeof(*in2));
+		in2->sin_family = AF_INET;
+		in2->sin_addr.s_addr = htonl((127<<24) | iface);
+		in2->sin_port = htons(prt);
+
+		*len = sizeof(*in2);
+		break;
+	}
+#ifdef HAVE_IPV6
+	case SOCKET_TYPE_CHAR_TCP_V6:
+	case SOCKET_TYPE_CHAR_UDP_V6: {
+		struct sockaddr_in6 *in2 = (struct sockaddr_in6 *)(void *)in;
+
+		if ((*len) < sizeof(*in2)) {
+			errno = EINVAL;
+			return -1;
+		}
+
+		memset(in2, 0, sizeof(*in2));
+		in2->sin6_family = AF_INET6;
+		in2->sin6_addr = *swrap_ipv6();
+		in2->sin6_addr.s6_addr[15] = iface;
+		in2->sin6_port = htons(prt);
+
+		*len = sizeof(*in2);
+		break;
+	}
+#endif
+	default:
+		errno = EINVAL;
+		return -1;
+	}
+
+	return 0;
+}
+
+static int convert_in_un_remote(struct socket_info *si, const struct sockaddr *inaddr, struct sockaddr_un *un,
+				int *bcast)
+{
+	char type = '\0';
+	unsigned int prt;
+	unsigned int iface;
+	int is_bcast = 0;
+
+	if (bcast) *bcast = 0;
+
+	switch (inaddr->sa_family) {
+	case AF_INET: {
+		const struct sockaddr_in *in = 
+		    (const struct sockaddr_in *)(const void *)inaddr;
+		unsigned int addr = ntohl(in->sin_addr.s_addr);
+		char u_type = '\0';
+		char b_type = '\0';
+		char a_type = '\0';
+
+		switch (si->type) {
+		case SOCK_STREAM:
+			u_type = SOCKET_TYPE_CHAR_TCP;
+			break;
+		case SOCK_DGRAM:
+			u_type = SOCKET_TYPE_CHAR_UDP;
+			a_type = SOCKET_TYPE_CHAR_UDP;
+			b_type = SOCKET_TYPE_CHAR_UDP;
+			break;
+		}
+
+		prt = ntohs(in->sin_port);
+		if (a_type && addr == 0xFFFFFFFF) {
+			/* 255.255.255.255 only udp */
+			is_bcast = 2;
+			type = a_type;
+			iface = socket_wrapper_default_iface();
+		} else if (b_type && addr == 0x7FFFFFFF) {
+			/* 127.255.255.255 only udp */
+			is_bcast = 1;
+			type = b_type;
+			iface = socket_wrapper_default_iface();
+		} else if ((addr & 0xFFFFFF00) == 0x7F000000) {
+			/* 127.0.0.X */
+			is_bcast = 0;
+			type = u_type;
+			iface = (addr & 0x000000FF);
+		} else {
+			errno = ENETUNREACH;
+			return -1;
+		}
+		if (bcast) *bcast = is_bcast;
+		break;
+	}
+#ifdef HAVE_IPV6
+	case AF_INET6: {
+		const struct sockaddr_in6 *in = 
+		    (const struct sockaddr_in6 *)(const void *)inaddr;
+		struct in6_addr cmp1, cmp2;
+
+		switch (si->type) {
+		case SOCK_STREAM:
+			type = SOCKET_TYPE_CHAR_TCP_V6;
+			break;
+		case SOCK_DGRAM:
+			type = SOCKET_TYPE_CHAR_UDP_V6;
+			break;
+		}
+
+		/* XXX no multicast/broadcast */
+
+		prt = ntohs(in->sin6_port);
+
+		cmp1 = *swrap_ipv6();
+		cmp2 = in->sin6_addr;
+		cmp2.s6_addr[15] = 0;
+		if (IN6_ARE_ADDR_EQUAL(&cmp1, &cmp2)) {
+			iface = in->sin6_addr.s6_addr[15];
+		} else {
+			errno = ENETUNREACH;
+			return -1;
+		}
+
+		break;
+	}
+#endif
+	default:
+		errno = ENETUNREACH;
+		return -1;
+	}
+
+	if (prt == 0) {
+		errno = EINVAL;
+		return -1;
+	}
+
+	if (is_bcast) {
+		snprintf(un->sun_path, sizeof(un->sun_path), "%s/EINVAL", 
+			 socket_wrapper_dir());
+		/* the caller need to do more processing */
+		return 0;
+	}
+
+	snprintf(un->sun_path, sizeof(un->sun_path), "%s/"SOCKET_FORMAT, 
+		 socket_wrapper_dir(), type, iface, prt);
+
+	return 0;
+}
+
+static int convert_in_un_alloc(struct socket_info *si, const struct sockaddr *inaddr, struct sockaddr_un *un,
+			       int *bcast)
+{
+	char type = '\0';
+	unsigned int prt;
+	unsigned int iface;
+	struct stat st;
+	int is_bcast = 0;
+
+	if (bcast) *bcast = 0;
+
+	switch (si->family) {
+	case AF_INET: {
+		const struct sockaddr_in *in = 
+		    (const struct sockaddr_in *)(const void *)inaddr;
+		unsigned int addr = ntohl(in->sin_addr.s_addr);
+		char u_type = '\0';
+		char d_type = '\0';
+		char b_type = '\0';
+		char a_type = '\0';
+
+		prt = ntohs(in->sin_port);
+
+		switch (si->type) {
+		case SOCK_STREAM:
+			u_type = SOCKET_TYPE_CHAR_TCP;
+			d_type = SOCKET_TYPE_CHAR_TCP;
+			break;
+		case SOCK_DGRAM:
+			u_type = SOCKET_TYPE_CHAR_UDP;
+			d_type = SOCKET_TYPE_CHAR_UDP;
+			a_type = SOCKET_TYPE_CHAR_UDP;
+			b_type = SOCKET_TYPE_CHAR_UDP;
+			break;
+		}
+
+		if (addr == 0) {
+			/* 0.0.0.0 */
+		 	is_bcast = 0;
+			type = d_type;
+			iface = socket_wrapper_default_iface();
+		} else if (a_type && addr == 0xFFFFFFFF) {
+			/* 255.255.255.255 only udp */
+			is_bcast = 2;
+			type = a_type;
+			iface = socket_wrapper_default_iface();
+		} else if (b_type && addr == 0x7FFFFFFF) {
+			/* 127.255.255.255 only udp */
+			is_bcast = 1;
+			type = b_type;
+			iface = socket_wrapper_default_iface();
+		} else if ((addr & 0xFFFFFF00) == 0x7F000000) {
+			/* 127.0.0.X */
+			is_bcast = 0;
+			type = u_type;
+			iface = (addr & 0x000000FF);
+		} else {
+			errno = EADDRNOTAVAIL;
+			return -1;
+		}
+		break;
+	}
+#ifdef HAVE_IPV6
+	case AF_INET6: {
+		const struct sockaddr_in6 *in = 
+		    (const struct sockaddr_in6 *)(const void *)inaddr;
+		struct in6_addr cmp1, cmp2;
+
+		switch (si->type) {
+		case SOCK_STREAM:
+			type = SOCKET_TYPE_CHAR_TCP_V6;
+			break;
+		case SOCK_DGRAM:
+			type = SOCKET_TYPE_CHAR_UDP_V6;
+			break;
+		}
+
+		/* XXX no multicast/broadcast */
+
+		prt = ntohs(in->sin6_port);
+
+		cmp1 = *swrap_ipv6();
+		cmp2 = in->sin6_addr;
+		cmp2.s6_addr[15] = 0;
+		if (IN6_IS_ADDR_UNSPECIFIED(&in->sin6_addr)) {
+			iface = socket_wrapper_default_iface();
+		} else if (IN6_ARE_ADDR_EQUAL(&cmp1, &cmp2)) {
+			iface = in->sin6_addr.s6_addr[15];
+		} else {
+			errno = EADDRNOTAVAIL;
+			return -1;
+		}
+
+		break;
+	}
+#endif
+	default:
+		errno = EADDRNOTAVAIL;
+		return -1;
+	}
+
+
+	if (bcast) *bcast = is_bcast;
+
+	if (iface == 0 || iface > MAX_WRAPPED_INTERFACES) {
+		errno = EINVAL;
+		return -1;
+	}
+
+	if (prt == 0) {
+		/* handle auto-allocation of ephemeral ports */
+		for (prt = 5001; prt < 10000; prt++) {
+			snprintf(un->sun_path, sizeof(un->sun_path), "%s/"SOCKET_FORMAT, 
+				 socket_wrapper_dir(), type, iface, prt);
+			if (stat(un->sun_path, &st) == 0) continue;
+
+			set_port(si->family, prt, si->myname);
+			break;
+		}
+		if (prt == 10000) {
+			errno = ENFILE;
+			return -1;
+		}
+	}
+
+	snprintf(un->sun_path, sizeof(un->sun_path), "%s/"SOCKET_FORMAT, 
+		 socket_wrapper_dir(), type, iface, prt);
+	return 0;
+}
+
+static struct socket_info *find_socket_info(int fd)
+{
+	struct socket_info *i;
+	for (i = sockets; i; i = i->next) {
+		struct socket_info_fd *f;
+		for (f = i->fds; f; f = f->next) {
+			if (f->fd == fd) {
+				return i;
+			}
+		}
+	}
+
+	return NULL;
+}
+
+static int sockaddr_convert_to_un(struct socket_info *si, const struct sockaddr *in_addr, socklen_t in_len, 
+				  struct sockaddr_un *out_addr, int alloc_sock, int *bcast)
+{
+	struct sockaddr *out = (struct sockaddr *)(void *)out_addr;
+	if (!out_addr)
+		return 0;
+
+	out->sa_family = AF_UNIX;
+#ifdef HAVE_STRUCT_SOCKADDR_SA_LEN
+	out->sa_len = sizeof(*out_addr);
+#endif
+
+	switch (in_addr->sa_family) {
+	case AF_INET:
+#ifdef HAVE_IPV6
+	case AF_INET6:
+#endif
+		switch (si->type) {
+		case SOCK_STREAM:
+		case SOCK_DGRAM:
+			break;
+		default:
+			errno = ESOCKTNOSUPPORT;
+			return -1;
+		}
+		if (alloc_sock) {
+			return convert_in_un_alloc(si, in_addr, out_addr, bcast);
+		} else {
+			return convert_in_un_remote(si, in_addr, out_addr, bcast);
+		}
+	default:
+		break;
+	}
+
+	errno = EAFNOSUPPORT;
+	return -1;
+}
+
+static int sockaddr_convert_from_un(const struct socket_info *si, 
+				    const struct sockaddr_un *in_addr, 
+				    socklen_t un_addrlen,
+				    int family,
+				    struct sockaddr *out_addr,
+				    socklen_t *out_addrlen)
+{
+	int ret;
+
+	if (out_addr == NULL || out_addrlen == NULL) 
+		return 0;
+
+	if (un_addrlen == 0) {
+		*out_addrlen = 0;
+		return 0;
+	}
+
+	switch (family) {
+	case AF_INET:
+#ifdef HAVE_IPV6
+	case AF_INET6:
+#endif
+		switch (si->type) {
+		case SOCK_STREAM:
+		case SOCK_DGRAM:
+			break;
+		default:
+			errno = ESOCKTNOSUPPORT;
+			return -1;
+		}
+		ret = convert_un_in(in_addr, out_addr, out_addrlen);
+#ifdef HAVE_STRUCT_SOCKADDR_SA_LEN
+		out_addr->sa_len = *out_addrlen;
+#endif
+		return ret;
+	default:
+		break;
+	}
+
+	errno = EAFNOSUPPORT;
+	return -1;
+}
+
+enum swrap_packet_type {
+	SWRAP_CONNECT_SEND,
+	SWRAP_CONNECT_UNREACH,
+	SWRAP_CONNECT_RECV,
+	SWRAP_CONNECT_ACK,
+	SWRAP_ACCEPT_SEND,
+	SWRAP_ACCEPT_RECV,
+	SWRAP_ACCEPT_ACK,
+	SWRAP_RECVFROM,
+	SWRAP_SENDTO,
+	SWRAP_SENDTO_UNREACH,
+	SWRAP_PENDING_RST,
+	SWRAP_RECV,
+	SWRAP_RECV_RST,
+	SWRAP_SEND,
+	SWRAP_SEND_RST,
+	SWRAP_CLOSE_SEND,
+	SWRAP_CLOSE_RECV,
+	SWRAP_CLOSE_ACK,
+};
+
+struct swrap_file_hdr {
+	uint32_t	magic;
+	uint16_t	version_major;
+	uint16_t	version_minor;
+	int32_t		timezone;
+	uint32_t	sigfigs;
+	uint32_t	frame_max_len;
+#define SWRAP_FRAME_LENGTH_MAX 0xFFFF
+	uint32_t	link_type;
+};
+#define SWRAP_FILE_HDR_SIZE 24
+
+struct swrap_packet_frame {
+	uint32_t seconds;
+	uint32_t micro_seconds;
+	uint32_t recorded_length;
+	uint32_t full_length;
+};
+#define SWRAP_PACKET_FRAME_SIZE 16
+
+union swrap_packet_ip {
+	struct {
+		uint8_t		ver_hdrlen;
+		uint8_t		tos;
+		uint16_t	packet_length;
+		uint16_t	identification;
+		uint8_t		flags;
+		uint8_t		fragment;
+		uint8_t		ttl;
+		uint8_t		protocol;
+		uint16_t	hdr_checksum;
+		uint32_t	src_addr;
+		uint32_t	dest_addr;
+	} v4;
+#define SWRAP_PACKET_IP_V4_SIZE 20
+	struct {
+		uint8_t		ver_prio;
+		uint8_t		flow_label_high;
+		uint16_t	flow_label_low;
+		uint16_t	payload_length;
+		uint8_t		next_header;
+		uint8_t		hop_limit;
+		uint8_t		src_addr[16];
+		uint8_t		dest_addr[16];
+	} v6;
+#define SWRAP_PACKET_IP_V6_SIZE 40
+};
+#define SWRAP_PACKET_IP_SIZE 40
+
+union swrap_packet_payload {
+	struct {
+		uint16_t	source_port;
+		uint16_t	dest_port;
+		uint32_t	seq_num;
+		uint32_t	ack_num;
+		uint8_t		hdr_length;
+		uint8_t		control;
+		uint16_t	window;
+		uint16_t	checksum;
+		uint16_t	urg;
+	} tcp;
+#define SWRAP_PACKET_PAYLOAD_TCP_SIZE 20
+	struct {
+		uint16_t	source_port;
+		uint16_t	dest_port;
+		uint16_t	length;
+		uint16_t	checksum;
+	} udp;
+#define SWRAP_PACKET_PAYLOAD_UDP_SIZE 8
+	struct {
+		uint8_t		type;
+		uint8_t		code;
+		uint16_t	checksum;
+		uint32_t	unused;
+	} icmp4;
+#define SWRAP_PACKET_PAYLOAD_ICMP4_SIZE 8
+	struct {
+		uint8_t		type;
+		uint8_t		code;
+		uint16_t	checksum;
+		uint32_t	unused;
+	} icmp6;
+#define SWRAP_PACKET_PAYLOAD_ICMP6_SIZE 8
+};
+#define SWRAP_PACKET_PAYLOAD_SIZE 20
+
+#define SWRAP_PACKET_MIN_ALLOC \
+	(SWRAP_PACKET_FRAME_SIZE + \
+	 SWRAP_PACKET_IP_SIZE + \
+	 SWRAP_PACKET_PAYLOAD_SIZE)
+
+static const char *socket_wrapper_pcap_file(void)
+{
+	static int initialized = 0;
+	static const char *s = NULL;
+	static const struct swrap_file_hdr h;
+	static const struct swrap_packet_frame f;
+	static const union swrap_packet_ip i;
+	static const union swrap_packet_payload p;
+
+	if (initialized == 1) {
+		return s;
+	}
+	initialized = 1;
+
+	/*
+	 * TODO: don't use the structs use plain buffer offsets
+	 *       and PUSH_U8(), PUSH_U16() and PUSH_U32()
+	 * 
+	 * for now make sure we disable PCAP support
+	 * if the struct has alignment!
+	 */
+	if (sizeof(h) != SWRAP_FILE_HDR_SIZE) {
+		return NULL;
+	}
+	if (sizeof(f) != SWRAP_PACKET_FRAME_SIZE) {
+		return NULL;
+	}
+	if (sizeof(i) != SWRAP_PACKET_IP_SIZE) {
+		return NULL;
+	}
+	if (sizeof(i.v4) != SWRAP_PACKET_IP_V4_SIZE) {
+		return NULL;
+	}
+	if (sizeof(i.v6) != SWRAP_PACKET_IP_V6_SIZE) {
+		return NULL;
+	}
+	if (sizeof(p) != SWRAP_PACKET_PAYLOAD_SIZE) {
+		return NULL;
+	}
+	if (sizeof(p.tcp) != SWRAP_PACKET_PAYLOAD_TCP_SIZE) {
+		return NULL;
+	}
+	if (sizeof(p.udp) != SWRAP_PACKET_PAYLOAD_UDP_SIZE) {
+		return NULL;
+	}
+	if (sizeof(p.icmp4) != SWRAP_PACKET_PAYLOAD_ICMP4_SIZE) {
+		return NULL;
+	}
+	if (sizeof(p.icmp6) != SWRAP_PACKET_PAYLOAD_ICMP6_SIZE) {
+		return NULL;
+	}
+
+	s = getenv("SOCKET_WRAPPER_PCAP_FILE");
+	if (s == NULL) {
+		return NULL;
+	}
+	if (strncmp(s, "./", 2) == 0) {
+		s += 2;
+	}
+	return s;
+}
+
+static uint8_t *swrap_packet_init(struct timeval *tval,
+				  const struct sockaddr *src,
+				  const struct sockaddr *dest,
+				  int socket_type,
+				  const uint8_t *payload,
+				  size_t payload_len,
+				  unsigned long tcp_seqno,
+				  unsigned long tcp_ack,
+				  unsigned char tcp_ctl,
+				  int unreachable,
+				  size_t *_packet_len)
+{
+	uint8_t *base;
+	uint8_t *buf;
+	struct swrap_packet_frame *frame;
+	union swrap_packet_ip *ip;
+	union swrap_packet_payload *pay;
+	size_t packet_len;
+	size_t alloc_len;
+	size_t nonwire_len = sizeof(*frame);
+	size_t wire_hdr_len = 0;
+	size_t wire_len = 0;
+	size_t ip_hdr_len = 0;
+	size_t icmp_hdr_len = 0;
+	size_t icmp_truncate_len = 0;
+	uint8_t protocol = 0, icmp_protocol = 0;
+	const struct sockaddr_in *src_in = NULL;
+	const struct sockaddr_in *dest_in = NULL;
+#ifdef HAVE_IPV6
+	const struct sockaddr_in6 *src_in6 = NULL;
+	const struct sockaddr_in6 *dest_in6 = NULL;
+#endif
+	uint16_t src_port;
+	uint16_t dest_port;
+
+	switch (src->sa_family) {
+	case AF_INET:
+		src_in = (const struct sockaddr_in *)src;
+		dest_in = (const struct sockaddr_in *)dest;
+		src_port = src_in->sin_port;
+		dest_port = dest_in->sin_port;
+		ip_hdr_len = sizeof(ip->v4);
+		break;
+#ifdef HAVE_IPV6
+	case AF_INET6:
+		src_in6 = (const struct sockaddr_in6 *)src;
+		dest_in6 = (const struct sockaddr_in6 *)dest;
+		src_port = src_in6->sin6_port;
+		dest_port = dest_in6->sin6_port;
+		ip_hdr_len = sizeof(ip->v6);
+		break;
+#endif
+	default:
+		return NULL;
+	}
+
+	switch (socket_type) {
+	case SOCK_STREAM:
+		protocol = 0x06; /* TCP */
+		wire_hdr_len = ip_hdr_len + sizeof(pay->tcp);
+		wire_len = wire_hdr_len + payload_len;
+		break;
+
+	case SOCK_DGRAM:
+		protocol = 0x11; /* UDP */
+		wire_hdr_len = ip_hdr_len + sizeof(pay->udp);
+		wire_len = wire_hdr_len + payload_len;
+		break;
+
+	default:
+		return NULL;
+	}
+
+	if (unreachable) {
+		icmp_protocol = protocol;
+		switch (src->sa_family) {
+		case AF_INET:
+			protocol = 0x01; /* ICMPv4 */
+			icmp_hdr_len = ip_hdr_len + sizeof(pay->icmp4);
+			break;
+#ifdef HAVE_IPV6
+		case AF_INET6:
+			protocol = 0x3A; /* ICMPv6 */
+			icmp_hdr_len = ip_hdr_len + sizeof(pay->icmp6);
+			break;
+#endif
+		}
+		if (wire_len > 64 ) {
+			icmp_truncate_len = wire_len - 64;
+		}
+		wire_hdr_len += icmp_hdr_len;
+		wire_len += icmp_hdr_len;
+	}
+
+	packet_len = nonwire_len + wire_len;
+	alloc_len = packet_len;
+	if (alloc_len < SWRAP_PACKET_MIN_ALLOC) {
+		alloc_len = SWRAP_PACKET_MIN_ALLOC;
+	}
+
+	base = (uint8_t *)malloc(alloc_len);
+	if (!base) return NULL;
+
+	buf = base;
+
+	frame = (struct swrap_packet_frame *)buf;
+	frame->seconds		= tval->tv_sec;
+	frame->micro_seconds	= tval->tv_usec;
+	frame->recorded_length	= wire_len - icmp_truncate_len;
+	frame->full_length	= wire_len - icmp_truncate_len;
+	buf += SWRAP_PACKET_FRAME_SIZE;
+
+	ip = (union swrap_packet_ip *)buf;
+	switch (src->sa_family) {
+	case AF_INET:
+		ip->v4.ver_hdrlen	= 0x45; /* version 4 and 5 * 32 bit words */
+		ip->v4.tos		= 0x00;
+		ip->v4.packet_length	= htons(wire_len - icmp_truncate_len);
+		ip->v4.identification	= htons(0xFFFF);
+		ip->v4.flags		= 0x40; /* BIT 1 set - means don't fraqment */
+		ip->v4.fragment		= htons(0x0000);
+		ip->v4.ttl		= 0xFF;
+		ip->v4.protocol		= protocol;
+		ip->v4.hdr_checksum	= htons(0x0000);
+		ip->v4.src_addr		= src_in->sin_addr.s_addr;
+		ip->v4.dest_addr	= dest_in->sin_addr.s_addr;
+		buf += SWRAP_PACKET_IP_V4_SIZE;
+		break;
+#ifdef HAVE_IPV6
+	case AF_INET6:
+		ip->v6.ver_prio		= 0x60; /* version 4 and 5 * 32 bit words */
+		ip->v6.flow_label_high	= 0x00;
+		ip->v6.flow_label_low	= 0x0000;
+		ip->v6.payload_length	= htons(wire_len - icmp_truncate_len); /* TODO */
+		ip->v6.next_header	= protocol;
+		memcpy(ip->v6.src_addr, src_in6->sin6_addr.s6_addr, 16);
+		memcpy(ip->v6.dest_addr, dest_in6->sin6_addr.s6_addr, 16);
+		buf += SWRAP_PACKET_IP_V6_SIZE;
+		break;
+#endif
+	}
+
+	if (unreachable) {
+		pay = (union swrap_packet_payload *)buf;
+		switch (src->sa_family) {
+		case AF_INET:
+			pay->icmp4.type		= 0x03; /* destination unreachable */
+			pay->icmp4.code		= 0x01; /* host unreachable */
+			pay->icmp4.checksum	= htons(0x0000);
+			pay->icmp4.unused	= htonl(0x00000000);
+			buf += SWRAP_PACKET_PAYLOAD_ICMP4_SIZE;
+
+			/* set the ip header in the ICMP payload */
+			ip = (union swrap_packet_ip *)buf;
+			ip->v4.ver_hdrlen	= 0x45; /* version 4 and 5 * 32 bit words */
+			ip->v4.tos		= 0x00;
+			ip->v4.packet_length	= htons(wire_len - icmp_hdr_len);
+			ip->v4.identification	= htons(0xFFFF);
+			ip->v4.flags		= 0x40; /* BIT 1 set - means don't fraqment */
+			ip->v4.fragment		= htons(0x0000);
+			ip->v4.ttl		= 0xFF;
+			ip->v4.protocol		= icmp_protocol;
+			ip->v4.hdr_checksum	= htons(0x0000);
+			ip->v4.src_addr		= dest_in->sin_addr.s_addr;
+			ip->v4.dest_addr	= src_in->sin_addr.s_addr;
+			buf += SWRAP_PACKET_IP_V4_SIZE;
+
+			src_port = dest_in->sin_port;
+			dest_port = src_in->sin_port;
+			break;
+#ifdef HAVE_IPV6
+		case AF_INET6:
+			pay->icmp6.type		= 0x01; /* destination unreachable */
+			pay->icmp6.code		= 0x03; /* address unreachable */
+			pay->icmp6.checksum	= htons(0x0000);
+			pay->icmp6.unused	= htonl(0x00000000);
+			buf += SWRAP_PACKET_PAYLOAD_ICMP6_SIZE;
+
+			/* set the ip header in the ICMP payload */
+			ip = (union swrap_packet_ip *)buf;
+			ip->v6.ver_prio		= 0x60; /* version 4 and 5 * 32 bit words */
+			ip->v6.flow_label_high	= 0x00;
+			ip->v6.flow_label_low	= 0x0000;
+			ip->v6.payload_length	= htons(wire_len - icmp_truncate_len); /* TODO */
+			ip->v6.next_header	= protocol;
+			memcpy(ip->v6.src_addr, dest_in6->sin6_addr.s6_addr, 16);
+			memcpy(ip->v6.dest_addr, src_in6->sin6_addr.s6_addr, 16);
+			buf += SWRAP_PACKET_IP_V6_SIZE;
+
+			src_port = dest_in6->sin6_port;
+			dest_port = src_in6->sin6_port;
+			break;
+#endif
+		}
+	}
+
+	pay = (union swrap_packet_payload *)buf;
+
+	switch (socket_type) {
+	case SOCK_STREAM:
+		pay->tcp.source_port	= src_port;
+		pay->tcp.dest_port	= dest_port;
+		pay->tcp.seq_num	= htonl(tcp_seqno);
+		pay->tcp.ack_num	= htonl(tcp_ack);
+		pay->tcp.hdr_length	= 0x50; /* 5 * 32 bit words */
+		pay->tcp.control	= tcp_ctl;
+		pay->tcp.window		= htons(0x7FFF);
+		pay->tcp.checksum	= htons(0x0000);
+		pay->tcp.urg		= htons(0x0000);
+		buf += SWRAP_PACKET_PAYLOAD_TCP_SIZE;
+
+		break;
+
+	case SOCK_DGRAM:
+		pay->udp.source_port	= src_port;
+		pay->udp.dest_port	= dest_port;
+		pay->udp.length		= htons(8 + payload_len);
+		pay->udp.checksum	= htons(0x0000);
+		buf += SWRAP_PACKET_PAYLOAD_UDP_SIZE;
+
+		break;
+	}
+
+	if (payload && payload_len > 0) {
+		memcpy(buf, payload, payload_len);
+	}
+
+	*_packet_len = packet_len - icmp_truncate_len;
+	return base;
+}
+
+static int swrap_get_pcap_fd(const char *fname)
+{
+	static int fd = -1;
+
+	if (fd != -1) return fd;
+
+	fd = open(fname, O_WRONLY|O_CREAT|O_EXCL|O_APPEND, 0644);
+	if (fd != -1) {
+		struct swrap_file_hdr file_hdr;
+		file_hdr.magic		= 0xA1B2C3D4;
+		file_hdr.version_major	= 0x0002;	
+		file_hdr.version_minor	= 0x0004;
+		file_hdr.timezone	= 0x00000000;
+		file_hdr.sigfigs	= 0x00000000;
+		file_hdr.frame_max_len	= SWRAP_FRAME_LENGTH_MAX;
+		file_hdr.link_type	= 0x0065; /* 101 RAW IP */
+
+		if (write(fd, &file_hdr, sizeof(file_hdr)) != sizeof(file_hdr)) {
+			close(fd);
+			fd = -1;
+		}
+		return fd;
+	}
+
+	fd = open(fname, O_WRONLY|O_APPEND, 0644);
+
+	return fd;
+}
+
+static uint8_t *swrap_marshall_packet(struct socket_info *si,
+				      const struct sockaddr *addr,
+				      enum swrap_packet_type type,
+				      const void *buf, size_t len,
+				      size_t *packet_len)
+{
+	const struct sockaddr *src_addr;
+	const struct sockaddr *dest_addr;
+	unsigned long tcp_seqno = 0;
+	unsigned long tcp_ack = 0;
+	unsigned char tcp_ctl = 0;
+	int unreachable = 0;
+
+	struct timeval tv;
+
+	switch (si->family) {
+	case AF_INET:
+		break;
+#ifdef HAVE_IPV6
+	case AF_INET6:
+		break;
+#endif
+	default:
+		return NULL;
+	}
+
+	switch (type) {
+	case SWRAP_CONNECT_SEND:
+		if (si->type != SOCK_STREAM) return NULL;
+
+		src_addr = si->myname;
+		dest_addr = addr;
+
+		tcp_seqno = si->io.pck_snd;
+		tcp_ack = si->io.pck_rcv;
+		tcp_ctl = 0x02; /* SYN */
+
+		si->io.pck_snd += 1;
+
+		break;
+
+	case SWRAP_CONNECT_RECV:
+		if (si->type != SOCK_STREAM) return NULL;
+
+		dest_addr = si->myname;
+		src_addr = addr;
+
+		tcp_seqno = si->io.pck_rcv;
+		tcp_ack = si->io.pck_snd;
+		tcp_ctl = 0x12; /** SYN,ACK */
+
+		si->io.pck_rcv += 1;
+
+		break;
+
+	case SWRAP_CONNECT_UNREACH:
+		if (si->type != SOCK_STREAM) return NULL;
+
+		dest_addr = si->myname;
+		src_addr = addr;
+
+		/* Unreachable: resend the data of SWRAP_CONNECT_SEND */
+		tcp_seqno = si->io.pck_snd - 1;
+		tcp_ack = si->io.pck_rcv;
+		tcp_ctl = 0x02; /* SYN */
+		unreachable = 1;
+
+		break;
+
+	case SWRAP_CONNECT_ACK:
+		if (si->type != SOCK_STREAM) return NULL;
+
+		src_addr = si->myname;
+		dest_addr = addr;
+
+		tcp_seqno = si->io.pck_snd;
+		tcp_ack = si->io.pck_rcv;
+		tcp_ctl = 0x10; /* ACK */
+
+		break;
+
+	case SWRAP_ACCEPT_SEND:
+		if (si->type != SOCK_STREAM) return NULL;
+
+		dest_addr = si->myname;
+		src_addr = addr;
+
+		tcp_seqno = si->io.pck_rcv;
+		tcp_ack = si->io.pck_snd;
+		tcp_ctl = 0x02; /* SYN */
+
+		si->io.pck_rcv += 1;
+
+		break;
+
+	case SWRAP_ACCEPT_RECV:
+		if (si->type != SOCK_STREAM) return NULL;
+
+		src_addr = si->myname;
+		dest_addr = addr;
+
+		tcp_seqno = si->io.pck_snd;
+		tcp_ack = si->io.pck_rcv;
+		tcp_ctl = 0x12; /* SYN,ACK */
+
+		si->io.pck_snd += 1;
+
+		break;
+
+	case SWRAP_ACCEPT_ACK:
+		if (si->type != SOCK_STREAM) return NULL;
+
+		dest_addr = si->myname;
+		src_addr = addr;
+
+		tcp_seqno = si->io.pck_rcv;
+		tcp_ack = si->io.pck_snd;
+		tcp_ctl = 0x10; /* ACK */
+
+		break;
+
+	case SWRAP_SEND:
+		src_addr = si->myname;
+		dest_addr = si->peername;
+
+		tcp_seqno = si->io.pck_snd;
+		tcp_ack = si->io.pck_rcv;
+		tcp_ctl = 0x18; /* PSH,ACK */
+
+		si->io.pck_snd += len;
+
+		break;
+
+	case SWRAP_SEND_RST:
+		dest_addr = si->myname;
+		src_addr = si->peername;
+
+		if (si->type == SOCK_DGRAM) {
+			return swrap_marshall_packet(si, si->peername,
+					  SWRAP_SENDTO_UNREACH,
+			      		  buf, len, packet_len);
+		}
+
+		tcp_seqno = si->io.pck_rcv;
+		tcp_ack = si->io.pck_snd;
+		tcp_ctl = 0x14; /** RST,ACK */
+
+		break;
+
+	case SWRAP_PENDING_RST:
+		dest_addr = si->myname;
+		src_addr = si->peername;
+
+		if (si->type == SOCK_DGRAM) {
+			return NULL;
+		}
+
+		tcp_seqno = si->io.pck_rcv;
+		tcp_ack = si->io.pck_snd;
+		tcp_ctl = 0x14; /* RST,ACK */
+
+		break;
+
+	case SWRAP_RECV:
+		dest_addr = si->myname;
+		src_addr = si->peername;
+
+		tcp_seqno = si->io.pck_rcv;
+		tcp_ack = si->io.pck_snd;
+		tcp_ctl = 0x18; /* PSH,ACK */
+
+		si->io.pck_rcv += len;
+
+		break;
+
+	case SWRAP_RECV_RST:
+		dest_addr = si->myname;
+		src_addr = si->peername;
+
+		if (si->type == SOCK_DGRAM) {
+			return NULL;
+		}
+
+		tcp_seqno = si->io.pck_rcv;
+		tcp_ack = si->io.pck_snd;
+		tcp_ctl = 0x14; /* RST,ACK */
+
+		break;
+
+	case SWRAP_SENDTO:
+		src_addr = si->myname;
+		dest_addr = addr;
+
+		si->io.pck_snd += len;
+
+		break;
+
+	case SWRAP_SENDTO_UNREACH:
+		dest_addr = si->myname;
+		src_addr = addr;
+
+		unreachable = 1;
+
+		break;
+
+	case SWRAP_RECVFROM:
+		dest_addr = si->myname;
+		src_addr = addr;
+
+		si->io.pck_rcv += len;
+
+		break;
+
+	case SWRAP_CLOSE_SEND:
+		if (si->type != SOCK_STREAM) return NULL;
+
+		src_addr = si->myname;
+		dest_addr = si->peername;
+
+		tcp_seqno = si->io.pck_snd;
+		tcp_ack = si->io.pck_rcv;
+		tcp_ctl = 0x11; /* FIN, ACK */
+
+		si->io.pck_snd += 1;
+
+		break;
+
+	case SWRAP_CLOSE_RECV:
+		if (si->type != SOCK_STREAM) return NULL;
+
+		dest_addr = si->myname;
+		src_addr = si->peername;
+
+		tcp_seqno = si->io.pck_rcv;
+		tcp_ack = si->io.pck_snd;
+		tcp_ctl = 0x11; /* FIN,ACK */
+
+		si->io.pck_rcv += 1;
+
+		break;
+
+	case SWRAP_CLOSE_ACK:
+		if (si->type != SOCK_STREAM) return NULL;
+
+		src_addr = si->myname;
+		dest_addr = si->peername;
+
+		tcp_seqno = si->io.pck_snd;
+		tcp_ack = si->io.pck_rcv;
+		tcp_ctl = 0x10; /* ACK */
+
+		break;
+	default:
+		return NULL;
+	}
+
+	swrapGetTimeOfDay(&tv);
+
+	return swrap_packet_init(&tv, src_addr, dest_addr, si->type,
+				 (const uint8_t *)buf, len,
+				 tcp_seqno, tcp_ack, tcp_ctl, unreachable,
+				 packet_len);
+}
+
+static void swrap_dump_packet(struct socket_info *si,
+			      const struct sockaddr *addr,
+			      enum swrap_packet_type type,
+			      const void *buf, size_t len)
+{
+	const char *file_name;
+	uint8_t *packet;
+	size_t packet_len = 0;
+	int fd;
+
+	file_name = socket_wrapper_pcap_file();
+	if (!file_name) {
+		return;
+	}
+
+	packet = swrap_marshall_packet(si, addr, type, buf, len, &packet_len);
+	if (!packet) {
+		return;
+	}
+
+	fd = swrap_get_pcap_fd(file_name);
+	if (fd != -1) {
+		if (write(fd, packet, packet_len) != packet_len) {
+			free(packet);
+			return;
+		}
+	}
+
+	free(packet);
+}
+
+_PUBLIC_ int swrap_socket(int family, int type, int protocol)
+{
+	struct socket_info *si;
+	struct socket_info_fd *fi;
+	int fd;
+	int real_type = type;
+#ifdef SOCK_CLOEXEC
+	real_type &= ~SOCK_CLOEXEC;
+#endif
+#ifdef SOCK_NONBLOCK
+	real_type &= ~SOCK_NONBLOCK;
+#endif
+
+	if (!socket_wrapper_dir()) {
+		return real_socket(family, type, protocol);
+	}
+
+	switch (family) {
+	case AF_INET:
+#ifdef HAVE_IPV6
+	case AF_INET6:
+#endif
+		break;
+	case AF_UNIX:
+		return real_socket(family, type, protocol);
+	default:
+		errno = EAFNOSUPPORT;
+		return -1;
+	}
+
+	switch (real_type) {
+	case SOCK_STREAM:
+		break;
+	case SOCK_DGRAM:
+		break;
+	default:
+		errno = EPROTONOSUPPORT;
+		return -1;
+	}
+
+	switch (protocol) {
+	case 0:
+		break;
+	case 6:
+		if (real_type == SOCK_STREAM) {
+			break;
+		}
+		/*fall through*/
+	case 17:
+		if (real_type == SOCK_DGRAM) {
+			break;
+		}
+		/*fall through*/
+	default:
+		errno = EPROTONOSUPPORT;
+		return -1;
+	}
+
+	/* We must call real_socket with type, from the caller, not the version we removed
+	   SOCK_CLOEXEC and SOCK_NONBLOCK from */
+	fd = real_socket(AF_UNIX, type, 0);
+
+	if (fd == -1) return -1;
+
+	si = (struct socket_info *)calloc(1, sizeof(struct socket_info));
+	if (si == NULL) {
+		errno = ENOMEM;
+		return -1;
+	}
+
+	si->family = family;
+
+	/* however, the rest of the socket_wrapper code expects just
+	 * the type, not the flags */
+	si->type = real_type;
+	si->protocol = protocol;
+
+	fi = (struct socket_info_fd *)calloc(1, sizeof(struct socket_info_fd));
+	if (fi == NULL) {
+		free(si);
+		errno = ENOMEM;
+		return -1;
+	}
+
+	fi->fd = fd;
+
+	SWRAP_DLIST_ADD(si->fds, fi);
+	SWRAP_DLIST_ADD(sockets, si);
+
+	return fd;
+}
+
+_PUBLIC_ int swrap_accept(int s, struct sockaddr *addr, socklen_t *addrlen)
+{
+	struct socket_info *parent_si, *child_si;
+	struct socket_info_fd *child_fi;
+	int fd;
+	struct sockaddr_un un_addr;
+	socklen_t un_addrlen = sizeof(un_addr);
+	struct sockaddr_un un_my_addr;
+	socklen_t un_my_addrlen = sizeof(un_my_addr);
+	struct sockaddr *my_addr;
+	socklen_t my_addrlen, len;
+	int ret;
+
+	parent_si = find_socket_info(s);
+	if (!parent_si) {
+		return real_accept(s, addr, addrlen);
+	}
+
+	/* 
+	 * assume out sockaddr have the same size as the in parent
+	 * socket family
+	 */
+	my_addrlen = socket_length(parent_si->family);
+	if (my_addrlen <= 0) {
+		errno = EINVAL;
+		return -1;
+	}
+
+	my_addr = (struct sockaddr *)malloc(my_addrlen);
+	if (my_addr == NULL) {
+		return -1;
+	}
+
+	memset(&un_addr, 0, sizeof(un_addr));
+	memset(&un_my_addr, 0, sizeof(un_my_addr));
+
+	ret = real_accept(s, (struct sockaddr *)(void *)&un_addr, &un_addrlen);
+	if (ret == -1) {
+		free(my_addr);
+		return ret;
+	}
+
+	fd = ret;
+
+	len = my_addrlen;
+	ret = sockaddr_convert_from_un(parent_si, &un_addr, un_addrlen,
+				       parent_si->family, my_addr, &len);
+	if (ret == -1) {
+		free(my_addr);
+		close(fd);
+		return ret;
+	}
+
+	child_si = (struct socket_info *)malloc(sizeof(struct socket_info));
+	memset(child_si, 0, sizeof(*child_si));
+
+	child_fi = (struct socket_info_fd *)calloc(1, sizeof(struct socket_info_fd));
+	if (child_fi == NULL) {
+		free(child_si);
+		free(my_addr);
+		close(fd);
+		errno = ENOMEM;
+		return -1;
+	}
+
+	child_fi->fd = fd;
+
+	SWRAP_DLIST_ADD(child_si->fds, child_fi);
+
+	child_si->family = parent_si->family;
+	child_si->type = parent_si->type;
+	child_si->protocol = parent_si->protocol;
+	child_si->bound = 1;
+	child_si->is_server = 1;
+	child_si->connected = 1;
+
+	child_si->peername_len = len;
+	child_si->peername = sockaddr_dup(my_addr, len);
+
+	if (addr != NULL && addrlen != NULL) {
+		size_t copy_len = MIN(*addrlen, len);
+		if (copy_len > 0) {
+			memcpy(addr, my_addr, copy_len);
+		}
+		*addrlen = len;
+	}
+
+	ret = real_getsockname(fd, (struct sockaddr *)(void *)&un_my_addr,
+			       &un_my_addrlen);
+	if (ret == -1) {
+		free(child_fi);
+		free(child_si);
+		close(fd);
+		return ret;
+	}
+
+	len = my_addrlen;
+	ret = sockaddr_convert_from_un(child_si, &un_my_addr, un_my_addrlen,
+				       child_si->family, my_addr, &len);
+	if (ret == -1) {
+		free(child_fi);
+		free(child_si);
+		free(my_addr);
+		close(fd);
+		return ret;
+	}
+
+	child_si->myname_len = len;
+	child_si->myname = sockaddr_dup(my_addr, len);
+	free(my_addr);
+
+	SWRAP_DLIST_ADD(sockets, child_si);
+
+	swrap_dump_packet(child_si, addr, SWRAP_ACCEPT_SEND, NULL, 0);
+	swrap_dump_packet(child_si, addr, SWRAP_ACCEPT_RECV, NULL, 0);
+	swrap_dump_packet(child_si, addr, SWRAP_ACCEPT_ACK, NULL, 0);
+
+	return fd;
+}
+
+static int autobind_start_init;
+static int autobind_start;
+
+/* using sendto() or connect() on an unbound socket would give the
+   recipient no way to reply, as unlike UDP and TCP, a unix domain
+   socket can't auto-assign emphemeral port numbers, so we need to
+   assign it here.
+   Note: this might change the family from ipv6 to ipv4
+*/
+static int swrap_auto_bind(int fd, struct socket_info *si, int family)
+{
+	struct sockaddr_un un_addr;
+	int i;
+	char type;
+	int ret;
+	int port;
+	struct stat st;
+
+	if (autobind_start_init != 1) {
+		autobind_start_init = 1;
+		autobind_start = getpid();
+		autobind_start %= 50000;
+		autobind_start += 10000;
+	}
+
+	un_addr.sun_family = AF_UNIX;
+
+	switch (family) {
+	case AF_INET: {
+		struct sockaddr_in in;
+
+		switch (si->type) {
+		case SOCK_STREAM:
+			type = SOCKET_TYPE_CHAR_TCP;
+			break;
+		case SOCK_DGRAM:
+		    	type = SOCKET_TYPE_CHAR_UDP;
+			break;
+		default:
+		    errno = ESOCKTNOSUPPORT;
+		    return -1;
+		}
+
+		memset(&in, 0, sizeof(in));
+		in.sin_family = AF_INET;
+		in.sin_addr.s_addr = htonl(127<<24 | 
+					   socket_wrapper_default_iface());
+
+		si->myname_len = sizeof(in);
+		si->myname = sockaddr_dup(&in, si->myname_len);
+		break;
+	}
+#ifdef HAVE_IPV6
+	case AF_INET6: {
+		struct sockaddr_in6 in6;
+
+		if (si->family != family) {
+			errno = ENETUNREACH;
+			return -1;
+		}
+
+		switch (si->type) {
+		case SOCK_STREAM:
+			type = SOCKET_TYPE_CHAR_TCP_V6;
+			break;
+		case SOCK_DGRAM:
+		    	type = SOCKET_TYPE_CHAR_UDP_V6;
+			break;
+		default:
+			errno = ESOCKTNOSUPPORT;
+			return -1;
+		}
+
+		memset(&in6, 0, sizeof(in6));
+		in6.sin6_family = AF_INET6;
+		in6.sin6_addr = *swrap_ipv6();
+		in6.sin6_addr.s6_addr[15] = socket_wrapper_default_iface();
+		si->myname_len = sizeof(in6);
+		si->myname = sockaddr_dup(&in6, si->myname_len);
+		break;
+	}
+#endif
+	default:
+		errno = ESOCKTNOSUPPORT;
+		return -1;
+	}
+
+	if (autobind_start > 60000) {
+		autobind_start = 10000;
+	}
+
+	for (i=0;i<1000;i++) {
+		port = autobind_start + i;
+		snprintf(un_addr.sun_path, sizeof(un_addr.sun_path), 
+			 "%s/"SOCKET_FORMAT, socket_wrapper_dir(),
+			 type, socket_wrapper_default_iface(), port);
+		if (stat(un_addr.sun_path, &st) == 0) continue;
+
+		ret = real_bind(fd, (struct sockaddr *)(void *)&un_addr,
+				sizeof(un_addr));
+		if (ret == -1) return ret;
+
+		si->tmp_path = strdup(un_addr.sun_path);
+		si->bound = 1;
+		autobind_start = port + 1;
+		break;
+	}
+	if (i == 1000) {
+		errno = ENFILE;
+		return -1;
+	}
+
+	si->family = family;
+	set_port(si->family, port, si->myname);
+
+	return 0;
+}
+
+
+_PUBLIC_ int swrap_connect(int s, const struct sockaddr *serv_addr, socklen_t addrlen)
+{
+	int ret;
+	struct sockaddr_un un_addr;
+	struct socket_info *si = find_socket_info(s);
+	int bcast = 0;
+
+	if (!si) {
+		return real_connect(s, serv_addr, addrlen);
+	}
+
+	if (si->bound == 0) {
+		ret = swrap_auto_bind(s, si, serv_addr->sa_family);
+		if (ret == -1) return -1;
+	}
+
+	if (si->family != serv_addr->sa_family) {
+		errno = EINVAL;
+		return -1;
+	}
+
+	ret = sockaddr_convert_to_un(si, serv_addr,
+				     addrlen, &un_addr, 0, &bcast);
+	if (ret == -1) return -1;
+
+	if (bcast) {
+		errno = ENETUNREACH;
+		return -1;
+	}
+
+	if (si->type == SOCK_DGRAM) {
+		si->defer_connect = 1;
+		ret = 0;
+	} else {
+		swrap_dump_packet(si, serv_addr, SWRAP_CONNECT_SEND, NULL, 0);
+
+		ret = real_connect(s, (struct sockaddr *)(void *)&un_addr,
+				   sizeof(struct sockaddr_un));
+	}
+
+	/* to give better errors */
+	if (ret == -1 && errno == ENOENT) {
+		errno = EHOSTUNREACH;
+	}
+
+	if (ret == 0) {
+		si->peername_len = addrlen;
+		si->peername = sockaddr_dup(serv_addr, addrlen);
+		si->connected = 1;
+
+		swrap_dump_packet(si, serv_addr, SWRAP_CONNECT_RECV, NULL, 0);
+		swrap_dump_packet(si, serv_addr, SWRAP_CONNECT_ACK, NULL, 0);
+	} else {
+		swrap_dump_packet(si, serv_addr, SWRAP_CONNECT_UNREACH, NULL, 0);
+	}
+
+	return ret;
+}
+
+_PUBLIC_ int swrap_bind(int s, const struct sockaddr *myaddr, socklen_t addrlen)
+{
+	int ret;
+	struct sockaddr_un un_addr;
+	struct socket_info *si = find_socket_info(s);
+
+	if (!si) {
+		return real_bind(s, myaddr, addrlen);
+	}
+
+	si->myname_len = addrlen;
+	si->myname = sockaddr_dup(myaddr, addrlen);
+
+	ret = sockaddr_convert_to_un(si, myaddr, addrlen, &un_addr, 1, &si->bcast);
+	if (ret == -1) return -1;
+
+	unlink(un_addr.sun_path);
+
+	ret = real_bind(s, (struct sockaddr *)(void *)&un_addr,
+			sizeof(struct sockaddr_un));
+
+	if (ret == 0) {
+		si->bound = 1;
+	}
+
+	return ret;
+}
+
+_PUBLIC_ int swrap_listen(int s, int backlog)
+{
+	int ret;
+	struct socket_info *si = find_socket_info(s);
+
+	if (!si) {
+		return real_listen(s, backlog);
+	}
+
+	ret = real_listen(s, backlog);
+
+	return ret;
+}
+
+_PUBLIC_ int swrap_getpeername(int s, struct sockaddr *name, socklen_t *addrlen)
+{
+	struct socket_info *si = find_socket_info(s);
+
+	if (!si) {
+		return real_getpeername(s, name, addrlen);
+	}
+
+	if (!si->peername)
+	{
+		errno = ENOTCONN;
+		return -1;
+	}
+
+	memcpy(name, si->peername, si->peername_len);
+	*addrlen = si->peername_len;
+
+	return 0;
+}
+
+_PUBLIC_ int swrap_getsockname(int s, struct sockaddr *name, socklen_t *addrlen)
+{
+	struct socket_info *si = find_socket_info(s);
+
+	if (!si) {
+		return real_getsockname(s, name, addrlen);
+	}
+
+	memcpy(name, si->myname, si->myname_len);
+	*addrlen = si->myname_len;
+
+	return 0;
+}
+
+_PUBLIC_ int swrap_getsockopt(int s, int level, int optname, void *optval, socklen_t *optlen)
+{
+	struct socket_info *si = find_socket_info(s);
+
+	if (!si) {
+		return real_getsockopt(s, level, optname, optval, optlen);
+	}
+
+	if (level == SOL_SOCKET) {
+		return real_getsockopt(s, level, optname, optval, optlen);
+	} 
+
+	errno = ENOPROTOOPT;
+	return -1;
+}
+
+_PUBLIC_ int swrap_setsockopt(int s, int  level,  int  optname,  const  void  *optval, socklen_t optlen)
+{
+	struct socket_info *si = find_socket_info(s);
+
+	if (!si) {
+		return real_setsockopt(s, level, optname, optval, optlen);
+	}
+
+	if (level == SOL_SOCKET) {
+		return real_setsockopt(s, level, optname, optval, optlen);
+	}
+
+	switch (si->family) {
+	case AF_INET:
+		return 0;
+#ifdef HAVE_IPV6
+	case AF_INET6:
+		return 0;
+#endif
+	default:
+		errno = ENOPROTOOPT;
+		return -1;
+	}
+}
+
+_PUBLIC_ int swrap_ioctl(int s, int r, void *p)
+{
+	int ret;
+	struct socket_info *si = find_socket_info(s);
+	int value;
+
+	if (!si) {
+		return real_ioctl(s, r, p);
+	}
+
+	ret = real_ioctl(s, r, p);
+
+	switch (r) {
+	case FIONREAD:
+		value = *((int *)p);
+		if (ret == -1 && errno != EAGAIN && errno != ENOBUFS) {
+			swrap_dump_packet(si, NULL, SWRAP_PENDING_RST, NULL, 0);
+		} else if (value == 0) { /* END OF FILE */
+			swrap_dump_packet(si, NULL, SWRAP_PENDING_RST, NULL, 0);
+		}
+		break;
+	}
+
+	return ret;
+}
+
+static ssize_t swrap_sendmsg_before(int fd,
+				    struct socket_info *si,
+				    struct msghdr *msg,
+				    struct iovec *tmp_iov,
+				    struct sockaddr_un *tmp_un,
+				    const struct sockaddr_un **to_un,
+				    const struct sockaddr **to,
+				    int *bcast)
+{
+	size_t i, len = 0;
+	ssize_t ret;
+
+	if (to_un) {
+		*to_un = NULL;
+	}
+	if (to) {
+		*to = NULL;
+	}
+	if (bcast) {
+		*bcast = 0;
+	}
+
+	switch (si->type) {
+	case SOCK_STREAM:
+		if (!si->connected) {
+			errno = ENOTCONN;
+			return -1;
+		}
+
+		if (msg->msg_iovlen == 0) {
+			break;
+		}
+
+		/*
+		 * cut down to 1500 byte packets for stream sockets,
+		 * which makes it easier to format PCAP capture files
+		 * (as the caller will simply continue from here)
+		 */
+
+		for (i=0; i < msg->msg_iovlen; i++) {
+			size_t nlen;
+			nlen = len + msg->msg_iov[i].iov_len;
+			if (nlen > 1500) {
+				break;
+			}
+		}
+		msg->msg_iovlen = i;
+		if (msg->msg_iovlen == 0) {
+			*tmp_iov = msg->msg_iov[0];
+			tmp_iov->iov_len = MIN(tmp_iov->iov_len, 1500);
+			msg->msg_iov = tmp_iov;
+			msg->msg_iovlen = 1;
+		}
+		break;
+
+	case SOCK_DGRAM:
+		if (si->connected) {
+			if (msg->msg_name) {
+				errno = EISCONN;
+				return -1;
+			}
+		} else {
+			const struct sockaddr *msg_name;
+			msg_name = (const struct sockaddr *)msg->msg_name;
+
+			if (msg_name == NULL) {
+				errno = ENOTCONN;
+				return -1;
+			}
+
+
+			ret = sockaddr_convert_to_un(si, msg_name, msg->msg_namelen,
+						     tmp_un, 0, bcast);
+			if (ret == -1) return -1;
+
+			if (to_un) {
+				*to_un = tmp_un;
+			}
+			if (to) {
+				*to = msg_name;
+			}
+			msg->msg_name = tmp_un;
+			msg->msg_namelen = sizeof(*tmp_un);
+		}
+
+		if (si->bound == 0) {
+			ret = swrap_auto_bind(fd, si, si->family);
+			if (ret == -1) return -1;
+		}
+
+		if (!si->defer_connect) {
+			break;
+		}
+
+		ret = sockaddr_convert_to_un(si, si->peername, si->peername_len,
+					     tmp_un, 0, NULL);
+		if (ret == -1) return -1;
+
+		ret = real_connect(fd, (struct sockaddr *)(void *)tmp_un,
+				   sizeof(*tmp_un));
+
+		/* to give better errors */
+		if (ret == -1 && errno == ENOENT) {
+			errno = EHOSTUNREACH;
+		}
+
+		if (ret == -1) {
+			return ret;
+		}
+
+		si->defer_connect = 0;
+		break;
+	default:
+		errno = EHOSTUNREACH;
+		return -1;
+	}
+
+	return 0;
+}
+
+static void swrap_sendmsg_after(struct socket_info *si,
+				struct msghdr *msg,
+				const struct sockaddr *to,
+				ssize_t ret)
+{
+	int saved_errno = errno;
+	size_t i, len = 0;
+	uint8_t *buf;
+	off_t ofs = 0;
+	size_t avail = 0;
+	size_t remain;
+
+	/* to give better errors */
+	if (ret == -1 && saved_errno == ENOENT) {
+		saved_errno = EHOSTUNREACH;
+	}
+
+	for (i=0; i < msg->msg_iovlen; i++) {
+		avail += msg->msg_iov[i].iov_len;
+	}
+
+	if (ret == -1) {
+		remain = MIN(80, avail);
+	} else {
+		remain = ret;
+	}
+
+	/* we capture it as one single packet */
+	buf = (uint8_t *)malloc(remain);
+	if (!buf) {
+		/* we just not capture the packet */
+		errno = saved_errno;
+		return;
+	}
+
+	for (i=0; i < msg->msg_iovlen; i++) {
+		size_t this_time = MIN(remain, msg->msg_iov[i].iov_len);
+		memcpy(buf + ofs,
+		       msg->msg_iov[i].iov_base,
+		       this_time);
+		ofs += this_time;
+		remain -= this_time;
+	}
+	len = ofs;
+
+	switch (si->type) {
+	case SOCK_STREAM:
+		if (ret == -1) {
+			swrap_dump_packet(si, NULL, SWRAP_SEND, buf, len);
+			swrap_dump_packet(si, NULL, SWRAP_SEND_RST, NULL, 0);
+		} else {
+			swrap_dump_packet(si, NULL, SWRAP_SEND, buf, len);
+		}
+		break;
+
+	case SOCK_DGRAM:
+		if (si->connected) {
+			to = si->peername;
+		}
+		if (ret == -1) {
+			swrap_dump_packet(si, to, SWRAP_SENDTO, buf, len);
+			swrap_dump_packet(si, to, SWRAP_SENDTO_UNREACH, buf, len);
+		} else {
+			swrap_dump_packet(si, to, SWRAP_SENDTO, buf, len);
+		}
+		break;
+	}
+
+	free(buf);
+	errno = saved_errno;
+}
+
+_PUBLIC_ ssize_t swrap_recvfrom(int s, void *buf, size_t len, int flags, struct sockaddr *from, socklen_t *fromlen)
+{
+	struct sockaddr_un un_addr;
+	socklen_t un_addrlen = sizeof(un_addr);
+	int ret;
+	struct socket_info *si = find_socket_info(s);
+	struct sockaddr_storage ss;
+	socklen_t ss_len = sizeof(ss);
+
+	if (!si) {
+		return real_recvfrom(s, buf, len, flags, from, fromlen);
+	}
+
+	if (!from) {
+		from = (struct sockaddr *)(void *)&ss;
+		fromlen = &ss_len;
+	}
+
+	if (si->type == SOCK_STREAM) {
+		/* cut down to 1500 byte packets for stream sockets,
+		 * which makes it easier to format PCAP capture files
+		 * (as the caller will simply continue from here) */
+		len = MIN(len, 1500);
+	}
+
+	/* irix 6.4 forgets to null terminate the sun_path string :-( */
+	memset(&un_addr, 0, sizeof(un_addr));
+	ret = real_recvfrom(s, buf, len, flags,
+			    (struct sockaddr *)(void *)&un_addr, &un_addrlen);
+	if (ret == -1) 
+		return ret;
+
+	if (sockaddr_convert_from_un(si, &un_addr, un_addrlen,
+				     si->family, from, fromlen) == -1) {
+		return -1;
+	}
+
+	swrap_dump_packet(si, from, SWRAP_RECVFROM, buf, ret);
+
+	return ret;
+}
+
+
+_PUBLIC_ ssize_t swrap_sendto(int s, const void *buf, size_t len, int flags, const struct sockaddr *to, socklen_t tolen)
+{
+	struct msghdr msg;
+	struct iovec tmp;
+	struct sockaddr_un un_addr;
+	const struct sockaddr_un *to_un = NULL;
+	ssize_t ret;
+	struct socket_info *si = find_socket_info(s);
+	int bcast = 0;
+
+	if (!si) {
+		return real_sendto(s, buf, len, flags, to, tolen);
+	}
+
+	tmp.iov_base = discard_const_p(char, buf);
+	tmp.iov_len = len;
+
+	ZERO_STRUCT(msg);
+	msg.msg_name = discard_const_p(struct sockaddr, to); /* optional address */
+	msg.msg_namelen = tolen;       /* size of address */
+	msg.msg_iov = &tmp;            /* scatter/gather array */
+	msg.msg_iovlen = 1;            /* # elements in msg_iov */
+#if 0 /* not available on solaris */
+	msg.msg_control = NULL;        /* ancillary data, see below */
+	msg.msg_controllen = 0;        /* ancillary data buffer len */
+	msg.msg_flags = 0;             /* flags on received message */
+#endif
+
+	ret = swrap_sendmsg_before(s, si, &msg, &tmp, &un_addr, &to_un, &to, &bcast);
+	if (ret == -1) return -1;
+
+	buf = msg.msg_iov[0].iov_base;
+	len = msg.msg_iov[0].iov_len;
+
+	if (bcast) {
+		struct stat st;
+		unsigned int iface;
+		unsigned int prt = ntohs(((const struct sockaddr_in *)to)->sin_port);
+		char type;
+
+		type = SOCKET_TYPE_CHAR_UDP;
+
+		for(iface=0; iface <= MAX_WRAPPED_INTERFACES; iface++) {
+			snprintf(un_addr.sun_path, sizeof(un_addr.sun_path), "%s/"SOCKET_FORMAT,
+				 socket_wrapper_dir(), type, iface, prt);
+			if (stat(un_addr.sun_path, &st) != 0) continue;
+
+			/* ignore the any errors in broadcast sends */
+			real_sendto(s, buf, len, flags,
+				    (struct sockaddr *)(void *)&un_addr,
+				    sizeof(un_addr));
+		}
+
+		swrap_dump_packet(si, to, SWRAP_SENDTO, buf, len);
+
+		return len;
+	}
+
+	ret = real_sendto(s, buf, len, flags, (struct sockaddr *)msg.msg_name,
+			  msg.msg_namelen);
+
+	swrap_sendmsg_after(si, &msg, to, ret);
+
+	return ret;
+}
+
+_PUBLIC_ ssize_t swrap_recv(int s, void *buf, size_t len, int flags)
+{
+	int ret;
+	struct socket_info *si = find_socket_info(s);
+
+	if (!si) {
+		return real_recv(s, buf, len, flags);
+	}
+
+	if (si->type == SOCK_STREAM) {
+		/* cut down to 1500 byte packets for stream sockets,
+		 * which makes it easier to format PCAP capture files
+		 * (as the caller will simply continue from here) */
+		len = MIN(len, 1500);
+	}
+
+	ret = real_recv(s, buf, len, flags);
+	if (ret == -1 && errno != EAGAIN && errno != ENOBUFS) {
+		swrap_dump_packet(si, NULL, SWRAP_RECV_RST, NULL, 0);
+	} else if (ret == 0) { /* END OF FILE */
+		swrap_dump_packet(si, NULL, SWRAP_RECV_RST, NULL, 0);
+	} else if (ret > 0) {
+		swrap_dump_packet(si, NULL, SWRAP_RECV, buf, ret);
+	}
+
+	return ret;
+}
+
+_PUBLIC_ ssize_t swrap_read(int s, void *buf, size_t len)
+{
+	int ret;
+	struct socket_info *si = find_socket_info(s);
+
+	if (!si) {
+		return real_read(s, buf, len);
+	}
+
+	if (si->type == SOCK_STREAM) {
+		/* cut down to 1500 byte packets for stream sockets,
+		 * which makes it easier to format PCAP capture files
+		 * (as the caller will simply continue from here) */
+		len = MIN(len, 1500);
+	}
+
+	ret = real_read(s, buf, len);
+	if (ret == -1 && errno != EAGAIN && errno != ENOBUFS) {
+		swrap_dump_packet(si, NULL, SWRAP_RECV_RST, NULL, 0);
+	} else if (ret == 0) { /* END OF FILE */
+		swrap_dump_packet(si, NULL, SWRAP_RECV_RST, NULL, 0);
+	} else if (ret > 0) {
+		swrap_dump_packet(si, NULL, SWRAP_RECV, buf, ret);
+	}
+
+	return ret;
+}
+
+
+_PUBLIC_ ssize_t swrap_send(int s, const void *buf, size_t len, int flags)
+{
+	struct msghdr msg;
+	struct iovec tmp;
+	struct sockaddr_un un_addr;
+	ssize_t ret;
+	struct socket_info *si = find_socket_info(s);
+
+	if (!si) {
+		return real_send(s, buf, len, flags);
+	}
+
+	tmp.iov_base = discard_const_p(char, buf);
+	tmp.iov_len = len;
+
+	ZERO_STRUCT(msg);
+	msg.msg_name = NULL;           /* optional address */
+	msg.msg_namelen = 0;           /* size of address */
+	msg.msg_iov = &tmp;            /* scatter/gather array */
+	msg.msg_iovlen = 1;            /* # elements in msg_iov */
+#if 0 /* not available on solaris */
+	msg.msg_control = NULL;        /* ancillary data, see below */
+	msg.msg_controllen = 0;        /* ancillary data buffer len */
+	msg.msg_flags = 0;             /* flags on received message */
+#endif
+
+	ret = swrap_sendmsg_before(s, si, &msg, &tmp, &un_addr, NULL, NULL, NULL);
+	if (ret == -1) return -1;
+
+	buf = msg.msg_iov[0].iov_base;
+	len = msg.msg_iov[0].iov_len;
+
+	ret = real_send(s, buf, len, flags);
+
+	swrap_sendmsg_after(si, &msg, NULL, ret);
+
+	return ret;
+}
+
+_PUBLIC_ ssize_t swrap_sendmsg(int s, const struct msghdr *omsg, int flags)
+{
+	struct msghdr msg;
+	struct iovec tmp;
+	struct sockaddr_un un_addr;
+	const struct sockaddr_un *to_un = NULL;
+	const struct sockaddr *to = NULL;
+	ssize_t ret;
+	struct socket_info *si = find_socket_info(s);
+	int bcast = 0;
+
+	if (!si) {
+		return real_sendmsg(s, omsg, flags);
+	}
+
+	tmp.iov_base = NULL;
+	tmp.iov_len = 0;
+
+	msg = *omsg;
+#if 0
+	msg.msg_name = omsg->msg_name;             /* optional address */
+	msg.msg_namelen = omsg->msg_namelen;       /* size of address */
+	msg.msg_iov = omsg->msg_iov;               /* scatter/gather array */
+	msg.msg_iovlen = omsg->msg_iovlen;         /* # elements in msg_iov */
+	/* the following is not available on solaris */
+	msg.msg_control = omsg->msg_control;       /* ancillary data, see below */
+	msg.msg_controllen = omsg->msg_controllen; /* ancillary data buffer len */
+	msg.msg_flags = omsg->msg_flags;           /* flags on received message */
+#endif
+
+	ret = swrap_sendmsg_before(s, si, &msg, &tmp, &un_addr, &to_un, &to, &bcast);
+	if (ret == -1) return -1;
+
+	if (bcast) {
+		struct stat st;
+		unsigned int iface;
+		unsigned int prt = ntohs(((const struct sockaddr_in *)to)->sin_port);
+		char type;
+		size_t i, len = 0;
+		uint8_t *buf;
+		off_t ofs = 0;
+		size_t avail = 0;
+		size_t remain;
+
+		for (i=0; i < msg.msg_iovlen; i++) {
+			avail += msg.msg_iov[i].iov_len;
+		}
+
+		len = avail;
+		remain = avail;
+
+		/* we capture it as one single packet */
+		buf = (uint8_t *)malloc(remain);
+		if (!buf) {
+			return -1;
+		}
+
+		for (i=0; i < msg.msg_iovlen; i++) {
+			size_t this_time = MIN(remain, msg.msg_iov[i].iov_len);
+			memcpy(buf + ofs,
+			       msg.msg_iov[i].iov_base,
+			       this_time);
+			ofs += this_time;
+			remain -= this_time;
+		}
+
+		type = SOCKET_TYPE_CHAR_UDP;
+
+		for(iface=0; iface <= MAX_WRAPPED_INTERFACES; iface++) {
+			snprintf(un_addr.sun_path, sizeof(un_addr.sun_path), "%s/"SOCKET_FORMAT,
+				 socket_wrapper_dir(), type, iface, prt);
+			if (stat(un_addr.sun_path, &st) != 0) continue;
+
+			msg.msg_name = &un_addr;           /* optional address */
+			msg.msg_namelen = sizeof(un_addr); /* size of address */
+
+			/* ignore the any errors in broadcast sends */
+			real_sendmsg(s, &msg, flags);
+		}
+
+		swrap_dump_packet(si, to, SWRAP_SENDTO, buf, len);
+		free(buf);
+
+		return len;
+	}
+
+	ret = real_sendmsg(s, &msg, flags);
+
+	swrap_sendmsg_after(si, &msg, to, ret);
+
+	return ret;
+}
+
+int swrap_readv(int s, const struct iovec *vector, size_t count)
+{
+	int ret;
+	struct socket_info *si = find_socket_info(s);
+	struct iovec v;
+
+	if (!si) {
+		return real_readv(s, vector, count);
+	}
+
+	if (!si->connected) {
+		errno = ENOTCONN;
+		return -1;
+	}
+
+	if (si->type == SOCK_STREAM && count > 0) {
+		/* cut down to 1500 byte packets for stream sockets,
+		 * which makes it easier to format PCAP capture files
+		 * (as the caller will simply continue from here) */
+		size_t i, len = 0;
+
+		for (i=0; i < count; i++) {
+			size_t nlen;
+			nlen = len + vector[i].iov_len;
+			if (nlen > 1500) {
+				break;
+			}
+		}
+		count = i;
+		if (count == 0) {
+			v = vector[0];
+			v.iov_len = MIN(v.iov_len, 1500);
+			vector = &v;
+			count = 1;
+		}
+	}
+
+	ret = real_readv(s, vector, count);
+	if (ret == -1 && errno != EAGAIN && errno != ENOBUFS) {
+		swrap_dump_packet(si, NULL, SWRAP_RECV_RST, NULL, 0);
+	} else if (ret == 0) { /* END OF FILE */
+		swrap_dump_packet(si, NULL, SWRAP_RECV_RST, NULL, 0);
+	} else if (ret > 0) {
+		uint8_t *buf;
+		off_t ofs = 0;
+		size_t i;
+		size_t remain = ret;
+
+		/* we capture it as one single packet */
+		buf = (uint8_t *)malloc(ret);
+		if (!buf) {
+			/* we just not capture the packet */
+			errno = 0;
+			return ret;
+		}
+
+		for (i=0; i < count; i++) {
+			size_t this_time = MIN(remain, vector[i].iov_len);
+			memcpy(buf + ofs,
+			       vector[i].iov_base,
+			       this_time);
+			ofs += this_time;
+			remain -= this_time;
+		}
+
+		swrap_dump_packet(si, NULL, SWRAP_RECV, buf, ret);
+		free(buf);
+	}
+
+	return ret;
+}
+
+int swrap_writev(int s, const struct iovec *vector, size_t count)
+{
+	struct msghdr msg;
+	struct iovec tmp;
+	struct sockaddr_un un_addr;
+	ssize_t ret;
+	struct socket_info *si = find_socket_info(s);
+
+	if (!si) {
+		return real_writev(s, vector, count);
+	}
+
+	tmp.iov_base = NULL;
+	tmp.iov_len = 0;
+
+	ZERO_STRUCT(msg);
+	msg.msg_name = NULL;           /* optional address */
+	msg.msg_namelen = 0;           /* size of address */
+	msg.msg_iov = discard_const_p(struct iovec, vector); /* scatter/gather array */
+	msg.msg_iovlen = count;        /* # elements in msg_iov */
+#if 0 /* not available on solaris */
+	msg.msg_control = NULL;        /* ancillary data, see below */
+	msg.msg_controllen = 0;        /* ancillary data buffer len */
+	msg.msg_flags = 0;             /* flags on received message */
+#endif
+
+	ret = swrap_sendmsg_before(s, si, &msg, &tmp, &un_addr, NULL, NULL, NULL);
+	if (ret == -1) return -1;
+
+	ret = real_writev(s, msg.msg_iov, msg.msg_iovlen);
+
+	swrap_sendmsg_after(si, &msg, NULL, ret);
+
+	return ret;
+}
+
+_PUBLIC_ int swrap_close(int fd)
+{
+	struct socket_info *si = find_socket_info(fd);
+	struct socket_info_fd *fi;
+	int ret;
+
+	if (!si) {
+		return real_close(fd);
+	}
+
+	for (fi = si->fds; fi; fi = fi->next) {
+		if (fi->fd == fd) {
+			SWRAP_DLIST_REMOVE(si->fds, fi);
+			free(fi);
+			break;
+		}
+	}
+
+	if (si->fds) {
+		/* there are still references left */
+		return real_close(fd);
+	}
+
+	SWRAP_DLIST_REMOVE(sockets, si);
+
+	if (si->myname && si->peername) {
+		swrap_dump_packet(si, NULL, SWRAP_CLOSE_SEND, NULL, 0);
+	}
+
+	ret = real_close(fd);
+
+	if (si->myname && si->peername) {
+		swrap_dump_packet(si, NULL, SWRAP_CLOSE_RECV, NULL, 0);
+		swrap_dump_packet(si, NULL, SWRAP_CLOSE_ACK, NULL, 0);
+	}
+
+	if (si->myname) free(si->myname);
+	if (si->peername) free(si->peername);
+	if (si->tmp_path) {
+		unlink(si->tmp_path);
+		free(si->tmp_path);
+	}
+	free(si);
+
+	return ret;
+}
+
+_PUBLIC_ int swrap_dup(int fd)
+{
+	struct socket_info *si;
+	struct socket_info_fd *fi;
+
+	si = find_socket_info(fd);
+
+	if (!si) {
+		return real_dup(fd);
+	}
+
+	fi = (struct socket_info_fd *)calloc(1, sizeof(struct socket_info_fd));
+	if (fi == NULL) {
+		errno = ENOMEM;
+		return -1;
+	}
+
+	fi->fd = real_dup(fd);
+	if (fi->fd == -1) {
+		int saved_errno = errno;
+		free(fi);
+		errno = saved_errno;
+		return -1;
+	}
+
+	SWRAP_DLIST_ADD(si->fds, fi);
+	return fi->fd;
+}
+
+_PUBLIC_ int swrap_dup2(int fd, int newfd)
+{
+	struct socket_info *si;
+	struct socket_info_fd *fi;
+
+	si = find_socket_info(fd);
+
+	if (!si) {
+		return real_dup2(fd, newfd);
+	}
+
+	if (find_socket_info(newfd)) {
+		/* dup2() does an implicit close of newfd, which we
+		 * need to emulate */
+		swrap_close(newfd);
+	}
+
+	fi = (struct socket_info_fd *)calloc(1, sizeof(struct socket_info_fd));
+	if (fi == NULL) {
+		errno = ENOMEM;
+		return -1;
+	}
+
+	fi->fd = real_dup2(fd, newfd);
+	if (fi->fd == -1) {
+		int saved_errno = errno;
+		free(fi);
+		errno = saved_errno;
+		return -1;
+	}
+
+	SWRAP_DLIST_ADD(si->fds, fi);
+	return fi->fd;
+}
diff --git a/ctdb/lib/socket_wrapper/socket_wrapper.h b/ctdb/lib/socket_wrapper/socket_wrapper.h
new file mode 100644
index 0000000000..32c9de6b68
--- /dev/null
+++ b/ctdb/lib/socket_wrapper/socket_wrapper.h
@@ -0,0 +1,177 @@
+/*
+ * Copyright (C) Jelmer Vernooij 2005 <jelmer@samba.org>
+ * Copyright (C) Stefan Metzmacher 2006 <metze@samba.org>
+ *
+ * All rights reserved.
+ * 
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ * 
+ * 3. Neither the name of the author nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ * 
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ *
+ */
+
+#ifndef __SOCKET_WRAPPER_H__
+#define __SOCKET_WRAPPER_H__
+
+const char *socket_wrapper_dir(void);
+unsigned int socket_wrapper_default_iface(void);
+int swrap_socket(int family, int type, int protocol);
+int swrap_accept(int s, struct sockaddr *addr, socklen_t *addrlen);
+int swrap_connect(int s, const struct sockaddr *serv_addr, socklen_t addrlen);
+int swrap_bind(int s, const struct sockaddr *myaddr, socklen_t addrlen);
+int swrap_listen(int s, int backlog);
+int swrap_getpeername(int s, struct sockaddr *name, socklen_t *addrlen);
+int swrap_getsockname(int s, struct sockaddr *name, socklen_t *addrlen);
+int swrap_getsockopt(int s, int level, int optname, void *optval, socklen_t *optlen);
+int swrap_setsockopt(int s, int  level,  int  optname,  const  void  *optval, socklen_t optlen);
+ssize_t swrap_recvfrom(int s, void *buf, size_t len, int flags, struct sockaddr *from, socklen_t *fromlen);
+ssize_t swrap_sendto(int s, const void *buf, size_t len, int flags, const struct sockaddr *to, socklen_t tolen);
+ssize_t swrap_sendmsg(int s, const struct msghdr *msg, int flags);
+ssize_t swrap_recvmsg(int s, struct msghdr *msg, int flags);
+int swrap_ioctl(int s, int req, void *ptr);
+ssize_t swrap_recv(int s, void *buf, size_t len, int flags);
+ssize_t swrap_read(int s, void *buf, size_t len);
+ssize_t swrap_send(int s, const void *buf, size_t len, int flags);
+int swrap_readv(int s, const struct iovec *vector, size_t count);
+int swrap_writev(int s, const struct iovec *vector, size_t count);
+int swrap_close(int);
+int swrap_dup(int oldfd);
+int swrap_dup2(int oldfd, int newfd);
+
+#ifdef SOCKET_WRAPPER_REPLACE
+
+#ifdef accept
+#undef accept
+#endif
+#define accept(s,addr,addrlen)		swrap_accept(s,addr,addrlen)
+
+#ifdef connect
+#undef connect
+#endif
+#define connect(s,serv_addr,addrlen)	swrap_connect(s,serv_addr,addrlen)
+
+#ifdef bind
+#undef bind
+#endif
+#define bind(s,myaddr,addrlen)		swrap_bind(s,myaddr,addrlen)
+
+#ifdef listen
+#undef listen
+#endif
+#define listen(s,blog)			swrap_listen(s,blog)
+
+#ifdef getpeername
+#undef getpeername
+#endif
+#define getpeername(s,name,addrlen)	swrap_getpeername(s,name,addrlen)
+
+#ifdef getsockname
+#undef getsockname
+#endif
+#define getsockname(s,name,addrlen)	swrap_getsockname(s,name,addrlen)
+
+#ifdef getsockopt
+#undef getsockopt
+#endif
+#define getsockopt(s,level,optname,optval,optlen) swrap_getsockopt(s,level,optname,optval,optlen)
+
+#ifdef setsockopt
+#undef setsockopt
+#endif
+#define setsockopt(s,level,optname,optval,optlen) swrap_setsockopt(s,level,optname,optval,optlen)
+
+#ifdef recvfrom
+#undef recvfrom
+#endif
+#define recvfrom(s,buf,len,flags,from,fromlen) 	  swrap_recvfrom(s,buf,len,flags,from,fromlen)
+
+#ifdef sendto
+#undef sendto
+#endif
+#define sendto(s,buf,len,flags,to,tolen)          swrap_sendto(s,buf,len,flags,to,tolen)
+
+#ifdef sendmsg
+#undef sendmsg
+#endif
+#define sendmsg(s,msg,flags)            swrap_sendmsg(s,msg,flags)
+
+#ifdef recvmsg
+#undef recvmsg
+#endif
+#define recvmsg(s,msg,flags)            swrap_recvmsg(s,msg,flags)
+
+#ifdef ioctl
+#undef ioctl
+#endif
+#define ioctl(s,req,ptr)		swrap_ioctl(s,req,ptr)
+
+#ifdef recv
+#undef recv
+#endif
+#define recv(s,buf,len,flags)		swrap_recv(s,buf,len,flags)
+
+#ifdef read
+#undef read
+#endif
+#define read(s,buf,len)		swrap_read(s,buf,len)
+
+#ifdef send
+#undef send
+#endif
+#define send(s,buf,len,flags)		swrap_send(s,buf,len,flags)
+
+#ifdef readv
+#undef readv
+#endif
+#define readv(s, vector, count)		swrap_readv(s,vector, count)
+
+#ifdef writev
+#undef writev
+#endif
+#define writev(s, vector, count)	swrap_writev(s,vector, count)
+
+#ifdef socket
+#undef socket
+#endif
+#define socket(domain,type,protocol)	swrap_socket(domain,type,protocol)
+
+#ifdef close
+#undef close
+#endif
+#define close(s)			swrap_close(s)
+
+#ifdef dup
+#undef dup
+#endif
+#define dup(s)			swrap_dup(s)
+
+#ifdef dup2
+#undef dup2
+#endif
+#define dup2(s, s2)		swrap_dup2(s, s2)
+
+#endif /* SOCKET_WRAPPER_REPLACE */
+#endif /* __SOCKET_WRAPPER_H__ */
diff --git a/ctdb/lib/socket_wrapper/testsuite.c b/ctdb/lib/socket_wrapper/testsuite.c
new file mode 100644
index 0000000000..9274e7f485
--- /dev/null
+++ b/ctdb/lib/socket_wrapper/testsuite.c
@@ -0,0 +1,105 @@
+/* 
+   Unix SMB/CIFS implementation.
+
+   local testing of the socket wrapper
+
+   Copyright (C) Jelmer Vernooij 2007
+   
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "includes.h"
+#include "system/network.h"
+#include "../socket_wrapper/socket_wrapper.h"
+#include "torture/torture.h"
+
+static char *old_dir = NULL;
+static char *old_iface = NULL;
+
+static void backup_env(void)
+{
+	old_dir = getenv("SOCKET_WRAPPER_DIR");
+	old_iface = getenv("SOCKET_WRAPPER_DEFAULT_IFACE");
+}
+
+static void restore_env(void)
+{
+	if (old_dir == NULL)
+		unsetenv("SOCKET_WRAPPER_DIR");
+	else
+		setenv("SOCKET_WRAPPER_DIR", old_dir, 1);
+	if (old_iface == NULL)
+		unsetenv("SOCKET_WRAPPER_DEFAULT_IFACE");
+	else
+		setenv("SOCKET_WRAPPER_DEFAULT_IFACE", old_iface, 1);
+}
+
+static bool test_socket_wrapper_dir(struct torture_context *tctx)
+{
+	backup_env();
+
+	setenv("SOCKET_WRAPPER_DIR", "foo", 1);
+	torture_assert_str_equal(tctx, socket_wrapper_dir(), "foo", "setting failed");
+	setenv("SOCKET_WRAPPER_DIR", "./foo", 1);
+	torture_assert_str_equal(tctx, socket_wrapper_dir(), "foo", "setting failed");
+	unsetenv("SOCKET_WRAPPER_DIR");
+	torture_assert_str_equal(tctx, socket_wrapper_dir(), NULL, "resetting failed");
+
+	restore_env();
+
+	return true;
+}
+
+static bool test_swrap_socket(struct torture_context *tctx)
+{
+	backup_env();
+	setenv("SOCKET_WRAPPER_DIR", "foo", 1);
+
+	torture_assert_int_equal(tctx, swrap_socket(1337, 1337, 0), -1, "unknown address family fails");
+	torture_assert_int_equal(tctx, errno, EAFNOSUPPORT, "correct errno set");
+	torture_assert_int_equal(tctx, swrap_socket(AF_INET, 1337, 0), -1, "unknown type fails");
+	torture_assert_int_equal(tctx, errno, EPROTONOSUPPORT, "correct errno set");
+	torture_assert_int_equal(tctx, swrap_socket(AF_INET, SOCK_DGRAM, 10), -1, "unknown protocol fails");
+	torture_assert_int_equal(tctx, errno, EPROTONOSUPPORT, "correct errno set");
+
+	restore_env();
+
+	return true;
+}
+
+unsigned int socket_wrapper_default_iface(void);
+static bool test_socket_wrapper_default_iface(struct torture_context *tctx)
+{
+	backup_env();
+	unsetenv("SOCKET_WRAPPER_DEFAULT_IFACE");
+	torture_assert_int_equal(tctx, socket_wrapper_default_iface(), 1, "unset");
+	setenv("SOCKET_WRAPPER_DEFAULT_IFACE", "2", 1);
+	torture_assert_int_equal(tctx, socket_wrapper_default_iface(), 2, "unset");
+	setenv("SOCKET_WRAPPER_DEFAULT_IFACE", "bla", 1);
+	torture_assert_int_equal(tctx, socket_wrapper_default_iface(), 1, "unset");
+	restore_env();
+	return true;
+}
+
+struct torture_suite *torture_local_socket_wrapper(TALLOC_CTX *mem_ctx)
+{
+	struct torture_suite *suite = torture_suite_create(mem_ctx, 
+													   "socket-wrapper");
+
+	torture_suite_add_simple_test(suite, "socket_wrapper_dir", test_socket_wrapper_dir);
+	torture_suite_add_simple_test(suite, "socket", test_swrap_socket);
+	torture_suite_add_simple_test(suite, "socket_wrapper_default_iface", test_socket_wrapper_default_iface);
+
+	return suite;
+}
diff --git a/ctdb/lib/socket_wrapper/wscript b/ctdb/lib/socket_wrapper/wscript
new file mode 100644
index 0000000000..9da578fb63
--- /dev/null
+++ b/ctdb/lib/socket_wrapper/wscript
@@ -0,0 +1,15 @@
+#!/usr/bin/env python
+
+import Options
+
+def set_options(opt):
+    gr = opt.option_group('developer options')
+    gr.add_option('--enable-socket-wrapper',
+                   help=("Turn on socket wrapper library (default=no)"),
+                   action="store_true", dest='enable_socket_wrapper', default=False)
+
+def configure(conf):
+    if (Options.options.enable_socket_wrapper or Options.options.developer or Options.options.enable_selftest):
+        conf.DEFINE('SOCKET_WRAPPER', 1)
+        conf.ADD_GLOBAL_DEPENDENCY('socket_wrapper')
+
diff --git a/ctdb/lib/socket_wrapper/wscript_build b/ctdb/lib/socket_wrapper/wscript_build
new file mode 100644
index 0000000000..a81c7aa61a
--- /dev/null
+++ b/ctdb/lib/socket_wrapper/wscript_build
@@ -0,0 +1,9 @@
+#!/usr/bin/env python
+
+bld.SAMBA_LIBRARY('socket_wrapper',
+                  source='socket_wrapper.c',
+                  group='base_libraries',
+                  private_library=True,
+                  enabled=bld.CONFIG_SET('SOCKET_WRAPPER'),
+                  deps='replace')
+
diff --git a/ctdb/lib/talloc/ABI/pytalloc-util-2.0.6.sigs b/ctdb/lib/talloc/ABI/pytalloc-util-2.0.6.sigs
new file mode 100644
index 0000000000..961c1a84d0
--- /dev/null
+++ b/ctdb/lib/talloc/ABI/pytalloc-util-2.0.6.sigs
@@ -0,0 +1,6 @@
+pytalloc_CObject_FromTallocPtr: PyObject *(void *)
+pytalloc_Check: int (PyObject *)
+pytalloc_GetObjectType: PyTypeObject *(void)
+pytalloc_reference_ex: PyObject *(PyTypeObject *, TALLOC_CTX *, void *)
+pytalloc_steal: PyObject *(PyTypeObject *, void *)
+pytalloc_steal_ex: PyObject *(PyTypeObject *, TALLOC_CTX *, void *)
diff --git a/ctdb/lib/talloc/ABI/pytalloc-util-2.0.7.sigs b/ctdb/lib/talloc/ABI/pytalloc-util-2.0.7.sigs
new file mode 100644
index 0000000000..961c1a84d0
--- /dev/null
+++ b/ctdb/lib/talloc/ABI/pytalloc-util-2.0.7.sigs
@@ -0,0 +1,6 @@
+pytalloc_CObject_FromTallocPtr: PyObject *(void *)
+pytalloc_Check: int (PyObject *)
+pytalloc_GetObjectType: PyTypeObject *(void)
+pytalloc_reference_ex: PyObject *(PyTypeObject *, TALLOC_CTX *, void *)
+pytalloc_steal: PyObject *(PyTypeObject *, void *)
+pytalloc_steal_ex: PyObject *(PyTypeObject *, TALLOC_CTX *, void *)
diff --git a/ctdb/lib/talloc/ABI/pytalloc-util-2.0.8.sigs b/ctdb/lib/talloc/ABI/pytalloc-util-2.0.8.sigs
new file mode 100644
index 0000000000..961c1a84d0
--- /dev/null
+++ b/ctdb/lib/talloc/ABI/pytalloc-util-2.0.8.sigs
@@ -0,0 +1,6 @@
+pytalloc_CObject_FromTallocPtr: PyObject *(void *)
+pytalloc_Check: int (PyObject *)
+pytalloc_GetObjectType: PyTypeObject *(void)
+pytalloc_reference_ex: PyObject *(PyTypeObject *, TALLOC_CTX *, void *)
+pytalloc_steal: PyObject *(PyTypeObject *, void *)
+pytalloc_steal_ex: PyObject *(PyTypeObject *, TALLOC_CTX *, void *)
diff --git a/ctdb/lib/talloc/ABI/talloc-2.0.2.sigs b/ctdb/lib/talloc/ABI/talloc-2.0.2.sigs
new file mode 100644
index 0000000000..6e236d52ac
--- /dev/null
+++ b/ctdb/lib/talloc/ABI/talloc-2.0.2.sigs
@@ -0,0 +1,62 @@
+_talloc: void *(const void *, size_t)
+_talloc_array: void *(const void *, size_t, unsigned int, const char *)
+_talloc_free: int (void *, const char *)
+_talloc_get_type_abort: void *(const void *, const char *, const char *)
+_talloc_memdup: void *(const void *, const void *, size_t, const char *)
+_talloc_move: void *(const void *, const void *)
+_talloc_realloc: void *(const void *, void *, size_t, const char *)
+_talloc_realloc_array: void *(const void *, void *, size_t, unsigned int, const char *)
+_talloc_reference_loc: void *(const void *, const void *, const char *)
+_talloc_set_destructor: void (const void *, int (*)(void *))
+_talloc_steal_loc: void *(const void *, const void *, const char *)
+_talloc_zero: void *(const void *, size_t, const char *)
+_talloc_zero_array: void *(const void *, size_t, unsigned int, const char *)
+talloc_asprintf: char *(const void *, const char *, ...)
+talloc_asprintf_append: char *(char *, const char *, ...)
+talloc_asprintf_append_buffer: char *(char *, const char *, ...)
+talloc_autofree_context: void *(void)
+talloc_check_name: void *(const void *, const char *)
+talloc_disable_null_tracking: void (void)
+talloc_enable_leak_report: void (void)
+talloc_enable_leak_report_full: void (void)
+talloc_enable_null_tracking: void (void)
+talloc_enable_null_tracking_no_autofree: void (void)
+talloc_find_parent_byname: void *(const void *, const char *)
+talloc_free_children: void (void *)
+talloc_get_name: const char *(const void *)
+talloc_get_size: size_t (const void *)
+talloc_increase_ref_count: int (const void *)
+talloc_init: void *(const char *, ...)
+talloc_is_parent: int (const void *, const void *)
+talloc_named: void *(const void *, size_t, const char *, ...)
+talloc_named_const: void *(const void *, size_t, const char *)
+talloc_parent: void *(const void *)
+talloc_parent_name: const char *(const void *)
+talloc_pool: void *(const void *, size_t)
+talloc_realloc_fn: void *(const void *, void *, size_t)
+talloc_reference_count: size_t (const void *)
+talloc_reparent: void *(const void *, const void *, const void *)
+talloc_report: void (const void *, FILE *)
+talloc_report_depth_cb: void (const void *, int, int, void (*)(const void *, int, int, int, void *), void *)
+talloc_report_depth_file: void (const void *, int, int, FILE *)
+talloc_report_full: void (const void *, FILE *)
+talloc_set_abort_fn: void (void (*)(const char *))
+talloc_set_log_fn: void (void (*)(const char *))
+talloc_set_log_stderr: void (void)
+talloc_set_name: const char *(const void *, const char *, ...)
+talloc_set_name_const: void (const void *, const char *)
+talloc_show_parents: void (const void *, FILE *)
+talloc_strdup: char *(const void *, const char *)
+talloc_strdup_append: char *(char *, const char *)
+talloc_strdup_append_buffer: char *(char *, const char *)
+talloc_strndup: char *(const void *, const char *, size_t)
+talloc_strndup_append: char *(char *, const char *, size_t)
+talloc_strndup_append_buffer: char *(char *, const char *, size_t)
+talloc_total_blocks: size_t (const void *)
+talloc_total_size: size_t (const void *)
+talloc_unlink: int (const void *, void *)
+talloc_vasprintf: char *(const void *, const char *, va_list)
+talloc_vasprintf_append: char *(char *, const char *, va_list)
+talloc_vasprintf_append_buffer: char *(char *, const char *, va_list)
+talloc_version_major: int (void)
+talloc_version_minor: int (void)
diff --git a/ctdb/lib/talloc/ABI/talloc-2.0.3.sigs b/ctdb/lib/talloc/ABI/talloc-2.0.3.sigs
new file mode 100644
index 0000000000..6e236d52ac
--- /dev/null
+++ b/ctdb/lib/talloc/ABI/talloc-2.0.3.sigs
@@ -0,0 +1,62 @@
+_talloc: void *(const void *, size_t)
+_talloc_array: void *(const void *, size_t, unsigned int, const char *)
+_talloc_free: int (void *, const char *)
+_talloc_get_type_abort: void *(const void *, const char *, const char *)
+_talloc_memdup: void *(const void *, const void *, size_t, const char *)
+_talloc_move: void *(const void *, const void *)
+_talloc_realloc: void *(const void *, void *, size_t, const char *)
+_talloc_realloc_array: void *(const void *, void *, size_t, unsigned int, const char *)
+_talloc_reference_loc: void *(const void *, const void *, const char *)
+_talloc_set_destructor: void (const void *, int (*)(void *))
+_talloc_steal_loc: void *(const void *, const void *, const char *)
+_talloc_zero: void *(const void *, size_t, const char *)
+_talloc_zero_array: void *(const void *, size_t, unsigned int, const char *)
+talloc_asprintf: char *(const void *, const char *, ...)
+talloc_asprintf_append: char *(char *, const char *, ...)
+talloc_asprintf_append_buffer: char *(char *, const char *, ...)
+talloc_autofree_context: void *(void)
+talloc_check_name: void *(const void *, const char *)
+talloc_disable_null_tracking: void (void)
+talloc_enable_leak_report: void (void)
+talloc_enable_leak_report_full: void (void)
+talloc_enable_null_tracking: void (void)
+talloc_enable_null_tracking_no_autofree: void (void)
+talloc_find_parent_byname: void *(const void *, const char *)
+talloc_free_children: void (void *)
+talloc_get_name: const char *(const void *)
+talloc_get_size: size_t (const void *)
+talloc_increase_ref_count: int (const void *)
+talloc_init: void *(const char *, ...)
+talloc_is_parent: int (const void *, const void *)
+talloc_named: void *(const void *, size_t, const char *, ...)
+talloc_named_const: void *(const void *, size_t, const char *)
+talloc_parent: void *(const void *)
+talloc_parent_name: const char *(const void *)
+talloc_pool: void *(const void *, size_t)
+talloc_realloc_fn: void *(const void *, void *, size_t)
+talloc_reference_count: size_t (const void *)
+talloc_reparent: void *(const void *, const void *, const void *)
+talloc_report: void (const void *, FILE *)
+talloc_report_depth_cb: void (const void *, int, int, void (*)(const void *, int, int, int, void *), void *)
+talloc_report_depth_file: void (const void *, int, int, FILE *)
+talloc_report_full: void (const void *, FILE *)
+talloc_set_abort_fn: void (void (*)(const char *))
+talloc_set_log_fn: void (void (*)(const char *))
+talloc_set_log_stderr: void (void)
+talloc_set_name: const char *(const void *, const char *, ...)
+talloc_set_name_const: void (const void *, const char *)
+talloc_show_parents: void (const void *, FILE *)
+talloc_strdup: char *(const void *, const char *)
+talloc_strdup_append: char *(char *, const char *)
+talloc_strdup_append_buffer: char *(char *, const char *)
+talloc_strndup: char *(const void *, const char *, size_t)
+talloc_strndup_append: char *(char *, const char *, size_t)
+talloc_strndup_append_buffer: char *(char *, const char *, size_t)
+talloc_total_blocks: size_t (const void *)
+talloc_total_size: size_t (const void *)
+talloc_unlink: int (const void *, void *)
+talloc_vasprintf: char *(const void *, const char *, va_list)
+talloc_vasprintf_append: char *(char *, const char *, va_list)
+talloc_vasprintf_append_buffer: char *(char *, const char *, va_list)
+talloc_version_major: int (void)
+talloc_version_minor: int (void)
diff --git a/ctdb/lib/talloc/ABI/talloc-2.0.4.sigs b/ctdb/lib/talloc/ABI/talloc-2.0.4.sigs
new file mode 100644
index 0000000000..6e236d52ac
--- /dev/null
+++ b/ctdb/lib/talloc/ABI/talloc-2.0.4.sigs
@@ -0,0 +1,62 @@
+_talloc: void *(const void *, size_t)
+_talloc_array: void *(const void *, size_t, unsigned int, const char *)
+_talloc_free: int (void *, const char *)
+_talloc_get_type_abort: void *(const void *, const char *, const char *)
+_talloc_memdup: void *(const void *, const void *, size_t, const char *)
+_talloc_move: void *(const void *, const void *)
+_talloc_realloc: void *(const void *, void *, size_t, const char *)
+_talloc_realloc_array: void *(const void *, void *, size_t, unsigned int, const char *)
+_talloc_reference_loc: void *(const void *, const void *, const char *)
+_talloc_set_destructor: void (const void *, int (*)(void *))
+_talloc_steal_loc: void *(const void *, const void *, const char *)
+_talloc_zero: void *(const void *, size_t, const char *)
+_talloc_zero_array: void *(const void *, size_t, unsigned int, const char *)
+talloc_asprintf: char *(const void *, const char *, ...)
+talloc_asprintf_append: char *(char *, const char *, ...)
+talloc_asprintf_append_buffer: char *(char *, const char *, ...)
+talloc_autofree_context: void *(void)
+talloc_check_name: void *(const void *, const char *)
+talloc_disable_null_tracking: void (void)
+talloc_enable_leak_report: void (void)
+talloc_enable_leak_report_full: void (void)
+talloc_enable_null_tracking: void (void)
+talloc_enable_null_tracking_no_autofree: void (void)
+talloc_find_parent_byname: void *(const void *, const char *)
+talloc_free_children: void (void *)
+talloc_get_name: const char *(const void *)
+talloc_get_size: size_t (const void *)
+talloc_increase_ref_count: int (const void *)
+talloc_init: void *(const char *, ...)
+talloc_is_parent: int (const void *, const void *)
+talloc_named: void *(const void *, size_t, const char *, ...)
+talloc_named_const: void *(const void *, size_t, const char *)
+talloc_parent: void *(const void *)
+talloc_parent_name: const char *(const void *)
+talloc_pool: void *(const void *, size_t)
+talloc_realloc_fn: void *(const void *, void *, size_t)
+talloc_reference_count: size_t (const void *)
+talloc_reparent: void *(const void *, const void *, const void *)
+talloc_report: void (const void *, FILE *)
+talloc_report_depth_cb: void (const void *, int, int, void (*)(const void *, int, int, int, void *), void *)
+talloc_report_depth_file: void (const void *, int, int, FILE *)
+talloc_report_full: void (const void *, FILE *)
+talloc_set_abort_fn: void (void (*)(const char *))
+talloc_set_log_fn: void (void (*)(const char *))
+talloc_set_log_stderr: void (void)
+talloc_set_name: const char *(const void *, const char *, ...)
+talloc_set_name_const: void (const void *, const char *)
+talloc_show_parents: void (const void *, FILE *)
+talloc_strdup: char *(const void *, const char *)
+talloc_strdup_append: char *(char *, const char *)
+talloc_strdup_append_buffer: char *(char *, const char *)
+talloc_strndup: char *(const void *, const char *, size_t)
+talloc_strndup_append: char *(char *, const char *, size_t)
+talloc_strndup_append_buffer: char *(char *, const char *, size_t)
+talloc_total_blocks: size_t (const void *)
+talloc_total_size: size_t (const void *)
+talloc_unlink: int (const void *, void *)
+talloc_vasprintf: char *(const void *, const char *, va_list)
+talloc_vasprintf_append: char *(char *, const char *, va_list)
+talloc_vasprintf_append_buffer: char *(char *, const char *, va_list)
+talloc_version_major: int (void)
+talloc_version_minor: int (void)
diff --git a/ctdb/lib/talloc/ABI/talloc-2.0.5.sigs b/ctdb/lib/talloc/ABI/talloc-2.0.5.sigs
new file mode 100644
index 0000000000..6e236d52ac
--- /dev/null
+++ b/ctdb/lib/talloc/ABI/talloc-2.0.5.sigs
@@ -0,0 +1,62 @@
+_talloc: void *(const void *, size_t)
+_talloc_array: void *(const void *, size_t, unsigned int, const char *)
+_talloc_free: int (void *, const char *)
+_talloc_get_type_abort: void *(const void *, const char *, const char *)
+_talloc_memdup: void *(const void *, const void *, size_t, const char *)
+_talloc_move: void *(const void *, const void *)
+_talloc_realloc: void *(const void *, void *, size_t, const char *)
+_talloc_realloc_array: void *(const void *, void *, size_t, unsigned int, const char *)
+_talloc_reference_loc: void *(const void *, const void *, const char *)
+_talloc_set_destructor: void (const void *, int (*)(void *))
+_talloc_steal_loc: void *(const void *, const void *, const char *)
+_talloc_zero: void *(const void *, size_t, const char *)
+_talloc_zero_array: void *(const void *, size_t, unsigned int, const char *)
+talloc_asprintf: char *(const void *, const char *, ...)
+talloc_asprintf_append: char *(char *, const char *, ...)
+talloc_asprintf_append_buffer: char *(char *, const char *, ...)
+talloc_autofree_context: void *(void)
+talloc_check_name: void *(const void *, const char *)
+talloc_disable_null_tracking: void (void)
+talloc_enable_leak_report: void (void)
+talloc_enable_leak_report_full: void (void)
+talloc_enable_null_tracking: void (void)
+talloc_enable_null_tracking_no_autofree: void (void)
+talloc_find_parent_byname: void *(const void *, const char *)
+talloc_free_children: void (void *)
+talloc_get_name: const char *(const void *)
+talloc_get_size: size_t (const void *)
+talloc_increase_ref_count: int (const void *)
+talloc_init: void *(const char *, ...)
+talloc_is_parent: int (const void *, const void *)
+talloc_named: void *(const void *, size_t, const char *, ...)
+talloc_named_const: void *(const void *, size_t, const char *)
+talloc_parent: void *(const void *)
+talloc_parent_name: const char *(const void *)
+talloc_pool: void *(const void *, size_t)
+talloc_realloc_fn: void *(const void *, void *, size_t)
+talloc_reference_count: size_t (const void *)
+talloc_reparent: void *(const void *, const void *, const void *)
+talloc_report: void (const void *, FILE *)
+talloc_report_depth_cb: void (const void *, int, int, void (*)(const void *, int, int, int, void *), void *)
+talloc_report_depth_file: void (const void *, int, int, FILE *)
+talloc_report_full: void (const void *, FILE *)
+talloc_set_abort_fn: void (void (*)(const char *))
+talloc_set_log_fn: void (void (*)(const char *))
+talloc_set_log_stderr: void (void)
+talloc_set_name: const char *(const void *, const char *, ...)
+talloc_set_name_const: void (const void *, const char *)
+talloc_show_parents: void (const void *, FILE *)
+talloc_strdup: char *(const void *, const char *)
+talloc_strdup_append: char *(char *, const char *)
+talloc_strdup_append_buffer: char *(char *, const char *)
+talloc_strndup: char *(const void *, const char *, size_t)
+talloc_strndup_append: char *(char *, const char *, size_t)
+talloc_strndup_append_buffer: char *(char *, const char *, size_t)
+talloc_total_blocks: size_t (const void *)
+talloc_total_size: size_t (const void *)
+talloc_unlink: int (const void *, void *)
+talloc_vasprintf: char *(const void *, const char *, va_list)
+talloc_vasprintf_append: char *(char *, const char *, va_list)
+talloc_vasprintf_append_buffer: char *(char *, const char *, va_list)
+talloc_version_major: int (void)
+talloc_version_minor: int (void)
diff --git a/ctdb/lib/talloc/ABI/talloc-2.0.6.sigs b/ctdb/lib/talloc/ABI/talloc-2.0.6.sigs
new file mode 100644
index 0000000000..6e236d52ac
--- /dev/null
+++ b/ctdb/lib/talloc/ABI/talloc-2.0.6.sigs
@@ -0,0 +1,62 @@
+_talloc: void *(const void *, size_t)
+_talloc_array: void *(const void *, size_t, unsigned int, const char *)
+_talloc_free: int (void *, const char *)
+_talloc_get_type_abort: void *(const void *, const char *, const char *)
+_talloc_memdup: void *(const void *, const void *, size_t, const char *)
+_talloc_move: void *(const void *, const void *)
+_talloc_realloc: void *(const void *, void *, size_t, const char *)
+_talloc_realloc_array: void *(const void *, void *, size_t, unsigned int, const char *)
+_talloc_reference_loc: void *(const void *, const void *, const char *)
+_talloc_set_destructor: void (const void *, int (*)(void *))
+_talloc_steal_loc: void *(const void *, const void *, const char *)
+_talloc_zero: void *(const void *, size_t, const char *)
+_talloc_zero_array: void *(const void *, size_t, unsigned int, const char *)
+talloc_asprintf: char *(const void *, const char *, ...)
+talloc_asprintf_append: char *(char *, const char *, ...)
+talloc_asprintf_append_buffer: char *(char *, const char *, ...)
+talloc_autofree_context: void *(void)
+talloc_check_name: void *(const void *, const char *)
+talloc_disable_null_tracking: void (void)
+talloc_enable_leak_report: void (void)
+talloc_enable_leak_report_full: void (void)
+talloc_enable_null_tracking: void (void)
+talloc_enable_null_tracking_no_autofree: void (void)
+talloc_find_parent_byname: void *(const void *, const char *)
+talloc_free_children: void (void *)
+talloc_get_name: const char *(const void *)
+talloc_get_size: size_t (const void *)
+talloc_increase_ref_count: int (const void *)
+talloc_init: void *(const char *, ...)
+talloc_is_parent: int (const void *, const void *)
+talloc_named: void *(const void *, size_t, const char *, ...)
+talloc_named_const: void *(const void *, size_t, const char *)
+talloc_parent: void *(const void *)
+talloc_parent_name: const char *(const void *)
+talloc_pool: void *(const void *, size_t)
+talloc_realloc_fn: void *(const void *, void *, size_t)
+talloc_reference_count: size_t (const void *)
+talloc_reparent: void *(const void *, const void *, const void *)
+talloc_report: void (const void *, FILE *)
+talloc_report_depth_cb: void (const void *, int, int, void (*)(const void *, int, int, int, void *), void *)
+talloc_report_depth_file: void (const void *, int, int, FILE *)
+talloc_report_full: void (const void *, FILE *)
+talloc_set_abort_fn: void (void (*)(const char *))
+talloc_set_log_fn: void (void (*)(const char *))
+talloc_set_log_stderr: void (void)
+talloc_set_name: const char *(const void *, const char *, ...)
+talloc_set_name_const: void (const void *, const char *)
+talloc_show_parents: void (const void *, FILE *)
+talloc_strdup: char *(const void *, const char *)
+talloc_strdup_append: char *(char *, const char *)
+talloc_strdup_append_buffer: char *(char *, const char *)
+talloc_strndup: char *(const void *, const char *, size_t)
+talloc_strndup_append: char *(char *, const char *, size_t)
+talloc_strndup_append_buffer: char *(char *, const char *, size_t)
+talloc_total_blocks: size_t (const void *)
+talloc_total_size: size_t (const void *)
+talloc_unlink: int (const void *, void *)
+talloc_vasprintf: char *(const void *, const char *, va_list)
+talloc_vasprintf_append: char *(char *, const char *, va_list)
+talloc_vasprintf_append_buffer: char *(char *, const char *, va_list)
+talloc_version_major: int (void)
+talloc_version_minor: int (void)
diff --git a/ctdb/lib/talloc/ABI/talloc-2.0.7.sigs b/ctdb/lib/talloc/ABI/talloc-2.0.7.sigs
new file mode 100644
index 0000000000..6e236d52ac
--- /dev/null
+++ b/ctdb/lib/talloc/ABI/talloc-2.0.7.sigs
@@ -0,0 +1,62 @@
+_talloc: void *(const void *, size_t)
+_talloc_array: void *(const void *, size_t, unsigned int, const char *)
+_talloc_free: int (void *, const char *)
+_talloc_get_type_abort: void *(const void *, const char *, const char *)
+_talloc_memdup: void *(const void *, const void *, size_t, const char *)
+_talloc_move: void *(const void *, const void *)
+_talloc_realloc: void *(const void *, void *, size_t, const char *)
+_talloc_realloc_array: void *(const void *, void *, size_t, unsigned int, const char *)
+_talloc_reference_loc: void *(const void *, const void *, const char *)
+_talloc_set_destructor: void (const void *, int (*)(void *))
+_talloc_steal_loc: void *(const void *, const void *, const char *)
+_talloc_zero: void *(const void *, size_t, const char *)
+_talloc_zero_array: void *(const void *, size_t, unsigned int, const char *)
+talloc_asprintf: char *(const void *, const char *, ...)
+talloc_asprintf_append: char *(char *, const char *, ...)
+talloc_asprintf_append_buffer: char *(char *, const char *, ...)
+talloc_autofree_context: void *(void)
+talloc_check_name: void *(const void *, const char *)
+talloc_disable_null_tracking: void (void)
+talloc_enable_leak_report: void (void)
+talloc_enable_leak_report_full: void (void)
+talloc_enable_null_tracking: void (void)
+talloc_enable_null_tracking_no_autofree: void (void)
+talloc_find_parent_byname: void *(const void *, const char *)
+talloc_free_children: void (void *)
+talloc_get_name: const char *(const void *)
+talloc_get_size: size_t (const void *)
+talloc_increase_ref_count: int (const void *)
+talloc_init: void *(const char *, ...)
+talloc_is_parent: int (const void *, const void *)
+talloc_named: void *(const void *, size_t, const char *, ...)
+talloc_named_const: void *(const void *, size_t, const char *)
+talloc_parent: void *(const void *)
+talloc_parent_name: const char *(const void *)
+talloc_pool: void *(const void *, size_t)
+talloc_realloc_fn: void *(const void *, void *, size_t)
+talloc_reference_count: size_t (const void *)
+talloc_reparent: void *(const void *, const void *, const void *)
+talloc_report: void (const void *, FILE *)
+talloc_report_depth_cb: void (const void *, int, int, void (*)(const void *, int, int, int, void *), void *)
+talloc_report_depth_file: void (const void *, int, int, FILE *)
+talloc_report_full: void (const void *, FILE *)
+talloc_set_abort_fn: void (void (*)(const char *))
+talloc_set_log_fn: void (void (*)(const char *))
+talloc_set_log_stderr: void (void)
+talloc_set_name: const char *(const void *, const char *, ...)
+talloc_set_name_const: void (const void *, const char *)
+talloc_show_parents: void (const void *, FILE *)
+talloc_strdup: char *(const void *, const char *)
+talloc_strdup_append: char *(char *, const char *)
+talloc_strdup_append_buffer: char *(char *, const char *)
+talloc_strndup: char *(const void *, const char *, size_t)
+talloc_strndup_append: char *(char *, const char *, size_t)
+talloc_strndup_append_buffer: char *(char *, const char *, size_t)
+talloc_total_blocks: size_t (const void *)
+talloc_total_size: size_t (const void *)
+talloc_unlink: int (const void *, void *)
+talloc_vasprintf: char *(const void *, const char *, va_list)
+talloc_vasprintf_append: char *(char *, const char *, va_list)
+talloc_vasprintf_append_buffer: char *(char *, const char *, va_list)
+talloc_version_major: int (void)
+talloc_version_minor: int (void)
diff --git a/ctdb/lib/talloc/ABI/talloc-2.0.8.sigs b/ctdb/lib/talloc/ABI/talloc-2.0.8.sigs
new file mode 100644
index 0000000000..15a9e9554f
--- /dev/null
+++ b/ctdb/lib/talloc/ABI/talloc-2.0.8.sigs
@@ -0,0 +1,63 @@
+_talloc: void *(const void *, size_t)
+_talloc_array: void *(const void *, size_t, unsigned int, const char *)
+_talloc_free: int (void *, const char *)
+_talloc_get_type_abort: void *(const void *, const char *, const char *)
+_talloc_memdup: void *(const void *, const void *, size_t, const char *)
+_talloc_move: void *(const void *, const void *)
+_talloc_realloc: void *(const void *, void *, size_t, const char *)
+_talloc_realloc_array: void *(const void *, void *, size_t, unsigned int, const char *)
+_talloc_reference_loc: void *(const void *, const void *, const char *)
+_talloc_set_destructor: void (const void *, int (*)(void *))
+_talloc_steal_loc: void *(const void *, const void *, const char *)
+_talloc_zero: void *(const void *, size_t, const char *)
+_talloc_zero_array: void *(const void *, size_t, unsigned int, const char *)
+talloc_asprintf: char *(const void *, const char *, ...)
+talloc_asprintf_append: char *(char *, const char *, ...)
+talloc_asprintf_append_buffer: char *(char *, const char *, ...)
+talloc_autofree_context: void *(void)
+talloc_check_name: void *(const void *, const char *)
+talloc_disable_null_tracking: void (void)
+talloc_enable_leak_report: void (void)
+talloc_enable_leak_report_full: void (void)
+talloc_enable_null_tracking: void (void)
+talloc_enable_null_tracking_no_autofree: void (void)
+talloc_find_parent_byname: void *(const void *, const char *)
+talloc_free_children: void (void *)
+talloc_get_name: const char *(const void *)
+talloc_get_size: size_t (const void *)
+talloc_increase_ref_count: int (const void *)
+talloc_init: void *(const char *, ...)
+talloc_is_parent: int (const void *, const void *)
+talloc_named: void *(const void *, size_t, const char *, ...)
+talloc_named_const: void *(const void *, size_t, const char *)
+talloc_parent: void *(const void *)
+talloc_parent_name: const char *(const void *)
+talloc_pool: void *(const void *, size_t)
+talloc_realloc_fn: void *(const void *, void *, size_t)
+talloc_reference_count: size_t (const void *)
+talloc_reparent: void *(const void *, const void *, const void *)
+talloc_report: void (const void *, FILE *)
+talloc_report_depth_cb: void (const void *, int, int, void (*)(const void *, int, int, int, void *), void *)
+talloc_report_depth_file: void (const void *, int, int, FILE *)
+talloc_report_full: void (const void *, FILE *)
+talloc_set_abort_fn: void (void (*)(const char *))
+talloc_set_log_fn: void (void (*)(const char *))
+talloc_set_log_stderr: void (void)
+talloc_set_memlimit: int (const void *, size_t)
+talloc_set_name: const char *(const void *, const char *, ...)
+talloc_set_name_const: void (const void *, const char *)
+talloc_show_parents: void (const void *, FILE *)
+talloc_strdup: char *(const void *, const char *)
+talloc_strdup_append: char *(char *, const char *)
+talloc_strdup_append_buffer: char *(char *, const char *)
+talloc_strndup: char *(const void *, const char *, size_t)
+talloc_strndup_append: char *(char *, const char *, size_t)
+talloc_strndup_append_buffer: char *(char *, const char *, size_t)
+talloc_total_blocks: size_t (const void *)
+talloc_total_size: size_t (const void *)
+talloc_unlink: int (const void *, void *)
+talloc_vasprintf: char *(const void *, const char *, va_list)
+talloc_vasprintf_append: char *(char *, const char *, va_list)
+talloc_vasprintf_append_buffer: char *(char *, const char *, va_list)
+talloc_version_major: int (void)
+talloc_version_minor: int (void)
diff --git a/ctdb/lib/talloc/NEWS b/ctdb/lib/talloc/NEWS
new file mode 100644
index 0000000000..e5b3aa0731
--- /dev/null
+++ b/ctdb/lib/talloc/NEWS
@@ -0,0 +1,13 @@
+1.0.1	26 May 2007
+
+ BUGS
+  
+   * Set name of correctly when using talloc_append_string() (metze)
+
+ LICENSE
+   
+   * Change license of files in lib/replace to LGPL (was GPL). (jelmer)
+
+1.0.0	30 April 2007
+ 
+ Initial release.
diff --git a/ctdb/lib/talloc/compat/talloc_compat1.c b/ctdb/lib/talloc/compat/talloc_compat1.c
new file mode 100644
index 0000000000..519e8c3a83
--- /dev/null
+++ b/ctdb/lib/talloc/compat/talloc_compat1.c
@@ -0,0 +1,51 @@
+/*
+   Samba trivial allocation library - compat functions
+
+   Copyright (C) Stefan Metzmacher 2009
+
+     ** NOTE! The following LGPL license applies to the talloc
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+/*
+ * This file contains only function to build a
+ * compat talloc.so.1 library on top of talloc.so.2
+ */
+
+#include "replace.h"
+#include "talloc.h"
+
+void *_talloc_reference(const void *context, const void *ptr);
+void *_talloc_reference(const void *context, const void *ptr) {
+	return _talloc_reference_loc(context, ptr,
+				     "Called from talloc compat1 "
+				     "_talloc_reference");
+}
+
+void *_talloc_steal(const void *new_ctx, const void *ptr);
+void *_talloc_steal(const void *new_ctx, const void *ptr)
+{
+	return talloc_reparent(talloc_parent(ptr), new_ctx, ptr);
+}
+
+#undef talloc_free
+int talloc_free(void *ptr);
+int talloc_free(void *ptr)
+{
+	return talloc_unlink(talloc_parent(ptr), ptr);
+}
+
diff --git a/ctdb/lib/talloc/compat/talloc_compat1.m4 b/ctdb/lib/talloc/compat/talloc_compat1.m4
new file mode 100644
index 0000000000..2ec530e079
--- /dev/null
+++ b/ctdb/lib/talloc/compat/talloc_compat1.m4
@@ -0,0 +1,14 @@
+TALLOC_COMPAT1_MK=""
+AC_SUBST(TALLOC_COMPAT1_MK)
+
+AC_ARG_ENABLE(talloc-compat1,
+	[AS_HELP_STRING([--enable-talloc-compat1],
+		[Build talloc 1.x.x compat library [default=no]])],
+	[ enable_talloc_compat1=$enableval ],
+	[ enable_talloc_compat1=no ]
+)
+
+if test "x$enable_talloc_compat1" = x"yes"; then
+	TALLOC_COMPAT1_MK='include $(tallocdir)/compat/talloc_compat1.mk'
+fi
+
diff --git a/ctdb/lib/talloc/compat/talloc_compat1.mk b/ctdb/lib/talloc/compat/talloc_compat1.mk
new file mode 100644
index 0000000000..d1817f0f71
--- /dev/null
+++ b/ctdb/lib/talloc/compat/talloc_compat1.mk
@@ -0,0 +1,21 @@
+talloccompatdir := $(tallocdir)/compat
+
+TALLOC_COMPAT1_VERSION_MAJOR = 1
+TALLOC_COMPAT1_OBJ = $(talloccompatdir)/talloc_compat1.o
+
+TALLOC_COMPAT1_SOLIB = libtalloc-compat1-$(TALLOC_VERSION).$(SHLIBEXT)
+TALLOC_COMPAT1_SONAME = libtalloc.$(SHLIBEXT).$(TALLOC_COMPAT1_VERSION_MAJOR)
+
+$(TALLOC_COMPAT1_SOLIB): $(TALLOC_COMPAT1_OBJ) $(TALLOC_SOLIB)
+	$(SHLD) $(SHLD_FLAGS) -o $@ $(TALLOC_COMPAT1_OBJ) \
+		$(TALLOC_SOLIB) $(SONAMEFLAG)$(TALLOC_COMPAT1_SONAME)
+
+all:: $(TALLOC_COMPAT1_SOLIB)
+
+install::
+	${INSTALLCMD} -d $(DESTDIR)$(libdir)
+	${INSTALLCMD} -m 755 $(TALLOC_COMPAT1_SOLIB) $(DESTDIR)$(libdir)
+
+clean::
+	rm -f $(TALLOC_COMPAT1_OBJ) $(TALLOC_COMPAT1_SOLIB)
+
diff --git a/ctdb/lib/talloc/doc/context.png b/ctdb/lib/talloc/doc/context.png
new file mode 100644
index 0000000000..48a6ca0e6a
--- /dev/null
+++ b/ctdb/lib/talloc/doc/context.png
diff --git a/ctdb/lib/talloc/doc/context_tree.png b/ctdb/lib/talloc/doc/context_tree.png
new file mode 100644
index 0000000000..9723459337
--- /dev/null
+++ b/ctdb/lib/talloc/doc/context_tree.png
diff --git a/ctdb/lib/talloc/doc/mainpage.dox b/ctdb/lib/talloc/doc/mainpage.dox
new file mode 100644
index 0000000000..3b56898a08
--- /dev/null
+++ b/ctdb/lib/talloc/doc/mainpage.dox
@@ -0,0 +1,110 @@
+/**
+ * @mainpage
+ *
+ * talloc is a hierarchical, reference counted memory pool system with
+ * destructors. It is the core memory allocator used in Samba.
+ *
+ * @section talloc_download Download
+ *
+ * You can download the latest releases of talloc from the
+ * <a href="http://samba.org/ftp/talloc" target="_blank">talloc directory</a>
+ * on the samba public source archive.
+ *
+ * @section main-tutorial Tutorial
+ *
+ * You should start by reading @subpage libtalloc_tutorial, then reading the documentation of
+ * the interesting functions as you go.
+
+ * @section talloc_bugs Discussion and bug reports
+ *
+ * talloc does not currently have its own mailing list or bug tracking system.
+ * For now, please use the
+ * <a href="https://lists.samba.org/mailman/listinfo/samba-technical" target="_blank">samba-technical</a>
+ * mailing list, and the
+ * <a href="http://bugzilla.samba.org/" target="_blank">Samba bugzilla</a>
+ * bug tracking system.
+ *
+ * @section talloc_devel Development
+ * You can download the latest code either via git or rsync.
+ *
+ * To fetch via git see the following guide:
+ *
+ * <a href="http://wiki.samba.org/index.php/Using_Git_for_Samba_Development" target="_blank">Using Git for Samba Development</a>
+ *
+ * Once you have cloned the tree switch to the master branch and cd into the
+ * lib/tevent directory.
+ *
+ * To fetch via rsync use this command:
+ *
+ * rsync -Pavz samba.org::ftp/unpacked/standalone_projects/lib/talloc .
+ *
+ * @section talloc_preample Preamble
+ *
+ * talloc is a hierarchical, reference counted memory pool system with
+ * destructors.
+ *
+ * Perhaps the biggest difference from other memory pool systems is that there
+ * is no distinction between a "talloc context" and a "talloc pointer". Any
+ * pointer returned from talloc() is itself a valid talloc context. This means
+ * you can do this:
+ *
+ * @code
+ *      struct foo *X = talloc(mem_ctx, struct foo);
+ *      X->name = talloc_strdup(X, "foo");
+ * @endcode
+ *
+ * The pointer X->name would be a "child" of the talloc context "X" which is
+ * itself a child of mem_ctx. So if you do talloc_free(mem_ctx) then it is all
+ * destroyed, whereas if you do talloc_free(X) then just X and X->name are
+ * destroyed, and if you do talloc_free(X->name) then just the name element of
+ * X is destroyed.
+ *
+ * If you think about this, then what this effectively gives you is an n-ary
+ * tree, where you can free any part of the tree with talloc_free().
+ *
+ * If you find this confusing, then run the testsuite to watch talloc in
+ * action. You may also like to add your own tests to testsuite.c to clarify
+ * how some particular situation is handled.
+ *
+ * @section talloc_performance Performance
+ *
+ * All the additional features of talloc() over malloc() do come at a price. We
+ * have a simple performance test in Samba4 that measures talloc() versus
+ * malloc() performance, and it seems that talloc() is about 4% slower than
+ * malloc() on my x86 Debian Linux box. For Samba, the great reduction in code
+ * complexity that we get by using talloc makes this worthwhile, especially as
+ * the total overhead of talloc/malloc in Samba is already quite small.
+ *
+ * @section talloc_named Named blocks
+ *
+ * Every talloc chunk has a name that can be used as a dynamic type-checking
+ * system. If for some reason like a callback function you had to cast a
+ * "struct foo *" to a "void *" variable, later you can safely reassign the
+ * "void *" pointer to a "struct foo *" by using the talloc_get_type() or
+ * talloc_get_type_abort() macros.
+ *
+ * @code
+ *      struct foo *X = talloc_get_type_abort(ptr, struct foo);
+ * @endcode
+ *
+ * This will abort if "ptr" does not contain a pointer that has been created
+ * with talloc(mem_ctx, struct foo).
+ *
+ * @section talloc_threading Multi-threading
+ *
+ * talloc itself does not deal with threads. It is thread-safe (assuming the
+ * underlying "malloc" is), as long as each thread uses different memory
+ * contexts.
+ *
+ * If two threads uses the same context then they need to synchronize in order
+ * to be safe. In particular:
+ *
+ *   - when using talloc_enable_leak_report(), giving directly NULL as a parent
+ *     context implicitly refers to a hidden "null context" global variable, so
+ *     this should not be used in a multi-threaded environment without proper
+ *     synchronization.
+ *   - the context returned by talloc_autofree_context() is also global so
+ *     shouldn't be used by several threads simultaneously without
+ *     synchronization.
+ *
+ */
diff --git a/ctdb/lib/talloc/doc/stealing.png b/ctdb/lib/talloc/doc/stealing.png
new file mode 100644
index 0000000000..8833e06a18
--- /dev/null
+++ b/ctdb/lib/talloc/doc/stealing.png
diff --git a/ctdb/lib/talloc/doc/tutorial_bestpractices.dox b/ctdb/lib/talloc/doc/tutorial_bestpractices.dox
new file mode 100644
index 0000000000..3634446743
--- /dev/null
+++ b/ctdb/lib/talloc/doc/tutorial_bestpractices.dox
@@ -0,0 +1,192 @@
+/**
+@page libtalloc_bestpractices Chapter 7: Best practises
+
+The following sections contain several best practices and good manners that were
+found by the <a href="http://www.samba.org">Samba</a> and
+<a href="https://fedorahosted.org/sssd">SSSD</a> developers over the years.
+These will help you to write code which is better, easier to debug and with as
+few (hopefully none) memory leaks as possible.
+
+@section bp-hierarchy Keep the context hierarchy steady
+
+The talloc is a hierarchy memory allocator. The hierarchy nature is what makes
+the programming more error proof. It makes the memory easier to manage and to
+free.  Therefore, the first thing we should have on our mind is: always project
+your data structures into the talloc context hierarchy.
+
+That means if we have a structure, we should always use it as a parent context
+for its elements. This way we will not encounter any troubles when freeing the
+structure or when changing its parent. The same rule applies for arrays.
+
+For example, the structure <code>user</code> from section @ref context-hierarchy
+should be created with the context hierarchy illustrated on the next image.
+
+@image html context_tree.png
+
+@section bp-tmpctx Every function should use its own context
+
+It is a good practice to create a temporary talloc context at the function
+beginning and free the context just before the return statement. All the data
+must be allocated on this context or on its children. This ensures that no
+memory leaks are created as long as we do not forget to free the temporary
+context.
+
+This pattern applies to both situations - when a function does not return any
+dynamically allocated value and when it does. However, it needs a little
+extension for the latter case.
+
+@subsection bp-tmpctx-1 Functions that do not return any dynamically allocated
+value
+
+If the function does not return any value created on the heap, we will just obey
+the aforementioned pattern.
+
+@code
+int bar()
+{
+  int ret;
+  TALLOC_CTX *tmp_ctx = talloc_new(NULL);
+  if (tmp_ctx == NULL) {
+    ret = ENOMEM;
+    goto done;
+  }
+  /* allocate data on tmp_ctx or on its descendants */
+  ret = EOK;
+done:
+  talloc_free(tmp_ctx);
+  return ret;
+}
+@endcode
+
+@subsection bp-tmpctx-2 Functions returning dynamically allocated values
+
+If our function returns any dynamically allocated data, its first parameter
+should always be the destination talloc context. This context serves as a parent
+for the output values. But again, we will create the output values as the
+descendants of the temporary context. If everything goes well, we will change
+the parent of the output values from the temporary to the destination talloc
+context.
+
+This pattern ensures that if an error occurs (e.g. I/O error or insufficient
+amount of the memory), all allocated data is freed and no garbage appears on
+the destination context.
+
+@code
+int struct_foo_init(TALLOC_CTX *mem_ctx, struct foo **_foo)
+{
+  int ret;
+  struct foo *foo = NULL;
+  TALLOC_CTX *tmp_ctx = talloc_new(NULL);
+  if (tmp_ctx == NULL) {
+    ret = ENOMEM;
+    goto done;
+  }
+  foo = talloc_zero(tmp_ctx, struct foo);
+  /* ... */
+  *_foo = talloc_steal(mem_ctx, foo);
+  ret = EOK;
+done:
+  talloc_free(tmp_ctx);
+  return ret;
+}
+@endcode
+
+@section bp-null Allocate temporary contexts on NULL
+
+As it can be seen on the previous listing, instead of allocating the temporary
+context directly on <code>mem_ctx</code>, we created a new top level context
+using <code>NULL</code> as the parameter for <code>talloc_new()</code> function.
+Take a look at the following example:
+
+@code
+char *create_user_filter(TALLOC_CTX *mem_ctx,
+                         uid_t uid, const char *username)
+{
+  char *filter = NULL;
+  char *sanitized_username = NULL;
+  /* tmp_ctx is a child of mem_ctx */
+  TALLOC_CTX *tmp_ctx = talloc_new(mem_ctx);
+  if (tmp_ctx == NULL) {
+    return NULL;
+  }
+
+  sanitized_username = sanitize_string(tmp_ctx, username);
+  if (sanitized_username == NULL) {
+    talloc_free(tmp_ctx);
+    return NULL;
+  }
+
+  filter = talloc_aprintf(tmp_ctx,"(|(uid=%llu)(uname=%s))",
+                          uid, sanitized_username);
+  if (filter == NULL) {
+    return NULL; /* tmp_ctx is not freed */ (*@\label{lst:tmp-ctx-3:leak}@*)
+  }
+
+  /* filter becomes a child of mem_ctx */
+  filter = talloc_steal(mem_ctx, filter);
+  talloc_free(tmp_ctx);
+  return filter;
+}
+@endcode
+
+We forgot to free <code>tmp_ctx</code> before the <code>return</code> statement
+in the <code>filter == NULL</code> condition. However, it is created as a child
+of <code>mem_ctx</code> context and as such it will be freed as soon as the
+<code>mem_ctx</code> is freed. Therefore, no detectable memory leak is created.
+
+On the other hand, we do not have any way to access the allocated data
+and for all we know <code>mem_ctx</code> may exist for the lifetime of our
+application. For these reasons this should be considered as a memory leak. How
+can we detect if it is unreferenced but still attached to its parent context?
+The only way is to notice the mistake in the source code.
+
+But if we create the temporary context as a top level context, it will not be
+freed and memory diagnostic tools
+(e.g. <a href="http://valgrind.org">valgrind</a>) are able to do their job.
+
+@section bp-pool Temporary contexts and the talloc pool
+
+If we want to take the advantage of the talloc pool but also keep to the
+pattern introduced in the previous section, we are unable to do it directly. The
+best thing to do is to create a conditional build where we can decide how do we
+want to create the temporary context. For example, we can create the following
+macros:
+
+@code
+#ifdef USE_POOL_CONTEXT
+  #define CREATE_POOL_CTX(ctx, size) talloc_pool(ctx, size)
+  #define CREATE_TMP_CTX(ctx)        talloc_new(ctx)
+#else
+  #define CREATE_POOL_CTX(ctx, size) talloc_new(ctx)
+  #define CREATE_TMP_CTX(ctx)        talloc_new(NULL)
+#endif
+@endcode
+
+Now if our application is under development, we will build it with macro
+<code>USE_POOL_CONTEXT</code> undefined. This way, we  can use memory diagnostic
+utilities to detect memory leaks.
+
+The release version will be compiled with the macro defined. This will  enable
+pool contexts and therefore reduce the <code>malloc()</code> calls, which will
+end up in a little bit faster processing.
+
+@code
+int struct_foo_init(TALLOC_CTX *mem_ctx, struct foo **_foo)
+{
+  int ret;
+  struct foo *foo = NULL;
+  TALLOC_CTX *tmp_ctx = CREATE_TMP_CTX(mem_ctx);
+  /* ... */
+}
+
+errno_t handle_request(TALLOC_CTX mem_ctx)
+{
+  int ret;
+  struct foo *foo = NULL;
+  TALLOC_CTX *pool_ctx = CREATE_POOL_CTX(NULL, 1024);
+  ret = struct_foo_init(mem_ctx, &foo);
+  /* ... */
+}
+@endcode
+
+*/
diff --git a/ctdb/lib/talloc/doc/tutorial_context.dox b/ctdb/lib/talloc/doc/tutorial_context.dox
new file mode 100644
index 0000000000..b8bfe26961
--- /dev/null
+++ b/ctdb/lib/talloc/doc/tutorial_context.dox
@@ -0,0 +1,198 @@
+/**
+@page libtalloc_context Chapter 1: Talloc context
+@section context Talloc context
+
+The talloc context is the most important part of this library and is
+responsible for every single feature of this memory allocator. It is a logical
+unit which represents a memory space managed by talloc.
+
+From the programmer's point of view, the talloc context is completely
+equivalent to a pointer that would be returned by the memory routines from the
+C standard library. This means that every context that is returned from the
+talloc library can be used directly in functions that do not use talloc
+internally. For example we can do the following:
+
+@code
+char *str1 = strdup("I am NOT a talloc context");
+char *str2 = talloc_strdup(NULL, "I AM a talloc context");
+
+printf("%d\n", strcmp(str1, str2) == 0);
+
+free(str1);
+talloc_free(str2); /* we can not use free() on str2 */
+@endcode
+
+This is possible because the context is internally handled as a special
+fixed-length structure called talloc chunk. Each chunk stores context metadata
+followed by the memory space requested by the programmer. When a talloc
+function returns a context (pointer), it will in fact return a pointer to the user
+space portion of the talloc chunk. If we to manipulate this context using
+talloc functions, the talloc library transforms the user-space pointer back to
+the starting address of the chunk. This is also the reason why we were unable
+to use <code>free(str2)</code> in the previous example - because
+<code>str2</code> does not point at the beginning of the allocated block of
+memory. This is illustrated on the next image:
+
+@image html context.png
+
+The type TALLOC_CTX is defined in talloc.h to identify a talloc context in
+function parameters. However, this type is just an alias for <code>void</code>
+and exists only for semantical reasons - thus we can differentiate between
+<code>void *</code> (arbitrary data) and <code>TALLOC_CTX *</code> (talloc
+context).
+
+@subsection metadata Context meta data
+
+Every talloc context carries several pieces of internal information along with
+the allocated memory:
+
+  - name - which is used in reports of context hierarchy and to simulate
+    a dynamic type system,
+  - size of the requested memory in bytes - this can be used to determine
+    the number of elements in arrays,
+  - attached destructor - which is executed just before the memory block is
+    about to be freed,
+  - references to the context
+  - children and parent contexts - create the hierarchical view on the
+    memory.
+
+@section context-hierarchy Hierarchy of talloc context
+
+Every talloc context contains information about its parent and children. Talloc
+uses this information to create a hierarchical model of memory or to be more
+precise, it creates an n-ary tree where each node represents a single talloc
+context. The root node of the tree is referred to as a top level context - a
+context without any parent.
+
+This approach has several advantages:
+
+  - as a consequence of freeing a talloc context, all of its children
+    will be properly deallocated as well,
+  - the parent of a context can be changed at any time, which
+    results in moving the whole subtree under another node,
+  - it creates a more natural way of managing data structures.
+
+@subsection Example
+
+We have a structure that stores basic information about a user - his/her name,
+identification number and groups he/she is a member of:
+
+@code
+struct user {
+  uid_t uid;
+  char *username;
+  size_t num_groups;
+  char **groups;
+};
+@endcode
+
+We will allocate this structure using talloc. The result will be the following
+context tree:
+
+@image html context_tree.png
+
+@code
+/* create new top level context */
+struct user *user = talloc(NULL, struct user);
+
+user->uid = 1000;
+user->num_groups = N;
+
+/* make user the parent of following contexts */
+user->username = talloc_strdup(user, "Test user");
+user->groups = talloc_array(user, char*, user->num_groups);
+
+for (i = 0; i < user->num_groups; i++) {
+  /* make user->groups the parent of following context */
+  user->groups[i] = talloc_asprintf(user->groups,
+                                    "Test group %d", i);
+}
+@endcode
+
+This way, we have gained a lot of additional capabilities, one of which is
+very simple deallocation of the structure and all of its elements.
+
+With the C standard library we need first to iterate over the array of groups
+and free every element separately. Then we must deallocate the array that stores
+them. Next we deallocate the username and as the last step free the structure
+itself. But with talloc, the only operation we need to execute is freeing the
+structure context. Its descendants will be freed automatically.
+
+@code
+talloc_free(user);
+@endcode
+
+@section keep-hierarchy Always keep the hieararchy steady!
+
+The talloc is a hierarchy memory allocator. The hierarchy nature is what makes
+the programming more error proof. It makes the memory easier to manage and to
+free.  Therefore, the first thing we should have on our mind is: <strong>always
+project our data structures into the talloc context hierarchy</strong>.
+
+That means if we have a structure, we should always use it as a parent context
+for its elements. This way we will not encounter any troubles when freeing this
+structure or when changing its parent. The same rule applies for arrays.
+
+@section creating-context Creating a talloc context
+
+Here are the most important functions that create a new talloc context.
+
+@subsection type-safe Type-safe functions
+
+It allocates the size that is necessary for the given type and returns a new,
+properly-casted pointer. This is the preferred way to create a new context as
+we can rely on the compiler to detect type mismatches.
+
+The name of the context is automatically set to the name of the data type which
+is used to simulate a dynamic type system.
+
+@code
+struct user *user = talloc(ctx, struct user);
+
+/* initialize to default values */
+user->uid = 0;
+user->name = NULL;
+user->num_groups = 0;
+user->groups = NULL;
+
+/* or we can achieve the same result with */
+struct user *user_zero = talloc_zero(ctx, struct user);
+@endcode
+
+@subsection zero-length Zero-length contexts
+
+The zero-length context is basically a context without any special semantical
+meaning. We can use it the same way as any other context. The only difference
+is that it consists only of the meta data about the context. Therefore, it is
+strictly of type <code>TALLOC_CTX*</code>. It is often used in cases where we
+want to aggregate several data structures under one parent (zero-length)
+context, such as a temporary context to contain memory needed within a single
+function that is not interesting to the caller. Allocating on a zero-length
+temporary context will make clean-up of the function simpler.
+
+@code
+TALLOC_CTX *tmp_ctx = NULL;
+struct foo *foo = NULL;
+struct bar *bar = NULL;
+
+/* new zero-length top level context */
+tmp_ctx = talloc_new(NULL);
+if (tmp_ctx == NULL) {
+  return ENOMEM;
+}
+
+foo = talloc(tmp_ctx, struct foo);
+bar = talloc(tmp_ctx, struct bar);
+
+/* free everything at once */
+talloc_free(tmp_ctx);
+@endcode
+
+@subsection context-see-also See also
+
+- talloc_size()
+- talloc_named()
+- @ref talloc_array
+- @ref talloc_string
+
+*/
diff --git a/ctdb/lib/talloc/doc/tutorial_debugging.dox b/ctdb/lib/talloc/doc/tutorial_debugging.dox
new file mode 100644
index 0000000000..aadbb0d12c
--- /dev/null
+++ b/ctdb/lib/talloc/doc/tutorial_debugging.dox
@@ -0,0 +1,116 @@
+/**
+@page libtalloc_debugging Chapter 6: Debugging
+
+Although talloc makes memory management significantly easier than the C standard
+library, developers are still only humans and can make mistakes. Therefore, it
+can be handy to know some tools for the inspection of talloc memory usage.
+
+@section log-abort Talloc log and abort
+
+We have already encountered the abort function in section @ref dts.
+In that case it was used when a type mismatch was detected. However, talloc
+calls this abort function in several more situations:
+
+- when the provided pointer is not a valid talloc context,
+- when the meta data is invalid - probably due to memory corruption,
+- and when an access after free is detected.
+
+The third one is probably the most interesting. It can help us with detecting
+an attempt to double-free a context or any other manipulation with it via
+talloc functions (using it as a parent, stealing it, etc.).
+
+Before the context is freed talloc sets a flag in the meta data. This is then
+used to detect the access after free. It basically works on the assumption that
+the memory stays unchanged (at least for a while) even when it is properly
+deallocated. This will work even if the memory is filled with the value
+specified in <code>TALLOC_FREE_FILL</code> environment variable, because it
+fills only the data part and leaves the meta data intact.
+
+Apart from the abort function, talloc uses a log function to provide additional
+information to the aforementioned violations. To enable logging we shall set the
+log function with one of:
+
+- talloc_set_log_fn()
+- talloc_set_log_stderr()
+
+The following code is a sample output of accessing a context after it has been
+freed:
+
+@code
+talloc_set_log_stderr();
+TALLOC_CTX *ctx = talloc_new(NULL);
+
+talloc_free(ctx);
+talloc_free(ctx);
+
+results in:
+talloc: access after free error - first free may be at ../src/main.c:55
+Bad talloc magic value - access after free
+@endcode
+
+Another example is an invalid context:
+
+@code
+talloc_set_log_stderr();
+TALLOC_CTX *ctx = talloc_new(NULL);
+char *str = strdup("not a talloc context");
+talloc_steal(ctx, str);
+
+results in:
+Bad talloc magic value - unknown value
+@endcode
+
+@section reports Memory usage reports
+
+Talloc can print reports of memory usage of a specified talloc context to a
+file (to <code>stdout</code> or <code>stderr</code>). The report can be
+simple or full. The simple report provides information only about the context
+itself and its direct descendants. The full report goes recursively through the
+entire context tree. See:
+
+- talloc_report()
+- talloc_report_full()
+
+We will use the following code to retrieve the sample report:
+
+@code
+struct foo {
+  char *str;
+};
+
+TALLOC_CTX *ctx = talloc_new(NULL);
+char *str =  talloc_strdup(ctx, "my string");
+struct foo *foo = talloc_zero(ctx, struct foo);
+foo->str = talloc_strdup(foo, "I am Foo");
+char *str2 = talloc_strdup(foo, "Foo is my parent");
+
+/* print full report */
+talloc_report_full(ctx, stdout);
+@endcode
+
+It will print a full report of <code>ctx</code> to the standard output.
+The message should be similar to:
+
+@code
+full talloc report on 'talloc_new: ../src/main.c:82' (total 46 bytes in 5 blocks)
+  struct foo contains 34 bytes in 3 blocks (ref 0) 0x1495130
+    Foo is my parent contains 17 bytes in 1 blocks (ref 0) 0x1495200
+    I am Foo contains 9 bytes in 1 blocks (ref 0) 0x1495190
+  my string contains 10 bytes in 1 blocks (ref 0) 0x14950c0
+@endcode
+
+We can notice in this report that something is wrong with the context containing
+<code>struct foo</code>. We know that the structure has only one string element.
+However, we can see in the report that it has two children. This indicates that
+we have either violated the memory hierarchy or forgotten to free it as
+temporary data. Looking into the code, we can see that <code>"Foo is my parent"
+</code> should be attached to <code>ctx</code>.
+
+See also:
+
+- talloc_enable_null_tracking()
+- talloc_disable_null_tracking()
+- talloc_enable_leak_report()
+- talloc_enable_leak_report_full()
+
+*/
diff --git a/ctdb/lib/talloc/doc/tutorial_destructors.dox b/ctdb/lib/talloc/doc/tutorial_destructors.dox
new file mode 100644
index 0000000000..ed063876a3
--- /dev/null
+++ b/ctdb/lib/talloc/doc/tutorial_destructors.dox
@@ -0,0 +1,82 @@
+/**
+@page libtalloc_destructors Chapter 4: Using destructors
+
+@section destructors Using destructors
+
+Destructors are well known methods in the world of object oriented programming.
+A destructor is a method of an object that is automatically run when the object
+is destroyed. It is usually used to return resources taken by the object back to
+the system (e.g. closing file descriptors, terminating connection to a database,
+deallocating memory).
+
+With talloc we can take the advantage of destructors even in C. We can easily
+attach our own destructor to a talloc context. When the context is freed, the
+destructor will run automatically.
+
+To attach/detach a destructor to a talloc context use: talloc_set_destructor().
+
+@section destructors-example Example
+
+Imagine that we have a dynamically created linked list. Before we deallocate an
+element of the list, we need to make sure that we have successfully removed it
+from the list. Normally, this would be done by two commands in the exact order:
+remove it from the list and then free the element. With talloc, we can do this
+at once by setting a destructor on the element which will remove it from the
+list and talloc_free() will do the rest.
+
+The destructor would be:
+
+@code
+int list_remove(void *ctx)
+{
+    struct list_el *el = NULL;
+    el = talloc_get_type_abort(ctx, struct list_el);
+    /* remove element from the list */
+}
+@endcode
+
+GCC version 3 and newer can check for the types during the compilation. So if
+it is our major compiler, we can use a more advanced destructor:
+
+@code
+int list_remove(struct list_el *el)
+{
+    /* remove element from the list */
+}
+@endcode
+
+Now we will assign the destructor to the list element. We can do this directly
+in the function that inserts it.
+
+@code
+struct list_el* list_insert(TALLOC_CTX *mem_ctx,
+                            struct list_el *where,
+                            void *ptr)
+{
+  struct list_el *el = talloc(mem_ctx, struct list_el);
+  el->data = ptr;
+  /* insert into list */
+
+  talloc_set_destructor(el, list_remove);
+  return el;
+}
+@endcode
+
+Because talloc is a hierarchical memory allocator, we can go a step further and
+free the data with the element as well:
+
+@code
+struct list_el* list_insert_free(TALLOC_CTX *mem_ctx,
+                                 struct list_el *where,
+                                 void *ptr)
+{
+  struct list_el *el = NULL;
+  el = list_insert(mem_ctx, where, ptr);
+
+  talloc_steal(el, ptr);
+
+  return el;
+}
+@endcode
+
+*/
diff --git a/ctdb/lib/talloc/doc/tutorial_dts.dox b/ctdb/lib/talloc/doc/tutorial_dts.dox
new file mode 100644
index 0000000000..75b5172bbe
--- /dev/null
+++ b/ctdb/lib/talloc/doc/tutorial_dts.dox
@@ -0,0 +1,109 @@
+/**
+@page libtalloc_dts Chapter 3: Dynamic type system
+
+@section dts Dynamic type system
+
+Generic programming in the C language is very difficult. There is no inheritance
+nor templates known from object oriented languages. There is no dynamic type
+system. Therefore, generic programming in this language is usually done by
+type-casting a variable to <code>void*</code> and transferring it through
+a generic function to a specialized callback as illustrated on the next listing.
+
+@code
+void generic_function(callback_fn cb, void *pvt)
+{
+  /* do some stuff and call the callback */
+  cb(pvt);
+}
+
+void specific_callback(void *pvt)
+{
+  struct specific_struct *data;
+  data = (struct specific_struct*)pvt;
+  /* ... */
+}
+
+void specific_function()
+{
+  struct specific_struct data;
+  generic_function(callback, &data);
+}
+@endcode
+
+Unfortunately, the type information is lost as a result of this type cast. The
+compiler cannot check the type during the compilation nor are we able to do it
+at runtime. Providing an invalid data type to the callback will result in
+unexpected behaviour (not necessarily a crash) of the application. This mistake
+is usually hard to detect because it is not the first thing which comes the
+mind.
+
+As we already know, every talloc context contains a name. This name is available
+at any time and it can be used to determine the type of a context even if we
+lose the type of a variable.
+
+Although the name of the context can be set to any arbitrary string, the best
+way of using it to simulate the dynamic type system is to set it directly to the
+type of the variable.
+
+It is recommended to use one of talloc() and talloc_array() (or its
+variants) to create the context as they set its name to the name of the
+given type automatically.
+
+If we have a context with such as a name, we can use two similar functions that
+do both the type check and the type cast for us:
+
+- talloc_get_type()
+- talloc_get_type_abort()
+
+@section dts-examples Examples
+
+The following example will show how generic programming with talloc is handled -
+if we provide invalid data to the callback, the program will be aborted. This
+is a sufficient reaction for such an error in most applications.
+
+@code
+void foo_callback(void *pvt)
+{
+  struct foo *data = talloc_get_type_abort(pvt, struct foo);
+  /* ... */
+}
+
+int do_foo()
+{
+  struct foo *data = talloc_zero(NULL, struct foo);
+  /* ... */
+  return generic_function(foo_callback, data);
+}
+@endcode
+
+But what if we are creating a service application that should be running for the
+uptime of a server, we may want to abort the application during the development
+process (to make sure the error is not overlooked) and try to recover from the
+error in the customer release. This can be achieved by creating a custom abort
+function with a conditional build.
+
+@code
+void my_abort(const char *reason)
+{
+  fprintf(stderr, "talloc abort: %s\n", reason);
+#ifdef ABORT_ON_TYPE_MISMATCH
+  abort();
+#endif
+}
+@endcode
+
+The usage of talloc_get_type_abort() would be then:
+
+@code
+talloc_set_abort_fn(my_abort);
+
+TALLOC_CTX *ctx = talloc_new(NULL);
+char *str = talloc_get_type_abort(ctx, char);
+if (str == NULL) {
+  /* recovery code */
+}
+/* talloc abort: ../src/main.c:25: Type mismatch:
+   name[talloc_new: ../src/main.c:24] expected[char] */
+@endcode
+
+*/
diff --git a/ctdb/lib/talloc/doc/tutorial_introduction.dox b/ctdb/lib/talloc/doc/tutorial_introduction.dox
new file mode 100644
index 0000000000..02777b9f77
--- /dev/null
+++ b/ctdb/lib/talloc/doc/tutorial_introduction.dox
@@ -0,0 +1,43 @@
+/**
+@page libtalloc_tutorial The Tutorial
+@section introduction Introduction
+
+Talloc is a hierarchical, reference counted memory pool system with destructors.
+It is built atop the C standard library and it defines a set of utility
+functions that altogether simplifies allocation and deallocation of data,
+especially for complex structures that contain many dynamically allocated
+elements such as strings and arrays.
+
+The main goals of this library are: removing the needs for creating a cleanup
+function for every complex structure, providing a logical organization of
+allocated memory blocks and reducing the likelihood of creating memory leaks in
+long-running applications. All of this is achieved by allocating memory in a
+hierarchical structure of talloc contexts such that deallocating one context
+recursively frees all of its descendants as well.
+
+@section main-features Main features
+- An open source project
+- A hierarchical memory model
+- Natural projection of data structures into the memory space
+- Simplifies memory management of large data structures
+- Automatic execution of a destructor before the memory is freed
+- Simulates a dynamic type system
+- Implements a transparent memory pool
+
+@section toc Table of contents:
+
+@subpage libtalloc_context
+
+@subpage libtalloc_stealing
+
+@subpage libtalloc_dts
+
+@subpage libtalloc_destructors
+
+@subpage libtalloc_pools
+
+@subpage libtalloc_debugging
+
+@subpage libtalloc_bestpractices
+
+*/
\ No newline at end of file
diff --git a/ctdb/lib/talloc/doc/tutorial_pools.dox b/ctdb/lib/talloc/doc/tutorial_pools.dox
new file mode 100644
index 0000000000..a0d1e1ac6f
--- /dev/null
+++ b/ctdb/lib/talloc/doc/tutorial_pools.dox
@@ -0,0 +1,93 @@
+/**
+@page libtalloc_pools Chapter 5: Memory pools
+
+@section pools Memory pools
+
+Allocation of a new memory is an expensive operation and large programs can
+contain thousands of calls of malloc() for a single computation, where every
+call allocates only a very small amount of the memory. This can result in an
+undesirable slowdown of the application. We can avoid this slowdown by
+decreasing the number of malloc() calls by using a memory pool.
+
+A memory pool is a preallocated memory space with a fixed size. If we need to
+allocate new data we will take the desired amount of the memory from the pool
+instead of requesting a new memory from the system. This is done by creating a
+pointer that points inside the preallocated memory. Such a pool must not be
+reallocated as it would change its location - pointers that were pointing
+inside the pool would become invalid. Therefore, a memory pool requires a very
+good estimate of the required memory space.
+
+The talloc library contains its own implementation of a memory pool. It is
+highly transparent for the programmer. The only thing that needs to be done is
+an initialization of a new pool context using talloc_pool() -
+which can be used in the same way as any other context.
+
+Refactoring of existing code (that uses talloc) to take the advantage of a
+memory pool is quite simple due to the following properties of the pool context:
+
+- if we are allocating data on a pool context, it takes the desired
+  amount of memory from the pool,
+- if the context is a descendant of the pool context, it takes the space
+  from the pool as well,
+- if the pool does not have sufficient portion of memory left, it will
+  create a new non-pool context, leaving the pool intact
+
+@code
+/* allocate 1KiB in a pool */
+TALLOC_CTX *pool_ctx = talloc_pool(NULL, 1024);
+
+/* Take 512B from the pool, 512B is left there */
+void *ptr = talloc_size(pool_ctx, 512);
+
+/* 1024B > 512B, this will create new talloc chunk outside
+   the pool */
+void *ptr2 = talloc_size(ptr, 1024);
+
+/* The pool still contains 512 free bytes
+ * this will take 200B from them. */
+void *ptr3 = talloc_size(ptr, 200);
+
+/* This will destroy context 'ptr3' but the memory
+ * is not freed, the available space in the pool
+ * will increase to 512B. */
+talloc_free(ptr3);
+
+/* This will free memory taken by 'pool_ctx'
+ * and 'ptr2' as well. */
+talloc_free(pool_ctx);
+@endcode
+
+The above given is very convenient, but there is one big issue to be kept in
+mind. If the parent of a talloc pool child is changed to a parent that is
+outside of this pool, the whole pool memory will not be freed until the child is
+freed. For this reason we must be very careful when stealing a descendant of a
+pool context.
+
+@code
+TALLOC_CTX *mem_ctx = talloc_new(NULL);
+TALLOC_CTX *pool_ctx = talloc_pool(NULL, 1024);
+struct foo *foo = talloc(pool_ctx, struct foo);
+
+/* mem_ctx is not in the pool */
+talloc_steal(mem_ctx, foo);
+
+/* pool_ctx is marked as freed but the memory is not
+   deallocated, accessing the pool_ctx again will cause
+   an error */
+talloc_free(pool_ctx);
+
+/* This deallocates the pool_ctx. */
+talloc_free(mem_ctx);
+@endcode
+
+It may often be better to copy the memory we want instead of stealing it to
+avoid this problem. If we do not need to retain the context name (to keep the
+type information), we can use talloc_memdup() to do this.
+
+Copying the memory out of the pool may, however, discard all the performance
+boost given by the pool, depending on the size of the copied memory. Therefore,
+the code should be well profiled before taking this path. In general, the
+golden rule is: if we need to steal from the pool context, we should not
+use a pool context.
+
+*/
diff --git a/ctdb/lib/talloc/doc/tutorial_stealing.dox b/ctdb/lib/talloc/doc/tutorial_stealing.dox
new file mode 100644
index 0000000000..67eae1dfc7
--- /dev/null
+++ b/ctdb/lib/talloc/doc/tutorial_stealing.dox
@@ -0,0 +1,55 @@
+/**
+@page libtalloc_stealing Chapter 2: Stealing a context
+
+@section stealing Stealing a context
+
+Talloc has the ability to change the parent of a talloc context to another
+one. This operation is commonly referred to as stealing and it is one of
+the most important actions performed with talloc contexts.
+
+Stealing a context is necessary if we want the pointer to outlive the context it
+is created on. This has many possible use cases, for instance stealing a result
+of a database search to an in-memory cache context, changing the parent of a
+field of a generic structure to a more specific one or vice-versa. The most
+common scenario, at least in Samba, is to steal output data from a function-specific
+context to the output context given as an argument of that function.
+
+@code
+struct foo {
+    char *a1;
+    char *a2;
+    char *a3;
+};
+
+struct bar {
+    char *wurst;
+    struct foo *foo;
+};
+
+struct foo *foo = talloc_zero(ctx, struct foo);
+foo->a1 = talloc_strdup(foo, "a1");
+foo->a2 = talloc_strdup(foo, "a2");
+foo->a3 = talloc_strdup(foo, "a3");
+
+struct bar *bar = talloc_zero(NULL, struct bar);
+/* change parent of foo from ctx to bar */
+bar->foo = talloc_steal(bar, foo);
+
+/* or do the same but assign foo = NULL */
+bar->foo = talloc_move(bar, &foo);
+@endcode
+
+The talloc_move() function is similar to the talloc_steal() function but
+additionally sets the source pointer to NULL.
+
+In general, the source pointer itself is not changed (it only replaces the
+parent in the meta data). But the common usage is that the result is
+assigned to another variable, thus further accessing the pointer from the
+original variable should be avoided unless it is necessary. In this case
+talloc_move() is the preferred way of stealing a context. Additionally sets the
+source pointer to NULL, thus.protects the pointer from being accidentally freed
+and accessed using the old variable after its parent has been changed.
+
+@image html stealing.png
+
+*/
diff --git a/ctdb/lib/talloc/doxy.config b/ctdb/lib/talloc/doxy.config
new file mode 100644
index 0000000000..0e27d61299
--- /dev/null
+++ b/ctdb/lib/talloc/doxy.config
@@ -0,0 +1,1807 @@
+# Doxyfile 1.8.0
+
+# This file describes the settings to be used by the documentation system
+# doxygen (www.doxygen.org) for a project.
+#
+# All text after a hash (#) is considered a comment and will be ignored.
+# The format is:
+#       TAG = value [value, ...]
+# For lists items can also be appended using:
+#       TAG += value [value, ...]
+# Values that contain spaces should be placed between quotes (" ").
+
+#---------------------------------------------------------------------------
+# Project related configuration options
+#---------------------------------------------------------------------------
+
+# This tag specifies the encoding used for all characters in the config file
+# that follow. The default is UTF-8 which is also the encoding used for all
+# text before the first occurrence of this tag. Doxygen uses libiconv (or the
+# iconv built into libc) for the transcoding. See
+# http://www.gnu.org/software/libiconv for the list of possible encodings.
+
+DOXYFILE_ENCODING      = UTF-8
+
+# The PROJECT_NAME tag is a single word (or sequence of words) that should
+# identify the project. Note that if you do not use Doxywizard you need
+# to put quotes around the project name if it contains spaces.
+
+PROJECT_NAME           = talloc
+
+# The PROJECT_NUMBER tag can be used to enter a project or revision number.
+# This could be handy for archiving the generated documentation or
+# if some version control system is used.
+
+PROJECT_NUMBER         = 2.0
+
+# Using the PROJECT_BRIEF tag one can provide an optional one line description
+# for a project that appears at the top of each page and should give viewer
+# a quick idea about the purpose of the project. Keep the description short.
+
+PROJECT_BRIEF          =
+
+# With the PROJECT_LOGO tag one can specify an logo or icon that is
+# included in the documentation. The maximum height of the logo should not
+# exceed 55 pixels and the maximum width should not exceed 200 pixels.
+# Doxygen will copy the logo to the output directory.
+
+PROJECT_LOGO           =
+
+# The OUTPUT_DIRECTORY tag is used to specify the (relative or absolute)
+# base path where the generated documentation will be put.
+# If a relative path is entered, it will be relative to the location
+# where doxygen was started. If left blank the current directory will be used.
+
+OUTPUT_DIRECTORY       = doc
+
+# If the CREATE_SUBDIRS tag is set to YES, then doxygen will create
+# 4096 sub-directories (in 2 levels) under the output directory of each output
+# format and will distribute the generated files over these directories.
+# Enabling this option can be useful when feeding doxygen a huge amount of
+# source files, where putting all generated files in the same directory would
+# otherwise cause performance problems for the file system.
+
+CREATE_SUBDIRS         = NO
+
+# The OUTPUT_LANGUAGE tag is used to specify the language in which all
+# documentation generated by doxygen is written. Doxygen will use this
+# information to generate all constant output in the proper language.
+# The default language is English, other supported languages are:
+# Afrikaans, Arabic, Brazilian, Catalan, Chinese, Chinese-Traditional,
+# Croatian, Czech, Danish, Dutch, Esperanto, Farsi, Finnish, French, German,
+# Greek, Hungarian, Italian, Japanese, Japanese-en (Japanese with English
+# messages), Korean, Korean-en, Lithuanian, Norwegian, Macedonian, Persian,
+# Polish, Portuguese, Romanian, Russian, Serbian, Serbian-Cyrillic, Slovak,
+# Slovene, Spanish, Swedish, Ukrainian, and Vietnamese.
+
+OUTPUT_LANGUAGE        = English
+
+# If the BRIEF_MEMBER_DESC tag is set to YES (the default) Doxygen will
+# include brief member descriptions after the members that are listed in
+# the file and class documentation (similar to JavaDoc).
+# Set to NO to disable this.
+
+BRIEF_MEMBER_DESC      = YES
+
+# If the REPEAT_BRIEF tag is set to YES (the default) Doxygen will prepend
+# the brief description of a member or function before the detailed description.
+# Note: if both HIDE_UNDOC_MEMBERS and BRIEF_MEMBER_DESC are set to NO, the
+# brief descriptions will be completely suppressed.
+
+REPEAT_BRIEF           = YES
+
+# This tag implements a quasi-intelligent brief description abbreviator
+# that is used to form the text in various listings. Each string
+# in this list, if found as the leading text of the brief description, will be
+# stripped from the text and the result after processing the whole list, is
+# used as the annotated text. Otherwise, the brief description is used as-is.
+# If left blank, the following values are used ("$name" is automatically
+# replaced with the name of the entity): "The $name class" "The $name widget"
+# "The $name file" "is" "provides" "specifies" "contains"
+# "represents" "a" "an" "the"
+
+ABBREVIATE_BRIEF       = "The $name class" \
+                         "The $name widget" \
+                         "The $name file" \
+                         is \
+                         provides \
+                         specifies \
+                         contains \
+                         represents \
+                         a \
+                         an \
+                         the
+
+# If the ALWAYS_DETAILED_SEC and REPEAT_BRIEF tags are both set to YES then
+# Doxygen will generate a detailed section even if there is only a brief
+# description.
+
+ALWAYS_DETAILED_SEC    = NO
+
+# If the INLINE_INHERITED_MEMB tag is set to YES, doxygen will show all
+# inherited members of a class in the documentation of that class as if those
+# members were ordinary class members. Constructors, destructors and assignment
+# operators of the base classes will not be shown.
+
+INLINE_INHERITED_MEMB  = NO
+
+# If the FULL_PATH_NAMES tag is set to YES then Doxygen will prepend the full
+# path before files name in the file list and in the header files. If set
+# to NO the shortest path that makes the file name unique will be used.
+
+FULL_PATH_NAMES        = YES
+
+# If the FULL_PATH_NAMES tag is set to YES then the STRIP_FROM_PATH tag
+# can be used to strip a user-defined part of the path. Stripping is
+# only done if one of the specified strings matches the left-hand part of
+# the path. The tag can be used to show relative paths in the file list.
+# If left blank the directory from which doxygen is run is used as the
+# path to strip.
+
+STRIP_FROM_PATH        =
+
+# The STRIP_FROM_INC_PATH tag can be used to strip a user-defined part of
+# the path mentioned in the documentation of a class, which tells
+# the reader which header file to include in order to use a class.
+# If left blank only the name of the header file containing the class
+# definition is used. Otherwise one should specify the include paths that
+# are normally passed to the compiler using the -I flag.
+
+STRIP_FROM_INC_PATH    =
+
+# If the SHORT_NAMES tag is set to YES, doxygen will generate much shorter
+# (but less readable) file names. This can be useful if your file system
+# doesn't support long names like on DOS, Mac, or CD-ROM.
+
+SHORT_NAMES            = NO
+
+# If the JAVADOC_AUTOBRIEF tag is set to YES then Doxygen
+# will interpret the first line (until the first dot) of a JavaDoc-style
+# comment as the brief description. If set to NO, the JavaDoc
+# comments will behave just like regular Qt-style comments
+# (thus requiring an explicit @brief command for a brief description.)
+
+JAVADOC_AUTOBRIEF      = YES
+
+# If the QT_AUTOBRIEF tag is set to YES then Doxygen will
+# interpret the first line (until the first dot) of a Qt-style
+# comment as the brief description. If set to NO, the comments
+# will behave just like regular Qt-style comments (thus requiring
+# an explicit \brief command for a brief description.)
+
+QT_AUTOBRIEF           = NO
+
+# The MULTILINE_CPP_IS_BRIEF tag can be set to YES to make Doxygen
+# treat a multi-line C++ special comment block (i.e. a block of //! or ///
+# comments) as a brief description. This used to be the default behaviour.
+# The new default is to treat a multi-line C++ comment block as a detailed
+# description. Set this tag to YES if you prefer the old behaviour instead.
+
+MULTILINE_CPP_IS_BRIEF = NO
+
+# If the INHERIT_DOCS tag is set to YES (the default) then an undocumented
+# member inherits the documentation from any documented member that it
+# re-implements.
+
+INHERIT_DOCS           = YES
+
+# If the SEPARATE_MEMBER_PAGES tag is set to YES, then doxygen will produce
+# a new page for each member. If set to NO, the documentation of a member will
+# be part of the file/class/namespace that contains it.
+
+SEPARATE_MEMBER_PAGES  = NO
+
+# The TAB_SIZE tag can be used to set the number of spaces in a tab.
+# Doxygen uses this value to replace tabs by spaces in code fragments.
+
+TAB_SIZE               = 8
+
+# This tag can be used to specify a number of aliases that acts
+# as commands in the documentation. An alias has the form "name=value".
+# For example adding "sideeffect=\par Side Effects:\n" will allow you to
+# put the command \sideeffect (or @sideeffect) in the documentation, which
+# will result in a user-defined paragraph with heading "Side Effects:".
+# You can put \n's in the value part of an alias to insert newlines.
+
+ALIASES                =
+
+# This tag can be used to specify a number of word-keyword mappings (TCL only).
+# A mapping has the form "name=value". For example adding
+# "class=itcl::class" will allow you to use the command class in the
+# itcl::class meaning.
+
+TCL_SUBST              =
+
+# Set the OPTIMIZE_OUTPUT_FOR_C tag to YES if your project consists of C
+# sources only. Doxygen will then generate output that is more tailored for C.
+# For instance, some of the names that are used will be different. The list
+# of all members will be omitted, etc.
+
+OPTIMIZE_OUTPUT_FOR_C  = YES
+
+# Set the OPTIMIZE_OUTPUT_JAVA tag to YES if your project consists of Java
+# sources only. Doxygen will then generate output that is more tailored for
+# Java. For instance, namespaces will be presented as packages, qualified
+# scopes will look different, etc.
+
+OPTIMIZE_OUTPUT_JAVA   = NO
+
+# Set the OPTIMIZE_FOR_FORTRAN tag to YES if your project consists of Fortran
+# sources only. Doxygen will then generate output that is more tailored for
+# Fortran.
+
+OPTIMIZE_FOR_FORTRAN   = NO
+
+# Set the OPTIMIZE_OUTPUT_VHDL tag to YES if your project consists of VHDL
+# sources. Doxygen will then generate output that is tailored for
+# VHDL.
+
+OPTIMIZE_OUTPUT_VHDL   = NO
+
+# Doxygen selects the parser to use depending on the extension of the files it
+# parses. With this tag you can assign which parser to use for a given extension.
+# Doxygen has a built-in mapping, but you can override or extend it using this
+# tag. The format is ext=language, where ext is a file extension, and language
+# is one of the parsers supported by doxygen: IDL, Java, Javascript, CSharp, C,
+# C++, D, PHP, Objective-C, Python, Fortran, VHDL, C, C++. For instance to make
+# doxygen treat .inc files as Fortran files (default is PHP), and .f files as C
+# (default is Fortran), use: inc=Fortran f=C. Note that for custom extensions
+# you also need to set FILE_PATTERNS otherwise the files are not read by doxygen.
+
+EXTENSION_MAPPING      =
+
+# If MARKDOWN_SUPPORT is enabled (the default) then doxygen pre-processes all
+# comments according to the Markdown format, which allows for more readable
+# documentation. See http://daringfireball.net/projects/markdown/ for details.
+# The output of markdown processing is further processed by doxygen, so you
+# can mix doxygen, HTML, and XML commands with Markdown formatting.
+# Disable only in case of backward compatibilities issues.
+
+MARKDOWN_SUPPORT       = YES
+
+# If you use STL classes (i.e. std::string, std::vector, etc.) but do not want
+# to include (a tag file for) the STL sources as input, then you should
+# set this tag to YES in order to let doxygen match functions declarations and
+# definitions whose arguments contain STL classes (e.g. func(std::string); v.s.
+# func(std::string) {}). This also makes the inheritance and collaboration
+# diagrams that involve STL classes more complete and accurate.
+
+BUILTIN_STL_SUPPORT    = NO
+
+# If you use Microsoft's C++/CLI language, you should set this option to YES to
+# enable parsing support.
+
+CPP_CLI_SUPPORT        = NO
+
+# Set the SIP_SUPPORT tag to YES if your project consists of sip sources only.
+# Doxygen will parse them like normal C++ but will assume all classes use public
+# instead of private inheritance when no explicit protection keyword is present.
+
+SIP_SUPPORT            = NO
+
+# For Microsoft's IDL there are propget and propput attributes to indicate getter
+# and setter methods for a property. Setting this option to YES (the default)
+# will make doxygen replace the get and set methods by a property in the
+# documentation. This will only work if the methods are indeed getting or
+# setting a simple type. If this is not the case, or you want to show the
+# methods anyway, you should set this option to NO.
+
+IDL_PROPERTY_SUPPORT   = YES
+
+# If member grouping is used in the documentation and the DISTRIBUTE_GROUP_DOC
+# tag is set to YES, then doxygen will reuse the documentation of the first
+# member in the group (if any) for the other members of the group. By default
+# all members of a group must be documented explicitly.
+
+DISTRIBUTE_GROUP_DOC   = NO
+
+# Set the SUBGROUPING tag to YES (the default) to allow class member groups of
+# the same type (for instance a group of public functions) to be put as a
+# subgroup of that type (e.g. under the Public Functions section). Set it to
+# NO to prevent subgrouping. Alternatively, this can be done per class using
+# the \nosubgrouping command.
+
+SUBGROUPING            = YES
+
+# When the INLINE_GROUPED_CLASSES tag is set to YES, classes, structs and
+# unions are shown inside the group in which they are included (e.g. using
+# @ingroup) instead of on a separate page (for HTML and Man pages) or
+# section (for LaTeX and RTF).
+
+INLINE_GROUPED_CLASSES = NO
+
+# When the INLINE_SIMPLE_STRUCTS tag is set to YES, structs, classes, and
+# unions with only public data fields will be shown inline in the documentation
+# of the scope in which they are defined (i.e. file, namespace, or group
+# documentation), provided this scope is documented. If set to NO (the default),
+# structs, classes, and unions are shown on a separate page (for HTML and Man
+# pages) or section (for LaTeX and RTF).
+
+INLINE_SIMPLE_STRUCTS  = NO
+
+# When TYPEDEF_HIDES_STRUCT is enabled, a typedef of a struct, union, or enum
+# is documented as struct, union, or enum with the name of the typedef. So
+# typedef struct TypeS {} TypeT, will appear in the documentation as a struct
+# with name TypeT. When disabled the typedef will appear as a member of a file,
+# namespace, or class. And the struct will be named TypeS. This can typically
+# be useful for C code in case the coding convention dictates that all compound
+# types are typedef'ed and only the typedef is referenced, never the tag name.
+
+TYPEDEF_HIDES_STRUCT   = NO
+
+# The SYMBOL_CACHE_SIZE determines the size of the internal cache use to
+# determine which symbols to keep in memory and which to flush to disk.
+# When the cache is full, less often used symbols will be written to disk.
+# For small to medium size projects (<1000 input files) the default value is
+# probably good enough. For larger projects a too small cache size can cause
+# doxygen to be busy swapping symbols to and from disk most of the time
+# causing a significant performance penalty.
+# If the system has enough physical memory increasing the cache will improve the
+# performance by keeping more symbols in memory. Note that the value works on
+# a logarithmic scale so increasing the size by one will roughly double the
+# memory usage. The cache size is given by this formula:
+# 2^(16+SYMBOL_CACHE_SIZE). The valid range is 0..9, the default is 0,
+# corresponding to a cache size of 2^16 = 65536 symbols.
+
+SYMBOL_CACHE_SIZE      = 0
+
+# Similar to the SYMBOL_CACHE_SIZE the size of the symbol lookup cache can be
+# set using LOOKUP_CACHE_SIZE. This cache is used to resolve symbols given
+# their name and scope. Since this can be an expensive process and often the
+# same symbol appear multiple times in the code, doxygen keeps a cache of
+# pre-resolved symbols. If the cache is too small doxygen will become slower.
+# If the cache is too large, memory is wasted. The cache size is given by this
+# formula: 2^(16+LOOKUP_CACHE_SIZE). The valid range is 0..9, the default is 0,
+# corresponding to a cache size of 2^16 = 65536 symbols.
+
+LOOKUP_CACHE_SIZE      = 0
+
+#---------------------------------------------------------------------------
+# Build related configuration options
+#---------------------------------------------------------------------------
+
+# If the EXTRACT_ALL tag is set to YES doxygen will assume all entities in
+# documentation are documented, even if no documentation was available.
+# Private class members and static file members will be hidden unless
+# the EXTRACT_PRIVATE and EXTRACT_STATIC tags are set to YES
+
+EXTRACT_ALL            = NO
+
+# If the EXTRACT_PRIVATE tag is set to YES all private members of a class
+# will be included in the documentation.
+
+EXTRACT_PRIVATE        = NO
+
+# If the EXTRACT_PACKAGE tag is set to YES all members with package or internal scope will be included in the documentation.
+
+EXTRACT_PACKAGE        = NO
+
+# If the EXTRACT_STATIC tag is set to YES all static members of a file
+# will be included in the documentation.
+
+EXTRACT_STATIC         = NO
+
+# If the EXTRACT_LOCAL_CLASSES tag is set to YES classes (and structs)
+# defined locally in source files will be included in the documentation.
+# If set to NO only classes defined in header files are included.
+
+EXTRACT_LOCAL_CLASSES  = NO
+
+# This flag is only useful for Objective-C code. When set to YES local
+# methods, which are defined in the implementation section but not in
+# the interface are included in the documentation.
+# If set to NO (the default) only methods in the interface are included.
+
+EXTRACT_LOCAL_METHODS  = NO
+
+# If this flag is set to YES, the members of anonymous namespaces will be
+# extracted and appear in the documentation as a namespace called
+# 'anonymous_namespace{file}', where file will be replaced with the base
+# name of the file that contains the anonymous namespace. By default
+# anonymous namespaces are hidden.
+
+EXTRACT_ANON_NSPACES   = NO
+
+# If the HIDE_UNDOC_MEMBERS tag is set to YES, Doxygen will hide all
+# undocumented members of documented classes, files or namespaces.
+# If set to NO (the default) these members will be included in the
+# various overviews, but no documentation section is generated.
+# This option has no effect if EXTRACT_ALL is enabled.
+
+HIDE_UNDOC_MEMBERS     = YES
+
+# If the HIDE_UNDOC_CLASSES tag is set to YES, Doxygen will hide all
+# undocumented classes that are normally visible in the class hierarchy.
+# If set to NO (the default) these classes will be included in the various
+# overviews. This option has no effect if EXTRACT_ALL is enabled.
+
+HIDE_UNDOC_CLASSES     = YES
+
+# If the HIDE_FRIEND_COMPOUNDS tag is set to YES, Doxygen will hide all
+# friend (class|struct|union) declarations.
+# If set to NO (the default) these declarations will be included in the
+# documentation.
+
+HIDE_FRIEND_COMPOUNDS  = NO
+
+# If the HIDE_IN_BODY_DOCS tag is set to YES, Doxygen will hide any
+# documentation blocks found inside the body of a function.
+# If set to NO (the default) these blocks will be appended to the
+# function's detailed documentation block.
+
+HIDE_IN_BODY_DOCS      = NO
+
+# The INTERNAL_DOCS tag determines if documentation
+# that is typed after a \internal command is included. If the tag is set
+# to NO (the default) then the documentation will be excluded.
+# Set it to YES to include the internal documentation.
+
+INTERNAL_DOCS          = NO
+
+# If the CASE_SENSE_NAMES tag is set to NO then Doxygen will only generate
+# file names in lower-case letters. If set to YES upper-case letters are also
+# allowed. This is useful if you have classes or files whose names only differ
+# in case and if your file system supports case sensitive file names. Windows
+# and Mac users are advised to set this option to NO.
+
+CASE_SENSE_NAMES       = YES
+
+# If the HIDE_SCOPE_NAMES tag is set to NO (the default) then Doxygen
+# will show members with their full class and namespace scopes in the
+# documentation. If set to YES the scope will be hidden.
+
+HIDE_SCOPE_NAMES       = NO
+
+# If the SHOW_INCLUDE_FILES tag is set to YES (the default) then Doxygen
+# will put a list of the files that are included by a file in the documentation
+# of that file.
+
+SHOW_INCLUDE_FILES     = YES
+
+# If the FORCE_LOCAL_INCLUDES tag is set to YES then Doxygen
+# will list include files with double quotes in the documentation
+# rather than with sharp brackets.
+
+FORCE_LOCAL_INCLUDES   = NO
+
+# If the INLINE_INFO tag is set to YES (the default) then a tag [inline]
+# is inserted in the documentation for inline members.
+
+INLINE_INFO            = YES
+
+# If the SORT_MEMBER_DOCS tag is set to YES (the default) then doxygen
+# will sort the (detailed) documentation of file and class members
+# alphabetically by member name. If set to NO the members will appear in
+# declaration order.
+
+SORT_MEMBER_DOCS       = YES
+
+# If the SORT_BRIEF_DOCS tag is set to YES then doxygen will sort the
+# brief documentation of file, namespace and class members alphabetically
+# by member name. If set to NO (the default) the members will appear in
+# declaration order.
+
+SORT_BRIEF_DOCS        = NO
+
+# If the SORT_MEMBERS_CTORS_1ST tag is set to YES then doxygen
+# will sort the (brief and detailed) documentation of class members so that
+# constructors and destructors are listed first. If set to NO (the default)
+# the constructors will appear in the respective orders defined by
+# SORT_MEMBER_DOCS and SORT_BRIEF_DOCS.
+# This tag will be ignored for brief docs if SORT_BRIEF_DOCS is set to NO
+# and ignored for detailed docs if SORT_MEMBER_DOCS is set to NO.
+
+SORT_MEMBERS_CTORS_1ST = NO
+
+# If the SORT_GROUP_NAMES tag is set to YES then doxygen will sort the
+# hierarchy of group names into alphabetical order. If set to NO (the default)
+# the group names will appear in their defined order.
+
+SORT_GROUP_NAMES       = NO
+
+# If the SORT_BY_SCOPE_NAME tag is set to YES, the class list will be
+# sorted by fully-qualified names, including namespaces. If set to
+# NO (the default), the class list will be sorted only by class name,
+# not including the namespace part.
+# Note: This option is not very useful if HIDE_SCOPE_NAMES is set to YES.
+# Note: This option applies only to the class list, not to the
+# alphabetical list.
+
+SORT_BY_SCOPE_NAME     = NO
+
+# If the STRICT_PROTO_MATCHING option is enabled and doxygen fails to
+# do proper type resolution of all parameters of a function it will reject a
+# match between the prototype and the implementation of a member function even
+# if there is only one candidate or it is obvious which candidate to choose
+# by doing a simple string match. By disabling STRICT_PROTO_MATCHING doxygen
+# will still accept a match between prototype and implementation in such cases.
+
+STRICT_PROTO_MATCHING  = NO
+
+# The GENERATE_TODOLIST tag can be used to enable (YES) or
+# disable (NO) the todo list. This list is created by putting \todo
+# commands in the documentation.
+
+GENERATE_TODOLIST      = YES
+
+# The GENERATE_TESTLIST tag can be used to enable (YES) or
+# disable (NO) the test list. This list is created by putting \test
+# commands in the documentation.
+
+GENERATE_TESTLIST      = YES
+
+# The GENERATE_BUGLIST tag can be used to enable (YES) or
+# disable (NO) the bug list. This list is created by putting \bug
+# commands in the documentation.
+
+GENERATE_BUGLIST       = YES
+
+# The GENERATE_DEPRECATEDLIST tag can be used to enable (YES) or
+# disable (NO) the deprecated list. This list is created by putting
+# \deprecated commands in the documentation.
+
+GENERATE_DEPRECATEDLIST= YES
+
+# The ENABLED_SECTIONS tag can be used to enable conditional
+# documentation sections, marked by \if sectionname ... \endif.
+
+ENABLED_SECTIONS       =
+
+# The MAX_INITIALIZER_LINES tag determines the maximum number of lines
+# the initial value of a variable or macro consists of for it to appear in
+# the documentation. If the initializer consists of more lines than specified
+# here it will be hidden. Use a value of 0 to hide initializers completely.
+# The appearance of the initializer of individual variables and macros in the
+# documentation can be controlled using \showinitializer or \hideinitializer
+# command in the documentation regardless of this setting.
+
+MAX_INITIALIZER_LINES  = 30
+
+# Set the SHOW_USED_FILES tag to NO to disable the list of files generated
+# at the bottom of the documentation of classes and structs. If set to YES the
+# list will mention the files that were used to generate the documentation.
+
+SHOW_USED_FILES        = YES
+
+# Set the SHOW_FILES tag to NO to disable the generation of the Files page.
+# This will remove the Files entry from the Quick Index and from the
+# Folder Tree View (if specified). The default is YES.
+
+SHOW_FILES             = YES
+
+# Set the SHOW_NAMESPACES tag to NO to disable the generation of the
+# Namespaces page.
+# This will remove the Namespaces entry from the Quick Index
+# and from the Folder Tree View (if specified). The default is YES.
+
+SHOW_NAMESPACES        = YES
+
+# The FILE_VERSION_FILTER tag can be used to specify a program or script that
+# doxygen should invoke to get the current version for each file (typically from
+# the version control system). Doxygen will invoke the program by executing (via
+# popen()) the command <command> <input-file>, where <command> is the value of
+# the FILE_VERSION_FILTER tag, and <input-file> is the name of an input file
+# provided by doxygen. Whatever the program writes to standard output
+# is used as the file version. See the manual for examples.
+
+FILE_VERSION_FILTER    =
+
+# The LAYOUT_FILE tag can be used to specify a layout file which will be parsed
+# by doxygen. The layout file controls the global structure of the generated
+# output files in an output format independent way. The create the layout file
+# that represents doxygen's defaults, run doxygen with the -l option.
+# You can optionally specify a file name after the option, if omitted
+# DoxygenLayout.xml will be used as the name of the layout file.
+
+LAYOUT_FILE            =
+
+# The CITE_BIB_FILES tag can be used to specify one or more bib files
+# containing the references data. This must be a list of .bib files. The
+# .bib extension is automatically appended if omitted. Using this command
+# requires the bibtex tool to be installed. See also
+# http://en.wikipedia.org/wiki/BibTeX for more info. For LaTeX the style
+# of the bibliography can be controlled using LATEX_BIB_STYLE. To use this
+# feature you need bibtex and perl available in the search path.
+
+CITE_BIB_FILES         =
+
+#---------------------------------------------------------------------------
+# configuration options related to warning and progress messages
+#---------------------------------------------------------------------------
+
+# The QUIET tag can be used to turn on/off the messages that are generated
+# by doxygen. Possible values are YES and NO. If left blank NO is used.
+
+QUIET                  = NO
+
+# The WARNINGS tag can be used to turn on/off the warning messages that are
+# generated by doxygen. Possible values are YES and NO. If left blank
+# NO is used.
+
+WARNINGS               = YES
+
+# If WARN_IF_UNDOCUMENTED is set to YES, then doxygen will generate warnings
+# for undocumented members. If EXTRACT_ALL is set to YES then this flag will
+# automatically be disabled.
+
+WARN_IF_UNDOCUMENTED   = YES
+
+# If WARN_IF_DOC_ERROR is set to YES, doxygen will generate warnings for
+# potential errors in the documentation, such as not documenting some
+# parameters in a documented function, or documenting parameters that
+# don't exist or using markup commands wrongly.
+
+WARN_IF_DOC_ERROR      = YES
+
+# The WARN_NO_PARAMDOC option can be enabled to get warnings for
+# functions that are documented, but have no documentation for their parameters
+# or return value. If set to NO (the default) doxygen will only warn about
+# wrong or incomplete parameter documentation, but not about the absence of
+# documentation.
+
+WARN_NO_PARAMDOC       = NO
+
+# The WARN_FORMAT tag determines the format of the warning messages that
+# doxygen can produce. The string should contain the $file, $line, and $text
+# tags, which will be replaced by the file and line number from which the
+# warning originated and the warning text. Optionally the format may contain
+# $version, which will be replaced by the version of the file (if it could
+# be obtained via FILE_VERSION_FILTER)
+
+WARN_FORMAT            = "$file:$line: $text"
+
+# The WARN_LOGFILE tag can be used to specify a file to which warning
+# and error messages should be written. If left blank the output is written
+# to stderr.
+
+WARN_LOGFILE           =
+
+#---------------------------------------------------------------------------
+# configuration options related to the input files
+#---------------------------------------------------------------------------
+
+# The INPUT tag can be used to specify the files and/or directories that contain
+# documented source files. You may enter file names like "myfile.cpp" or
+# directories like "/usr/src/myproject". Separate the files or directories
+# with spaces.
+
+INPUT                  = . \
+                         doc
+
+# This tag can be used to specify the character encoding of the source files
+# that doxygen parses. Internally doxygen uses the UTF-8 encoding, which is
+# also the default input encoding. Doxygen uses libiconv (or the iconv built
+# into libc) for the transcoding. See http://www.gnu.org/software/libiconv for
+# the list of possible encodings.
+
+INPUT_ENCODING         = UTF-8
+
+# If the value of the INPUT tag contains directories, you can use the
+# FILE_PATTERNS tag to specify one or more wildcard pattern (like *.cpp
+# and *.h) to filter out the source-files in the directories. If left
+# blank the following patterns are tested:
+# *.c *.cc *.cxx *.cpp *.c++ *.d *.java *.ii *.ixx *.ipp *.i++ *.inl *.h *.hh
+# *.hxx *.hpp *.h++ *.idl *.odl *.cs *.php *.php3 *.inc *.m *.mm *.dox *.py
+# *.f90 *.f *.for *.vhd *.vhdl
+
+FILE_PATTERNS          = *.cpp \
+                         *.cc \
+                         *.c \
+                         *.h \
+                         *.hh \
+                         *.hpp \
+                         *.dox
+
+# The RECURSIVE tag can be used to turn specify whether or not subdirectories
+# should be searched for input files as well. Possible values are YES and NO.
+# If left blank NO is used.
+
+RECURSIVE              = NO
+
+# The EXCLUDE tag can be used to specify files and/or directories that should be
+# excluded from the INPUT source files. This way you can easily exclude a
+# subdirectory from a directory tree whose root is specified with the INPUT tag.
+# Note that relative paths are relative to the directory from which doxygen is
+# run.
+
+EXCLUDE                =
+
+# The EXCLUDE_SYMLINKS tag can be used to select whether or not files or
+# directories that are symbolic links (a Unix file system feature) are excluded
+# from the input.
+
+EXCLUDE_SYMLINKS       = NO
+
+# If the value of the INPUT tag contains directories, you can use the
+# EXCLUDE_PATTERNS tag to specify one or more wildcard patterns to exclude
+# certain files from those directories. Note that the wildcards are matched
+# against the file with absolute path, so to exclude all test directories
+# for example use the pattern */test/*
+
+EXCLUDE_PATTERNS       = */.git/*
+
+# The EXCLUDE_SYMBOLS tag can be used to specify one or more symbol names
+# (namespaces, classes, functions, etc.) that should be excluded from the
+# output. The symbol name can be a fully qualified name, a word, or if the
+# wildcard * is used, a substring. Examples: ANamespace, AClass,
+# AClass::ANamespace, ANamespace::*Test
+
+EXCLUDE_SYMBOLS        =
+
+# The EXAMPLE_PATH tag can be used to specify one or more files or
+# directories that contain example code fragments that are included (see
+# the \include command).
+
+EXAMPLE_PATH           =
+
+# If the value of the EXAMPLE_PATH tag contains directories, you can use the
+# EXAMPLE_PATTERNS tag to specify one or more wildcard pattern (like *.cpp
+# and *.h) to filter out the source-files in the directories. If left
+# blank all files are included.
+
+EXAMPLE_PATTERNS       =
+
+# If the EXAMPLE_RECURSIVE tag is set to YES then subdirectories will be
+# searched for input files to be used with the \include or \dontinclude
+# commands irrespective of the value of the RECURSIVE tag.
+# Possible values are YES and NO. If left blank NO is used.
+
+EXAMPLE_RECURSIVE      = NO
+
+# The IMAGE_PATH tag can be used to specify one or more files or
+# directories that contain image that are included in the documentation (see
+# the \image command).
+
+IMAGE_PATH             = doc
+
+# The INPUT_FILTER tag can be used to specify a program that doxygen should
+# invoke to filter for each input file. Doxygen will invoke the filter program
+# by executing (via popen()) the command <filter> <input-file>, where <filter>
+# is the value of the INPUT_FILTER tag, and <input-file> is the name of an
+# input file. Doxygen will then use the output that the filter program writes
+# to standard output.
+# If FILTER_PATTERNS is specified, this tag will be
+# ignored.
+
+INPUT_FILTER           =
+
+# The FILTER_PATTERNS tag can be used to specify filters on a per file pattern
+# basis.
+# Doxygen will compare the file name with each pattern and apply the
+# filter if there is a match.
+# The filters are a list of the form:
+# pattern=filter (like *.cpp=my_cpp_filter). See INPUT_FILTER for further
+# info on how filters are used. If FILTER_PATTERNS is empty or if
+# non of the patterns match the file name, INPUT_FILTER is applied.
+
+FILTER_PATTERNS        =
+
+# If the FILTER_SOURCE_FILES tag is set to YES, the input filter (if set using
+# INPUT_FILTER) will be used to filter the input files when producing source
+# files to browse (i.e. when SOURCE_BROWSER is set to YES).
+
+FILTER_SOURCE_FILES    = NO
+
+# The FILTER_SOURCE_PATTERNS tag can be used to specify source filters per file
+# pattern. A pattern will override the setting for FILTER_PATTERN (if any)
+# and it is also possible to disable source filtering for a specific pattern
+# using *.ext= (so without naming a filter). This option only has effect when
+# FILTER_SOURCE_FILES is enabled.
+
+FILTER_SOURCE_PATTERNS =
+
+#---------------------------------------------------------------------------
+# configuration options related to source browsing
+#---------------------------------------------------------------------------
+
+# If the SOURCE_BROWSER tag is set to YES then a list of source files will
+# be generated. Documented entities will be cross-referenced with these sources.
+# Note: To get rid of all source code in the generated output, make sure also
+# VERBATIM_HEADERS is set to NO.
+
+SOURCE_BROWSER         = NO
+
+# Setting the INLINE_SOURCES tag to YES will include the body
+# of functions and classes directly in the documentation.
+
+INLINE_SOURCES         = NO
+
+# Setting the STRIP_CODE_COMMENTS tag to YES (the default) will instruct
+# doxygen to hide any special comment blocks from generated source code
+# fragments. Normal C and C++ comments will always remain visible.
+
+STRIP_CODE_COMMENTS    = YES
+
+# If the REFERENCED_BY_RELATION tag is set to YES
+# then for each documented function all documented
+# functions referencing it will be listed.
+
+REFERENCED_BY_RELATION = NO
+
+# If the REFERENCES_RELATION tag is set to YES
+# then for each documented function all documented entities
+# called/used by that function will be listed.
+
+REFERENCES_RELATION    = NO
+
+# If the REFERENCES_LINK_SOURCE tag is set to YES (the default)
+# and SOURCE_BROWSER tag is set to YES, then the hyperlinks from
+# functions in REFERENCES_RELATION and REFERENCED_BY_RELATION lists will
+# link to the source code.
+# Otherwise they will link to the documentation.
+
+REFERENCES_LINK_SOURCE = YES
+
+# If the USE_HTAGS tag is set to YES then the references to source code
+# will point to the HTML generated by the htags(1) tool instead of doxygen
+# built-in source browser. The htags tool is part of GNU's global source
+# tagging system (see http://www.gnu.org/software/global/global.html). You
+# will need version 4.8.6 or higher.
+
+USE_HTAGS              = NO
+
+# If the VERBATIM_HEADERS tag is set to YES (the default) then Doxygen
+# will generate a verbatim copy of the header file for each class for
+# which an include is specified. Set to NO to disable this.
+
+VERBATIM_HEADERS       = YES
+
+#---------------------------------------------------------------------------
+# configuration options related to the alphabetical class index
+#---------------------------------------------------------------------------
+
+# If the ALPHABETICAL_INDEX tag is set to YES, an alphabetical index
+# of all compounds will be generated. Enable this if the project
+# contains a lot of classes, structs, unions or interfaces.
+
+ALPHABETICAL_INDEX     = NO
+
+# If the alphabetical index is enabled (see ALPHABETICAL_INDEX) then
+# the COLS_IN_ALPHA_INDEX tag can be used to specify the number of columns
+# in which this list will be split (can be a number in the range [1..20])
+
+COLS_IN_ALPHA_INDEX    = 5
+
+# In case all classes in a project start with a common prefix, all
+# classes will be put under the same header in the alphabetical index.
+# The IGNORE_PREFIX tag can be used to specify one or more prefixes that
+# should be ignored while generating the index headers.
+
+IGNORE_PREFIX          =
+
+#---------------------------------------------------------------------------
+# configuration options related to the HTML output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_HTML tag is set to YES (the default) Doxygen will
+# generate HTML output.
+
+GENERATE_HTML          = YES
+
+# The HTML_OUTPUT tag is used to specify where the HTML docs will be put.
+# If a relative path is entered the value of OUTPUT_DIRECTORY will be
+# put in front of it. If left blank `html' will be used as the default path.
+
+HTML_OUTPUT            = html
+
+# The HTML_FILE_EXTENSION tag can be used to specify the file extension for
+# each generated HTML page (for example: .htm,.php,.asp). If it is left blank
+# doxygen will generate files with .html extension.
+
+HTML_FILE_EXTENSION    = .html
+
+# The HTML_HEADER tag can be used to specify a personal HTML header for
+# each generated HTML page. If it is left blank doxygen will generate a
+# standard header. Note that when using a custom header you are responsible
+#  for the proper inclusion of any scripts and style sheets that doxygen
+# needs, which is dependent on the configuration options used.
+# It is advised to generate a default header using "doxygen -w html
+# header.html footer.html stylesheet.css YourConfigFile" and then modify
+# that header. Note that the header is subject to change so you typically
+# have to redo this when upgrading to a newer version of doxygen or when
+# changing the value of configuration settings such as GENERATE_TREEVIEW!
+
+HTML_HEADER            =
+
+# The HTML_FOOTER tag can be used to specify a personal HTML footer for
+# each generated HTML page. If it is left blank doxygen will generate a
+# standard footer.
+
+HTML_FOOTER            =
+
+# The HTML_STYLESHEET tag can be used to specify a user-defined cascading
+# style sheet that is used by each HTML page. It can be used to
+# fine-tune the look of the HTML output. If the tag is left blank doxygen
+# will generate a default style sheet. Note that doxygen will try to copy
+# the style sheet file to the HTML output directory, so don't put your own
+# style sheet in the HTML output directory as well, or it will be erased!
+
+HTML_STYLESHEET        =
+
+# The HTML_EXTRA_FILES tag can be used to specify one or more extra images or
+# other source files which should be copied to the HTML output directory. Note
+# that these files will be copied to the base HTML output directory. Use the
+# $relpath$ marker in the HTML_HEADER and/or HTML_FOOTER files to load these
+# files. In the HTML_STYLESHEET file, use the file name only. Also note that
+# the files will be copied as-is; there are no commands or markers available.
+
+HTML_EXTRA_FILES       =
+
+# The HTML_COLORSTYLE_HUE tag controls the color of the HTML output.
+# Doxygen will adjust the colors in the style sheet and background images
+# according to this color. Hue is specified as an angle on a colorwheel,
+# see http://en.wikipedia.org/wiki/Hue for more information.
+# For instance the value 0 represents red, 60 is yellow, 120 is green,
+# 180 is cyan, 240 is blue, 300 purple, and 360 is red again.
+# The allowed range is 0 to 359.
+
+HTML_COLORSTYLE_HUE    = 220
+
+# The HTML_COLORSTYLE_SAT tag controls the purity (or saturation) of
+# the colors in the HTML output. For a value of 0 the output will use
+# grayscales only. A value of 255 will produce the most vivid colors.
+
+HTML_COLORSTYLE_SAT    = 100
+
+# The HTML_COLORSTYLE_GAMMA tag controls the gamma correction applied to
+# the luminance component of the colors in the HTML output. Values below
+# 100 gradually make the output lighter, whereas values above 100 make
+# the output darker. The value divided by 100 is the actual gamma applied,
+# so 80 represents a gamma of 0.8, The value 220 represents a gamma of 2.2,
+# and 100 does not change the gamma.
+
+HTML_COLORSTYLE_GAMMA  = 80
+
+# If the HTML_TIMESTAMP tag is set to YES then the footer of each generated HTML
+# page will contain the date and time when the page was generated. Setting
+# this to NO can help when comparing the output of multiple runs.
+
+HTML_TIMESTAMP         = NO
+
+# If the HTML_ALIGN_MEMBERS tag is set to YES, the members of classes,
+# files or namespaces will be aligned in HTML using tables. If set to
+# NO a bullet list will be used.
+
+HTML_ALIGN_MEMBERS     = YES
+
+# If the HTML_DYNAMIC_SECTIONS tag is set to YES then the generated HTML
+# documentation will contain sections that can be hidden and shown after the
+# page has loaded. For this to work a browser that supports
+# JavaScript and DHTML is required (for instance Mozilla 1.0+, Firefox
+# Netscape 6.0+, Internet explorer 5.0+, Konqueror, or Safari).
+
+HTML_DYNAMIC_SECTIONS  = NO
+
+# If the GENERATE_DOCSET tag is set to YES, additional index files
+# will be generated that can be used as input for Apple's Xcode 3
+# integrated development environment, introduced with OSX 10.5 (Leopard).
+# To create a documentation set, doxygen will generate a Makefile in the
+# HTML output directory. Running make will produce the docset in that
+# directory and running "make install" will install the docset in
+# ~/Library/Developer/Shared/Documentation/DocSets so that Xcode will find
+# it at startup.
+# See http://developer.apple.com/tools/creatingdocsetswithdoxygen.html
+# for more information.
+
+GENERATE_DOCSET        = NO
+
+# When GENERATE_DOCSET tag is set to YES, this tag determines the name of the
+# feed. A documentation feed provides an umbrella under which multiple
+# documentation sets from a single provider (such as a company or product suite)
+# can be grouped.
+
+DOCSET_FEEDNAME        = "Doxygen generated docs"
+
+# When GENERATE_DOCSET tag is set to YES, this tag specifies a string that
+# should uniquely identify the documentation set bundle. This should be a
+# reverse domain-name style string, e.g. com.mycompany.MyDocSet. Doxygen
+# will append .docset to the name.
+
+DOCSET_BUNDLE_ID       = org.doxygen.Project
+
+# When GENERATE_PUBLISHER_ID tag specifies a string that should uniquely identify
+# the documentation publisher. This should be a reverse domain-name style
+# string, e.g. com.mycompany.MyDocSet.documentation.
+
+DOCSET_PUBLISHER_ID    = org.doxygen.Publisher
+
+# The GENERATE_PUBLISHER_NAME tag identifies the documentation publisher.
+
+DOCSET_PUBLISHER_NAME  = Publisher
+
+# If the GENERATE_HTMLHELP tag is set to YES, additional index files
+# will be generated that can be used as input for tools like the
+# Microsoft HTML help workshop to generate a compiled HTML help file (.chm)
+# of the generated HTML documentation.
+
+GENERATE_HTMLHELP      = NO
+
+# If the GENERATE_HTMLHELP tag is set to YES, the CHM_FILE tag can
+# be used to specify the file name of the resulting .chm file. You
+# can add a path in front of the file if the result should not be
+# written to the html output directory.
+
+CHM_FILE               =
+
+# If the GENERATE_HTMLHELP tag is set to YES, the HHC_LOCATION tag can
+# be used to specify the location (absolute path including file name) of
+# the HTML help compiler (hhc.exe). If non-empty doxygen will try to run
+# the HTML help compiler on the generated index.hhp.
+
+HHC_LOCATION           =
+
+# If the GENERATE_HTMLHELP tag is set to YES, the GENERATE_CHI flag
+# controls if a separate .chi index file is generated (YES) or that
+# it should be included in the master .chm file (NO).
+
+GENERATE_CHI           = NO
+
+# If the GENERATE_HTMLHELP tag is set to YES, the CHM_INDEX_ENCODING
+# is used to encode HtmlHelp index (hhk), content (hhc) and project file
+# content.
+
+CHM_INDEX_ENCODING     =
+
+# If the GENERATE_HTMLHELP tag is set to YES, the BINARY_TOC flag
+# controls whether a binary table of contents is generated (YES) or a
+# normal table of contents (NO) in the .chm file.
+
+BINARY_TOC             = NO
+
+# The TOC_EXPAND flag can be set to YES to add extra items for group members
+# to the contents of the HTML help documentation and to the tree view.
+
+TOC_EXPAND             = NO
+
+# If the GENERATE_QHP tag is set to YES and both QHP_NAMESPACE and
+# QHP_VIRTUAL_FOLDER are set, an additional index file will be generated
+# that can be used as input for Qt's qhelpgenerator to generate a
+# Qt Compressed Help (.qch) of the generated HTML documentation.
+
+GENERATE_QHP           = NO
+
+# If the QHG_LOCATION tag is specified, the QCH_FILE tag can
+# be used to specify the file name of the resulting .qch file.
+# The path specified is relative to the HTML output folder.
+
+QCH_FILE               =
+
+# The QHP_NAMESPACE tag specifies the namespace to use when generating
+# Qt Help Project output. For more information please see
+# http://doc.trolltech.com/qthelpproject.html#namespace
+
+QHP_NAMESPACE          =
+
+# The QHP_VIRTUAL_FOLDER tag specifies the namespace to use when generating
+# Qt Help Project output. For more information please see
+# http://doc.trolltech.com/qthelpproject.html#virtual-folders
+
+QHP_VIRTUAL_FOLDER     = doc
+
+# If QHP_CUST_FILTER_NAME is set, it specifies the name of a custom filter to
+# add. For more information please see
+# http://doc.trolltech.com/qthelpproject.html#custom-filters
+
+QHP_CUST_FILTER_NAME   =
+
+# The QHP_CUST_FILT_ATTRS tag specifies the list of the attributes of the
+# custom filter to add. For more information please see
+# <a href="http://doc.trolltech.com/qthelpproject.html#custom-filters">
+# Qt Help Project / Custom Filters</a>.
+
+QHP_CUST_FILTER_ATTRS  =
+
+# The QHP_SECT_FILTER_ATTRS tag specifies the list of the attributes this
+# project's
+# filter section matches.
+# <a href="http://doc.trolltech.com/qthelpproject.html#filter-attributes">
+# Qt Help Project / Filter Attributes</a>.
+
+QHP_SECT_FILTER_ATTRS  =
+
+# If the GENERATE_QHP tag is set to YES, the QHG_LOCATION tag can
+# be used to specify the location of Qt's qhelpgenerator.
+# If non-empty doxygen will try to run qhelpgenerator on the generated
+# .qhp file.
+
+QHG_LOCATION           =
+
+# If the GENERATE_ECLIPSEHELP tag is set to YES, additional index files
+#  will be generated, which together with the HTML files, form an Eclipse help
+# plugin. To install this plugin and make it available under the help contents
+# menu in Eclipse, the contents of the directory containing the HTML and XML
+# files needs to be copied into the plugins directory of eclipse. The name of
+# the directory within the plugins directory should be the same as
+# the ECLIPSE_DOC_ID value. After copying Eclipse needs to be restarted before
+# the help appears.
+
+GENERATE_ECLIPSEHELP   = NO
+
+# A unique identifier for the eclipse help plugin. When installing the plugin
+# the directory name containing the HTML and XML files should also have
+# this name.
+
+ECLIPSE_DOC_ID         = org.doxygen.Project
+
+# The DISABLE_INDEX tag can be used to turn on/off the condensed index (tabs)
+# at top of each HTML page. The value NO (the default) enables the index and
+# the value YES disables it. Since the tabs have the same information as the
+# navigation tree you can set this option to NO if you already set
+# GENERATE_TREEVIEW to YES.
+
+DISABLE_INDEX          = NO
+
+# The GENERATE_TREEVIEW tag is used to specify whether a tree-like index
+# structure should be generated to display hierarchical information.
+# If the tag value is set to YES, a side panel will be generated
+# containing a tree-like index structure (just like the one that
+# is generated for HTML Help). For this to work a browser that supports
+# JavaScript, DHTML, CSS and frames is required (i.e. any modern browser).
+# Windows users are probably better off using the HTML help feature.
+# Since the tree basically has the same information as the tab index you
+# could consider to set DISABLE_INDEX to NO when enabling this option.
+
+GENERATE_TREEVIEW      = NONE
+
+# The ENUM_VALUES_PER_LINE tag can be used to set the number of enum values
+# (range [0,1..20]) that doxygen will group on one line in the generated HTML
+# documentation. Note that a value of 0 will completely suppress the enum
+# values from appearing in the overview section.
+
+ENUM_VALUES_PER_LINE   = 4
+
+# If the treeview is enabled (see GENERATE_TREEVIEW) then this tag can be
+# used to set the initial width (in pixels) of the frame in which the tree
+# is shown.
+
+TREEVIEW_WIDTH         = 250
+
+# When the EXT_LINKS_IN_WINDOW option is set to YES doxygen will open
+# links to external symbols imported via tag files in a separate window.
+
+EXT_LINKS_IN_WINDOW    = NO
+
+# Use this tag to change the font size of Latex formulas included
+# as images in the HTML documentation. The default is 10. Note that
+# when you change the font size after a successful doxygen run you need
+# to manually remove any form_*.png images from the HTML output directory
+# to force them to be regenerated.
+
+FORMULA_FONTSIZE       = 10
+
+# Use the FORMULA_TRANPARENT tag to determine whether or not the images
+# generated for formulas are transparent PNGs. Transparent PNGs are
+# not supported properly for IE 6.0, but are supported on all modern browsers.
+# Note that when changing this option you need to delete any form_*.png files
+# in the HTML output before the changes have effect.
+
+FORMULA_TRANSPARENT    = YES
+
+# Enable the USE_MATHJAX option to render LaTeX formulas using MathJax
+# (see http://www.mathjax.org) which uses client side Javascript for the
+# rendering instead of using prerendered bitmaps. Use this if you do not
+# have LaTeX installed or if you want to formulas look prettier in the HTML
+# output. When enabled you may also need to install MathJax separately and
+# configure the path to it using the MATHJAX_RELPATH option.
+
+USE_MATHJAX            = NO
+
+# When MathJax is enabled you need to specify the location relative to the
+# HTML output directory using the MATHJAX_RELPATH option. The destination
+# directory should contain the MathJax.js script. For instance, if the mathjax
+# directory is located at the same level as the HTML output directory, then
+# MATHJAX_RELPATH should be ../mathjax. The default value points to
+# the MathJax Content Delivery Network so you can quickly see the result without
+# installing MathJax.
+# However, it is strongly recommended to install a local
+# copy of MathJax from http://www.mathjax.org before deployment.
+
+MATHJAX_RELPATH        = http://cdn.mathjax.org/mathjax/latest
+
+# The MATHJAX_EXTENSIONS tag can be used to specify one or MathJax extension
+# names that should be enabled during MathJax rendering.
+
+MATHJAX_EXTENSIONS     =
+
+# When the SEARCHENGINE tag is enabled doxygen will generate a search box
+# for the HTML output. The underlying search engine uses javascript
+# and DHTML and should work on any modern browser. Note that when using
+# HTML help (GENERATE_HTMLHELP), Qt help (GENERATE_QHP), or docsets
+# (GENERATE_DOCSET) there is already a search function so this one should
+# typically be disabled. For large projects the javascript based search engine
+# can be slow, then enabling SERVER_BASED_SEARCH may provide a better solution.
+
+SEARCHENGINE           = NO
+
+# When the SERVER_BASED_SEARCH tag is enabled the search engine will be
+# implemented using a PHP enabled web server instead of at the web client
+# using Javascript. Doxygen will generate the search PHP script and index
+# file to put on the web server. The advantage of the server
+# based approach is that it scales better to large projects and allows
+# full text search. The disadvantages are that it is more difficult to setup
+# and does not have live searching capabilities.
+
+SERVER_BASED_SEARCH    = NO
+
+#---------------------------------------------------------------------------
+# configuration options related to the LaTeX output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_LATEX tag is set to YES (the default) Doxygen will
+# generate Latex output.
+
+GENERATE_LATEX         = YES
+
+# The LATEX_OUTPUT tag is used to specify where the LaTeX docs will be put.
+# If a relative path is entered the value of OUTPUT_DIRECTORY will be
+# put in front of it. If left blank `latex' will be used as the default path.
+
+LATEX_OUTPUT           = latex
+
+# The LATEX_CMD_NAME tag can be used to specify the LaTeX command name to be
+# invoked. If left blank `latex' will be used as the default command name.
+# Note that when enabling USE_PDFLATEX this option is only used for
+# generating bitmaps for formulas in the HTML output, but not in the
+# Makefile that is written to the output directory.
+
+LATEX_CMD_NAME         = latex
+
+# The MAKEINDEX_CMD_NAME tag can be used to specify the command name to
+# generate index for LaTeX. If left blank `makeindex' will be used as the
+# default command name.
+
+MAKEINDEX_CMD_NAME     = makeindex
+
+# If the COMPACT_LATEX tag is set to YES Doxygen generates more compact
+# LaTeX documents. This may be useful for small projects and may help to
+# save some trees in general.
+
+COMPACT_LATEX          = NO
+
+# The PAPER_TYPE tag can be used to set the paper type that is used
+# by the printer. Possible values are: a4, letter, legal and
+# executive. If left blank a4wide will be used.
+
+PAPER_TYPE             = a4wide
+
+# The EXTRA_PACKAGES tag can be to specify one or more names of LaTeX
+# packages that should be included in the LaTeX output.
+
+EXTRA_PACKAGES         =
+
+# The LATEX_HEADER tag can be used to specify a personal LaTeX header for
+# the generated latex document. The header should contain everything until
+# the first chapter. If it is left blank doxygen will generate a
+# standard header. Notice: only use this tag if you know what you are doing!
+
+LATEX_HEADER           =
+
+# The LATEX_FOOTER tag can be used to specify a personal LaTeX footer for
+# the generated latex document. The footer should contain everything after
+# the last chapter. If it is left blank doxygen will generate a
+# standard footer. Notice: only use this tag if you know what you are doing!
+
+LATEX_FOOTER           =
+
+# If the PDF_HYPERLINKS tag is set to YES, the LaTeX that is generated
+# is prepared for conversion to pdf (using ps2pdf). The pdf file will
+# contain links (just like the HTML output) instead of page references
+# This makes the output suitable for online browsing using a pdf viewer.
+
+PDF_HYPERLINKS         = YES
+
+# If the USE_PDFLATEX tag is set to YES, pdflatex will be used instead of
+# plain latex in the generated Makefile. Set this option to YES to get a
+# higher quality PDF documentation.
+
+USE_PDFLATEX           = YES
+
+# If the LATEX_BATCHMODE tag is set to YES, doxygen will add the \\batchmode.
+# command to the generated LaTeX files. This will instruct LaTeX to keep
+# running if errors occur, instead of asking the user for help.
+# This option is also used when generating formulas in HTML.
+
+LATEX_BATCHMODE        = NO
+
+# If LATEX_HIDE_INDICES is set to YES then doxygen will not
+# include the index chapters (such as File Index, Compound Index, etc.)
+# in the output.
+
+LATEX_HIDE_INDICES     = NO
+
+# If LATEX_SOURCE_CODE is set to YES then doxygen will include
+# source code with syntax highlighting in the LaTeX output.
+# Note that which sources are shown also depends on other settings
+# such as SOURCE_BROWSER.
+
+LATEX_SOURCE_CODE      = NO
+
+# The LATEX_BIB_STYLE tag can be used to specify the style to use for the
+# bibliography, e.g. plainnat, or ieeetr. The default style is "plain". See
+# http://en.wikipedia.org/wiki/BibTeX for more info.
+
+LATEX_BIB_STYLE        = plain
+
+#---------------------------------------------------------------------------
+# configuration options related to the RTF output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_RTF tag is set to YES Doxygen will generate RTF output
+# The RTF output is optimized for Word 97 and may not look very pretty with
+# other RTF readers or editors.
+
+GENERATE_RTF           = NO
+
+# The RTF_OUTPUT tag is used to specify where the RTF docs will be put.
+# If a relative path is entered the value of OUTPUT_DIRECTORY will be
+# put in front of it. If left blank `rtf' will be used as the default path.
+
+RTF_OUTPUT             = rtf
+
+# If the COMPACT_RTF tag is set to YES Doxygen generates more compact
+# RTF documents. This may be useful for small projects and may help to
+# save some trees in general.
+
+COMPACT_RTF            = NO
+
+# If the RTF_HYPERLINKS tag is set to YES, the RTF that is generated
+# will contain hyperlink fields. The RTF file will
+# contain links (just like the HTML output) instead of page references.
+# This makes the output suitable for online browsing using WORD or other
+# programs which support those fields.
+# Note: wordpad (write) and others do not support links.
+
+RTF_HYPERLINKS         = NO
+
+# Load style sheet definitions from file. Syntax is similar to doxygen's
+# config file, i.e. a series of assignments. You only have to provide
+# replacements, missing definitions are set to their default value.
+
+RTF_STYLESHEET_FILE    =
+
+# Set optional variables used in the generation of an rtf document.
+# Syntax is similar to doxygen's config file.
+
+RTF_EXTENSIONS_FILE    =
+
+#---------------------------------------------------------------------------
+# configuration options related to the man page output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_MAN tag is set to YES (the default) Doxygen will
+# generate man pages
+
+GENERATE_MAN           = YES
+
+# The MAN_OUTPUT tag is used to specify where the man pages will be put.
+# If a relative path is entered the value of OUTPUT_DIRECTORY will be
+# put in front of it. If left blank `man' will be used as the default path.
+
+MAN_OUTPUT             = man
+
+# The MAN_EXTENSION tag determines the extension that is added to
+# the generated man pages (default is the subroutine's section .3)
+
+MAN_EXTENSION          = .3
+
+# If the MAN_LINKS tag is set to YES and Doxygen generates man output,
+# then it will generate one additional man file for each entity
+# documented in the real man page(s). These additional files
+# only source the real man page, but without them the man command
+# would be unable to find the correct page. The default is NO.
+
+MAN_LINKS              = NO
+
+#---------------------------------------------------------------------------
+# configuration options related to the XML output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_XML tag is set to YES Doxygen will
+# generate an XML file that captures the structure of
+# the code including all documentation.
+
+GENERATE_XML           = NO
+
+# The XML_OUTPUT tag is used to specify where the XML pages will be put.
+# If a relative path is entered the value of OUTPUT_DIRECTORY will be
+# put in front of it. If left blank `xml' will be used as the default path.
+
+XML_OUTPUT             = xml
+
+# The XML_SCHEMA tag can be used to specify an XML schema,
+# which can be used by a validating XML parser to check the
+# syntax of the XML files.
+
+XML_SCHEMA             =
+
+# The XML_DTD tag can be used to specify an XML DTD,
+# which can be used by a validating XML parser to check the
+# syntax of the XML files.
+
+XML_DTD                =
+
+# If the XML_PROGRAMLISTING tag is set to YES Doxygen will
+# dump the program listings (including syntax highlighting
+# and cross-referencing information) to the XML output. Note that
+# enabling this will significantly increase the size of the XML output.
+
+XML_PROGRAMLISTING     = YES
+
+#---------------------------------------------------------------------------
+# configuration options for the AutoGen Definitions output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_AUTOGEN_DEF tag is set to YES Doxygen will
+# generate an AutoGen Definitions (see autogen.sf.net) file
+# that captures the structure of the code including all
+# documentation. Note that this feature is still experimental
+# and incomplete at the moment.
+
+GENERATE_AUTOGEN_DEF   = NO
+
+#---------------------------------------------------------------------------
+# configuration options related to the Perl module output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_PERLMOD tag is set to YES Doxygen will
+# generate a Perl module file that captures the structure of
+# the code including all documentation. Note that this
+# feature is still experimental and incomplete at the
+# moment.
+
+GENERATE_PERLMOD       = NO
+
+# If the PERLMOD_LATEX tag is set to YES Doxygen will generate
+# the necessary Makefile rules, Perl scripts and LaTeX code to be able
+# to generate PDF and DVI output from the Perl module output.
+
+PERLMOD_LATEX          = NO
+
+# If the PERLMOD_PRETTY tag is set to YES the Perl module output will be
+# nicely formatted so it can be parsed by a human reader.
+# This is useful
+# if you want to understand what is going on.
+# On the other hand, if this
+# tag is set to NO the size of the Perl module output will be much smaller
+# and Perl will parse it just the same.
+
+PERLMOD_PRETTY         = YES
+
+# The names of the make variables in the generated doxyrules.make file
+# are prefixed with the string contained in PERLMOD_MAKEVAR_PREFIX.
+# This is useful so different doxyrules.make files included by the same
+# Makefile don't overwrite each other's variables.
+
+PERLMOD_MAKEVAR_PREFIX =
+
+#---------------------------------------------------------------------------
+# Configuration options related to the preprocessor
+#---------------------------------------------------------------------------
+
+# If the ENABLE_PREPROCESSING tag is set to YES (the default) Doxygen will
+# evaluate all C-preprocessor directives found in the sources and include
+# files.
+
+ENABLE_PREPROCESSING   = YES
+
+# If the MACRO_EXPANSION tag is set to YES Doxygen will expand all macro
+# names in the source code. If set to NO (the default) only conditional
+# compilation will be performed. Macro expansion can be done in a controlled
+# way by setting EXPAND_ONLY_PREDEF to YES.
+
+MACRO_EXPANSION        = YES
+
+# If the EXPAND_ONLY_PREDEF and MACRO_EXPANSION tags are both set to YES
+# then the macro expansion is limited to the macros specified with the
+# PREDEFINED and EXPAND_AS_DEFINED tags.
+
+EXPAND_ONLY_PREDEF     = YES
+
+# If the SEARCH_INCLUDES tag is set to YES (the default) the includes files
+# pointed to by INCLUDE_PATH will be searched when a #include is found.
+
+SEARCH_INCLUDES        = YES
+
+# The INCLUDE_PATH tag can be used to specify one or more directories that
+# contain include files that are not input files but should be processed by
+# the preprocessor.
+
+INCLUDE_PATH           =
+
+# You can use the INCLUDE_FILE_PATTERNS tag to specify one or more wildcard
+# patterns (like *.h and *.hpp) to filter out the header-files in the
+# directories. If left blank, the patterns specified with FILE_PATTERNS will
+# be used.
+
+INCLUDE_FILE_PATTERNS  =
+
+# The PREDEFINED tag can be used to specify one or more macro names that
+# are defined before the preprocessor is started (similar to the -D option of
+# gcc). The argument of the tag is a list of macros of the form: name
+# or name=definition (no spaces). If the definition and the = are
+# omitted =1 is assumed. To prevent a macro definition from being
+# undefined via #undef or recursively expanded use the := operator
+# instead of the = operator.
+
+PREDEFINED             = DOXYGEN \
+                         PRINTF_ATTRIBUTE(x,y)=
+
+# If the MACRO_EXPANSION and EXPAND_ONLY_PREDEF tags are set to YES then
+# this tag can be used to specify a list of macro names that should be expanded.
+# The macro definition that is found in the sources will be used.
+# Use the PREDEFINED tag if you want to use a different macro definition that
+# overrules the definition found in the source code.
+
+EXPAND_AS_DEFINED      =
+
+# If the SKIP_FUNCTION_MACROS tag is set to YES (the default) then
+# doxygen's preprocessor will remove all references to function-like macros
+# that are alone on a line, have an all uppercase name, and do not end with a
+# semicolon, because these will confuse the parser if not removed.
+
+SKIP_FUNCTION_MACROS   = YES
+
+#---------------------------------------------------------------------------
+# Configuration::additions related to external references
+#---------------------------------------------------------------------------
+
+# The TAGFILES option can be used to specify one or more tagfiles. For each
+# tag file the location of the external documentation should be added. The
+# format of a tag file without this location is as follows:
+#
+# TAGFILES = file1 file2 ...
+# Adding location for the tag files is done as follows:
+#
+# TAGFILES = file1=loc1 "file2 = loc2" ...
+# where "loc1" and "loc2" can be relative or absolute paths
+# or URLs. Note that each tag file must have a unique name (where the name does
+# NOT include the path). If a tag file is not located in the directory in which
+# doxygen is run, you must also specify the path to the tagfile here.
+
+TAGFILES               =
+
+# When a file name is specified after GENERATE_TAGFILE, doxygen will create
+# a tag file that is based on the input files it reads.
+
+GENERATE_TAGFILE       =
+
+# If the ALLEXTERNALS tag is set to YES all external classes will be listed
+# in the class index. If set to NO only the inherited external classes
+# will be listed.
+
+ALLEXTERNALS           = NO
+
+# If the EXTERNAL_GROUPS tag is set to YES all external groups will be listed
+# in the modules index. If set to NO, only the current project's groups will
+# be listed.
+
+EXTERNAL_GROUPS        = YES
+
+# The PERL_PATH should be the absolute path and name of the perl script
+# interpreter (i.e. the result of `which perl').
+
+PERL_PATH              = /usr/bin/perl
+
+#---------------------------------------------------------------------------
+# Configuration options related to the dot tool
+#---------------------------------------------------------------------------
+
+# If the CLASS_DIAGRAMS tag is set to YES (the default) Doxygen will
+# generate a inheritance diagram (in HTML, RTF and LaTeX) for classes with base
+# or super classes. Setting the tag to NO turns the diagrams off. Note that
+# this option also works with HAVE_DOT disabled, but it is recommended to
+# install and use dot, since it yields more powerful graphs.
+
+CLASS_DIAGRAMS         = YES
+
+# You can define message sequence charts within doxygen comments using the \msc
+# command. Doxygen will then run the mscgen tool (see
+# http://www.mcternan.me.uk/mscgen/) to produce the chart and insert it in the
+# documentation. The MSCGEN_PATH tag allows you to specify the directory where
+# the mscgen tool resides. If left empty the tool is assumed to be found in the
+# default search path.
+
+MSCGEN_PATH            =
+
+# If set to YES, the inheritance and collaboration graphs will hide
+# inheritance and usage relations if the target is undocumented
+# or is not a class.
+
+HIDE_UNDOC_RELATIONS   = YES
+
+# If you set the HAVE_DOT tag to YES then doxygen will assume the dot tool is
+# available from the path. This tool is part of Graphviz, a graph visualization
+# toolkit from AT&T and Lucent Bell Labs. The other options in this section
+# have no effect if this option is set to NO (the default)
+
+HAVE_DOT               = NO
+
+# The DOT_NUM_THREADS specifies the number of dot invocations doxygen is
+# allowed to run in parallel. When set to 0 (the default) doxygen will
+# base this on the number of processors available in the system. You can set it
+# explicitly to a value larger than 0 to get control over the balance
+# between CPU load and processing speed.
+
+DOT_NUM_THREADS        = 0
+
+# By default doxygen will use the Helvetica font for all dot files that
+# doxygen generates. When you want a differently looking font you can specify
+# the font name using DOT_FONTNAME. You need to make sure dot is able to find
+# the font, which can be done by putting it in a standard location or by setting
+# the DOTFONTPATH environment variable or by setting DOT_FONTPATH to the
+# directory containing the font.
+
+DOT_FONTNAME           = FreeSans
+
+# The DOT_FONTSIZE tag can be used to set the size of the font of dot graphs.
+# The default size is 10pt.
+
+DOT_FONTSIZE           = 10
+
+# By default doxygen will tell dot to use the Helvetica font.
+# If you specify a different font using DOT_FONTNAME you can use DOT_FONTPATH to
+# set the path where dot can find it.
+
+DOT_FONTPATH           =
+
+# If the CLASS_GRAPH and HAVE_DOT tags are set to YES then doxygen
+# will generate a graph for each documented class showing the direct and
+# indirect inheritance relations. Setting this tag to YES will force the
+# CLASS_DIAGRAMS tag to NO.
+
+CLASS_GRAPH            = YES
+
+# If the COLLABORATION_GRAPH and HAVE_DOT tags are set to YES then doxygen
+# will generate a graph for each documented class showing the direct and
+# indirect implementation dependencies (inheritance, containment, and
+# class references variables) of the class with other documented classes.
+
+COLLABORATION_GRAPH    = YES
+
+# If the GROUP_GRAPHS and HAVE_DOT tags are set to YES then doxygen
+# will generate a graph for groups, showing the direct groups dependencies
+
+GROUP_GRAPHS           = YES
+
+# If the UML_LOOK tag is set to YES doxygen will generate inheritance and
+# collaboration diagrams in a style similar to the OMG's Unified Modeling
+# Language.
+
+UML_LOOK               = NO
+
+# If the UML_LOOK tag is enabled, the fields and methods are shown inside
+# the class node. If there are many fields or methods and many nodes the
+# graph may become too big to be useful. The UML_LIMIT_NUM_FIELDS
+# threshold limits the number of items for each type to make the size more
+# managable. Set this to 0 for no limit. Note that the threshold may be
+# exceeded by 50% before the limit is enforced.
+
+UML_LIMIT_NUM_FIELDS   = 10
+
+# If set to YES, the inheritance and collaboration graphs will show the
+# relations between templates and their instances.
+
+TEMPLATE_RELATIONS     = NO
+
+# If the ENABLE_PREPROCESSING, SEARCH_INCLUDES, INCLUDE_GRAPH, and HAVE_DOT
+# tags are set to YES then doxygen will generate a graph for each documented
+# file showing the direct and indirect include dependencies of the file with
+# other documented files.
+
+INCLUDE_GRAPH          = YES
+
+# If the ENABLE_PREPROCESSING, SEARCH_INCLUDES, INCLUDED_BY_GRAPH, and
+# HAVE_DOT tags are set to YES then doxygen will generate a graph for each
+# documented header file showing the documented files that directly or
+# indirectly include this file.
+
+INCLUDED_BY_GRAPH      = YES
+
+# If the CALL_GRAPH and HAVE_DOT options are set to YES then
+# doxygen will generate a call dependency graph for every global function
+# or class method. Note that enabling this option will significantly increase
+# the time of a run. So in most cases it will be better to enable call graphs
+# for selected functions only using the \callgraph command.
+
+CALL_GRAPH             = NO
+
+# If the CALLER_GRAPH and HAVE_DOT tags are set to YES then
+# doxygen will generate a caller dependency graph for every global function
+# or class method. Note that enabling this option will significantly increase
+# the time of a run. So in most cases it will be better to enable caller
+# graphs for selected functions only using the \callergraph command.
+
+CALLER_GRAPH           = NO
+
+# If the GRAPHICAL_HIERARCHY and HAVE_DOT tags are set to YES then doxygen
+# will generate a graphical hierarchy of all classes instead of a textual one.
+
+GRAPHICAL_HIERARCHY    = YES
+
+# If the DIRECTORY_GRAPH and HAVE_DOT tags are set to YES
+# then doxygen will show the dependencies a directory has on other directories
+# in a graphical way. The dependency relations are determined by the #include
+# relations between the files in the directories.
+
+DIRECTORY_GRAPH        = YES
+
+# The DOT_IMAGE_FORMAT tag can be used to set the image format of the images
+# generated by dot. Possible values are svg, png, jpg, or gif.
+# If left blank png will be used. If you choose svg you need to set
+# HTML_FILE_EXTENSION to xhtml in order to make the SVG files
+# visible in IE 9+ (other browsers do not have this requirement).
+
+DOT_IMAGE_FORMAT       = png
+
+# If DOT_IMAGE_FORMAT is set to svg, then this option can be set to YES to
+# enable generation of interactive SVG images that allow zooming and panning.
+# Note that this requires a modern browser other than Internet Explorer.
+# Tested and working are Firefox, Chrome, Safari, and Opera. For IE 9+ you
+# need to set HTML_FILE_EXTENSION to xhtml in order to make the SVG files
+# visible. Older versions of IE do not have SVG support.
+
+INTERACTIVE_SVG        = NO
+
+# The tag DOT_PATH can be used to specify the path where the dot tool can be
+# found. If left blank, it is assumed the dot tool can be found in the path.
+
+DOT_PATH               =
+
+# The DOTFILE_DIRS tag can be used to specify one or more directories that
+# contain dot files that are included in the documentation (see the
+# \dotfile command).
+
+DOTFILE_DIRS           =
+
+# The MSCFILE_DIRS tag can be used to specify one or more directories that
+# contain msc files that are included in the documentation (see the
+# \mscfile command).
+
+MSCFILE_DIRS           =
+
+# The DOT_GRAPH_MAX_NODES tag can be used to set the maximum number of
+# nodes that will be shown in the graph. If the number of nodes in a graph
+# becomes larger than this value, doxygen will truncate the graph, which is
+# visualized by representing a node as a red box. Note that doxygen if the
+# number of direct children of the root node in a graph is already larger than
+# DOT_GRAPH_MAX_NODES then the graph will not be shown at all. Also note
+# that the size of a graph can be further restricted by MAX_DOT_GRAPH_DEPTH.
+
+DOT_GRAPH_MAX_NODES    = 50
+
+# The MAX_DOT_GRAPH_DEPTH tag can be used to set the maximum depth of the
+# graphs generated by dot. A depth value of 3 means that only nodes reachable
+# from the root by following a path via at most 3 edges will be shown. Nodes
+# that lay further from the root node will be omitted. Note that setting this
+# option to 1 or 2 may greatly reduce the computation time needed for large
+# code bases. Also note that the size of a graph can be further restricted by
+# DOT_GRAPH_MAX_NODES. Using a depth of 0 means no depth restriction.
+
+MAX_DOT_GRAPH_DEPTH    = 0
+
+# Set the DOT_TRANSPARENT tag to YES to generate images with a transparent
+# background. This is disabled by default, because dot on Windows does not
+# seem to support this out of the box. Warning: Depending on the platform used,
+# enabling this option may lead to badly anti-aliased labels on the edges of
+# a graph (i.e. they become hard to read).
+
+DOT_TRANSPARENT        = YES
+
+# Set the DOT_MULTI_TARGETS tag to YES allow dot to generate multiple output
+# files in one run (i.e. multiple -o and -T options on the command line). This
+# makes dot run faster, but since only newer versions of dot (>1.8.10)
+# support this, this feature is disabled by default.
+
+DOT_MULTI_TARGETS      = NO
+
+# If the GENERATE_LEGEND tag is set to YES (the default) Doxygen will
+# generate a legend page explaining the meaning of the various boxes and
+# arrows in the dot generated graphs.
+
+GENERATE_LEGEND        = YES
+
+# If the DOT_CLEANUP tag is set to YES (the default) Doxygen will
+# remove the intermediate dot files that are used to generate
+# the various graphs.
+
+DOT_CLEANUP            = YES
diff --git a/ctdb/lib/talloc/install-sh b/ctdb/lib/talloc/install-sh
new file mode 100755
index 0000000000..58719246f0
--- /dev/null
+++ b/ctdb/lib/talloc/install-sh
@@ -0,0 +1,238 @@
+#! /bin/sh
+#
+# install - install a program, script, or datafile
+# This comes from X11R5.
+#
+# Calling this script install-sh is preferred over install.sh, to prevent
+# `make' implicit rules from creating a file called install from it
+# when there is no Makefile.
+#
+# This script is compatible with the BSD install script, but was written
+# from scratch.
+#
+
+
+# set DOITPROG to echo to test this script
+
+# Don't use :- since 4.3BSD and earlier shells don't like it.
+doit="${DOITPROG-}"
+
+
+# put in absolute paths if you don't have them in your path; or use env. vars.
+
+mvprog="${MVPROG-mv}"
+cpprog="${CPPROG-cp}"
+chmodprog="${CHMODPROG-chmod}"
+chownprog="${CHOWNPROG-chown}"
+chgrpprog="${CHGRPPROG-chgrp}"
+stripprog="${STRIPPROG-strip}"
+rmprog="${RMPROG-rm}"
+mkdirprog="${MKDIRPROG-mkdir}"
+
+transformbasename=""
+transform_arg=""
+instcmd="$mvprog"
+chmodcmd="$chmodprog 0755"
+chowncmd=""
+chgrpcmd=""
+stripcmd=""
+rmcmd="$rmprog -f"
+mvcmd="$mvprog"
+src=""
+dst=""
+dir_arg=""
+
+while [ x"$1" != x ]; do
+    case $1 in
+	-c) instcmd="$cpprog"
+	    shift
+	    continue;;
+
+	-d) dir_arg=true
+	    shift
+	    continue;;
+
+	-m) chmodcmd="$chmodprog $2"
+	    shift
+	    shift
+	    continue;;
+
+	-o) chowncmd="$chownprog $2"
+	    shift
+	    shift
+	    continue;;
+
+	-g) chgrpcmd="$chgrpprog $2"
+	    shift
+	    shift
+	    continue;;
+
+	-s) stripcmd="$stripprog"
+	    shift
+	    continue;;
+
+	-t=*) transformarg=`echo $1 | sed 's/-t=//'`
+	    shift
+	    continue;;
+
+	-b=*) transformbasename=`echo $1 | sed 's/-b=//'`
+	    shift
+	    continue;;
+
+	*)  if [ x"$src" = x ]
+	    then
+		src=$1
+	    else
+		# this colon is to work around a 386BSD /bin/sh bug
+		:
+		dst=$1
+	    fi
+	    shift
+	    continue;;
+    esac
+done
+
+if [ x"$src" = x ]
+then
+	echo "install:	no input file specified"
+	exit 1
+else
+	true
+fi
+
+if [ x"$dir_arg" != x ]; then
+	dst=$src
+	src=""
+	
+	if [ -d $dst ]; then
+		instcmd=:
+	else
+		instcmd=mkdir
+	fi
+else
+
+# Waiting for this to be detected by the "$instcmd $src $dsttmp" command
+# might cause directories to be created, which would be especially bad 
+# if $src (and thus $dsttmp) contains '*'.
+
+	if [ -f $src -o -d $src ]
+	then
+		true
+	else
+		echo "install:  $src does not exist"
+		exit 1
+	fi
+	
+	if [ x"$dst" = x ]
+	then
+		echo "install:	no destination specified"
+		exit 1
+	else
+		true
+	fi
+
+# If destination is a directory, append the input filename; if your system
+# does not like double slashes in filenames, you may need to add some logic
+
+	if [ -d $dst ]
+	then
+		dst="$dst"/`basename $src`
+	else
+		true
+	fi
+fi
+
+## this sed command emulates the dirname command
+dstdir=`echo $dst | sed -e 's,[^/]*$,,;s,/$,,;s,^$,.,'`
+
+# Make sure that the destination directory exists.
+#  this part is taken from Noah Friedman's mkinstalldirs script
+
+# Skip lots of stat calls in the usual case.
+if [ ! -d "$dstdir" ]; then
+defaultIFS='	
+'
+IFS="${IFS-${defaultIFS}}"
+
+oIFS="${IFS}"
+# Some sh's can't handle IFS=/ for some reason.
+IFS='%'
+set - `echo ${dstdir} | sed -e 's@/@%@g' -e 's@^%@/@'`
+IFS="${oIFS}"
+
+pathcomp=''
+
+while [ $# -ne 0 ] ; do
+	pathcomp="${pathcomp}${1}"
+	shift
+
+	if [ ! -d "${pathcomp}" ] ;
+        then
+		$mkdirprog "${pathcomp}"
+	else
+		true
+	fi
+
+	pathcomp="${pathcomp}/"
+done
+fi
+
+if [ x"$dir_arg" != x ]
+then
+	$doit $instcmd $dst &&
+
+	if [ x"$chowncmd" != x ]; then $doit $chowncmd $dst; else true ; fi &&
+	if [ x"$chgrpcmd" != x ]; then $doit $chgrpcmd $dst; else true ; fi &&
+	if [ x"$stripcmd" != x ]; then $doit $stripcmd $dst; else true ; fi &&
+	if [ x"$chmodcmd" != x ]; then $doit $chmodcmd $dst; else true ; fi
+else
+
+# If we're going to rename the final executable, determine the name now.
+
+	if [ x"$transformarg" = x ] 
+	then
+		dstfile=`basename $dst`
+	else
+		dstfile=`basename $dst $transformbasename | 
+			sed $transformarg`$transformbasename
+	fi
+
+# don't allow the sed command to completely eliminate the filename
+
+	if [ x"$dstfile" = x ] 
+	then
+		dstfile=`basename $dst`
+	else
+		true
+	fi
+
+# Make a temp file name in the proper directory.
+
+	dsttmp=$dstdir/#inst.$$#
+
+# Move or copy the file name to the temp name
+
+	$doit $instcmd $src $dsttmp &&
+
+	trap "rm -f ${dsttmp}" 0 &&
+
+# and set any options; do chmod last to preserve setuid bits
+
+# If any of these fail, we abort the whole thing.  If we want to
+# ignore errors from any of these, just make sure not to ignore
+# errors from the above "$doit $instcmd $src $dsttmp" command.
+
+	if [ x"$chowncmd" != x ]; then $doit $chowncmd $dsttmp; else true;fi &&
+	if [ x"$chgrpcmd" != x ]; then $doit $chgrpcmd $dsttmp; else true;fi &&
+	if [ x"$stripcmd" != x ]; then $doit $stripcmd $dsttmp; else true;fi &&
+	if [ x"$chmodcmd" != x ]; then $doit $chmodcmd $dsttmp; else true;fi &&
+
+# Now rename the file to the real destination.
+
+	$doit $rmcmd -f $dstdir/$dstfile &&
+	$doit $mvcmd $dsttmp $dstdir/$dstfile 
+
+fi &&
+
+
+exit 0
diff --git a/ctdb/lib/talloc/libtalloc.m4 b/ctdb/lib/talloc/libtalloc.m4
new file mode 100644
index 0000000000..65227d89f6
--- /dev/null
+++ b/ctdb/lib/talloc/libtalloc.m4
@@ -0,0 +1,63 @@
+dnl Check to see if we should use the included talloc
+
+INCLUDED_TALLOC=auto
+AC_ARG_WITH(included-talloc,
+    [AC_HELP_STRING([--with-included-talloc], [use bundled talloc library, not from system])],
+    [ INCLUDED_TALLOC=$withval ])
+
+AC_SUBST(TALLOC_LIBS)
+AC_SUBST(TALLOC_CFLAGS)
+
+if test x"$INCLUDED_TALLOC" != x"yes" ; then
+    AC_CHECK_HEADERS(talloc.h)
+    AC_CHECK_LIB(talloc, talloc_init, [ TALLOC_LIBS="-ltalloc" ])
+    if test x"$ac_cv_header_talloc_h" = x"no" -o x"$ac_cv_lib_talloc_talloc_init" = x"no" ; then
+        INCLUDED_TALLOC=yes
+        TALLOC_CFLAGS=""
+    else
+        INCLUDED_TALLOC=no
+    fi
+fi
+
+AC_MSG_CHECKING(whether to use included talloc)
+AC_MSG_RESULT($INCLUDED_TALLOC)
+if test x"$INCLUDED_TALLOC" != x"no" ; then
+    dnl find the talloc sources. This is meant to work both for 
+    dnl talloc standalone builds, and builds of packages using talloc
+    tallocdir=""
+    tallocpaths=". lib/talloc talloc ../talloc ../lib/talloc"
+    for d in $tallocpaths; do
+    	if test -f "$srcdir/$d/talloc.c"; then
+    		tallocdir="$d"
+    		AC_SUBST(tallocdir)
+    		break
+    	fi
+    done
+    if test x"$tallocdir" = "x"; then
+        AC_MSG_ERROR([cannot find talloc source in $tallocpaths])
+    fi
+    TALLOC_OBJ="talloc.o"
+    AC_SUBST(TALLOC_OBJ)
+
+    TALLOC_CFLAGS="-I$srcdir/$tallocdir"
+    AC_SUBST(TALLOC_CFLAGS)
+
+    TALLOC_LIBS=""
+    AC_SUBST(TALLOC_LIBS)
+fi
+
+AC_CHECK_SIZEOF(size_t,cross)
+AC_CHECK_SIZEOF(void *,cross)
+
+if test $ac_cv_sizeof_size_t -lt $ac_cv_sizeof_void_p; then
+	AC_WARN([size_t cannot represent the amount of used memory of a process])
+	AC_WARN([please report this to <samba-technical@samba.org>])
+	AC_WARN([sizeof(size_t) = $ac_cv_sizeof_size_t])
+	AC_WARN([sizeof(void *) = $ac_cv_sizeof_void_p])
+	AC_ERROR([sizeof(size_t) < sizeof(void *)])
+fi
+
+if test x"$VERSIONSCRIPT" != "x"; then
+    EXPORTSFILE=talloc.exports
+    AC_SUBST(EXPORTSFILE)
+fi
diff --git a/ctdb/lib/talloc/pytalloc-util.pc.in b/ctdb/lib/talloc/pytalloc-util.pc.in
new file mode 100644
index 0000000000..bc704b4929
--- /dev/null
+++ b/ctdb/lib/talloc/pytalloc-util.pc.in
@@ -0,0 +1,11 @@
+prefix=@prefix@
+exec_prefix=@exec_prefix@
+libdir=@libdir@
+includedir=@includedir@
+
+Name: pytalloc-util
+Description: Utility functions for using talloc objects with Python
+Version: @TALLOC_VERSION@
+Libs: -L${libdir} -lpytalloc-util
+Cflags: @LIB_RPATH@ -I${includedir}
+URL: http://talloc.samba.org/
diff --git a/ctdb/lib/talloc/pytalloc.c b/ctdb/lib/talloc/pytalloc.c
new file mode 100644
index 0000000000..80196c6c77
--- /dev/null
+++ b/ctdb/lib/talloc/pytalloc.c
@@ -0,0 +1,134 @@
+/* 
+   Unix SMB/CIFS implementation.
+   Python Talloc Module
+   Copyright (C) Jelmer Vernooij <jelmer@samba.org> 2010-2011
+
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include <Python.h>
+#include <talloc.h>
+#include <pytalloc.h>
+
+void inittalloc(void);
+
+/* print a talloc tree report for a talloc python object */
+static PyObject *pytalloc_report_full(PyObject *self, PyObject *args)
+{
+	PyObject *py_obj = Py_None;
+
+	if (!PyArg_ParseTuple(args, "|O", &py_obj))
+		return NULL;
+
+	if (py_obj == Py_None) {
+		talloc_report_full(NULL, stdout);
+	} else {
+		talloc_report_full(pytalloc_get_mem_ctx(py_obj), stdout);
+	}
+	return Py_None;
+}
+
+/* enable null tracking */
+static PyObject *pytalloc_enable_null_tracking(PyObject *self)
+{
+	talloc_enable_null_tracking();
+	return Py_None;
+}
+
+/* return the number of talloc blocks */
+static PyObject *pytalloc_total_blocks(PyObject *self, PyObject *args)
+{
+	PyObject *py_obj = Py_None;
+
+	if (!PyArg_ParseTuple(args, "|O", &py_obj))
+		return NULL;
+
+	if (py_obj == Py_None) {
+		return PyLong_FromLong(talloc_total_blocks(NULL));
+	}
+
+	return PyLong_FromLong(talloc_total_blocks(pytalloc_get_mem_ctx(py_obj)));
+}
+
+static PyMethodDef talloc_methods[] = {
+	{ "report_full", (PyCFunction)pytalloc_report_full, METH_VARARGS,
+		"show a talloc tree for an object"},
+	{ "enable_null_tracking", (PyCFunction)pytalloc_enable_null_tracking, METH_NOARGS,
+		"enable tracking of the NULL object"},
+	{ "total_blocks", (PyCFunction)pytalloc_total_blocks, METH_VARARGS,
+		"return talloc block count"},
+	{ NULL }
+};
+
+/**
+ * Default (but only slightly more useful than the default) implementation of Repr().
+ */
+static PyObject *pytalloc_default_repr(PyObject *obj)
+{
+	pytalloc_Object *talloc_obj = (pytalloc_Object *)obj;
+	PyTypeObject *type = (PyTypeObject*)PyObject_Type(obj);
+
+	return PyString_FromFormat("<%s talloc object at 0x%p>", 
+				   type->tp_name, talloc_obj->ptr);
+}
+
+/**
+ * Simple dealloc for talloc-wrapping PyObjects
+ */
+static void pytalloc_dealloc(PyObject* self)
+{
+	pytalloc_Object *obj = (pytalloc_Object *)self;
+	assert(talloc_unlink(NULL, obj->talloc_ctx) != -1);
+	obj->talloc_ctx = NULL;
+	self->ob_type->tp_free(self);
+}
+
+/**
+ * Default (but only slightly more useful than the default) implementation of cmp.
+ */
+static int pytalloc_default_cmp(PyObject *_obj1, PyObject *_obj2)
+{
+	pytalloc_Object *obj1 = (pytalloc_Object *)_obj1,
+					 *obj2 = (pytalloc_Object *)_obj2;
+	if (obj1->ob_type != obj2->ob_type)
+		return (obj1->ob_type - obj2->ob_type);
+
+	return ((char *)pytalloc_get_ptr(obj1) - (char *)pytalloc_get_ptr(obj2));
+}
+
+static PyTypeObject TallocObject_Type = {
+	.tp_name = "talloc.Object",
+	.tp_doc = "Python wrapper for a talloc-maintained object.",
+	.tp_basicsize = sizeof(pytalloc_Object),
+	.tp_dealloc = (destructor)pytalloc_dealloc,
+	.tp_flags = Py_TPFLAGS_DEFAULT | Py_TPFLAGS_BASETYPE,
+	.tp_repr = pytalloc_default_repr,
+	.tp_compare = pytalloc_default_cmp,
+};
+
+void inittalloc(void)
+{
+	PyObject *m;
+
+	if (PyType_Ready(&TallocObject_Type) < 0)
+		return;
+
+	m = Py_InitModule3("talloc", talloc_methods,
+					   "Python wrapping of talloc-maintained objects.");
+	if (m == NULL)
+		return;
+
+	Py_INCREF(&TallocObject_Type);
+	PyModule_AddObject(m, "Object", (PyObject *)&TallocObject_Type);
+}
diff --git a/ctdb/lib/talloc/pytalloc.h b/ctdb/lib/talloc/pytalloc.h
new file mode 100644
index 0000000000..2d2c57b789
--- /dev/null
+++ b/ctdb/lib/talloc/pytalloc.h
@@ -0,0 +1,54 @@
+/* 
+   Unix SMB/CIFS implementation.
+   Samba utility functions
+   Copyright (C) Jelmer Vernooij <jelmer@samba.org> 2008
+   
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.
+*/
+
+#ifndef _PYTALLOC_H_
+#define _PYTALLOC_H_
+
+#include <Python.h>
+#include <talloc.h>
+
+typedef struct {
+	PyObject_HEAD
+	TALLOC_CTX *talloc_ctx;
+	void *ptr;
+} pytalloc_Object;
+
+PyTypeObject *pytalloc_GetObjectType(void);
+int pytalloc_Check(PyObject *);
+
+/* Retrieve the pointer for a pytalloc_object. Like talloc_get_type() 
+ * but for pytalloc_Objects. */
+
+/* FIXME: Call PyErr_SetString(PyExc_TypeError, "expected " __STR(type) ") 
+ * when talloc_get_type() returns NULL. */
+#define pytalloc_get_type(py_obj, type) (talloc_get_type(pytalloc_get_ptr(py_obj), type))
+
+#define pytalloc_get_ptr(py_obj) (((pytalloc_Object *)py_obj)->ptr)
+#define pytalloc_get_mem_ctx(py_obj)  ((pytalloc_Object *)py_obj)->talloc_ctx
+
+PyObject *pytalloc_steal_ex(PyTypeObject *py_type, TALLOC_CTX *mem_ctx, void *ptr);
+PyObject *pytalloc_steal(PyTypeObject *py_type, void *ptr);
+PyObject *pytalloc_reference_ex(PyTypeObject *py_type, TALLOC_CTX *mem_ctx, void *ptr);
+#define pytalloc_reference(py_type, talloc_ptr) pytalloc_reference_ex(py_type, talloc_ptr, talloc_ptr)
+
+#define pytalloc_new(type, typeobj) pytalloc_steal(typeobj, talloc_zero(NULL, type))
+
+PyObject *pytalloc_CObject_FromTallocPtr(void *);
+
+#endif /* _PYTALLOC_H_ */
diff --git a/ctdb/lib/talloc/pytalloc_util.c b/ctdb/lib/talloc/pytalloc_util.c
new file mode 100644
index 0000000000..89a093b1ce
--- /dev/null
+++ b/ctdb/lib/talloc/pytalloc_util.c
@@ -0,0 +1,118 @@
+/* 
+   Unix SMB/CIFS implementation.
+   Python/Talloc glue
+   Copyright (C) Jelmer Vernooij <jelmer@samba.org> 2008
+   
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include <Python.h>
+#include "replace.h"
+#include <talloc.h>
+#include "pytalloc.h"
+#include <assert.h>
+
+_PUBLIC_ PyTypeObject *pytalloc_GetObjectType(void)
+{
+	static PyTypeObject *type = NULL;
+	PyObject *mod;
+
+	if (type != NULL) {
+		return type;
+	}
+
+	mod = PyImport_ImportModule("talloc");
+	if (mod == NULL) {
+		return NULL;
+	}
+
+	type = (PyTypeObject *)PyObject_GetAttrString(mod, "Object");
+	Py_DECREF(mod);
+
+	return type;
+}
+
+/**
+ * Import an existing talloc pointer into a Python object.
+ */
+_PUBLIC_ PyObject *pytalloc_steal_ex(PyTypeObject *py_type, TALLOC_CTX *mem_ctx,
+						   void *ptr)
+{
+	pytalloc_Object *ret = (pytalloc_Object *)py_type->tp_alloc(py_type, 0);
+	ret->talloc_ctx = talloc_new(NULL);
+	if (ret->talloc_ctx == NULL) {
+		return NULL;
+	}
+	if (talloc_steal(ret->talloc_ctx, mem_ctx) == NULL) {
+		return NULL;
+	}
+	talloc_set_name_const(ret->talloc_ctx, py_type->tp_name);
+	ret->ptr = ptr;
+	return (PyObject *)ret;
+}
+
+/**
+ * Import an existing talloc pointer into a Python object.
+ */
+_PUBLIC_ PyObject *pytalloc_steal(PyTypeObject *py_type, void *ptr)
+{
+	return pytalloc_steal_ex(py_type, ptr, ptr);
+}
+
+
+/**
+ * Import an existing talloc pointer into a Python object, leaving the
+ * original parent, and creating a reference to the object in the python
+ * object
+ */
+_PUBLIC_ PyObject *pytalloc_reference_ex(PyTypeObject *py_type, TALLOC_CTX *mem_ctx, void *ptr)
+{
+	pytalloc_Object *ret;
+
+	if (ptr == NULL) {
+		Py_RETURN_NONE;
+	}
+
+	ret = (pytalloc_Object *)py_type->tp_alloc(py_type, 0);
+	ret->talloc_ctx = talloc_new(NULL);
+	if (ret->talloc_ctx == NULL) {
+		return NULL;
+	}
+	if (talloc_reference(ret->talloc_ctx, mem_ctx) == NULL) {
+		return NULL;
+	}
+	talloc_set_name_const(ret->talloc_ctx, py_type->tp_name);
+	ret->ptr = ptr;
+	return (PyObject *)ret;
+}
+
+static void py_cobject_talloc_free(void *ptr)
+{
+	talloc_free(ptr);
+}
+
+_PUBLIC_ PyObject *pytalloc_CObject_FromTallocPtr(void *ptr)
+{
+	if (ptr == NULL) {
+		Py_RETURN_NONE;
+	}
+	return PyCObject_FromVoidPtr(ptr, py_cobject_talloc_free);
+}
+
+_PUBLIC_ int pytalloc_Check(PyObject *obj)
+{
+	PyTypeObject *tp = pytalloc_GetObjectType();
+
+	return PyObject_TypeCheck(obj, tp);
+}
diff --git a/ctdb/lib/talloc/talloc.3.xml b/ctdb/lib/talloc/talloc.3.xml
new file mode 100644
index 0000000000..99e8bcdb2f
--- /dev/null
+++ b/ctdb/lib/talloc/talloc.3.xml
@@ -0,0 +1,801 @@
+<?xml version="1.0"?>
+<!DOCTYPE refentry PUBLIC "-//OASIS//DTD DocBook XML V4.2//EN" "http://www.oasis-open.org/docbook/xml/4.2/docbookx.dtd">
+<refentry>
+  <refmeta>
+    <refentrytitle>talloc</refentrytitle>
+    <manvolnum>3</manvolnum>
+  </refmeta>
+  <refnamediv>
+    <refname>talloc</refname>
+<refpurpose>hierarchical reference counted memory pool system with destructors</refpurpose>
+  </refnamediv>
+  <refsynopsisdiv>
+<synopsis>#include &lt;talloc.h&gt;</synopsis>
+  </refsynopsisdiv>
+  <refsect1><title>DESCRIPTION</title>
+    <para>
+      If you are used to talloc from Samba3 then please read this
+      carefully, as talloc has changed a lot.
+    </para>
+    <para>
+      The new talloc is a hierarchical, reference counted memory pool
+      system with destructors.	Quite a mouthful really, but not too bad
+      once you get used to it.
+    </para>
+    <para>
+      Perhaps the biggest change from Samba3 is that there is no
+      distinction between a "talloc context" and a "talloc pointer".  Any
+      pointer returned from talloc() is itself a valid talloc context. 
+      This means you can do this:
+    </para>
+    <programlisting>
+    struct foo *X = talloc(mem_ctx, struct foo);
+    X->name = talloc_strdup(X, "foo");
+    </programlisting>
+    <para>
+      and the pointer <literal role="code">X-&gt;name</literal>
+      would be a "child" of the talloc context <literal
+      role="code">X</literal> which is itself a child of
+      <literal role="code">mem_ctx</literal>.  So if you do
+      <literal role="code">talloc_free(mem_ctx)</literal> then
+      it is all destroyed, whereas if you do <literal
+      role="code">talloc_free(X)</literal> then just <literal
+      role="code">X</literal> and <literal
+      role="code">X-&gt;name</literal> are destroyed, and if
+      you do <literal
+      role="code">talloc_free(X-&gt;name)</literal> then just
+      the name element of <literal role="code">X</literal> is
+      destroyed.
+    </para>
+    <para>
+      If you think about this, then what this effectively gives you is an
+      n-ary tree, where you can free any part of the tree with
+      talloc_free().
+    </para>
+    <para>
+      If you find this confusing, then I suggest you run the <literal
+      role="code">testsuite</literal> program to watch talloc
+      in action.  You may also like to add your own tests to <literal
+      role="code">testsuite.c</literal> to clarify how some
+      particular situation is handled.
+    </para>
+  </refsect1>
+  <refsect1><title>TALLOC API</title>
+    <para>
+      The following is a complete guide to the talloc API. Read it all at
+      least twice.
+    </para>
+    <refsect2><title>(type *)talloc(const void *ctx, type);</title>
+        <para>
+	  The talloc() macro is the core of the talloc library.  It takes a
+	  memory <emphasis role="italic">ctx</emphasis> and a <emphasis
+	  role="italic">type</emphasis>, and returns a pointer to a new
+	  area of memory of the given <emphasis
+	  role="italic">type</emphasis>.
+        </para>
+        <para>
+	  The returned pointer is itself a talloc context, so you can use
+	  it as the <emphasis role="italic">ctx</emphasis> argument to more
+	  calls to talloc() if you wish.
+        </para>
+        <para>
+	  The returned pointer is a "child" of the supplied context.  This
+	  means that if you talloc_free() the <emphasis
+	  role="italic">ctx</emphasis> then the new child disappears as
+	  well.  Alternatively you can free just the child.
+        </para>
+        <para>
+	  The <emphasis role="italic">ctx</emphasis> argument to talloc()
+	  can be NULL, in which case a new top level context is created.
+        </para>
+    </refsect2>
+    <refsect2><title>void *talloc_size(const void *ctx, size_t size);</title>
+        <para>
+	  The function talloc_size() should be used when you don't have a
+	  convenient type to pass to talloc().	Unlike talloc(), it is not
+	  type safe (as it returns a void *), so you are on your own for
+	  type checking.
+        </para>
+    </refsect2>
+    <refsect2><title>(typeof(ptr)) talloc_ptrtype(const void *ctx, ptr);</title>
+        <para>
+	  The talloc_ptrtype() macro should be used when you have a pointer and
+	  want to allocate memory to point at with this pointer. When compiling
+	  with gcc >= 3 it is typesafe. Note this is a wrapper of talloc_size()
+	  and talloc_get_name() will return the current location in the source file.
+	  and not the type.
+        </para>
+    </refsect2>
+    <refsect2><title>int talloc_free(void *ptr);</title>
+        <para>
+	  The talloc_free() function frees a piece of talloc memory, and
+	  all its children.  You can call talloc_free() on any pointer
+	  returned by talloc().
+        </para>
+        <para>
+	  The return value of talloc_free() indicates success or failure,
+	  with 0 returned for success and -1 for failure.  The only
+	  possible failure condition is if <emphasis
+	  role="italic">ptr</emphasis> had a destructor attached to it and
+	  the destructor returned -1.  See <link
+	  linkend="talloc_set_destructor"><quote>talloc_set_destructor()</quote></link>
+	  for details on destructors.
+        </para>
+        <para>
+	  If this pointer has an additional parent when talloc_free() is
+	  called then the memory is not actually released, but instead the
+	  most recently established parent is destroyed.  See <link
+	  linkend="talloc_reference"><quote>talloc_reference()</quote></link>
+	  for details on establishing additional parents.
+        </para>
+        <para>
+	  For more control on which parent is removed, see <link
+	  linkend="talloc_unlink"><quote>talloc_unlink()</quote></link>.
+        </para>
+        <para>
+	  talloc_free() operates recursively on its children.
+        </para>
+	<para>
+	  From the 2.0 version of talloc, as a special case,
+	  talloc_free() is refused on pointers that have more than one
+	  parent, as talloc would have no way of knowing which parent
+	  should be removed. To free a pointer that has more than one
+	  parent please use talloc_unlink().
+	</para>
+	<para>
+	  To help you find problems in your code caused by this behaviour, if
+	  you do try and free a pointer with more than one parent then the
+	  talloc logging function will be called to give output like this:
+	</para>
+	<para>
+	  <screen format="linespecific">
+	    ERROR: talloc_free with references at some_dir/source/foo.c:123
+  	  	reference at some_dir/source/other.c:325
+  	  	reference at some_dir/source/third.c:121
+	  </screen>
+	</para>
+	<para>
+	  Please see the documentation for talloc_set_log_fn() and
+	  talloc_set_log_stderr() for more information on talloc logging
+	  functions.
+	</para>
+    </refsect2>
+    <refsect2 id="talloc_reference"><title>void *talloc_reference(const void *ctx, const void *ptr);</title>
+        <para>
+	  The talloc_reference() function makes <emphasis
+	  role="italic">ctx</emphasis> an additional parent of <emphasis
+	  role="italic">ptr</emphasis>.
+        </para>
+        <para>
+	  The return value of talloc_reference() is always the original
+	  pointer <emphasis role="italic">ptr</emphasis>, unless talloc ran
+	  out of memory in creating the reference in which case it will
+	  return NULL (each additional reference consumes around 48 bytes
+	  of memory on intel x86 platforms).
+        </para>
+        <para>
+	  If <emphasis role="italic">ptr</emphasis> is NULL, then the
+	  function is a no-op, and simply returns NULL.
+        </para>
+        <para>
+	  After creating a reference you can free it in one of the
+	  following ways:
+        </para>
+      <para>
+        <itemizedlist>
+          <listitem>
+            <para>
+	      you can talloc_free() any parent of the original pointer. 
+	      That will reduce the number of parents of this pointer by 1,
+	      and will cause this pointer to be freed if it runs out of
+	      parents.
+            </para>
+          </listitem>
+          <listitem>
+            <para>
+	      you can talloc_free() the pointer itself.  That will destroy
+	      the most recently established parent to the pointer and leave
+	      the pointer as a child of its current parent.
+            </para>
+          </listitem>
+        </itemizedlist>
+      </para>
+      <para>
+	For more control on which parent to remove, see <link
+	linkend="talloc_unlink"><quote>talloc_unlink()</quote></link>.
+      </para>
+    </refsect2>
+    <refsect2 id="talloc_unlink"><title>int talloc_unlink(const void *ctx, const void *ptr);</title>
+        <para>
+	  The talloc_unlink() function removes a specific parent from
+	  <emphasis role="italic">ptr</emphasis>. The <emphasis
+	  role="italic">ctx</emphasis> passed must either be a context used
+	  in talloc_reference() with this pointer, or must be a direct
+	  parent of ptr.
+        </para>
+        <para>
+	  Note that if the parent has already been removed using
+	  talloc_free() then this function will fail and will return -1. 
+	  Likewise, if <emphasis role="italic">ptr</emphasis> is NULL, then
+	  the function will make no modifications and return -1.
+        </para>
+        <para>
+	  Usually you can just use talloc_free() instead of
+	  talloc_unlink(), but sometimes it is useful to have the
+	  additional control on which parent is removed.
+        </para>
+    </refsect2>
+    <refsect2 id="talloc_set_destructor"><title>void talloc_set_destructor(const void *ptr, int (*destructor)(void *));</title>
+        <para>
+	  The function talloc_set_destructor() sets the <emphasis
+	  role="italic">destructor</emphasis> for the pointer <emphasis
+	  role="italic">ptr</emphasis>.  A <emphasis
+	  role="italic">destructor</emphasis> is a function that is called
+	  when the memory used by a pointer is about to be released.  The
+	  destructor receives <emphasis role="italic">ptr</emphasis> as an
+	  argument, and should return 0 for success and -1 for failure.
+        </para>
+        <para>
+	  The <emphasis role="italic">destructor</emphasis> can do anything
+	  it wants to, including freeing other pieces of memory.  A common
+	  use for destructors is to clean up operating system resources
+	  (such as open file descriptors) contained in the structure the
+	  destructor is placed on.
+        </para>
+        <para>
+	  You can only place one destructor on a pointer.  If you need more
+	  than one destructor then you can create a zero-length child of
+	  the pointer and place an additional destructor on that.
+        </para>
+        <para>
+	  To remove a destructor call talloc_set_destructor() with NULL for
+	  the destructor.
+        </para>
+        <para>
+	  If your destructor attempts to talloc_free() the pointer that it
+	  is the destructor for then talloc_free() will return -1 and the
+	  free will be ignored.  This would be a pointless operation
+	  anyway, as the destructor is only called when the memory is just
+	  about to go away.
+        </para>
+    </refsect2>
+    <refsect2><title>int talloc_increase_ref_count(const void *<emphasis role="italic">ptr</emphasis>);</title>
+        <para>
+	  The talloc_increase_ref_count(<emphasis
+	  role="italic">ptr</emphasis>) function is exactly equivalent to:
+        </para>
+        <programlisting>talloc_reference(NULL, ptr);</programlisting>
+        <para>
+	  You can use either syntax, depending on which you think is
+	  clearer in your code.
+        </para>
+        <para>
+	  It returns 0 on success and -1 on failure.
+        </para>
+    </refsect2>
+    <refsect2><title>size_t talloc_reference_count(const void *<emphasis role="italic">ptr</emphasis>);</title>
+        <para>
+	  Return the number of references to the pointer.
+        </para>
+    </refsect2>
+    <refsect2 id="talloc_set_name"><title>void talloc_set_name(const void *ptr, const char *fmt, ...);</title>
+        <para>
+	  Each talloc pointer has a "name".  The name is used principally
+	  for debugging purposes, although it is also possible to set and
+	  get the name on a pointer in as a way of "marking" pointers in
+	  your code.
+        </para>
+        <para>
+	  The main use for names on pointer is for "talloc reports".  See
+	  <link
+	  linkend="talloc_report"><quote>talloc_report_depth_cb()</quote></link>,
+	  <link
+	  linkend="talloc_report"><quote>talloc_report_depth_file()</quote></link>,
+	  <link
+	  linkend="talloc_report"><quote>talloc_report()</quote></link>
+	  <link
+	  linkend="talloc_report"><quote>talloc_report()</quote></link>
+	  and <link
+	  linkend="talloc_report_full"><quote>talloc_report_full()</quote></link>
+	  for details.	Also see <link
+	  linkend="talloc_enable_leak_report"><quote>talloc_enable_leak_report()</quote></link>
+	  and <link
+	  linkend="talloc_enable_leak_report_full"><quote>talloc_enable_leak_report_full()</quote></link>.
+        </para>
+        <para>
+	  The talloc_set_name() function allocates memory as a child of the
+	  pointer.  It is logically equivalent to:
+        </para>
+        <programlisting>talloc_set_name_const(ptr, talloc_asprintf(ptr, fmt, ...));</programlisting>
+        <para>
+	  Note that multiple calls to talloc_set_name() will allocate more
+	  memory without releasing the name.  All of the memory is released
+	  when the ptr is freed using talloc_free().
+        </para>
+    </refsect2>
+    <refsect2><title>void talloc_set_name_const(const void *<emphasis role="italic">ptr</emphasis>, const char *<emphasis role="italic">name</emphasis>);</title>
+        <para>
+	  The function talloc_set_name_const() is just like
+	  talloc_set_name(), but it takes a string constant, and is much
+	  faster.  It is extensively used by the "auto naming" macros, such
+	  as talloc_p().
+        </para>
+        <para>
+	  This function does not allocate any memory.  It just copies the
+	  supplied pointer into the internal representation of the talloc
+	  ptr. This means you must not pass a <emphasis
+	  role="italic">name</emphasis> pointer to memory that will
+	  disappear before <emphasis role="italic">ptr</emphasis> is freed
+	  with talloc_free().
+        </para>
+    </refsect2>
+    <refsect2><title>void *talloc_named(const void *<emphasis role="italic">ctx</emphasis>, size_t <emphasis role="italic">size</emphasis>, const char *<emphasis role="italic">fmt</emphasis>, ...);</title>
+        <para>
+	  The talloc_named() function creates a named talloc pointer.  It
+	  is equivalent to:
+        </para>
+        <programlisting>ptr = talloc_size(ctx, size);
+talloc_set_name(ptr, fmt, ....);</programlisting>
+    </refsect2>
+    <refsect2><title>void *talloc_named_const(const void *<emphasis role="italic">ctx</emphasis>, size_t <emphasis role="italic">size</emphasis>, const char *<emphasis role="italic">name</emphasis>);</title>
+        <para>
+	  This is equivalent to:
+        </para>
+        <programlisting>ptr = talloc_size(ctx, size);
+talloc_set_name_const(ptr, name);</programlisting>
+    </refsect2>
+    <refsect2><title>const char *talloc_get_name(const void *<emphasis role="italic">ptr</emphasis>);</title>
+        <para>
+	  This returns the current name for the given talloc pointer,
+	  <emphasis role="italic">ptr</emphasis>. See <link
+	  linkend="talloc_set_name"><quote>talloc_set_name()</quote></link>
+	  for details.
+        </para>
+    </refsect2>
+    <refsect2><title>void *talloc_init(const char *<emphasis role="italic">fmt</emphasis>, ...);</title>
+        <para>
+	  This function creates a zero length named talloc context as a top
+	  level context.  It is equivalent to:
+        </para>
+        <programlisting>talloc_named(NULL, 0, fmt, ...);</programlisting>
+    </refsect2>
+    <refsect2><title>void *talloc_new(void *<emphasis role="italic">ctx</emphasis>);</title>
+        <para>
+	  This is a utility macro that creates a new memory context hanging
+	  off an existing context, automatically naming it "talloc_new:
+	  __location__" where __location__ is the source line it is called
+	  from.  It is particularly useful for creating a new temporary
+	  working context.
+        </para>
+    </refsect2>
+    <refsect2><title>(<emphasis role="italic">type</emphasis> *)talloc_realloc(const void *<emphasis role="italic">ctx</emphasis>, void *<emphasis role="italic">ptr</emphasis>, <emphasis role="italic">type</emphasis>, <emphasis role="italic">count</emphasis>);</title>
+        <para>
+	  The talloc_realloc() macro changes the size of a talloc pointer. 
+	  It has the following equivalences:
+        </para>
+        <programlisting>talloc_realloc(ctx, NULL, type, 1) ==> talloc(ctx, type);
+talloc_realloc(ctx, ptr, type, 0)  ==> talloc_free(ptr);</programlisting>
+        <para>
+	  The <emphasis role="italic">ctx</emphasis> argument is only used
+	  if <emphasis role="italic">ptr</emphasis> is not NULL, otherwise
+	  it is ignored.
+        </para>
+        <para>
+	  talloc_realloc() returns the new pointer, or NULL on failure. 
+	  The call will fail either due to a lack of memory, or because the
+	  pointer has more than one parent (see <link
+	  linkend="talloc_reference"><quote>talloc_reference()</quote></link>).
+        </para>
+    </refsect2>
+    <refsect2><title>void *talloc_realloc_size(const void *ctx, void *ptr, size_t size);</title>
+        <para>
+	  the talloc_realloc_size() function is useful when the type is not
+	  known so the type-safe talloc_realloc() cannot be used.
+        </para>
+    </refsect2>
+    <refsect2><title>TYPE *talloc_steal(const void *<emphasis role="italic">new_ctx</emphasis>, const TYPE *<emphasis role="italic">ptr</emphasis>);</title>
+        <para>
+	  The talloc_steal() function changes the parent context of a
+	  talloc pointer.  It is typically used when the context that the
+	  pointer is currently a child of is going to be freed and you wish
+	  to keep the memory for a longer time.
+        </para>
+        <para>
+	  The talloc_steal() function returns the pointer that you pass it.
+	   It does not have any failure modes.
+        </para>
+        <para>
+	  It is possible to produce loops in the parent/child
+	  relationship if you are not careful with talloc_steal().  No
+	  guarantees are provided as to your sanity or the safety of your
+	  data if you do this.
+        </para>
+        <para>
+	  Note that if you try and call talloc_steal() on a pointer that has
+	  more than one parent then the result is ambiguous. Talloc will choose
+	  to remove the parent that is currently indicated by talloc_parent()
+	  and replace it with the chosen parent. You will also get a message
+	  like this via the talloc logging functions:
+        </para>
+        <para>
+	  <screen format="linespecific">
+	  WARNING: talloc_steal with references at some_dir/source/foo.c:123
+  	  	reference at some_dir/source/other.c:325
+  	  	reference at some_dir/source/third.c:121
+	  </screen>
+        </para>
+        <para>
+	  To unambiguously change the parent of a pointer please see
+	  the
+	  function <link linkend="talloc_reference"><quote>talloc_reparent()</quote></link>. See
+	  the talloc_set_log_fn() documentation for more information
+	  on talloc logging.  
+	</para>
+    </refsect2>
+    <refsect2><title>TYPE *talloc_reparent(const void *<emphasis role="italic">old_parent</emphasis>, const void *<emphasis role="italic">new_parent</emphasis>, const TYPE *<emphasis role="italic">ptr</emphasis>);</title>
+        <para>
+	  The talloc_reparent() function changes the parent context of a talloc
+	  pointer. It is typically used when the context that the pointer is
+	  currently a child of is going to be freed and you wish to keep the
+	  memory for a longer time.
+        </para>
+        <para>
+	  The talloc_reparent() function returns the pointer that you pass it. It
+	  does not have any failure modes.
+        </para>
+        <para>
+	  The difference between talloc_reparent() and talloc_steal() is that
+	  talloc_reparent() can specify which parent you wish to change. This is
+	  useful when a pointer has multiple parents via references.
+        </para>
+    </refsect2>
+    <refsect2><title>TYPE *talloc_move(const void *<emphasis role="italic">new_ctx</emphasis>, TYPE **<emphasis role="italic">ptr</emphasis>);</title>
+        <para>
+	  The talloc_move() function is a wrapper around
+	  talloc_steal() which zeros the source pointer after the
+	  move. This avoids a potential source of bugs where a
+	  programmer leaves a pointer in two structures, and uses the
+	  pointer from the old structure after it has been moved to a
+	  new one.
+        </para>
+    </refsect2>
+    <refsect2><title>size_t talloc_total_size(const void *<emphasis role="italic">ptr</emphasis>);</title>
+        <para>
+	  The talloc_total_size() function returns the total size in bytes
+	  used by this pointer and all child pointers.	Mostly useful for
+	  debugging.
+        </para>
+        <para>
+	  Passing NULL is allowed, but it will only give a meaningful
+	  result if talloc_enable_leak_report() or
+	  talloc_enable_leak_report_full() has been called.
+        </para>
+    </refsect2>
+    <refsect2><title>size_t talloc_total_blocks(const void *<emphasis role="italic">ptr</emphasis>);</title>
+        <para>
+	  The talloc_total_blocks() function returns the total memory block
+	  count used by this pointer and all child pointers.  Mostly useful
+	  for debugging.
+        </para>
+        <para>
+	  Passing NULL is allowed, but it will only give a meaningful
+	  result if talloc_enable_leak_report() or
+	  talloc_enable_leak_report_full() has been called.
+        </para>
+    </refsect2>
+    <refsect2 id="talloc_report"><title>void talloc_report(const void *ptr, FILE *f);</title>
+        <para>
+	  The talloc_report() function prints a summary report of all
+	  memory used by <emphasis role="italic">ptr</emphasis>.  One line
+	  of report is printed for each immediate child of ptr, showing the
+	  total memory and number of blocks used by that child.
+        </para>
+        <para>
+	  You can pass NULL for the pointer, in which case a report is
+	  printed for the top level memory context, but only if
+	  talloc_enable_leak_report() or talloc_enable_leak_report_full()
+	  has been called.
+        </para>
+    </refsect2>
+    <refsect2 id="talloc_report_full"><title>void talloc_report_full(const void *<emphasis role="italic">ptr</emphasis>, FILE *<emphasis role="italic">f</emphasis>);</title>
+        <para>
+	  This provides a more detailed report than talloc_report().  It
+	  will recursively print the entire tree of memory referenced by
+	  the pointer. References in the tree are shown by giving the name
+	  of the pointer that is referenced.
+        </para>
+        <para>
+	  You can pass NULL for the pointer, in which case a report is
+	  printed for the top level memory context, but only if
+	  talloc_enable_leak_report() or talloc_enable_leak_report_full()
+	  has been called.
+        </para>
+    </refsect2>
+    <refsect2 id="talloc_report_depth_cb">
+     <funcsynopsis><funcprototype>
+      <funcdef>void <function>talloc_report_depth_cb</function></funcdef>
+      <paramdef><parameter>const void *ptr</parameter></paramdef>
+      <paramdef><parameter>int depth</parameter></paramdef>
+      <paramdef><parameter>int max_depth</parameter></paramdef>
+      <paramdef><parameter>void (*callback)(const void *ptr, int depth, int max_depth, int is_ref, void *priv)</parameter></paramdef>
+      <paramdef><parameter>void *priv</parameter></paramdef>
+     </funcprototype></funcsynopsis>
+        <para>
+	  This provides a more flexible reports than talloc_report(). It
+	  will recursively call the callback for the entire tree of memory
+	  referenced by the pointer. References in the tree are passed with
+	  <emphasis role="italic">is_ref = 1</emphasis> and the pointer that is referenced.
+        </para>
+        <para>
+	  You can pass NULL for the pointer, in which case a report is
+	  printed for the top level memory context, but only if
+	  talloc_enable_leak_report() or talloc_enable_leak_report_full()
+	  has been called.
+        </para>
+        <para>
+	  The recursion is stopped when depth >= max_depth.
+	  max_depth = -1 means only stop at leaf nodes.
+        </para>
+    </refsect2>
+    <refsect2 id="talloc_report_depth_file">
+     <funcsynopsis><funcprototype>
+      <funcdef>void <function>talloc_report_depth_file</function></funcdef>
+      <paramdef><parameter>const void *ptr</parameter></paramdef>
+      <paramdef><parameter>int depth</parameter></paramdef>
+      <paramdef><parameter>int max_depth</parameter></paramdef>
+      <paramdef><parameter>FILE *f</parameter></paramdef>
+     </funcprototype></funcsynopsis>
+        <para>
+	  This provides a more flexible reports than talloc_report(). It
+	  will let you specify the depth and max_depth.
+        </para>
+    </refsect2>
+    <refsect2 id="talloc_enable_leak_report"><title>void talloc_enable_leak_report(void);</title>
+        <para>
+	  This enables calling of talloc_report(NULL, stderr) when the
+	  program exits.  In Samba4 this is enabled by using the
+	  --leak-report command line option.
+        </para>
+        <para>
+	  For it to be useful, this function must be called before any
+	  other talloc function as it establishes a "null context" that
+	  acts as the top of the tree.	If you don't call this function
+	  first then passing NULL to talloc_report() or
+	  talloc_report_full() won't give you the full tree printout.
+        </para>
+        <para>
+	  Here is a typical talloc report:
+        </para>
+        <screen format="linespecific">talloc report on 'null_context' (total 267 bytes in 15 blocks)
+libcli/auth/spnego_parse.c:55  contains   31 bytes in   2 blocks
+libcli/auth/spnego_parse.c:55  contains   31 bytes in   2 blocks
+iconv(UTF8,CP850)              contains   42 bytes in   2 blocks
+libcli/auth/spnego_parse.c:55  contains   31 bytes in   2 blocks
+iconv(CP850,UTF8)              contains   42 bytes in   2 blocks
+iconv(UTF8,UTF-16LE)           contains   45 bytes in   2 blocks
+iconv(UTF-16LE,UTF8)           contains   45 bytes in   2 blocks
+      </screen>
+    </refsect2>
+    <refsect2 id="talloc_enable_leak_report_full"><title>void talloc_enable_leak_report_full(void);</title>
+        <para>
+	  This enables calling of talloc_report_full(NULL, stderr) when the
+	  program exits.  In Samba4 this is enabled by using the
+	  --leak-report-full command line option.
+        </para>
+        <para>
+	  For it to be useful, this function must be called before any
+	  other talloc function as it establishes a "null context" that
+	  acts as the top of the tree.	If you don't call this function
+	  first then passing NULL to talloc_report() or
+	  talloc_report_full() won't give you the full tree printout.
+        </para>
+        <para>
+	  Here is a typical full report:
+        </para>
+        <screen format="linespecific">full talloc report on 'root' (total 18 bytes in 8 blocks)
+p1               contains     18 bytes in   7 blocks (ref 0)
+    r1               contains     13 bytes in   2 blocks (ref 0)
+        reference to: p2
+    p2               contains      1 bytes in   1 blocks (ref 1)
+    x3               contains      1 bytes in   1 blocks (ref 0)
+    x2               contains      1 bytes in   1 blocks (ref 0)
+    x1               contains      1 bytes in   1 blocks (ref 0)
+      </screen>
+    </refsect2>
+    <refsect2><title>(<emphasis role="italic">type</emphasis> *)talloc_zero(const void *<emphasis role="italic">ctx</emphasis>, <emphasis role="italic">type</emphasis>);</title>
+        <para>
+	  The talloc_zero() macro is equivalent to:
+        </para>
+        <programlisting>ptr = talloc(ctx, type);
+if (ptr) memset(ptr, 0, sizeof(type));</programlisting>
+    </refsect2>
+    <refsect2><title>void *talloc_zero_size(const void *<emphasis role="italic">ctx</emphasis>, size_t <emphasis role="italic">size</emphasis>)</title>
+        <para>
+	  The talloc_zero_size() function is useful when you don't have a
+	  known type.
+        </para>
+    </refsect2>
+    <refsect2><title>void *talloc_memdup(const void *<emphasis role="italic">ctx</emphasis>, const void *<emphasis role="italic">p</emphasis>, size_t size);</title>
+        <para>
+	  The talloc_memdup() function is equivalent to:
+        </para>
+        <programlisting>ptr = talloc_size(ctx, size);
+if (ptr) memcpy(ptr, p, size);</programlisting>
+    </refsect2>
+    <refsect2><title>char *talloc_strdup(const void *<emphasis role="italic">ctx</emphasis>, const char *<emphasis role="italic">p</emphasis>);</title>
+        <para>
+	  The talloc_strdup() function is equivalent to:
+        </para>
+        <programlisting>ptr = talloc_size(ctx, strlen(p)+1);
+if (ptr) memcpy(ptr, p, strlen(p)+1);</programlisting>
+        <para>
+	  This function sets the name of the new pointer to the passed
+	  string. This is equivalent to:
+        </para>
+        <programlisting>talloc_set_name_const(ptr, ptr)</programlisting>
+    </refsect2>
+    <refsect2><title>char *talloc_strndup(const void *<emphasis role="italic">t</emphasis>, const char *<emphasis role="italic">p</emphasis>, size_t <emphasis role="italic">n</emphasis>);</title>
+        <para>
+	  The talloc_strndup() function is the talloc equivalent of the C
+	  library function strndup(3).
+        </para>
+        <para>
+	  This function sets the name of the new pointer to the passed
+	  string. This is equivalent to:
+        </para>
+        <programlisting>talloc_set_name_const(ptr, ptr)</programlisting>
+    </refsect2>
+    <refsect2><title>char *talloc_vasprintf(const void *<emphasis role="italic">t</emphasis>, const char *<emphasis role="italic">fmt</emphasis>, va_list <emphasis role="italic">ap</emphasis>);</title>
+        <para>
+	  The talloc_vasprintf() function is the talloc equivalent of the C
+	  library function vasprintf(3).
+        </para>
+        <para>
+	  This function sets the name of the new pointer to the new
+	  string. This is equivalent to:
+        </para>
+        <programlisting>talloc_set_name_const(ptr, ptr)</programlisting>
+    </refsect2>
+    <refsect2><title>char *talloc_asprintf(const void *<emphasis role="italic">t</emphasis>, const char *<emphasis role="italic">fmt</emphasis>, ...);</title>
+        <para>
+	  The talloc_asprintf() function is the talloc equivalent of the C
+	  library function asprintf(3).
+        </para>
+        <para>
+	  This function sets the name of the new pointer to the passed
+	  string. This is equivalent to:
+        </para>
+        <programlisting>talloc_set_name_const(ptr, ptr)</programlisting>
+    </refsect2>
+    <refsect2><title>char *talloc_asprintf_append(char *s, const char *fmt, ...);</title>
+        <para>
+	  The talloc_asprintf_append() function appends the given formatted
+	  string to the given string.
+        </para>
+        <para>
+	  This function sets the name of the new pointer to the new
+	  string. This is equivalent to:
+        </para>
+        <programlisting>talloc_set_name_const(ptr, ptr)</programlisting>
+    </refsect2>
+    <refsect2><title>(type *)talloc_array(const void *ctx, type, unsigned int count);</title>
+        <para>
+	  The talloc_array() macro is equivalent to:
+        </para>
+        <programlisting>(type *)talloc_size(ctx, sizeof(type) * count);</programlisting>
+        <para>
+	  except that it provides integer overflow protection for the
+	  multiply, returning NULL if the multiply overflows.
+        </para>
+    </refsect2>
+    <refsect2><title>void *talloc_array_size(const void *ctx, size_t size, unsigned int count);</title>
+        <para>
+	  The talloc_array_size() function is useful when the type is not
+	  known. It operates in the same way as talloc_array(), but takes a
+	  size instead of a type.
+        </para>
+    </refsect2>
+    <refsect2><title>(typeof(ptr)) talloc_array_ptrtype(const void *ctx, ptr, unsigned int count);</title>
+        <para>
+	  The talloc_ptrtype() macro should be used when you have a pointer to an array
+	  and want to allocate memory of an array to point at with this pointer. When compiling
+	  with gcc >= 3 it is typesafe. Note this is a wrapper of talloc_array_size()
+	  and talloc_get_name() will return the current location in the source file.
+	  and not the type.
+        </para>
+    </refsect2>
+    <refsect2><title>void *talloc_realloc_fn(const void *ctx, void *ptr, size_t size)</title>
+        <para>
+	  This is a non-macro version of talloc_realloc(), which is useful
+	  as libraries sometimes want a realloc function pointer.  A
+	  realloc(3) implementation encapsulates the functionality of
+	  malloc(3), free(3) and realloc(3) in one call, which is why it is
+	  useful to be able to pass around a single function pointer.
+        </para>
+    </refsect2>
+    <refsect2><title>void *talloc_autofree_context(void);</title>
+        <para>
+	  This is a handy utility function that returns a talloc context
+	  which will be automatically freed on program exit.  This can be
+	  used to reduce the noise in memory leak reports.
+        </para>
+    </refsect2>
+    <refsect2><title>void *talloc_check_name(const void *ptr, const char *name);</title>
+        <para>
+	  This function checks if a pointer has the specified <emphasis
+	  role="italic">name</emphasis>.  If it does then the pointer is
+	  returned.  It it doesn't then NULL is returned.
+        </para>
+    </refsect2>
+    <refsect2><title>(type *)talloc_get_type(const void *ptr, type);</title>
+        <para>
+	  This macro allows you to do type checking on talloc pointers.  It
+	  is particularly useful for void* private pointers.  It is
+	  equivalent to this:
+        </para>
+        <programlisting>(type *)talloc_check_name(ptr, #type)</programlisting>
+    </refsect2>
+    <refsect2><title>talloc_set_type(const void *ptr, type);</title>
+        <para>
+	  This macro allows you to force the name of a pointer to be a
+	  particular <emphasis>type</emphasis>.  This can be
+	  used in conjunction with talloc_get_type() to do type checking on
+	  void* pointers.
+        </para>
+        <para>
+	  It is equivalent to this:
+        </para>
+        <programlisting>talloc_set_name_const(ptr, #type)</programlisting>
+    </refsect2>
+    <refsect2><title>talloc_set_log_fn(void (*log_fn)(const char *message));</title>
+        <para>
+	  This function sets a logging function that talloc will use for
+	  warnings and errors. By default talloc will not print any warnings or
+	  errors.
+	</para>
+    </refsect2>
+    <refsect2><title>talloc_set_log_stderr(void);</title>
+        <para>
+	  This sets the talloc log function to write log messages to stderr
+	</para>
+    </refsect2>
+  </refsect1>
+  <refsect1><title>PERFORMANCE</title>
+    <para>
+      All the additional features of talloc(3) over malloc(3) do come at a
+      price.  We have a simple performance test in Samba4 that measures
+      talloc() versus malloc() performance, and it seems that talloc() is
+      about 10% slower than malloc() on my x86 Debian Linux box.  For
+      Samba, the great reduction in code complexity that we get by using
+      talloc makes this worthwhile, especially as the total overhead of
+      talloc/malloc in Samba is already quite small.
+    </para>
+  </refsect1>
+  <refsect1><title>SEE ALSO</title>
+    <para>
+      malloc(3), strndup(3), vasprintf(3), asprintf(3), 
+      <ulink url="http://talloc.samba.org/"/>
+    </para>
+  </refsect1>
+  <refsect1><title>COPYRIGHT/LICENSE</title>
+    <para>
+      Copyright (C) Andrew Tridgell 2004
+    </para>
+    <para>
+      This program is free software; you can redistribute it and/or modify
+      it under the terms of the GNU Lesser General Public License as 
+      published by the Free Software Foundation; either version 3 of the
+      License, or (at your option) any later version.
+    </para>
+    <para>
+      This program is distributed in the hope that it will be useful, but
+      WITHOUT ANY WARRANTY; without even the implied warranty of
+      MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+      General Public License for more details.
+    </para>
+    <para>
+      You should have received a copy of the GNU General Public License
+      along with this program; if not, see http://www.gnu.org/licenses/.
+    </para>
+  </refsect1>
+</refentry>
diff --git a/ctdb/lib/talloc/talloc.c b/ctdb/lib/talloc/talloc.c
new file mode 100644
index 0000000000..3e33fc0fad
--- /dev/null
+++ b/ctdb/lib/talloc/talloc.c
@@ -0,0 +1,2597 @@
+/* 
+   Samba Unix SMB/CIFS implementation.
+
+   Samba trivial allocation library - new interface
+
+   NOTE: Please read talloc_guide.txt for full documentation
+
+   Copyright (C) Andrew Tridgell 2004
+   Copyright (C) Stefan Metzmacher 2006
+   
+     ** NOTE! The following LGPL license applies to the talloc
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+/*
+  inspired by http://swapped.cc/halloc/
+*/
+
+#include "replace.h"
+#include "talloc.h"
+
+#ifdef TALLOC_BUILD_VERSION_MAJOR
+#if (TALLOC_VERSION_MAJOR != TALLOC_BUILD_VERSION_MAJOR)
+#error "TALLOC_VERSION_MAJOR != TALLOC_BUILD_VERSION_MAJOR"
+#endif
+#endif
+
+#ifdef TALLOC_BUILD_VERSION_MINOR
+#if (TALLOC_VERSION_MINOR != TALLOC_BUILD_VERSION_MINOR)
+#error "TALLOC_VERSION_MINOR != TALLOC_BUILD_VERSION_MINOR"
+#endif
+#endif
+
+/* Special macros that are no-ops except when run under Valgrind on
+ * x86.  They've moved a little bit from valgrind 1.0.4 to 1.9.4 */
+#ifdef HAVE_VALGRIND_MEMCHECK_H
+        /* memcheck.h includes valgrind.h */
+#include <valgrind/memcheck.h>
+#elif defined(HAVE_VALGRIND_H)
+#include <valgrind.h>
+#endif
+
+/* use this to force every realloc to change the pointer, to stress test
+   code that might not cope */
+#define ALWAYS_REALLOC 0
+
+
+#define MAX_TALLOC_SIZE 0x10000000
+#define TALLOC_MAGIC_BASE 0xe814ec70
+#define TALLOC_MAGIC ( \
+	TALLOC_MAGIC_BASE + \
+	(TALLOC_VERSION_MAJOR << 12) + \
+	(TALLOC_VERSION_MINOR << 4) \
+)
+
+#define TALLOC_FLAG_FREE 0x01
+#define TALLOC_FLAG_LOOP 0x02
+#define TALLOC_FLAG_POOL 0x04		/* This is a talloc pool */
+#define TALLOC_FLAG_POOLMEM 0x08	/* This is allocated in a pool */
+
+#define TALLOC_MAGIC_REFERENCE ((const char *)1)
+
+/* by default we abort when given a bad pointer (such as when talloc_free() is called 
+   on a pointer that came from malloc() */
+#ifndef TALLOC_ABORT
+#define TALLOC_ABORT(reason) abort()
+#endif
+
+#ifndef discard_const_p
+#if defined(__intptr_t_defined) || defined(HAVE_INTPTR_T)
+# define discard_const_p(type, ptr) ((type *)((intptr_t)(ptr)))
+#else
+# define discard_const_p(type, ptr) ((type *)(ptr))
+#endif
+#endif
+
+/* these macros gain us a few percent of speed on gcc */
+#if (__GNUC__ >= 3)
+/* the strange !! is to ensure that __builtin_expect() takes either 0 or 1
+   as its first argument */
+#ifndef likely
+#define likely(x)   __builtin_expect(!!(x), 1)
+#endif
+#ifndef unlikely
+#define unlikely(x) __builtin_expect(!!(x), 0)
+#endif
+#else
+#ifndef likely
+#define likely(x) (x)
+#endif
+#ifndef unlikely
+#define unlikely(x) (x)
+#endif
+#endif
+
+/* this null_context is only used if talloc_enable_leak_report() or
+   talloc_enable_leak_report_full() is called, otherwise it remains
+   NULL
+*/
+static void *null_context;
+static void *autofree_context;
+
+/* used to enable fill of memory on free, which can be useful for
+ * catching use after free errors when valgrind is too slow
+ */
+static struct {
+	bool initialised;
+	bool enabled;
+	uint8_t fill_value;
+} talloc_fill;
+
+#define TALLOC_FILL_ENV "TALLOC_FREE_FILL"
+
+/*
+ * do not wipe the header, to allow the
+ * double-free logic to still work
+ */
+#define TC_INVALIDATE_FULL_FILL_CHUNK(_tc) do { \
+	if (unlikely(talloc_fill.enabled)) { \
+		size_t _flen = (_tc)->size; \
+		char *_fptr = (char *)TC_PTR_FROM_CHUNK(_tc); \
+		memset(_fptr, talloc_fill.fill_value, _flen); \
+	} \
+} while (0)
+
+#if defined(DEVELOPER) && defined(VALGRIND_MAKE_MEM_NOACCESS)
+/* Mark the whole chunk as not accessable */
+#define TC_INVALIDATE_FULL_VALGRIND_CHUNK(_tc) do { \
+	size_t _flen = TC_HDR_SIZE + (_tc)->size; \
+	char *_fptr = (char *)(_tc); \
+	VALGRIND_MAKE_MEM_NOACCESS(_fptr, _flen); \
+} while(0)
+#else
+#define TC_INVALIDATE_FULL_VALGRIND_CHUNK(_tc) do { } while (0)
+#endif
+
+#define TC_INVALIDATE_FULL_CHUNK(_tc) do { \
+	TC_INVALIDATE_FULL_FILL_CHUNK(_tc); \
+	TC_INVALIDATE_FULL_VALGRIND_CHUNK(_tc); \
+} while (0)
+
+#define TC_INVALIDATE_SHRINK_FILL_CHUNK(_tc, _new_size) do { \
+	if (unlikely(talloc_fill.enabled)) { \
+		size_t _flen = (_tc)->size - (_new_size); \
+		char *_fptr = (char *)TC_PTR_FROM_CHUNK(_tc); \
+		_fptr += (_new_size); \
+		memset(_fptr, talloc_fill.fill_value, _flen); \
+	} \
+} while (0)
+
+#if defined(DEVELOPER) && defined(VALGRIND_MAKE_MEM_NOACCESS)
+/* Mark the unused bytes not accessable */
+#define TC_INVALIDATE_SHRINK_VALGRIND_CHUNK(_tc, _new_size) do { \
+	size_t _flen = (_tc)->size - (_new_size); \
+	char *_fptr = (char *)TC_PTR_FROM_CHUNK(_tc); \
+	_fptr += (_new_size); \
+	VALGRIND_MAKE_MEM_NOACCESS(_fptr, _flen); \
+} while (0)
+#else
+#define TC_INVALIDATE_SHRINK_VALGRIND_CHUNK(_tc, _new_size) do { } while (0)
+#endif
+
+#define TC_INVALIDATE_SHRINK_CHUNK(_tc, _new_size) do { \
+	TC_INVALIDATE_SHRINK_FILL_CHUNK(_tc, _new_size); \
+	TC_INVALIDATE_SHRINK_VALGRIND_CHUNK(_tc, _new_size); \
+} while (0)
+
+#define TC_UNDEFINE_SHRINK_FILL_CHUNK(_tc, _new_size) do { \
+	if (unlikely(talloc_fill.enabled)) { \
+		size_t _flen = (_tc)->size - (_new_size); \
+		char *_fptr = (char *)TC_PTR_FROM_CHUNK(_tc); \
+		_fptr += (_new_size); \
+		memset(_fptr, talloc_fill.fill_value, _flen); \
+	} \
+} while (0)
+
+#if defined(DEVELOPER) && defined(VALGRIND_MAKE_MEM_UNDEFINED)
+/* Mark the unused bytes as undefined */
+#define TC_UNDEFINE_SHRINK_VALGRIND_CHUNK(_tc, _new_size) do { \
+	size_t _flen = (_tc)->size - (_new_size); \
+	char *_fptr = (char *)TC_PTR_FROM_CHUNK(_tc); \
+	_fptr += (_new_size); \
+	VALGRIND_MAKE_MEM_UNDEFINED(_fptr, _flen); \
+} while (0)
+#else
+#define TC_UNDEFINE_SHRINK_VALGRIND_CHUNK(_tc, _new_size) do { } while (0)
+#endif
+
+#define TC_UNDEFINE_SHRINK_CHUNK(_tc, _new_size) do { \
+	TC_UNDEFINE_SHRINK_FILL_CHUNK(_tc, _new_size); \
+	TC_UNDEFINE_SHRINK_VALGRIND_CHUNK(_tc, _new_size); \
+} while (0)
+
+#if defined(DEVELOPER) && defined(VALGRIND_MAKE_MEM_UNDEFINED)
+/* Mark the new bytes as undefined */
+#define TC_UNDEFINE_GROW_VALGRIND_CHUNK(_tc, _new_size) do { \
+	size_t _old_used = TC_HDR_SIZE + (_tc)->size; \
+	size_t _new_used = TC_HDR_SIZE + (_new_size); \
+	size_t _flen = _new_used - _old_used; \
+	char *_fptr = _old_used + (char *)(_tc); \
+	VALGRIND_MAKE_MEM_UNDEFINED(_fptr, _flen); \
+} while (0)
+#else
+#define TC_UNDEFINE_GROW_VALGRIND_CHUNK(_tc, _new_size) do { } while (0)
+#endif
+
+#define TC_UNDEFINE_GROW_CHUNK(_tc, _new_size) do { \
+	TC_UNDEFINE_GROW_VALGRIND_CHUNK(_tc, _new_size); \
+} while (0)
+
+struct talloc_reference_handle {
+	struct talloc_reference_handle *next, *prev;
+	void *ptr;
+	const char *location;
+};
+
+struct talloc_memlimit {
+	struct talloc_chunk *parent;
+	struct talloc_memlimit *upper;
+	size_t max_size;
+	size_t cur_size;
+};
+
+static bool talloc_memlimit_check(struct talloc_memlimit *limit, size_t size);
+static bool talloc_memlimit_update(struct talloc_memlimit *limit,
+				   size_t old_size, size_t new_size);
+
+typedef int (*talloc_destructor_t)(void *);
+
+struct talloc_chunk {
+	struct talloc_chunk *next, *prev;
+	struct talloc_chunk *parent, *child;
+	struct talloc_reference_handle *refs;
+	talloc_destructor_t destructor;
+	const char *name;
+	size_t size;
+	unsigned flags;
+
+	/*
+	 * limit semantics:
+	 * if 'limit' is set it means all *new* children of the context will
+	 * be limited to a total aggregate size ox max_size for memory
+	 * allocations.
+	 * cur_size is used to kep track of the current use
+	 */
+	struct talloc_memlimit *limit;
+
+	/*
+	 * "pool" has dual use:
+	 *
+	 * For the talloc pool itself (i.e. TALLOC_FLAG_POOL is set), "pool"
+	 * marks the end of the currently allocated area.
+	 *
+	 * For members of the pool (i.e. TALLOC_FLAG_POOLMEM is set), "pool"
+	 * is a pointer to the struct talloc_chunk of the pool that it was
+	 * allocated from. This way children can quickly find the pool to chew
+	 * from.
+	 */
+	void *pool;
+};
+
+/* 16 byte alignment seems to keep everyone happy */
+#define TC_ALIGN16(s) (((s)+15)&~15)
+#define TC_HDR_SIZE TC_ALIGN16(sizeof(struct talloc_chunk))
+#define TC_PTR_FROM_CHUNK(tc) ((void *)(TC_HDR_SIZE + (char*)tc))
+
+_PUBLIC_ int talloc_version_major(void)
+{
+	return TALLOC_VERSION_MAJOR;
+}
+
+_PUBLIC_ int talloc_version_minor(void)
+{
+	return TALLOC_VERSION_MINOR;
+}
+
+static void (*talloc_log_fn)(const char *message);
+
+_PUBLIC_ void talloc_set_log_fn(void (*log_fn)(const char *message))
+{
+	talloc_log_fn = log_fn;
+}
+
+static void talloc_log(const char *fmt, ...) PRINTF_ATTRIBUTE(1,2);
+static void talloc_log(const char *fmt, ...)
+{
+	va_list ap;
+	char *message;
+
+	if (!talloc_log_fn) {
+		return;
+	}
+
+	va_start(ap, fmt);
+	message = talloc_vasprintf(NULL, fmt, ap);
+	va_end(ap);
+
+	talloc_log_fn(message);
+	talloc_free(message);
+}
+
+static void talloc_log_stderr(const char *message)
+{
+	fprintf(stderr, "%s", message);
+}
+
+_PUBLIC_ void talloc_set_log_stderr(void)
+{
+	talloc_set_log_fn(talloc_log_stderr);
+}
+
+static void (*talloc_abort_fn)(const char *reason);
+
+_PUBLIC_ void talloc_set_abort_fn(void (*abort_fn)(const char *reason))
+{
+	talloc_abort_fn = abort_fn;
+}
+
+static void talloc_abort(const char *reason)
+{
+	talloc_log("%s\n", reason);
+
+	if (!talloc_abort_fn) {
+		TALLOC_ABORT(reason);
+	}
+
+	talloc_abort_fn(reason);
+}
+
+static void talloc_abort_magic(unsigned magic)
+{
+	unsigned striped = magic - TALLOC_MAGIC_BASE;
+	unsigned major = (striped & 0xFFFFF000) >> 12;
+	unsigned minor = (striped & 0x00000FF0) >> 4;
+	talloc_log("Bad talloc magic[0x%08X/%u/%u] expected[0x%08X/%u/%u]\n",
+		   magic, major, minor,
+		   TALLOC_MAGIC, TALLOC_VERSION_MAJOR, TALLOC_VERSION_MINOR);
+	talloc_abort("Bad talloc magic value - wrong talloc version used/mixed");
+}
+
+static void talloc_abort_access_after_free(void)
+{
+	talloc_abort("Bad talloc magic value - access after free");
+}
+
+static void talloc_abort_unknown_value(void)
+{
+	talloc_abort("Bad talloc magic value - unknown value");
+}
+
+/* panic if we get a bad magic value */
+static inline struct talloc_chunk *talloc_chunk_from_ptr(const void *ptr)
+{
+	const char *pp = (const char *)ptr;
+	struct talloc_chunk *tc = discard_const_p(struct talloc_chunk, pp - TC_HDR_SIZE);
+	if (unlikely((tc->flags & (TALLOC_FLAG_FREE | ~0xF)) != TALLOC_MAGIC)) { 
+		if ((tc->flags & (~0xFFF)) == TALLOC_MAGIC_BASE) {
+			talloc_abort_magic(tc->flags & (~0xF));
+			return NULL;
+		}
+
+		if (tc->flags & TALLOC_FLAG_FREE) {
+			talloc_log("talloc: access after free error - first free may be at %s\n", tc->name);
+			talloc_abort_access_after_free();
+			return NULL;
+		} else {
+			talloc_abort_unknown_value();
+			return NULL;
+		}
+	}
+	return tc;
+}
+
+/* hook into the front of the list */
+#define _TLIST_ADD(list, p) \
+do { \
+        if (!(list)) { \
+		(list) = (p); \
+		(p)->next = (p)->prev = NULL; \
+	} else { \
+		(list)->prev = (p); \
+		(p)->next = (list); \
+		(p)->prev = NULL; \
+		(list) = (p); \
+	}\
+} while (0)
+
+/* remove an element from a list - element doesn't have to be in list. */
+#define _TLIST_REMOVE(list, p) \
+do { \
+	if ((p) == (list)) { \
+		(list) = (p)->next; \
+		if (list) (list)->prev = NULL; \
+	} else { \
+		if ((p)->prev) (p)->prev->next = (p)->next; \
+		if ((p)->next) (p)->next->prev = (p)->prev; \
+	} \
+	if ((p) && ((p) != (list))) (p)->next = (p)->prev = NULL; \
+} while (0)
+
+
+/*
+  return the parent chunk of a pointer
+*/
+static inline struct talloc_chunk *talloc_parent_chunk(const void *ptr)
+{
+	struct talloc_chunk *tc;
+
+	if (unlikely(ptr == NULL)) {
+		return NULL;
+	}
+
+	tc = talloc_chunk_from_ptr(ptr);
+	while (tc->prev) tc=tc->prev;
+
+	return tc->parent;
+}
+
+_PUBLIC_ void *talloc_parent(const void *ptr)
+{
+	struct talloc_chunk *tc = talloc_parent_chunk(ptr);
+	return tc? TC_PTR_FROM_CHUNK(tc) : NULL;
+}
+
+/*
+  find parents name
+*/
+_PUBLIC_ const char *talloc_parent_name(const void *ptr)
+{
+	struct talloc_chunk *tc = talloc_parent_chunk(ptr);
+	return tc? tc->name : NULL;
+}
+
+/*
+  A pool carries an in-pool object count count in the first 16 bytes.
+  bytes. This is done to support talloc_steal() to a parent outside of the
+  pool. The count includes the pool itself, so a talloc_free() on a pool will
+  only destroy the pool if the count has dropped to zero. A talloc_free() of a
+  pool member will reduce the count, and eventually also call free(3) on the
+  pool memory.
+
+  The object count is not put into "struct talloc_chunk" because it is only
+  relevant for talloc pools and the alignment to 16 bytes would increase the
+  memory footprint of each talloc chunk by those 16 bytes.
+*/
+
+union talloc_pool_chunk {
+	/* This lets object_count nestle into 16-byte padding of talloc_chunk,
+	 * on 32-bit platforms. */
+	struct tc_pool_hdr {
+		struct talloc_chunk c;
+		unsigned int object_count;
+	} hdr;
+	/* This makes it always 16 byte aligned. */
+	char pad[TC_ALIGN16(sizeof(struct tc_pool_hdr))];
+};
+
+static void *tc_pool_end(union talloc_pool_chunk *pool_tc)
+{
+	return (char *)pool_tc + TC_HDR_SIZE + pool_tc->hdr.c.size;
+}
+
+static size_t tc_pool_space_left(union talloc_pool_chunk *pool_tc)
+{
+	return (char *)tc_pool_end(pool_tc) - (char *)pool_tc->hdr.c.pool;
+}
+
+static void *tc_pool_first_chunk(union talloc_pool_chunk *pool_tc)
+{
+	return pool_tc + 1;
+}
+
+/* If tc is inside a pool, this gives the next neighbour. */
+static void *tc_next_chunk(struct talloc_chunk *tc)
+{
+	return (char *)tc + TC_ALIGN16(TC_HDR_SIZE + tc->size);
+}
+
+/* Mark the whole remaining pool as not accessable */
+static void tc_invalidate_pool(union talloc_pool_chunk *pool_tc)
+{
+	size_t flen = tc_pool_space_left(pool_tc);
+
+	if (unlikely(talloc_fill.enabled)) {
+		memset(pool_tc->hdr.c.pool, talloc_fill.fill_value, flen);
+	}
+
+#if defined(DEVELOPER) && defined(VALGRIND_MAKE_MEM_NOACCESS)
+	VALGRIND_MAKE_MEM_NOACCESS(pool_tc->hdr.c.pool, flen);
+#endif
+}
+
+/*
+  Allocate from a pool
+*/
+
+static struct talloc_chunk *talloc_alloc_pool(struct talloc_chunk *parent,
+					      size_t size)
+{
+	union talloc_pool_chunk *pool_ctx = NULL;
+	size_t space_left;
+	struct talloc_chunk *result;
+	size_t chunk_size;
+
+	if (parent == NULL) {
+		return NULL;
+	}
+
+	if (parent->flags & TALLOC_FLAG_POOL) {
+		pool_ctx = (union talloc_pool_chunk *)parent;
+	}
+	else if (parent->flags & TALLOC_FLAG_POOLMEM) {
+		pool_ctx = (union talloc_pool_chunk *)parent->pool;
+	}
+
+	if (pool_ctx == NULL) {
+		return NULL;
+	}
+
+	space_left = tc_pool_space_left(pool_ctx);
+
+	/*
+	 * Align size to 16 bytes
+	 */
+	chunk_size = TC_ALIGN16(size);
+
+	if (space_left < chunk_size) {
+		return NULL;
+	}
+
+	result = (struct talloc_chunk *)pool_ctx->hdr.c.pool;
+
+#if defined(DEVELOPER) && defined(VALGRIND_MAKE_MEM_UNDEFINED)
+	VALGRIND_MAKE_MEM_UNDEFINED(result, size);
+#endif
+
+	pool_ctx->hdr.c.pool = (void *)((char *)result + chunk_size);
+
+	result->flags = TALLOC_MAGIC | TALLOC_FLAG_POOLMEM;
+	result->pool = pool_ctx;
+
+	pool_ctx->hdr.object_count++;
+
+	return result;
+}
+
+/* 
+   Allocate a bit of memory as a child of an existing pointer
+*/
+static inline void *__talloc(const void *context, size_t size)
+{
+	struct talloc_chunk *tc = NULL;
+	struct talloc_memlimit *limit = NULL;
+
+	if (unlikely(context == NULL)) {
+		context = null_context;
+	}
+
+	if (unlikely(size >= MAX_TALLOC_SIZE)) {
+		return NULL;
+	}
+
+	if (context != NULL) {
+		struct talloc_chunk *ptc = talloc_chunk_from_ptr(context);
+
+		if (ptc->limit != NULL) {
+			limit = ptc->limit;
+		}
+
+		if (!talloc_memlimit_check(limit, (TC_HDR_SIZE+size))) {
+			errno = ENOMEM;
+			return NULL;
+		}
+
+		tc = talloc_alloc_pool(ptc, TC_HDR_SIZE+size);
+	}
+
+	if (tc == NULL) {
+		tc = (struct talloc_chunk *)malloc(TC_HDR_SIZE+size);
+		if (unlikely(tc == NULL)) return NULL;
+		tc->flags = TALLOC_MAGIC;
+		tc->pool  = NULL;
+	}
+
+	if (limit != NULL) {
+		struct talloc_memlimit *l;
+
+		for (l = limit; l != NULL; l = l->upper) {
+			l->cur_size += TC_HDR_SIZE+size;
+		}
+	}
+
+	tc->limit = limit;
+	tc->size = size;
+	tc->destructor = NULL;
+	tc->child = NULL;
+	tc->name = NULL;
+	tc->refs = NULL;
+
+	if (likely(context)) {
+		struct talloc_chunk *parent = talloc_chunk_from_ptr(context);
+
+		if (parent->child) {
+			parent->child->parent = NULL;
+			tc->next = parent->child;
+			tc->next->prev = tc;
+		} else {
+			tc->next = NULL;
+		}
+		tc->parent = parent;
+		tc->prev = NULL;
+		parent->child = tc;
+	} else {
+		tc->next = tc->prev = tc->parent = NULL;
+	}
+
+	return TC_PTR_FROM_CHUNK(tc);
+}
+
+/*
+ * Create a talloc pool
+ */
+
+_PUBLIC_ void *talloc_pool(const void *context, size_t size)
+{
+	union talloc_pool_chunk *pool_tc;
+	void *result = __talloc(context, sizeof(*pool_tc) - TC_HDR_SIZE + size);
+
+	if (unlikely(result == NULL)) {
+		return NULL;
+	}
+
+	pool_tc = (union talloc_pool_chunk *)talloc_chunk_from_ptr(result);
+	if (unlikely(pool_tc->hdr.c.flags & TALLOC_FLAG_POOLMEM)) {
+		/* We don't handle this correctly, so fail. */
+		talloc_log("talloc: cannot allocate pool off another pool %s\n",
+			   talloc_get_name(context));
+		talloc_free(result);
+		return NULL;
+	}
+	pool_tc->hdr.c.flags |= TALLOC_FLAG_POOL;
+	pool_tc->hdr.c.pool = tc_pool_first_chunk(pool_tc);
+
+	pool_tc->hdr.object_count = 1;
+
+	tc_invalidate_pool(pool_tc);
+
+	return result;
+}
+
+/*
+  setup a destructor to be called on free of a pointer
+  the destructor should return 0 on success, or -1 on failure.
+  if the destructor fails then the free is failed, and the memory can
+  be continued to be used
+*/
+_PUBLIC_ void _talloc_set_destructor(const void *ptr, int (*destructor)(void *))
+{
+	struct talloc_chunk *tc = talloc_chunk_from_ptr(ptr);
+	tc->destructor = destructor;
+}
+
+/*
+  increase the reference count on a piece of memory. 
+*/
+_PUBLIC_ int talloc_increase_ref_count(const void *ptr)
+{
+	if (unlikely(!talloc_reference(null_context, ptr))) {
+		return -1;
+	}
+	return 0;
+}
+
+/*
+  helper for talloc_reference()
+
+  this is referenced by a function pointer and should not be inline
+*/
+static int talloc_reference_destructor(struct talloc_reference_handle *handle)
+{
+	struct talloc_chunk *ptr_tc = talloc_chunk_from_ptr(handle->ptr);
+	_TLIST_REMOVE(ptr_tc->refs, handle);
+	return 0;
+}
+
+/*
+   more efficient way to add a name to a pointer - the name must point to a 
+   true string constant
+*/
+static inline void _talloc_set_name_const(const void *ptr, const char *name)
+{
+	struct talloc_chunk *tc = talloc_chunk_from_ptr(ptr);
+	tc->name = name;
+}
+
+/*
+  internal talloc_named_const()
+*/
+static inline void *_talloc_named_const(const void *context, size_t size, const char *name)
+{
+	void *ptr;
+
+	ptr = __talloc(context, size);
+	if (unlikely(ptr == NULL)) {
+		return NULL;
+	}
+
+	_talloc_set_name_const(ptr, name);
+
+	return ptr;
+}
+
+/*
+  make a secondary reference to a pointer, hanging off the given context.
+  the pointer remains valid until both the original caller and this given
+  context are freed.
+  
+  the major use for this is when two different structures need to reference the 
+  same underlying data, and you want to be able to free the two instances separately,
+  and in either order
+*/
+_PUBLIC_ void *_talloc_reference_loc(const void *context, const void *ptr, const char *location)
+{
+	struct talloc_chunk *tc;
+	struct talloc_reference_handle *handle;
+	if (unlikely(ptr == NULL)) return NULL;
+
+	tc = talloc_chunk_from_ptr(ptr);
+	handle = (struct talloc_reference_handle *)_talloc_named_const(context,
+						   sizeof(struct talloc_reference_handle),
+						   TALLOC_MAGIC_REFERENCE);
+	if (unlikely(handle == NULL)) return NULL;
+
+	/* note that we hang the destructor off the handle, not the
+	   main context as that allows the caller to still setup their
+	   own destructor on the context if they want to */
+	talloc_set_destructor(handle, talloc_reference_destructor);
+	handle->ptr = discard_const_p(void, ptr);
+	handle->location = location;
+	_TLIST_ADD(tc->refs, handle);
+	return handle->ptr;
+}
+
+static void *_talloc_steal_internal(const void *new_ctx, const void *ptr);
+
+static inline void _talloc_free_poolmem(struct talloc_chunk *tc,
+					const char *location)
+{
+	union talloc_pool_chunk *pool;
+	void *next_tc;
+
+	pool = (union talloc_pool_chunk *)tc->pool;
+	next_tc = tc_next_chunk(tc);
+
+	tc->flags |= TALLOC_FLAG_FREE;
+
+	/* we mark the freed memory with where we called the free
+	 * from. This means on a double free error we can report where
+	 * the first free came from
+	 */
+	tc->name = location;
+
+	TC_INVALIDATE_FULL_CHUNK(tc);
+
+	if (unlikely(pool->hdr.object_count == 0)) {
+		talloc_abort("Pool object count zero!");
+		return;
+	}
+
+	pool->hdr.object_count--;
+
+	if (unlikely(pool->hdr.object_count == 1
+		     && !(pool->hdr.c.flags & TALLOC_FLAG_FREE))) {
+		/*
+		 * if there is just one object left in the pool
+		 * and pool->flags does not have TALLOC_FLAG_FREE,
+		 * it means this is the pool itself and
+		 * the rest is available for new objects
+		 * again.
+		 */
+		pool->hdr.c.pool = tc_pool_first_chunk(pool);
+		tc_invalidate_pool(pool);
+	} else if (unlikely(pool->hdr.object_count == 0)) {
+		/*
+		 * we mark the freed memory with where we called the free
+		 * from. This means on a double free error we can report where
+		 * the first free came from
+		 */
+		pool->hdr.c.name = location;
+
+		TC_INVALIDATE_FULL_CHUNK(&pool->hdr.c);
+		free(pool);
+	} else if (pool->hdr.c.pool == next_tc) {
+		/*
+		 * if pool->pool still points to end of
+		 * 'tc' (which is stored in the 'next_tc' variable),
+		 * we can reclaim the memory of 'tc'.
+		 */
+		pool->hdr.c.pool = tc;
+	}
+}
+
+static inline void _talloc_free_children_internal(struct talloc_chunk *tc,
+						  void *ptr,
+						  const char *location);
+
+/* 
+   internal talloc_free call
+*/
+static inline int _talloc_free_internal(void *ptr, const char *location)
+{
+	struct talloc_chunk *tc;
+
+	if (unlikely(ptr == NULL)) {
+		return -1;
+	}
+
+	/* possibly initialised the talloc fill value */
+	if (unlikely(!talloc_fill.initialised)) {
+		const char *fill = getenv(TALLOC_FILL_ENV);
+		if (fill != NULL) {
+			talloc_fill.enabled = true;
+			talloc_fill.fill_value = strtoul(fill, NULL, 0);
+		}
+		talloc_fill.initialised = true;
+	}
+
+	tc = talloc_chunk_from_ptr(ptr);
+
+	if (unlikely(tc->refs)) {
+		int is_child;
+		/* check if this is a reference from a child or
+		 * grandchild back to it's parent or grandparent
+		 *
+		 * in that case we need to remove the reference and
+		 * call another instance of talloc_free() on the current
+		 * pointer.
+		 */
+		is_child = talloc_is_parent(tc->refs, ptr);
+		_talloc_free_internal(tc->refs, location);
+		if (is_child) {
+			return _talloc_free_internal(ptr, location);
+		}
+		return -1;
+	}
+
+	if (unlikely(tc->flags & TALLOC_FLAG_LOOP)) {
+		/* we have a free loop - stop looping */
+		return 0;
+	}
+
+	if (unlikely(tc->destructor)) {
+		talloc_destructor_t d = tc->destructor;
+		if (d == (talloc_destructor_t)-1) {
+			return -1;
+		}
+		tc->destructor = (talloc_destructor_t)-1;
+		if (d(ptr) == -1) {
+			tc->destructor = d;
+			return -1;
+		}
+		tc->destructor = NULL;
+	}
+
+	if (tc->parent) {
+		_TLIST_REMOVE(tc->parent->child, tc);
+		if (tc->parent->child) {
+			tc->parent->child->parent = tc->parent;
+		}
+	} else {
+		if (tc->prev) tc->prev->next = tc->next;
+		if (tc->next) tc->next->prev = tc->prev;
+		tc->prev = tc->next = NULL;
+	}
+
+	tc->flags |= TALLOC_FLAG_LOOP;
+
+	_talloc_free_children_internal(tc, ptr, location);
+
+	tc->flags |= TALLOC_FLAG_FREE;
+
+	/*
+	 * If we are part of a memory limited context hierarchy
+	 * we need to subtract the memory used from the counters
+	 */
+	if (tc->limit) {
+		struct talloc_memlimit *l;
+
+		for (l = tc->limit; l != NULL; l = l->upper) {
+			if (l->cur_size >= tc->size+TC_HDR_SIZE) {
+				l->cur_size -= tc->size+TC_HDR_SIZE;
+			} else {
+				talloc_abort("cur_size memlimit counter not correct!");
+				return 0;
+			}
+		}
+
+		if (tc->limit->parent == tc) {
+			free(tc->limit);
+		}
+
+		tc->limit = NULL;
+	}
+
+	/* we mark the freed memory with where we called the free
+	 * from. This means on a double free error we can report where
+	 * the first free came from 
+	 */	 
+	tc->name = location;
+
+	if (tc->flags & TALLOC_FLAG_POOL) {
+		union talloc_pool_chunk *pool = (union talloc_pool_chunk *)tc;
+
+		if (unlikely(pool->hdr.object_count == 0)) {
+			talloc_abort("Pool object count zero!");
+			return 0;
+		}
+
+		pool->hdr.object_count--;
+		if (unlikely(pool->hdr.object_count == 0)) {
+			TC_INVALIDATE_FULL_CHUNK(tc);
+			free(tc);
+		}
+	} else if (tc->flags & TALLOC_FLAG_POOLMEM) {
+		_talloc_free_poolmem(tc, location);
+	} else {
+		TC_INVALIDATE_FULL_CHUNK(tc);
+		free(tc);
+	}
+	return 0;
+}
+
+static size_t _talloc_total_limit_size(const void *ptr,
+					struct talloc_memlimit *old_limit,
+					struct talloc_memlimit *new_limit);
+
+/* 
+   move a lump of memory from one talloc context to another return the
+   ptr on success, or NULL if it could not be transferred.
+   passing NULL as ptr will always return NULL with no side effects.
+*/
+static void *_talloc_steal_internal(const void *new_ctx, const void *ptr)
+{
+	struct talloc_chunk *tc, *new_tc;
+	size_t ctx_size = 0;
+
+	if (unlikely(!ptr)) {
+		return NULL;
+	}
+
+	if (unlikely(new_ctx == NULL)) {
+		new_ctx = null_context;
+	}
+
+	tc = talloc_chunk_from_ptr(ptr);
+
+	if (tc->limit != NULL) {
+
+		ctx_size = _talloc_total_limit_size(ptr, NULL, NULL);
+
+		if (!talloc_memlimit_update(tc->limit->upper, ctx_size, 0)) {
+			talloc_abort("cur_size memlimit counter not correct!");
+			errno = EINVAL;
+			return NULL;
+		}
+
+		if (tc->limit->parent == tc) {
+			tc->limit->upper = NULL;
+		} else {
+			tc->limit = NULL;
+		}
+	}
+
+	if (unlikely(new_ctx == NULL)) {
+		if (tc->parent) {
+			_TLIST_REMOVE(tc->parent->child, tc);
+			if (tc->parent->child) {
+				tc->parent->child->parent = tc->parent;
+			}
+		} else {
+			if (tc->prev) tc->prev->next = tc->next;
+			if (tc->next) tc->next->prev = tc->prev;
+		}
+
+		tc->parent = tc->next = tc->prev = NULL;
+		return discard_const_p(void, ptr);
+	}
+
+	new_tc = talloc_chunk_from_ptr(new_ctx);
+
+	if (unlikely(tc == new_tc || tc->parent == new_tc)) {
+		return discard_const_p(void, ptr);
+	}
+
+	if (tc->parent) {
+		_TLIST_REMOVE(tc->parent->child, tc);
+		if (tc->parent->child) {
+			tc->parent->child->parent = tc->parent;
+		}
+	} else {
+		if (tc->prev) tc->prev->next = tc->next;
+		if (tc->next) tc->next->prev = tc->prev;
+		tc->prev = tc->next = NULL;
+	}
+
+	tc->parent = new_tc;
+	if (new_tc->child) new_tc->child->parent = NULL;
+	_TLIST_ADD(new_tc->child, tc);
+
+	if (tc->limit || new_tc->limit) {
+		ctx_size = _talloc_total_limit_size(ptr, tc->limit,
+						    new_tc->limit);
+	}
+
+	if (new_tc->limit) {
+		struct talloc_memlimit *l;
+
+		for (l = new_tc->limit; l != NULL; l = l->upper) {
+			l->cur_size += ctx_size;
+		}
+	}
+
+	return discard_const_p(void, ptr);
+}
+
+/* 
+   move a lump of memory from one talloc context to another return the
+   ptr on success, or NULL if it could not be transferred.
+   passing NULL as ptr will always return NULL with no side effects.
+*/
+_PUBLIC_ void *_talloc_steal_loc(const void *new_ctx, const void *ptr, const char *location)
+{
+	struct talloc_chunk *tc;
+
+	if (unlikely(ptr == NULL)) {
+		return NULL;
+	}
+	
+	tc = talloc_chunk_from_ptr(ptr);
+	
+	if (unlikely(tc->refs != NULL) && talloc_parent(ptr) != new_ctx) {
+		struct talloc_reference_handle *h;
+
+		talloc_log("WARNING: talloc_steal with references at %s\n",
+			   location);
+
+		for (h=tc->refs; h; h=h->next) {
+			talloc_log("\treference at %s\n",
+				   h->location);
+		}
+	}
+
+#if 0
+	/* this test is probably too expensive to have on in the
+	   normal build, but it useful for debugging */
+	if (talloc_is_parent(new_ctx, ptr)) {
+		talloc_log("WARNING: stealing into talloc child at %s\n", location);
+	}
+#endif
+	
+	return _talloc_steal_internal(new_ctx, ptr);
+}
+
+/* 
+   this is like a talloc_steal(), but you must supply the old
+   parent. This resolves the ambiguity in a talloc_steal() which is
+   called on a context that has more than one parent (via references)
+
+   The old parent can be either a reference or a parent
+*/
+_PUBLIC_ void *talloc_reparent(const void *old_parent, const void *new_parent, const void *ptr)
+{
+	struct talloc_chunk *tc;
+	struct talloc_reference_handle *h;
+
+	if (unlikely(ptr == NULL)) {
+		return NULL;
+	}
+
+	if (old_parent == talloc_parent(ptr)) {
+		return _talloc_steal_internal(new_parent, ptr);
+	}
+
+	tc = talloc_chunk_from_ptr(ptr);
+	for (h=tc->refs;h;h=h->next) {
+		if (talloc_parent(h) == old_parent) {
+			if (_talloc_steal_internal(new_parent, h) != h) {
+				return NULL;
+			}
+			return discard_const_p(void, ptr);
+		}
+	}	
+
+	/* it wasn't a parent */
+	return NULL;
+}
+
+/*
+  remove a secondary reference to a pointer. This undo's what
+  talloc_reference() has done. The context and pointer arguments
+  must match those given to a talloc_reference()
+*/
+static inline int talloc_unreference(const void *context, const void *ptr)
+{
+	struct talloc_chunk *tc = talloc_chunk_from_ptr(ptr);
+	struct talloc_reference_handle *h;
+
+	if (unlikely(context == NULL)) {
+		context = null_context;
+	}
+
+	for (h=tc->refs;h;h=h->next) {
+		struct talloc_chunk *p = talloc_parent_chunk(h);
+		if (p == NULL) {
+			if (context == NULL) break;
+		} else if (TC_PTR_FROM_CHUNK(p) == context) {
+			break;
+		}
+	}
+	if (h == NULL) {
+		return -1;
+	}
+
+	return _talloc_free_internal(h, __location__);
+}
+
+/*
+  remove a specific parent context from a pointer. This is a more
+  controlled variant of talloc_free()
+*/
+_PUBLIC_ int talloc_unlink(const void *context, void *ptr)
+{
+	struct talloc_chunk *tc_p, *new_p, *tc_c;
+	void *new_parent;
+
+	if (ptr == NULL) {
+		return -1;
+	}
+
+	if (context == NULL) {
+		context = null_context;
+	}
+
+	if (talloc_unreference(context, ptr) == 0) {
+		return 0;
+	}
+
+	if (context != NULL) {
+		tc_c = talloc_chunk_from_ptr(context);
+	} else {
+		tc_c = NULL;
+	}
+	if (tc_c != talloc_parent_chunk(ptr)) {
+		return -1;
+	}
+
+	tc_p = talloc_chunk_from_ptr(ptr);
+
+	if (tc_p->refs == NULL) {
+		return _talloc_free_internal(ptr, __location__);
+	}
+
+	new_p = talloc_parent_chunk(tc_p->refs);
+	if (new_p) {
+		new_parent = TC_PTR_FROM_CHUNK(new_p);
+	} else {
+		new_parent = NULL;
+	}
+
+	if (talloc_unreference(new_parent, ptr) != 0) {
+		return -1;
+	}
+
+	_talloc_steal_internal(new_parent, ptr);
+
+	return 0;
+}
+
+/*
+  add a name to an existing pointer - va_list version
+*/
+static inline const char *talloc_set_name_v(const void *ptr, const char *fmt, va_list ap) PRINTF_ATTRIBUTE(2,0);
+
+static inline const char *talloc_set_name_v(const void *ptr, const char *fmt, va_list ap)
+{
+	struct talloc_chunk *tc = talloc_chunk_from_ptr(ptr);
+	tc->name = talloc_vasprintf(ptr, fmt, ap);
+	if (likely(tc->name)) {
+		_talloc_set_name_const(tc->name, ".name");
+	}
+	return tc->name;
+}
+
+/*
+  add a name to an existing pointer
+*/
+_PUBLIC_ const char *talloc_set_name(const void *ptr, const char *fmt, ...)
+{
+	const char *name;
+	va_list ap;
+	va_start(ap, fmt);
+	name = talloc_set_name_v(ptr, fmt, ap);
+	va_end(ap);
+	return name;
+}
+
+
+/*
+  create a named talloc pointer. Any talloc pointer can be named, and
+  talloc_named() operates just like talloc() except that it allows you
+  to name the pointer.
+*/
+_PUBLIC_ void *talloc_named(const void *context, size_t size, const char *fmt, ...)
+{
+	va_list ap;
+	void *ptr;
+	const char *name;
+
+	ptr = __talloc(context, size);
+	if (unlikely(ptr == NULL)) return NULL;
+
+	va_start(ap, fmt);
+	name = talloc_set_name_v(ptr, fmt, ap);
+	va_end(ap);
+
+	if (unlikely(name == NULL)) {
+		_talloc_free_internal(ptr, __location__);
+		return NULL;
+	}
+
+	return ptr;
+}
+
+/*
+  return the name of a talloc ptr, or "UNNAMED"
+*/
+_PUBLIC_ const char *talloc_get_name(const void *ptr)
+{
+	struct talloc_chunk *tc = talloc_chunk_from_ptr(ptr);
+	if (unlikely(tc->name == TALLOC_MAGIC_REFERENCE)) {
+		return ".reference";
+	}
+	if (likely(tc->name)) {
+		return tc->name;
+	}
+	return "UNNAMED";
+}
+
+
+/*
+  check if a pointer has the given name. If it does, return the pointer,
+  otherwise return NULL
+*/
+_PUBLIC_ void *talloc_check_name(const void *ptr, const char *name)
+{
+	const char *pname;
+	if (unlikely(ptr == NULL)) return NULL;
+	pname = talloc_get_name(ptr);
+	if (likely(pname == name || strcmp(pname, name) == 0)) {
+		return discard_const_p(void, ptr);
+	}
+	return NULL;
+}
+
+static void talloc_abort_type_mismatch(const char *location,
+					const char *name,
+					const char *expected)
+{
+	const char *reason;
+
+	reason = talloc_asprintf(NULL,
+				 "%s: Type mismatch: name[%s] expected[%s]",
+				 location,
+				 name?name:"NULL",
+				 expected);
+	if (!reason) {
+		reason = "Type mismatch";
+	}
+
+	talloc_abort(reason);
+}
+
+_PUBLIC_ void *_talloc_get_type_abort(const void *ptr, const char *name, const char *location)
+{
+	const char *pname;
+
+	if (unlikely(ptr == NULL)) {
+		talloc_abort_type_mismatch(location, NULL, name);
+		return NULL;
+	}
+
+	pname = talloc_get_name(ptr);
+	if (likely(pname == name || strcmp(pname, name) == 0)) {
+		return discard_const_p(void, ptr);
+	}
+
+	talloc_abort_type_mismatch(location, pname, name);
+	return NULL;
+}
+
+/*
+  this is for compatibility with older versions of talloc
+*/
+_PUBLIC_ void *talloc_init(const char *fmt, ...)
+{
+	va_list ap;
+	void *ptr;
+	const char *name;
+
+	ptr = __talloc(NULL, 0);
+	if (unlikely(ptr == NULL)) return NULL;
+
+	va_start(ap, fmt);
+	name = talloc_set_name_v(ptr, fmt, ap);
+	va_end(ap);
+
+	if (unlikely(name == NULL)) {
+		_talloc_free_internal(ptr, __location__);
+		return NULL;
+	}
+
+	return ptr;
+}
+
+static inline void _talloc_free_children_internal(struct talloc_chunk *tc,
+						  void *ptr,
+						  const char *location)
+{
+	while (tc->child) {
+		/* we need to work out who will own an abandoned child
+		   if it cannot be freed. In priority order, the first
+		   choice is owner of any remaining reference to this
+		   pointer, the second choice is our parent, and the
+		   final choice is the null context. */
+		void *child = TC_PTR_FROM_CHUNK(tc->child);
+		const void *new_parent = null_context;
+		if (unlikely(tc->child->refs)) {
+			struct talloc_chunk *p = talloc_parent_chunk(tc->child->refs);
+			if (p) new_parent = TC_PTR_FROM_CHUNK(p);
+		}
+		if (unlikely(_talloc_free_internal(child, location) == -1)) {
+			if (new_parent == null_context) {
+				struct talloc_chunk *p = talloc_parent_chunk(ptr);
+				if (p) new_parent = TC_PTR_FROM_CHUNK(p);
+			}
+			_talloc_steal_internal(new_parent, child);
+		}
+	}
+}
+
+/*
+  this is a replacement for the Samba3 talloc_destroy_pool functionality. It
+  should probably not be used in new code. It's in here to keep the talloc
+  code consistent across Samba 3 and 4.
+*/
+_PUBLIC_ void talloc_free_children(void *ptr)
+{
+	struct talloc_chunk *tc_name = NULL;
+	struct talloc_chunk *tc;
+
+	if (unlikely(ptr == NULL)) {
+		return;
+	}
+
+	tc = talloc_chunk_from_ptr(ptr);
+
+	/* we do not want to free the context name if it is a child .. */
+	if (likely(tc->child)) {
+		for (tc_name = tc->child; tc_name; tc_name = tc_name->next) {
+			if (tc->name == TC_PTR_FROM_CHUNK(tc_name)) break;
+		}
+		if (tc_name) {
+			_TLIST_REMOVE(tc->child, tc_name);
+			if (tc->child) {
+				tc->child->parent = tc;
+			}
+		}
+	}
+
+	_talloc_free_children_internal(tc, ptr, __location__);
+
+	/* .. so we put it back after all other children have been freed */
+	if (tc_name) {
+		if (tc->child) {
+			tc->child->parent = NULL;
+		}
+		tc_name->parent = tc;
+		_TLIST_ADD(tc->child, tc_name);
+	}
+}
+
+/* 
+   Allocate a bit of memory as a child of an existing pointer
+*/
+_PUBLIC_ void *_talloc(const void *context, size_t size)
+{
+	return __talloc(context, size);
+}
+
+/*
+  externally callable talloc_set_name_const()
+*/
+_PUBLIC_ void talloc_set_name_const(const void *ptr, const char *name)
+{
+	_talloc_set_name_const(ptr, name);
+}
+
+/*
+  create a named talloc pointer. Any talloc pointer can be named, and
+  talloc_named() operates just like talloc() except that it allows you
+  to name the pointer.
+*/
+_PUBLIC_ void *talloc_named_const(const void *context, size_t size, const char *name)
+{
+	return _talloc_named_const(context, size, name);
+}
+
+/* 
+   free a talloc pointer. This also frees all child pointers of this 
+   pointer recursively
+
+   return 0 if the memory is actually freed, otherwise -1. The memory
+   will not be freed if the ref_count is > 1 or the destructor (if
+   any) returns non-zero
+*/
+_PUBLIC_ int _talloc_free(void *ptr, const char *location)
+{
+	struct talloc_chunk *tc;
+
+	if (unlikely(ptr == NULL)) {
+		return -1;
+	}
+	
+	tc = talloc_chunk_from_ptr(ptr);
+	
+	if (unlikely(tc->refs != NULL)) {
+		struct talloc_reference_handle *h;
+
+		if (talloc_parent(ptr) == null_context && tc->refs->next == NULL) {
+			/* in this case we do know which parent should
+			   get this pointer, as there is really only
+			   one parent */
+			return talloc_unlink(null_context, ptr);
+		}
+
+		talloc_log("ERROR: talloc_free with references at %s\n",
+			   location);
+
+		for (h=tc->refs; h; h=h->next) {
+			talloc_log("\treference at %s\n",
+				   h->location);
+		}
+		return -1;
+	}
+	
+	return _talloc_free_internal(ptr, location);
+}
+
+
+
+/*
+  A talloc version of realloc. The context argument is only used if
+  ptr is NULL
+*/
+_PUBLIC_ void *_talloc_realloc(const void *context, void *ptr, size_t size, const char *name)
+{
+	struct talloc_chunk *tc;
+	void *new_ptr;
+	bool malloced = false;
+	union talloc_pool_chunk *pool_tc = NULL;
+
+	/* size zero is equivalent to free() */
+	if (unlikely(size == 0)) {
+		talloc_unlink(context, ptr);
+		return NULL;
+	}
+
+	if (unlikely(size >= MAX_TALLOC_SIZE)) {
+		return NULL;
+	}
+
+	/* realloc(NULL) is equivalent to malloc() */
+	if (ptr == NULL) {
+		return _talloc_named_const(context, size, name);
+	}
+
+	tc = talloc_chunk_from_ptr(ptr);
+
+	/* don't allow realloc on referenced pointers */
+	if (unlikely(tc->refs)) {
+		return NULL;
+	}
+
+	/* don't let anybody try to realloc a talloc_pool */
+	if (unlikely(tc->flags & TALLOC_FLAG_POOL)) {
+		return NULL;
+	}
+
+	if (tc->limit && (size - tc->size > 0)) {
+		if (!talloc_memlimit_check(tc->limit, (size - tc->size))) {
+			errno = ENOMEM;
+			return NULL;
+		}
+	}
+
+	/* handle realloc inside a talloc_pool */
+	if (unlikely(tc->flags & TALLOC_FLAG_POOLMEM)) {
+		pool_tc = (union talloc_pool_chunk *)tc->pool;
+	}
+
+#if (ALWAYS_REALLOC == 0)
+	/* don't shrink if we have less than 1k to gain */
+	if (size < tc->size && tc->limit == NULL) {
+		if (pool_tc) {
+			void *next_tc = tc_next_chunk(tc);
+			TC_INVALIDATE_SHRINK_CHUNK(tc, size);
+			tc->size = size;
+			if (next_tc == pool_tc->hdr.c.pool) {
+				/* note: tc->size has changed, so this works */
+				pool_tc->hdr.c.pool = tc_next_chunk(tc);
+			}
+			return ptr;
+		} else if ((tc->size - size) < 1024) {
+			/*
+			 * if we call TC_INVALIDATE_SHRINK_CHUNK() here
+			 * we would need to call TC_UNDEFINE_GROW_CHUNK()
+			 * after each realloc call, which slows down
+			 * testing a lot :-(.
+			 *
+			 * That is why we only mark memory as undefined here.
+			 */
+			TC_UNDEFINE_SHRINK_CHUNK(tc, size);
+
+			/* do not shrink if we have less than 1k to gain */
+			tc->size = size;
+			return ptr;
+		}
+	} else if (tc->size == size) {
+		/*
+		 * do not change the pointer if it is exactly
+		 * the same size.
+		 */
+		return ptr;
+	}
+#endif
+
+	/* by resetting magic we catch users of the old memory */
+	tc->flags |= TALLOC_FLAG_FREE;
+
+#if ALWAYS_REALLOC
+	if (pool_tc) {
+		new_ptr = talloc_alloc_pool(tc, size + TC_HDR_SIZE);
+		pool_tc->hdr.object_count--;
+
+		if (new_ptr == NULL) {
+			new_ptr = malloc(TC_HDR_SIZE+size);
+			malloced = true;
+		}
+
+		if (new_ptr) {
+			memcpy(new_ptr, tc, MIN(tc->size,size) + TC_HDR_SIZE);
+			TC_INVALIDATE_FULL_CHUNK(tc);
+		}
+	} else {
+		new_ptr = malloc(size + TC_HDR_SIZE);
+		if (new_ptr) {
+			memcpy(new_ptr, tc, MIN(tc->size, size) + TC_HDR_SIZE);
+			free(tc);
+		}
+	}
+#else
+	if (pool_tc) {
+		void *next_tc = tc_next_chunk(tc);
+		size_t old_chunk_size = TC_ALIGN16(TC_HDR_SIZE + tc->size);
+		size_t new_chunk_size = TC_ALIGN16(TC_HDR_SIZE + size);
+		size_t space_needed;
+		size_t space_left;
+		unsigned int chunk_count = pool_tc->hdr.object_count;
+
+		if (!(pool_tc->hdr.c.flags & TALLOC_FLAG_FREE)) {
+			chunk_count -= 1;
+		}
+
+		if (chunk_count == 1) {
+			/*
+			 * optimize for the case where 'tc' is the only
+			 * chunk in the pool.
+			 */
+			char *start = tc_pool_first_chunk(pool_tc);
+			space_needed = new_chunk_size;
+			space_left = (char *)tc_pool_end(pool_tc) - start;
+
+			if (space_left >= space_needed) {
+				size_t old_used = TC_HDR_SIZE + tc->size;
+				size_t new_used = TC_HDR_SIZE + size;
+				new_ptr = start;
+				memmove(new_ptr, tc, old_used);
+
+				tc = (struct talloc_chunk *)new_ptr;
+				TC_UNDEFINE_GROW_CHUNK(tc, size);
+
+				/*
+				 * first we do not align the pool pointer
+				 * because we want to invalidate the padding
+				 * too.
+				 */
+				pool_tc->hdr.c.pool = new_used + (char *)new_ptr;
+				tc_invalidate_pool(pool_tc);
+
+				/* now the aligned pointer */
+				pool_tc->hdr.c.pool = new_chunk_size + (char *)new_ptr;
+				goto got_new_ptr;
+			}
+
+			next_tc = NULL;
+		}
+
+		if (new_chunk_size == old_chunk_size) {
+			TC_UNDEFINE_GROW_CHUNK(tc, size);
+			tc->flags &= ~TALLOC_FLAG_FREE;
+			if (!talloc_memlimit_update(tc->limit,
+							tc->size, size)) {
+				talloc_abort("cur_size memlimit counter not"
+					     " correct!");
+				errno = EINVAL;
+				return NULL;
+			}
+
+			tc->size = size;
+			return ptr;
+		}
+
+		if (next_tc == pool_tc->hdr.c.pool) {
+			/*
+			 * optimize for the case where 'tc' is the last
+			 * chunk in the pool.
+			 */
+			space_needed = new_chunk_size - old_chunk_size;
+			space_left = tc_pool_space_left(pool_tc);
+
+			if (space_left >= space_needed) {
+				TC_UNDEFINE_GROW_CHUNK(tc, size);
+				tc->flags &= ~TALLOC_FLAG_FREE;
+				if (!talloc_memlimit_update(tc->limit,
+							tc->size, size)) {
+					talloc_abort("cur_size memlimit "
+						     "counter not correct!");
+					errno = EINVAL;
+					return NULL;
+				}
+				tc->size = size;
+				pool_tc->hdr.c.pool = tc_next_chunk(tc);
+				return ptr;
+			}
+		}
+
+		new_ptr = talloc_alloc_pool(tc, size + TC_HDR_SIZE);
+
+		if (new_ptr == NULL) {
+			new_ptr = malloc(TC_HDR_SIZE+size);
+			malloced = true;
+		}
+
+		if (new_ptr) {
+			memcpy(new_ptr, tc, MIN(tc->size,size) + TC_HDR_SIZE);
+
+			_talloc_free_poolmem(tc, __location__ "_talloc_realloc");
+		}
+	}
+	else {
+		new_ptr = realloc(tc, size + TC_HDR_SIZE);
+	}
+got_new_ptr:
+#endif
+	if (unlikely(!new_ptr)) {	
+		tc->flags &= ~TALLOC_FLAG_FREE; 
+		return NULL; 
+	}
+
+	tc = (struct talloc_chunk *)new_ptr;
+	tc->flags &= ~TALLOC_FLAG_FREE;
+	if (malloced) {
+		tc->flags &= ~TALLOC_FLAG_POOLMEM;
+	}
+	if (tc->parent) {
+		tc->parent->child = tc;
+	}
+	if (tc->child) {
+		tc->child->parent = tc;
+	}
+
+	if (tc->prev) {
+		tc->prev->next = tc;
+	}
+	if (tc->next) {
+		tc->next->prev = tc;
+	}
+
+	if (!talloc_memlimit_update(tc->limit, tc->size, size)) {
+		talloc_abort("cur_size memlimit counter not correct!");
+		errno = EINVAL;
+		return NULL;
+	}
+	tc->size = size;
+	_talloc_set_name_const(TC_PTR_FROM_CHUNK(tc), name);
+
+	return TC_PTR_FROM_CHUNK(tc);
+}
+
+/*
+  a wrapper around talloc_steal() for situations where you are moving a pointer
+  between two structures, and want the old pointer to be set to NULL
+*/
+_PUBLIC_ void *_talloc_move(const void *new_ctx, const void *_pptr)
+{
+	const void **pptr = discard_const_p(const void *,_pptr);
+	void *ret = talloc_steal(new_ctx, discard_const_p(void, *pptr));
+	(*pptr) = NULL;
+	return ret;
+}
+
+enum talloc_mem_count_type {
+	TOTAL_MEM_SIZE,
+	TOTAL_MEM_BLOCKS,
+	TOTAL_MEM_LIMIT,
+};
+
+static size_t _talloc_total_mem_internal(const void *ptr,
+					 enum talloc_mem_count_type type,
+					 struct talloc_memlimit *old_limit,
+					 struct talloc_memlimit *new_limit)
+{
+	size_t total = 0;
+	struct talloc_chunk *c, *tc;
+
+	if (ptr == NULL) {
+		ptr = null_context;
+	}
+	if (ptr == NULL) {
+		return 0;
+	}
+
+	tc = talloc_chunk_from_ptr(ptr);
+
+	if (old_limit || new_limit) {
+		if (tc->limit && tc->limit->upper == old_limit) {
+			tc->limit->upper = new_limit;
+		}
+	}
+
+	/* optimize in the memlimits case */
+	if (type == TOTAL_MEM_LIMIT &&
+	    tc->limit != NULL &&
+	    tc->limit != old_limit &&
+	    tc->limit->parent == tc) {
+		return tc->limit->cur_size;
+	}
+
+	if (tc->flags & TALLOC_FLAG_LOOP) {
+		return 0;
+	}
+
+	tc->flags |= TALLOC_FLAG_LOOP;
+
+	if (old_limit || new_limit) {
+		if (old_limit == tc->limit) {
+			tc->limit = new_limit;
+		}
+	}
+
+	switch (type) {
+	case TOTAL_MEM_SIZE:
+		if (likely(tc->name != TALLOC_MAGIC_REFERENCE)) {
+			total = tc->size;
+		}
+		break;
+	case TOTAL_MEM_BLOCKS:
+		total++;
+		break;
+	case TOTAL_MEM_LIMIT:
+		if (likely(tc->name != TALLOC_MAGIC_REFERENCE)) {
+			total = tc->size + TC_HDR_SIZE;
+		}
+		break;
+	}
+	for (c = tc->child; c; c = c->next) {
+		total += _talloc_total_mem_internal(TC_PTR_FROM_CHUNK(c), type,
+						    old_limit, new_limit);
+	}
+
+	tc->flags &= ~TALLOC_FLAG_LOOP;
+
+	return total;
+}
+
+/*
+  return the total size of a talloc pool (subtree)
+*/
+_PUBLIC_ size_t talloc_total_size(const void *ptr)
+{
+	return _talloc_total_mem_internal(ptr, TOTAL_MEM_SIZE, NULL, NULL);
+}
+
+/*
+  return the total number of blocks in a talloc pool (subtree)
+*/
+_PUBLIC_ size_t talloc_total_blocks(const void *ptr)
+{
+	return _talloc_total_mem_internal(ptr, TOTAL_MEM_BLOCKS, NULL, NULL);
+}
+
+/*
+  return the number of external references to a pointer
+*/
+_PUBLIC_ size_t talloc_reference_count(const void *ptr)
+{
+	struct talloc_chunk *tc = talloc_chunk_from_ptr(ptr);
+	struct talloc_reference_handle *h;
+	size_t ret = 0;
+
+	for (h=tc->refs;h;h=h->next) {
+		ret++;
+	}
+	return ret;
+}
+
+/*
+  report on memory usage by all children of a pointer, giving a full tree view
+*/
+_PUBLIC_ void talloc_report_depth_cb(const void *ptr, int depth, int max_depth,
+			    void (*callback)(const void *ptr,
+			  		     int depth, int max_depth,
+					     int is_ref,
+					     void *private_data),
+			    void *private_data)
+{
+	struct talloc_chunk *c, *tc;
+
+	if (ptr == NULL) {
+		ptr = null_context;
+	}
+	if (ptr == NULL) return;
+
+	tc = talloc_chunk_from_ptr(ptr);
+
+	if (tc->flags & TALLOC_FLAG_LOOP) {
+		return;
+	}
+
+	callback(ptr, depth, max_depth, 0, private_data);
+
+	if (max_depth >= 0 && depth >= max_depth) {
+		return;
+	}
+
+	tc->flags |= TALLOC_FLAG_LOOP;
+	for (c=tc->child;c;c=c->next) {
+		if (c->name == TALLOC_MAGIC_REFERENCE) {
+			struct talloc_reference_handle *h = (struct talloc_reference_handle *)TC_PTR_FROM_CHUNK(c);
+			callback(h->ptr, depth + 1, max_depth, 1, private_data);
+		} else {
+			talloc_report_depth_cb(TC_PTR_FROM_CHUNK(c), depth + 1, max_depth, callback, private_data);
+		}
+	}
+	tc->flags &= ~TALLOC_FLAG_LOOP;
+}
+
+static void talloc_report_depth_FILE_helper(const void *ptr, int depth, int max_depth, int is_ref, void *_f)
+{
+	const char *name = talloc_get_name(ptr);
+	struct talloc_chunk *tc;
+	FILE *f = (FILE *)_f;
+
+	if (is_ref) {
+		fprintf(f, "%*sreference to: %s\n", depth*4, "", name);
+		return;
+	}
+
+	tc = talloc_chunk_from_ptr(ptr);
+	if (tc->limit && tc->limit->parent == tc) {
+		fprintf(f, "%*s%-30s is a memlimit context"
+			" (max_size = %lu bytes, cur_size = %lu bytes)\n",
+			depth*4, "",
+			name,
+			(unsigned long)tc->limit->max_size,
+			(unsigned long)tc->limit->cur_size);
+	}
+
+	if (depth == 0) {
+		fprintf(f,"%stalloc report on '%s' (total %6lu bytes in %3lu blocks)\n", 
+			(max_depth < 0 ? "full " :""), name,
+			(unsigned long)talloc_total_size(ptr),
+			(unsigned long)talloc_total_blocks(ptr));
+		return;
+	}
+
+	fprintf(f, "%*s%-30s contains %6lu bytes in %3lu blocks (ref %d) %p\n", 
+		depth*4, "",
+		name,
+		(unsigned long)talloc_total_size(ptr),
+		(unsigned long)talloc_total_blocks(ptr),
+		(int)talloc_reference_count(ptr), ptr);
+
+#if 0
+	fprintf(f, "content: ");
+	if (talloc_total_size(ptr)) {
+		int tot = talloc_total_size(ptr);
+		int i;
+
+		for (i = 0; i < tot; i++) {
+			if ((((char *)ptr)[i] > 31) && (((char *)ptr)[i] < 126)) {
+				fprintf(f, "%c", ((char *)ptr)[i]);
+			} else {
+				fprintf(f, "~%02x", ((char *)ptr)[i]);
+			}
+		}
+	}
+	fprintf(f, "\n");
+#endif
+}
+
+/*
+  report on memory usage by all children of a pointer, giving a full tree view
+*/
+_PUBLIC_ void talloc_report_depth_file(const void *ptr, int depth, int max_depth, FILE *f)
+{
+	if (f) {
+		talloc_report_depth_cb(ptr, depth, max_depth, talloc_report_depth_FILE_helper, f);
+		fflush(f);
+	}
+}
+
+/*
+  report on memory usage by all children of a pointer, giving a full tree view
+*/
+_PUBLIC_ void talloc_report_full(const void *ptr, FILE *f)
+{
+	talloc_report_depth_file(ptr, 0, -1, f);
+}
+
+/*
+  report on memory usage by all children of a pointer
+*/
+_PUBLIC_ void talloc_report(const void *ptr, FILE *f)
+{
+	talloc_report_depth_file(ptr, 0, 1, f);
+}
+
+/*
+  report on any memory hanging off the null context
+*/
+static void talloc_report_null(void)
+{
+	if (talloc_total_size(null_context) != 0) {
+		talloc_report(null_context, stderr);
+	}
+}
+
+/*
+  report on any memory hanging off the null context
+*/
+static void talloc_report_null_full(void)
+{
+	if (talloc_total_size(null_context) != 0) {
+		talloc_report_full(null_context, stderr);
+	}
+}
+
+/*
+  enable tracking of the NULL context
+*/
+_PUBLIC_ void talloc_enable_null_tracking(void)
+{
+	if (null_context == NULL) {
+		null_context = _talloc_named_const(NULL, 0, "null_context");
+		if (autofree_context != NULL) {
+			talloc_reparent(NULL, null_context, autofree_context);
+		}
+	}
+}
+
+/*
+  enable tracking of the NULL context, not moving the autofree context
+  into the NULL context. This is needed for the talloc testsuite
+*/
+_PUBLIC_ void talloc_enable_null_tracking_no_autofree(void)
+{
+	if (null_context == NULL) {
+		null_context = _talloc_named_const(NULL, 0, "null_context");
+	}
+}
+
+/*
+  disable tracking of the NULL context
+*/
+_PUBLIC_ void talloc_disable_null_tracking(void)
+{
+	if (null_context != NULL) {
+		/* we have to move any children onto the real NULL
+		   context */
+		struct talloc_chunk *tc, *tc2;
+		tc = talloc_chunk_from_ptr(null_context);
+		for (tc2 = tc->child; tc2; tc2=tc2->next) {
+			if (tc2->parent == tc) tc2->parent = NULL;
+			if (tc2->prev == tc) tc2->prev = NULL;
+		}
+		for (tc2 = tc->next; tc2; tc2=tc2->next) {
+			if (tc2->parent == tc) tc2->parent = NULL;
+			if (tc2->prev == tc) tc2->prev = NULL;
+		}
+		tc->child = NULL;
+		tc->next = NULL;
+	}
+	talloc_free(null_context);
+	null_context = NULL;
+}
+
+/*
+  enable leak reporting on exit
+*/
+_PUBLIC_ void talloc_enable_leak_report(void)
+{
+	talloc_enable_null_tracking();
+	atexit(talloc_report_null);
+}
+
+/*
+  enable full leak reporting on exit
+*/
+_PUBLIC_ void talloc_enable_leak_report_full(void)
+{
+	talloc_enable_null_tracking();
+	atexit(talloc_report_null_full);
+}
+
+/* 
+   talloc and zero memory. 
+*/
+_PUBLIC_ void *_talloc_zero(const void *ctx, size_t size, const char *name)
+{
+	void *p = _talloc_named_const(ctx, size, name);
+
+	if (p) {
+		memset(p, '\0', size);
+	}
+
+	return p;
+}
+
+/*
+  memdup with a talloc. 
+*/
+_PUBLIC_ void *_talloc_memdup(const void *t, const void *p, size_t size, const char *name)
+{
+	void *newp = _talloc_named_const(t, size, name);
+
+	if (likely(newp)) {
+		memcpy(newp, p, size);
+	}
+
+	return newp;
+}
+
+static inline char *__talloc_strlendup(const void *t, const char *p, size_t len)
+{
+	char *ret;
+
+	ret = (char *)__talloc(t, len + 1);
+	if (unlikely(!ret)) return NULL;
+
+	memcpy(ret, p, len);
+	ret[len] = 0;
+
+	_talloc_set_name_const(ret, ret);
+	return ret;
+}
+
+/*
+  strdup with a talloc
+*/
+_PUBLIC_ char *talloc_strdup(const void *t, const char *p)
+{
+	if (unlikely(!p)) return NULL;
+	return __talloc_strlendup(t, p, strlen(p));
+}
+
+/*
+  strndup with a talloc
+*/
+_PUBLIC_ char *talloc_strndup(const void *t, const char *p, size_t n)
+{
+	if (unlikely(!p)) return NULL;
+	return __talloc_strlendup(t, p, strnlen(p, n));
+}
+
+static inline char *__talloc_strlendup_append(char *s, size_t slen,
+					      const char *a, size_t alen)
+{
+	char *ret;
+
+	ret = talloc_realloc(NULL, s, char, slen + alen + 1);
+	if (unlikely(!ret)) return NULL;
+
+	/* append the string and the trailing \0 */
+	memcpy(&ret[slen], a, alen);
+	ret[slen+alen] = 0;
+
+	_talloc_set_name_const(ret, ret);
+	return ret;
+}
+
+/*
+ * Appends at the end of the string.
+ */
+_PUBLIC_ char *talloc_strdup_append(char *s, const char *a)
+{
+	if (unlikely(!s)) {
+		return talloc_strdup(NULL, a);
+	}
+
+	if (unlikely(!a)) {
+		return s;
+	}
+
+	return __talloc_strlendup_append(s, strlen(s), a, strlen(a));
+}
+
+/*
+ * Appends at the end of the talloc'ed buffer,
+ * not the end of the string.
+ */
+_PUBLIC_ char *talloc_strdup_append_buffer(char *s, const char *a)
+{
+	size_t slen;
+
+	if (unlikely(!s)) {
+		return talloc_strdup(NULL, a);
+	}
+
+	if (unlikely(!a)) {
+		return s;
+	}
+
+	slen = talloc_get_size(s);
+	if (likely(slen > 0)) {
+		slen--;
+	}
+
+	return __talloc_strlendup_append(s, slen, a, strlen(a));
+}
+
+/*
+ * Appends at the end of the string.
+ */
+_PUBLIC_ char *talloc_strndup_append(char *s, const char *a, size_t n)
+{
+	if (unlikely(!s)) {
+		return talloc_strndup(NULL, a, n);
+	}
+
+	if (unlikely(!a)) {
+		return s;
+	}
+
+	return __talloc_strlendup_append(s, strlen(s), a, strnlen(a, n));
+}
+
+/*
+ * Appends at the end of the talloc'ed buffer,
+ * not the end of the string.
+ */
+_PUBLIC_ char *talloc_strndup_append_buffer(char *s, const char *a, size_t n)
+{
+	size_t slen;
+
+	if (unlikely(!s)) {
+		return talloc_strndup(NULL, a, n);
+	}
+
+	if (unlikely(!a)) {
+		return s;
+	}
+
+	slen = talloc_get_size(s);
+	if (likely(slen > 0)) {
+		slen--;
+	}
+
+	return __talloc_strlendup_append(s, slen, a, strnlen(a, n));
+}
+
+#ifndef HAVE_VA_COPY
+#ifdef HAVE___VA_COPY
+#define va_copy(dest, src) __va_copy(dest, src)
+#else
+#define va_copy(dest, src) (dest) = (src)
+#endif
+#endif
+
+_PUBLIC_ char *talloc_vasprintf(const void *t, const char *fmt, va_list ap)
+{
+	int len;
+	char *ret;
+	va_list ap2;
+	char c;
+
+	/* this call looks strange, but it makes it work on older solaris boxes */
+	va_copy(ap2, ap);
+	len = vsnprintf(&c, 1, fmt, ap2);
+	va_end(ap2);
+	if (unlikely(len < 0)) {
+		return NULL;
+	}
+
+	ret = (char *)__talloc(t, len+1);
+	if (unlikely(!ret)) return NULL;
+
+	va_copy(ap2, ap);
+	vsnprintf(ret, len+1, fmt, ap2);
+	va_end(ap2);
+
+	_talloc_set_name_const(ret, ret);
+	return ret;
+}
+
+
+/*
+  Perform string formatting, and return a pointer to newly allocated
+  memory holding the result, inside a memory pool.
+ */
+_PUBLIC_ char *talloc_asprintf(const void *t, const char *fmt, ...)
+{
+	va_list ap;
+	char *ret;
+
+	va_start(ap, fmt);
+	ret = talloc_vasprintf(t, fmt, ap);
+	va_end(ap);
+	return ret;
+}
+
+static inline char *__talloc_vaslenprintf_append(char *s, size_t slen,
+						 const char *fmt, va_list ap)
+						 PRINTF_ATTRIBUTE(3,0);
+
+static inline char *__talloc_vaslenprintf_append(char *s, size_t slen,
+						 const char *fmt, va_list ap)
+{
+	ssize_t alen;
+	va_list ap2;
+	char c;
+
+	va_copy(ap2, ap);
+	alen = vsnprintf(&c, 1, fmt, ap2);
+	va_end(ap2);
+
+	if (alen <= 0) {
+		/* Either the vsnprintf failed or the format resulted in
+		 * no characters being formatted. In the former case, we
+		 * ought to return NULL, in the latter we ought to return
+		 * the original string. Most current callers of this
+		 * function expect it to never return NULL.
+		 */
+		return s;
+	}
+
+	s = talloc_realloc(NULL, s, char, slen + alen + 1);
+	if (!s) return NULL;
+
+	va_copy(ap2, ap);
+	vsnprintf(s + slen, alen + 1, fmt, ap2);
+	va_end(ap2);
+
+	_talloc_set_name_const(s, s);
+	return s;
+}
+
+/**
+ * Realloc @p s to append the formatted result of @p fmt and @p ap,
+ * and return @p s, which may have moved.  Good for gradually
+ * accumulating output into a string buffer. Appends at the end
+ * of the string.
+ **/
+_PUBLIC_ char *talloc_vasprintf_append(char *s, const char *fmt, va_list ap)
+{
+	if (unlikely(!s)) {
+		return talloc_vasprintf(NULL, fmt, ap);
+	}
+
+	return __talloc_vaslenprintf_append(s, strlen(s), fmt, ap);
+}
+
+/**
+ * Realloc @p s to append the formatted result of @p fmt and @p ap,
+ * and return @p s, which may have moved. Always appends at the
+ * end of the talloc'ed buffer, not the end of the string.
+ **/
+_PUBLIC_ char *talloc_vasprintf_append_buffer(char *s, const char *fmt, va_list ap)
+{
+	size_t slen;
+
+	if (unlikely(!s)) {
+		return talloc_vasprintf(NULL, fmt, ap);
+	}
+
+	slen = talloc_get_size(s);
+	if (likely(slen > 0)) {
+		slen--;
+	}
+
+	return __talloc_vaslenprintf_append(s, slen, fmt, ap);
+}
+
+/*
+  Realloc @p s to append the formatted result of @p fmt and return @p
+  s, which may have moved.  Good for gradually accumulating output
+  into a string buffer.
+ */
+_PUBLIC_ char *talloc_asprintf_append(char *s, const char *fmt, ...)
+{
+	va_list ap;
+
+	va_start(ap, fmt);
+	s = talloc_vasprintf_append(s, fmt, ap);
+	va_end(ap);
+	return s;
+}
+
+/*
+  Realloc @p s to append the formatted result of @p fmt and return @p
+  s, which may have moved.  Good for gradually accumulating output
+  into a buffer.
+ */
+_PUBLIC_ char *talloc_asprintf_append_buffer(char *s, const char *fmt, ...)
+{
+	va_list ap;
+
+	va_start(ap, fmt);
+	s = talloc_vasprintf_append_buffer(s, fmt, ap);
+	va_end(ap);
+	return s;
+}
+
+/*
+  alloc an array, checking for integer overflow in the array size
+*/
+_PUBLIC_ void *_talloc_array(const void *ctx, size_t el_size, unsigned count, const char *name)
+{
+	if (count >= MAX_TALLOC_SIZE/el_size) {
+		return NULL;
+	}
+	return _talloc_named_const(ctx, el_size * count, name);
+}
+
+/*
+  alloc an zero array, checking for integer overflow in the array size
+*/
+_PUBLIC_ void *_talloc_zero_array(const void *ctx, size_t el_size, unsigned count, const char *name)
+{
+	if (count >= MAX_TALLOC_SIZE/el_size) {
+		return NULL;
+	}
+	return _talloc_zero(ctx, el_size * count, name);
+}
+
+/*
+  realloc an array, checking for integer overflow in the array size
+*/
+_PUBLIC_ void *_talloc_realloc_array(const void *ctx, void *ptr, size_t el_size, unsigned count, const char *name)
+{
+	if (count >= MAX_TALLOC_SIZE/el_size) {
+		return NULL;
+	}
+	return _talloc_realloc(ctx, ptr, el_size * count, name);
+}
+
+/*
+  a function version of talloc_realloc(), so it can be passed as a function pointer
+  to libraries that want a realloc function (a realloc function encapsulates
+  all the basic capabilities of an allocation library, which is why this is useful)
+*/
+_PUBLIC_ void *talloc_realloc_fn(const void *context, void *ptr, size_t size)
+{
+	return _talloc_realloc(context, ptr, size, NULL);
+}
+
+
+static int talloc_autofree_destructor(void *ptr)
+{
+	autofree_context = NULL;
+	return 0;
+}
+
+static void talloc_autofree(void)
+{
+	talloc_free(autofree_context);
+}
+
+/*
+  return a context which will be auto-freed on exit
+  this is useful for reducing the noise in leak reports
+*/
+_PUBLIC_ void *talloc_autofree_context(void)
+{
+	if (autofree_context == NULL) {
+		autofree_context = _talloc_named_const(NULL, 0, "autofree_context");
+		talloc_set_destructor(autofree_context, talloc_autofree_destructor);
+		atexit(talloc_autofree);
+	}
+	return autofree_context;
+}
+
+_PUBLIC_ size_t talloc_get_size(const void *context)
+{
+	struct talloc_chunk *tc;
+
+	if (context == NULL) {
+		context = null_context;
+	}
+	if (context == NULL) {
+		return 0;
+	}
+
+	tc = talloc_chunk_from_ptr(context);
+
+	return tc->size;
+}
+
+/*
+  find a parent of this context that has the given name, if any
+*/
+_PUBLIC_ void *talloc_find_parent_byname(const void *context, const char *name)
+{
+	struct talloc_chunk *tc;
+
+	if (context == NULL) {
+		return NULL;
+	}
+
+	tc = talloc_chunk_from_ptr(context);
+	while (tc) {
+		if (tc->name && strcmp(tc->name, name) == 0) {
+			return TC_PTR_FROM_CHUNK(tc);
+		}
+		while (tc && tc->prev) tc = tc->prev;
+		if (tc) {
+			tc = tc->parent;
+		}
+	}
+	return NULL;
+}
+
+/*
+  show the parentage of a context
+*/
+_PUBLIC_ void talloc_show_parents(const void *context, FILE *file)
+{
+	struct talloc_chunk *tc;
+
+	if (context == NULL) {
+		fprintf(file, "talloc no parents for NULL\n");
+		return;
+	}
+
+	tc = talloc_chunk_from_ptr(context);
+	fprintf(file, "talloc parents of '%s'\n", talloc_get_name(context));
+	while (tc) {
+		fprintf(file, "\t'%s'\n", talloc_get_name(TC_PTR_FROM_CHUNK(tc)));
+		while (tc && tc->prev) tc = tc->prev;
+		if (tc) {
+			tc = tc->parent;
+		}
+	}
+	fflush(file);
+}
+
+/*
+  return 1 if ptr is a parent of context
+*/
+static int _talloc_is_parent(const void *context, const void *ptr, int depth)
+{
+	struct talloc_chunk *tc;
+
+	if (context == NULL) {
+		return 0;
+	}
+
+	tc = talloc_chunk_from_ptr(context);
+	while (tc && depth > 0) {
+		if (TC_PTR_FROM_CHUNK(tc) == ptr) return 1;
+		while (tc && tc->prev) tc = tc->prev;
+		if (tc) {
+			tc = tc->parent;
+			depth--;
+		}
+	}
+	return 0;
+}
+
+/*
+  return 1 if ptr is a parent of context
+*/
+_PUBLIC_ int talloc_is_parent(const void *context, const void *ptr)
+{
+	return _talloc_is_parent(context, ptr, TALLOC_MAX_DEPTH);
+}
+
+/*
+  return the total size of memory used by this context and all children
+*/
+static size_t _talloc_total_limit_size(const void *ptr,
+					struct talloc_memlimit *old_limit,
+					struct talloc_memlimit *new_limit)
+{
+	return _talloc_total_mem_internal(ptr, TOTAL_MEM_LIMIT,
+					  old_limit, new_limit);
+}
+
+static bool talloc_memlimit_check(struct talloc_memlimit *limit, size_t size)
+{
+	struct talloc_memlimit *l;
+
+	for (l = limit; l != NULL; l = l->upper) {
+		if (l->max_size != 0 &&
+		    ((l->max_size <= l->cur_size) ||
+		     (l->max_size - l->cur_size < TC_HDR_SIZE+size))) {
+			return false;
+		}
+	}
+
+	return true;
+}
+
+static bool talloc_memlimit_update(struct talloc_memlimit *limit,
+				   size_t old_size, size_t new_size)
+{
+	struct talloc_memlimit *l;
+	ssize_t d;
+
+	if (old_size == 0) {
+		d = new_size + TC_HDR_SIZE;
+	} else {
+		d = new_size - old_size;
+	}
+	for (l = limit; l != NULL; l = l->upper) {
+		ssize_t new_cur_size = l->cur_size + d;
+		if (new_cur_size < 0) {
+			return false;
+		}
+		l->cur_size = new_cur_size;
+	}
+
+	return true;
+}
+
+_PUBLIC_ int talloc_set_memlimit(const void *ctx, size_t max_size)
+{
+	struct talloc_chunk *tc = talloc_chunk_from_ptr(ctx);
+	struct talloc_memlimit *orig_limit;
+	struct talloc_memlimit *limit = NULL;
+
+	if (tc->limit && tc->limit->parent == tc) {
+		tc->limit->max_size = max_size;
+		return 0;
+	}
+	orig_limit = tc->limit;
+
+	limit = malloc(sizeof(struct talloc_memlimit));
+	if (limit == NULL) {
+		return 1;
+	}
+	limit->parent = tc;
+	limit->max_size = max_size;
+	limit->cur_size = _talloc_total_limit_size(ctx, tc->limit, limit);
+
+	if (orig_limit) {
+		limit->upper = orig_limit;
+	} else {
+		limit->upper = NULL;
+	}
+
+	return 0;
+}
diff --git a/ctdb/lib/talloc/talloc.h b/ctdb/lib/talloc/talloc.h
new file mode 100644
index 0000000000..f3cbcd0e7c
--- /dev/null
+++ b/ctdb/lib/talloc/talloc.h
@@ -0,0 +1,1883 @@
+#ifndef _TALLOC_H_
+#define _TALLOC_H_
+/*
+   Unix SMB/CIFS implementation.
+   Samba temporary memory allocation functions
+
+   Copyright (C) Andrew Tridgell 2004-2005
+   Copyright (C) Stefan Metzmacher 2006
+
+     ** NOTE! The following LGPL license applies to the talloc
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <stdarg.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * @defgroup talloc The talloc API
+ *
+ * talloc is a hierarchical, reference counted memory pool system with
+ * destructors. It is the core memory allocator used in Samba.
+ *
+ * @{
+ */
+
+#define TALLOC_VERSION_MAJOR 2
+#define TALLOC_VERSION_MINOR 0
+
+int talloc_version_major(void);
+int talloc_version_minor(void);
+
+/**
+ * @brief Define a talloc parent type
+ *
+ * As talloc is a hierarchial memory allocator, every talloc chunk is a
+ * potential parent to other talloc chunks. So defining a separate type for a
+ * talloc chunk is not strictly necessary. TALLOC_CTX is defined nevertheless,
+ * as it provides an indicator for function arguments. You will frequently
+ * write code like
+ *
+ * @code
+ *      struct foo *foo_create(TALLOC_CTX *mem_ctx)
+ *      {
+ *              struct foo *result;
+ *              result = talloc(mem_ctx, struct foo);
+ *              if (result == NULL) return NULL;
+ *                      ... initialize foo ...
+ *              return result;
+ *      }
+ * @endcode
+ *
+ * In this type of allocating functions it is handy to have a general
+ * TALLOC_CTX type to indicate which parent to put allocated structures on.
+ */
+typedef void TALLOC_CTX;
+
+/*
+  this uses a little trick to allow __LINE__ to be stringified
+*/
+#ifndef __location__
+#define __TALLOC_STRING_LINE1__(s)    #s
+#define __TALLOC_STRING_LINE2__(s)   __TALLOC_STRING_LINE1__(s)
+#define __TALLOC_STRING_LINE3__  __TALLOC_STRING_LINE2__(__LINE__)
+#define __location__ __FILE__ ":" __TALLOC_STRING_LINE3__
+#endif
+
+#ifndef TALLOC_DEPRECATED
+#define TALLOC_DEPRECATED 0
+#endif
+
+#ifndef PRINTF_ATTRIBUTE
+#if (__GNUC__ >= 3)
+/** Use gcc attribute to check printf fns.  a1 is the 1-based index of
+ * the parameter containing the format, and a2 the index of the first
+ * argument. Note that some gcc 2.x versions don't handle this
+ * properly **/
+#define PRINTF_ATTRIBUTE(a1, a2) __attribute__ ((format (__printf__, a1, a2)))
+#else
+#define PRINTF_ATTRIBUTE(a1, a2)
+#endif
+#endif
+
+#ifdef DOXYGEN
+/**
+ * @brief Create a new talloc context.
+ *
+ * The talloc() macro is the core of the talloc library. It takes a memory
+ * context and a type, and returns a pointer to a new area of memory of the
+ * given type.
+ *
+ * The returned pointer is itself a talloc context, so you can use it as the
+ * context argument to more calls to talloc if you wish.
+ *
+ * The returned pointer is a "child" of the supplied context. This means that if
+ * you talloc_free() the context then the new child disappears as well.
+ * Alternatively you can free just the child.
+ *
+ * @param[in]  ctx      A talloc context to create a new reference on or NULL to
+ *                      create a new top level context.
+ *
+ * @param[in]  type     The type of memory to allocate.
+ *
+ * @return              A type casted talloc context or NULL on error.
+ *
+ * @code
+ *      unsigned int *a, *b;
+ *
+ *      a = talloc(NULL, unsigned int);
+ *      b = talloc(a, unsigned int);
+ * @endcode
+ *
+ * @see talloc_zero
+ * @see talloc_array
+ * @see talloc_steal
+ * @see talloc_free
+ */
+void *talloc(const void *ctx, #type);
+#else
+#define talloc(ctx, type) (type *)talloc_named_const(ctx, sizeof(type), #type)
+void *_talloc(const void *context, size_t size);
+#endif
+
+/**
+ * @brief Create a new top level talloc context.
+ *
+ * This function creates a zero length named talloc context as a top level
+ * context. It is equivalent to:
+ *
+ * @code
+ *      talloc_named(NULL, 0, fmt, ...);
+ * @endcode
+ * @param[in]  fmt      Format string for the name.
+ *
+ * @param[in]  ...      Additional printf-style arguments.
+ *
+ * @return              The allocated memory chunk, NULL on error.
+ *
+ * @see talloc_named()
+ */
+void *talloc_init(const char *fmt, ...) PRINTF_ATTRIBUTE(1,2);
+
+#ifdef DOXYGEN
+/**
+ * @brief Free a chunk of talloc memory.
+ *
+ * The talloc_free() function frees a piece of talloc memory, and all its
+ * children. You can call talloc_free() on any pointer returned by
+ * talloc().
+ *
+ * The return value of talloc_free() indicates success or failure, with 0
+ * returned for success and -1 for failure. A possible failure condition
+ * is if the pointer had a destructor attached to it and the destructor
+ * returned -1. See talloc_set_destructor() for details on
+ * destructors. Likewise, if "ptr" is NULL, then the function will make
+ * no modifications and return -1.
+ *
+ * From version 2.0 and onwards, as a special case, talloc_free() is
+ * refused on pointers that have more than one parent associated, as talloc
+ * would have no way of knowing which parent should be removed. This is
+ * different from older versions in the sense that always the reference to
+ * the most recently established parent has been destroyed. Hence to free a
+ * pointer that has more than one parent please use talloc_unlink().
+ *
+ * To help you find problems in your code caused by this behaviour, if
+ * you do try and free a pointer with more than one parent then the
+ * talloc logging function will be called to give output like this:
+ *
+ * @code
+ *   ERROR: talloc_free with references at some_dir/source/foo.c:123
+ *     reference at some_dir/source/other.c:325
+ *     reference at some_dir/source/third.c:121
+ * @endcode
+ *
+ * Please see the documentation for talloc_set_log_fn() and
+ * talloc_set_log_stderr() for more information on talloc logging
+ * functions.
+ *
+ * If <code>TALLOC_FREE_FILL</code> environment variable is set,
+ * the memory occupied by the context is filled with the value of this variable.
+ * The value should be a numeric representation of the character you want to
+ * use.
+ *
+ * talloc_free() operates recursively on its children.
+ *
+ * @param[in]  ptr      The chunk to be freed.
+ *
+ * @return              Returns 0 on success and -1 on error. A possible
+ *                      failure condition is if the pointer had a destructor
+ *                      attached to it and the destructor returned -1. Likewise,
+ *                      if "ptr" is NULL, then the function will make no
+ *                      modifications and returns -1.
+ *
+ * Example:
+ * @code
+ *      unsigned int *a, *b;
+ *      a = talloc(NULL, unsigned int);
+ *      b = talloc(a, unsigned int);
+ *
+ *      talloc_free(a); // Frees a and b
+ * @endcode
+ *
+ * @see talloc_set_destructor()
+ * @see talloc_unlink()
+ */
+int talloc_free(void *ptr);
+#else
+#define talloc_free(ctx) _talloc_free(ctx, __location__)
+int _talloc_free(void *ptr, const char *location);
+#endif
+
+/**
+ * @brief Free a talloc chunk's children.
+ *
+ * The function walks along the list of all children of a talloc context and
+ * talloc_free()s only the children, not the context itself.
+ *
+ * A NULL argument is handled as no-op.
+ *
+ * @param[in]  ptr      The chunk that you want to free the children of
+ *                      (NULL is allowed too)
+ */
+void talloc_free_children(void *ptr);
+
+#ifdef DOXYGEN
+/**
+ * @brief Assign a destructor function to be called when a chunk is freed.
+ *
+ * The function talloc_set_destructor() sets the "destructor" for the pointer
+ * "ptr". A destructor is a function that is called when the memory used by a
+ * pointer is about to be released. The destructor receives the pointer as an
+ * argument, and should return 0 for success and -1 for failure.
+ *
+ * The destructor can do anything it wants to, including freeing other pieces
+ * of memory. A common use for destructors is to clean up operating system
+ * resources (such as open file descriptors) contained in the structure the
+ * destructor is placed on.
+ *
+ * You can only place one destructor on a pointer. If you need more than one
+ * destructor then you can create a zero-length child of the pointer and place
+ * an additional destructor on that.
+ *
+ * To remove a destructor call talloc_set_destructor() with NULL for the
+ * destructor.
+ *
+ * If your destructor attempts to talloc_free() the pointer that it is the
+ * destructor for then talloc_free() will return -1 and the free will be
+ * ignored. This would be a pointless operation anyway, as the destructor is
+ * only called when the memory is just about to go away.
+ *
+ * @param[in]  ptr      The talloc chunk to add a destructor to.
+ *
+ * @param[in]  destructor  The destructor function to be called. NULL to remove
+ *                         it.
+ *
+ * Example:
+ * @code
+ *      static int destroy_fd(int *fd) {
+ *              close(*fd);
+ *              return 0;
+ *      }
+ *
+ *      int *open_file(const char *filename) {
+ *              int *fd = talloc(NULL, int);
+ *              *fd = open(filename, O_RDONLY);
+ *              if (*fd < 0) {
+ *                      talloc_free(fd);
+ *                      return NULL;
+ *              }
+ *              // Whenever they free this, we close the file.
+ *              talloc_set_destructor(fd, destroy_fd);
+ *              return fd;
+ *      }
+ * @endcode
+ *
+ * @see talloc()
+ * @see talloc_free()
+ */
+void talloc_set_destructor(const void *ptr, int (*destructor)(void *));
+
+/**
+ * @brief Change a talloc chunk's parent.
+ *
+ * The talloc_steal() function changes the parent context of a talloc
+ * pointer. It is typically used when the context that the pointer is
+ * currently a child of is going to be freed and you wish to keep the
+ * memory for a longer time.
+ *
+ * To make the changed hierarchy less error-prone, you might consider to use
+ * talloc_move().
+ *
+ * If you try and call talloc_steal() on a pointer that has more than one
+ * parent then the result is ambiguous. Talloc will choose to remove the
+ * parent that is currently indicated by talloc_parent() and replace it with
+ * the chosen parent. You will also get a message like this via the talloc
+ * logging functions:
+ *
+ * @code
+ *   WARNING: talloc_steal with references at some_dir/source/foo.c:123
+ *     reference at some_dir/source/other.c:325
+ *     reference at some_dir/source/third.c:121
+ * @endcode
+ *
+ * To unambiguously change the parent of a pointer please see the function
+ * talloc_reparent(). See the talloc_set_log_fn() documentation for more
+ * information on talloc logging.
+ *
+ * @param[in]  new_ctx  The new parent context.
+ *
+ * @param[in]  ptr      The talloc chunk to move.
+ *
+ * @return              Returns the pointer that you pass it. It does not have
+ *                      any failure modes.
+ *
+ * @note It is possible to produce loops in the parent/child relationship
+ * if you are not careful with talloc_steal(). No guarantees are provided
+ * as to your sanity or the safety of your data if you do this.
+ */
+void *talloc_steal(const void *new_ctx, const void *ptr);
+#else /* DOXYGEN */
+/* try to make talloc_set_destructor() and talloc_steal() type safe,
+   if we have a recent gcc */
+#if (__GNUC__ >= 3)
+#define _TALLOC_TYPEOF(ptr) __typeof__(ptr)
+#define talloc_set_destructor(ptr, function)				      \
+	do {								      \
+		int (*_talloc_destructor_fn)(_TALLOC_TYPEOF(ptr)) = (function);	      \
+		_talloc_set_destructor((ptr), (int (*)(void *))_talloc_destructor_fn); \
+	} while(0)
+/* this extremely strange macro is to avoid some braindamaged warning
+   stupidity in gcc 4.1.x */
+#define talloc_steal(ctx, ptr) ({ _TALLOC_TYPEOF(ptr) __talloc_steal_ret = (_TALLOC_TYPEOF(ptr))_talloc_steal_loc((ctx),(ptr), __location__); __talloc_steal_ret; })
+#else /* __GNUC__ >= 3 */
+#define talloc_set_destructor(ptr, function) \
+	_talloc_set_destructor((ptr), (int (*)(void *))(function))
+#define _TALLOC_TYPEOF(ptr) void *
+#define talloc_steal(ctx, ptr) (_TALLOC_TYPEOF(ptr))_talloc_steal_loc((ctx),(ptr), __location__)
+#endif /* __GNUC__ >= 3 */
+void _talloc_set_destructor(const void *ptr, int (*_destructor)(void *));
+void *_talloc_steal_loc(const void *new_ctx, const void *ptr, const char *location);
+#endif /* DOXYGEN */
+
+/**
+ * @brief Assign a name to a talloc chunk.
+ *
+ * Each talloc pointer has a "name". The name is used principally for
+ * debugging purposes, although it is also possible to set and get the name on
+ * a pointer in as a way of "marking" pointers in your code.
+ *
+ * The main use for names on pointer is for "talloc reports". See
+ * talloc_report() and talloc_report_full() for details. Also see
+ * talloc_enable_leak_report() and talloc_enable_leak_report_full().
+ *
+ * The talloc_set_name() function allocates memory as a child of the
+ * pointer. It is logically equivalent to:
+ *
+ * @code
+ *      talloc_set_name_const(ptr, talloc_asprintf(ptr, fmt, ...));
+ * @endcode
+ *
+ * @param[in]  ptr      The talloc chunk to assign a name to.
+ *
+ * @param[in]  fmt      Format string for the name.
+ *
+ * @param[in]  ...      Add printf-style additional arguments.
+ *
+ * @return              The assigned name, NULL on error.
+ *
+ * @note Multiple calls to talloc_set_name() will allocate more memory without
+ * releasing the name. All of the memory is released when the ptr is freed
+ * using talloc_free().
+ */
+const char *talloc_set_name(const void *ptr, const char *fmt, ...) PRINTF_ATTRIBUTE(2,3);
+
+#ifdef DOXYGEN
+/**
+ * @brief Change a talloc chunk's parent.
+ *
+ * This function has the same effect as talloc_steal(), and additionally sets
+ * the source pointer to NULL. You would use it like this:
+ *
+ * @code
+ *      struct foo *X = talloc(tmp_ctx, struct foo);
+ *      struct foo *Y;
+ *      Y = talloc_move(new_ctx, &X);
+ * @endcode
+ *
+ * @param[in]  new_ctx  The new parent context.
+ *
+ * @param[in]  pptr     Pointer to the talloc chunk to move.
+ *
+ * @return              The pointer of the talloc chunk it has been moved to,
+ *                      NULL on error.
+ */
+void *talloc_move(const void *new_ctx, void **pptr);
+#else
+#define talloc_move(ctx, pptr) (_TALLOC_TYPEOF(*(pptr)))_talloc_move((ctx),(void *)(pptr))
+void *_talloc_move(const void *new_ctx, const void *pptr);
+#endif
+
+/**
+ * @brief Assign a name to a talloc chunk.
+ *
+ * The function is just like talloc_set_name(), but it takes a string constant,
+ * and is much faster. It is extensively used by the "auto naming" macros, such
+ * as talloc_p().
+ *
+ * This function does not allocate any memory. It just copies the supplied
+ * pointer into the internal representation of the talloc ptr. This means you
+ * must not pass a name pointer to memory that will disappear before the ptr
+ * is freed with talloc_free().
+ *
+ * @param[in]  ptr      The talloc chunk to assign a name to.
+ *
+ * @param[in]  name     Format string for the name.
+ */
+void talloc_set_name_const(const void *ptr, const char *name);
+
+/**
+ * @brief Create a named talloc chunk.
+ *
+ * The talloc_named() function creates a named talloc pointer. It is
+ * equivalent to:
+ *
+ * @code
+ *      ptr = talloc_size(context, size);
+ *      talloc_set_name(ptr, fmt, ....);
+ * @endcode
+ *
+ * @param[in]  context  The talloc context to hang the result off.
+ *
+ * @param[in]  size     Number of char's that you want to allocate.
+ *
+ * @param[in]  fmt      Format string for the name.
+ *
+ * @param[in]  ...      Additional printf-style arguments.
+ *
+ * @return              The allocated memory chunk, NULL on error.
+ *
+ * @see talloc_set_name()
+ */
+void *talloc_named(const void *context, size_t size,
+		   const char *fmt, ...) PRINTF_ATTRIBUTE(3,4);
+
+/**
+ * @brief Basic routine to allocate a chunk of memory.
+ *
+ * This is equivalent to:
+ *
+ * @code
+ *      ptr = talloc_size(context, size);
+ *      talloc_set_name_const(ptr, name);
+ * @endcode
+ *
+ * @param[in]  context  The parent context.
+ *
+ * @param[in]  size     The number of char's that we want to allocate.
+ *
+ * @param[in]  name     The name the talloc block has.
+ *
+ * @return             The allocated memory chunk, NULL on error.
+ */
+void *talloc_named_const(const void *context, size_t size, const char *name);
+
+#ifdef DOXYGEN
+/**
+ * @brief Untyped allocation.
+ *
+ * The function should be used when you don't have a convenient type to pass to
+ * talloc(). Unlike talloc(), it is not type safe (as it returns a void *), so
+ * you are on your own for type checking.
+ *
+ * Best to use talloc() or talloc_array() instead.
+ *
+ * @param[in]  ctx     The talloc context to hang the result off.
+ *
+ * @param[in]  size    Number of char's that you want to allocate.
+ *
+ * @return             The allocated memory chunk, NULL on error.
+ *
+ * Example:
+ * @code
+ *      void *mem = talloc_size(NULL, 100);
+ * @endcode
+ */
+void *talloc_size(const void *ctx, size_t size);
+#else
+#define talloc_size(ctx, size) talloc_named_const(ctx, size, __location__)
+#endif
+
+#ifdef DOXYGEN
+/**
+ * @brief Allocate into a typed pointer.
+ *
+ * The talloc_ptrtype() macro should be used when you have a pointer and want
+ * to allocate memory to point at with this pointer. When compiling with
+ * gcc >= 3 it is typesafe. Note this is a wrapper of talloc_size() and
+ * talloc_get_name() will return the current location in the source file and
+ * not the type.
+ *
+ * @param[in]  ctx      The talloc context to hang the result off.
+ *
+ * @param[in]  type     The pointer you want to assign the result to.
+ *
+ * @return              The properly casted allocated memory chunk, NULL on
+ *                      error.
+ *
+ * Example:
+ * @code
+ *       unsigned int *a = talloc_ptrtype(NULL, a);
+ * @endcode
+ */
+void *talloc_ptrtype(const void *ctx, #type);
+#else
+#define talloc_ptrtype(ctx, ptr) (_TALLOC_TYPEOF(ptr))talloc_size(ctx, sizeof(*(ptr)))
+#endif
+
+#ifdef DOXYGEN
+/**
+ * @brief Allocate a new 0-sized talloc chunk.
+ *
+ * This is a utility macro that creates a new memory context hanging off an
+ * existing context, automatically naming it "talloc_new: __location__" where
+ * __location__ is the source line it is called from. It is particularly
+ * useful for creating a new temporary working context.
+ *
+ * @param[in]  ctx      The talloc parent context.
+ *
+ * @return              A new talloc chunk, NULL on error.
+ */
+void *talloc_new(const void *ctx);
+#else
+#define talloc_new(ctx) talloc_named_const(ctx, 0, "talloc_new: " __location__)
+#endif
+
+#ifdef DOXYGEN
+/**
+ * @brief Allocate a 0-initizialized structure.
+ *
+ * The macro is equivalent to:
+ *
+ * @code
+ *      ptr = talloc(ctx, type);
+ *      if (ptr) memset(ptr, 0, sizeof(type));
+ * @endcode
+ *
+ * @param[in]  ctx      The talloc context to hang the result off.
+ *
+ * @param[in]  type     The type that we want to allocate.
+ *
+ * @return              Pointer to a piece of memory, properly cast to 'type *',
+ *                      NULL on error.
+ *
+ * Example:
+ * @code
+ *      unsigned int *a, *b;
+ *      a = talloc_zero(NULL, unsigned int);
+ *      b = talloc_zero(a, unsigned int);
+ * @endcode
+ *
+ * @see talloc()
+ * @see talloc_zero_size()
+ * @see talloc_zero_array()
+ */
+void *talloc_zero(const void *ctx, #type);
+
+/**
+ * @brief Allocate untyped, 0-initialized memory.
+ *
+ * @param[in]  ctx      The talloc context to hang the result off.
+ *
+ * @param[in]  size     Number of char's that you want to allocate.
+ *
+ * @return              The allocated memory chunk.
+ */
+void *talloc_zero_size(const void *ctx, size_t size);
+#else
+#define talloc_zero(ctx, type) (type *)_talloc_zero(ctx, sizeof(type), #type)
+#define talloc_zero_size(ctx, size) _talloc_zero(ctx, size, __location__)
+void *_talloc_zero(const void *ctx, size_t size, const char *name);
+#endif
+
+/**
+ * @brief Return the name of a talloc chunk.
+ *
+ * @param[in]  ptr      The talloc chunk.
+ *
+ * @return              The current name for the given talloc pointer.
+ *
+ * @see talloc_set_name()
+ */
+const char *talloc_get_name(const void *ptr);
+
+/**
+ * @brief Verify that a talloc chunk carries a specified name.
+ *
+ * This function checks if a pointer has the specified name. If it does
+ * then the pointer is returned.
+ *
+ * @param[in]  ptr       The talloc chunk to check.
+ *
+ * @param[in]  name      The name to check against.
+ *
+ * @return               The pointer if the name matches, NULL if it doesn't.
+ */
+void *talloc_check_name(const void *ptr, const char *name);
+
+/**
+ * @brief Get the parent chunk of a pointer.
+ *
+ * @param[in]  ptr      The talloc pointer to inspect.
+ *
+ * @return              The talloc parent of ptr, NULL on error.
+ */
+void *talloc_parent(const void *ptr);
+
+/**
+ * @brief Get a talloc chunk's parent name.
+ *
+ * @param[in]  ptr      The talloc pointer to inspect.
+ *
+ * @return              The name of ptr's parent chunk.
+ */
+const char *talloc_parent_name(const void *ptr);
+
+/**
+ * @brief Get the total size of a talloc chunk including its children.
+ *
+ * The function returns the total size in bytes used by this pointer and all
+ * child pointers. Mostly useful for debugging.
+ *
+ * Passing NULL is allowed, but it will only give a meaningful result if
+ * talloc_enable_leak_report() or talloc_enable_leak_report_full() has
+ * been called.
+ *
+ * @param[in]  ptr      The talloc chunk.
+ *
+ * @return              The total size.
+ */
+size_t talloc_total_size(const void *ptr);
+
+/**
+ * @brief Get the number of talloc chunks hanging off a chunk.
+ *
+ * The talloc_total_blocks() function returns the total memory block
+ * count used by this pointer and all child pointers. Mostly useful for
+ * debugging.
+ *
+ * Passing NULL is allowed, but it will only give a meaningful result if
+ * talloc_enable_leak_report() or talloc_enable_leak_report_full() has
+ * been called.
+ *
+ * @param[in]  ptr      The talloc chunk.
+ *
+ * @return              The total size.
+ */
+size_t talloc_total_blocks(const void *ptr);
+
+#ifdef DOXYGEN
+/**
+ * @brief Duplicate a memory area into a talloc chunk.
+ *
+ * The function is equivalent to:
+ *
+ * @code
+ *      ptr = talloc_size(ctx, size);
+ *      if (ptr) memcpy(ptr, p, size);
+ * @endcode
+ *
+ * @param[in]  t        The talloc context to hang the result off.
+ *
+ * @param[in]  p        The memory chunk you want to duplicate.
+ *
+ * @param[in]  size     Number of char's that you want copy.
+ *
+ * @return              The allocated memory chunk.
+ *
+ * @see talloc_size()
+ */
+void *talloc_memdup(const void *t, const void *p, size_t size);
+#else
+#define talloc_memdup(t, p, size) _talloc_memdup(t, p, size, __location__)
+void *_talloc_memdup(const void *t, const void *p, size_t size, const char *name);
+#endif
+
+#ifdef DOXYGEN
+/**
+ * @brief Assign a type to a talloc chunk.
+ *
+ * This macro allows you to force the name of a pointer to be of a particular
+ * type. This can be used in conjunction with talloc_get_type() to do type
+ * checking on void* pointers.
+ *
+ * It is equivalent to this:
+ *
+ * @code
+ *      talloc_set_name_const(ptr, #type)
+ * @endcode
+ *
+ * @param[in]  ptr      The talloc chunk to assign the type to.
+ *
+ * @param[in]  type     The type to assign.
+ */
+void talloc_set_type(const char *ptr, #type);
+
+/**
+ * @brief Get a typed pointer out of a talloc pointer.
+ *
+ * This macro allows you to do type checking on talloc pointers. It is
+ * particularly useful for void* private pointers. It is equivalent to
+ * this:
+ *
+ * @code
+ *      (type *)talloc_check_name(ptr, #type)
+ * @endcode
+ *
+ * @param[in]  ptr      The talloc pointer to check.
+ *
+ * @param[in]  type     The type to check against.
+ *
+ * @return              The properly casted pointer given by ptr, NULL on error.
+ */
+type *talloc_get_type(const void *ptr, #type);
+#else
+#define talloc_set_type(ptr, type) talloc_set_name_const(ptr, #type)
+#define talloc_get_type(ptr, type) (type *)talloc_check_name(ptr, #type)
+#endif
+
+#ifdef DOXYGEN
+/**
+ * @brief Safely turn a void pointer into a typed pointer.
+ *
+ * This macro is used together with talloc(mem_ctx, struct foo). If you had to
+ * assing the talloc chunk pointer to some void pointer variable,
+ * talloc_get_type_abort() is the recommended way to get the convert the void
+ * pointer back to a typed pointer.
+ *
+ * @param[in]  ptr      The void pointer to convert.
+ *
+ * @param[in]  type     The type that this chunk contains
+ *
+ * @return              The same value as ptr, type-checked and properly cast.
+ */
+void *talloc_get_type_abort(const void *ptr, #type);
+#else
+#define talloc_get_type_abort(ptr, type) (type *)_talloc_get_type_abort(ptr, #type, __location__)
+void *_talloc_get_type_abort(const void *ptr, const char *name, const char *location);
+#endif
+
+/**
+ * @brief Find a parent context by name.
+ *
+ * Find a parent memory context of the current context that has the given
+ * name. This can be very useful in complex programs where it may be
+ * difficult to pass all information down to the level you need, but you
+ * know the structure you want is a parent of another context.
+ *
+ * @param[in]  ctx      The talloc chunk to start from.
+ *
+ * @param[in]  name     The name of the parent we look for.
+ *
+ * @return              The memory context we are looking for, NULL if not
+ *                      found.
+ */
+void *talloc_find_parent_byname(const void *ctx, const char *name);
+
+#ifdef DOXYGEN
+/**
+ * @brief Find a parent context by type.
+ *
+ * Find a parent memory context of the current context that has the given
+ * name. This can be very useful in complex programs where it may be
+ * difficult to pass all information down to the level you need, but you
+ * know the structure you want is a parent of another context.
+ *
+ * Like talloc_find_parent_byname() but takes a type, making it typesafe.
+ *
+ * @param[in]  ptr      The talloc chunk to start from.
+ *
+ * @param[in]  type     The type of the parent to look for.
+ *
+ * @return              The memory context we are looking for, NULL if not
+ *                      found.
+ */
+void *talloc_find_parent_bytype(const void *ptr, #type);
+#else
+#define talloc_find_parent_bytype(ptr, type) (type *)talloc_find_parent_byname(ptr, #type)
+#endif
+
+/**
+ * @brief Allocate a talloc pool.
+ *
+ * A talloc pool is a pure optimization for specific situations. In the
+ * release process for Samba 3.2 we found out that we had become considerably
+ * slower than Samba 3.0 was. Profiling showed that malloc(3) was a large CPU
+ * consumer in benchmarks. For Samba 3.2 we have internally converted many
+ * static buffers to dynamically allocated ones, so malloc(3) being beaten
+ * more was no surprise. But it made us slower.
+ *
+ * talloc_pool() is an optimization to call malloc(3) a lot less for the use
+ * pattern Samba has: The SMB protocol is mainly a request/response protocol
+ * where we have to allocate a certain amount of memory per request and free
+ * that after the SMB reply is sent to the client.
+ *
+ * talloc_pool() creates a talloc chunk that you can use as a talloc parent
+ * exactly as you would use any other ::TALLOC_CTX. The difference is that
+ * when you talloc a child of this pool, no malloc(3) is done. Instead, talloc
+ * just increments a pointer inside the talloc_pool. This also works
+ * recursively. If you use the child of the talloc pool as a parent for
+ * grand-children, their memory is also taken from the talloc pool.
+ *
+ * If there is not enough memory in the pool to allocate the new child,
+ * it will create a new talloc chunk as if the parent was a normal talloc
+ * context.
+ *
+ * If you talloc_free() children of a talloc pool, the memory is not given
+ * back to the system. Instead, free(3) is only called if the talloc_pool()
+ * itself is released with talloc_free().
+ *
+ * The downside of a talloc pool is that if you talloc_move() a child of a
+ * talloc pool to a talloc parent outside the pool, the whole pool memory is
+ * not free(3)'ed until that moved chunk is also talloc_free()ed.
+ *
+ * @param[in]  context  The talloc context to hang the result off (must not
+ *			be another pool).
+ *
+ * @param[in]  size     Size of the talloc pool.
+ *
+ * @return              The allocated talloc pool, NULL on error.
+ */
+void *talloc_pool(const void *context, size_t size);
+
+/**
+ * @brief Free a talloc chunk and NULL out the pointer.
+ *
+ * TALLOC_FREE() frees a pointer and sets it to NULL. Use this if you want
+ * immediate feedback (i.e. crash) if you use a pointer after having free'ed
+ * it.
+ *
+ * @param[in]  ctx      The chunk to be freed.
+ */
+#define TALLOC_FREE(ctx) do { talloc_free(ctx); ctx=NULL; } while(0)
+
+/* @} ******************************************************************/
+
+/**
+ * \defgroup talloc_ref The talloc reference function.
+ * @ingroup talloc
+ *
+ * This module contains the definitions around talloc references
+ *
+ * @{
+ */
+
+/**
+ * @brief Increase the reference count of a talloc chunk.
+ *
+ * The talloc_increase_ref_count(ptr) function is exactly equivalent to:
+ *
+ * @code
+ *      talloc_reference(NULL, ptr);
+ * @endcode
+ *
+ * You can use either syntax, depending on which you think is clearer in
+ * your code.
+ *
+ * @param[in]  ptr      The pointer to increase the reference count.
+ *
+ * @return              0 on success, -1 on error.
+ */
+int talloc_increase_ref_count(const void *ptr);
+
+/**
+ * @brief Get the number of references to a talloc chunk.
+ *
+ * @param[in]  ptr      The pointer to retrieve the reference count from.
+ *
+ * @return              The number of references.
+ */
+size_t talloc_reference_count(const void *ptr);
+
+#ifdef DOXYGEN
+/**
+ * @brief Create an additional talloc parent to a pointer.
+ *
+ * The talloc_reference() function makes "context" an additional parent of
+ * ptr. Each additional reference consumes around 48 bytes of memory on intel
+ * x86 platforms.
+ *
+ * If ptr is NULL, then the function is a no-op, and simply returns NULL.
+ *
+ * After creating a reference you can free it in one of the following ways:
+ *
+ * - you can talloc_free() any parent of the original pointer. That
+ *   will reduce the number of parents of this pointer by 1, and will
+ *   cause this pointer to be freed if it runs out of parents.
+ *
+ * - you can talloc_free() the pointer itself if it has at maximum one
+ *   parent. This behaviour has been changed since the release of version
+ *   2.0. Further informations in the description of "talloc_free".
+ *
+ * For more control on which parent to remove, see talloc_unlink()
+ * @param[in]  ctx      The additional parent.
+ *
+ * @param[in]  ptr      The pointer you want to create an additional parent for.
+ *
+ * @return              The original pointer 'ptr', NULL if talloc ran out of
+ *                      memory in creating the reference.
+ *
+ * Example:
+ * @code
+ *      unsigned int *a, *b, *c;
+ *      a = talloc(NULL, unsigned int);
+ *      b = talloc(NULL, unsigned int);
+ *      c = talloc(a, unsigned int);
+ *      // b also serves as a parent of c.
+ *      talloc_reference(b, c);
+ * @endcode
+ *
+ * @see talloc_unlink()
+ */
+void *talloc_reference(const void *ctx, const void *ptr);
+#else
+#define talloc_reference(ctx, ptr) (_TALLOC_TYPEOF(ptr))_talloc_reference_loc((ctx),(ptr), __location__)
+void *_talloc_reference_loc(const void *context, const void *ptr, const char *location);
+#endif
+
+/**
+ * @brief Remove a specific parent from a talloc chunk.
+ *
+ * The function removes a specific parent from ptr. The context passed must
+ * either be a context used in talloc_reference() with this pointer, or must be
+ * a direct parent of ptr.
+ *
+ * You can just use talloc_free() instead of talloc_unlink() if there
+ * is at maximum one parent. This behaviour has been changed since the
+ * release of version 2.0. Further informations in the description of
+ * "talloc_free".
+ *
+ * @param[in]  context  The talloc parent to remove.
+ *
+ * @param[in]  ptr      The talloc ptr you want to remove the parent from.
+ *
+ * @return              0 on success, -1 on error.
+ *
+ * @note If the parent has already been removed using talloc_free() then
+ * this function will fail and will return -1.  Likewise, if ptr is NULL,
+ * then the function will make no modifications and return -1.
+ *
+ * Example:
+ * @code
+ *      unsigned int *a, *b, *c;
+ *      a = talloc(NULL, unsigned int);
+ *      b = talloc(NULL, unsigned int);
+ *      c = talloc(a, unsigned int);
+ *      // b also serves as a parent of c.
+ *      talloc_reference(b, c);
+ *      talloc_unlink(b, c);
+ * @endcode
+ */
+int talloc_unlink(const void *context, void *ptr);
+
+/**
+ * @brief Provide a talloc context that is freed at program exit.
+ *
+ * This is a handy utility function that returns a talloc context
+ * which will be automatically freed on program exit. This can be used
+ * to reduce the noise in memory leak reports.
+ *
+ * Never use this in code that might be used in objects loaded with
+ * dlopen and unloaded with dlclose. talloc_autofree_context()
+ * internally uses atexit(3). Some platforms like modern Linux handles
+ * this fine, but for example FreeBSD does not deal well with dlopen()
+ * and atexit() used simultaneously: dlclose() does not clean up the
+ * list of atexit-handlers, so when the program exits the code that
+ * was registered from within talloc_autofree_context() is gone, the
+ * program crashes at exit.
+ *
+ * @return              A talloc context, NULL on error.
+ */
+void *talloc_autofree_context(void);
+
+/**
+ * @brief Get the size of a talloc chunk.
+ *
+ * This function lets you know the amount of memory allocated so far by
+ * this context. It does NOT account for subcontext memory.
+ * This can be used to calculate the size of an array.
+ *
+ * @param[in]  ctx      The talloc chunk.
+ *
+ * @return              The size of the talloc chunk.
+ */
+size_t talloc_get_size(const void *ctx);
+
+/**
+ * @brief Show the parentage of a context.
+ *
+ * @param[in]  context            The talloc context to look at.
+ *
+ * @param[in]  file               The output to use, a file, stdout or stderr.
+ */
+void talloc_show_parents(const void *context, FILE *file);
+
+/**
+ * @brief Check if a context is parent of a talloc chunk.
+ *
+ * This checks if context is referenced in the talloc hierarchy above ptr.
+ *
+ * @param[in]  context  The assumed talloc context.
+ *
+ * @param[in]  ptr      The talloc chunk to check.
+ *
+ * @return              Return 1 if this is the case, 0 if not.
+ */
+int talloc_is_parent(const void *context, const void *ptr);
+
+/**
+ * @brief Change the parent context of a talloc pointer.
+ *
+ * The function changes the parent context of a talloc pointer. It is typically
+ * used when the context that the pointer is currently a child of is going to be
+ * freed and you wish to keep the memory for a longer time.
+ *
+ * The difference between talloc_reparent() and talloc_steal() is that
+ * talloc_reparent() can specify which parent you wish to change. This is
+ * useful when a pointer has multiple parents via references.
+ *
+ * @param[in]  old_parent
+ * @param[in]  new_parent
+ * @param[in]  ptr
+ *
+ * @return              Return the pointer you passed. It does not have any
+ *                      failure modes.
+ */
+void *talloc_reparent(const void *old_parent, const void *new_parent, const void *ptr);
+
+/* @} ******************************************************************/
+
+/**
+ * @defgroup talloc_array The talloc array functions
+ * @ingroup talloc
+ *
+ * Talloc contains some handy helpers for handling Arrays conveniently
+ *
+ * @{
+ */
+
+#ifdef DOXYGEN
+/**
+ * @brief Allocate an array.
+ *
+ * The macro is equivalent to:
+ *
+ * @code
+ *      (type *)talloc_size(ctx, sizeof(type) * count);
+ * @endcode
+ *
+ * except that it provides integer overflow protection for the multiply,
+ * returning NULL if the multiply overflows.
+ *
+ * @param[in]  ctx      The talloc context to hang the result off.
+ *
+ * @param[in]  type     The type that we want to allocate.
+ *
+ * @param[in]  count    The number of 'type' elements you want to allocate.
+ *
+ * @return              The allocated result, properly cast to 'type *', NULL on
+ *                      error.
+ *
+ * Example:
+ * @code
+ *      unsigned int *a, *b;
+ *      a = talloc_zero(NULL, unsigned int);
+ *      b = talloc_array(a, unsigned int, 100);
+ * @endcode
+ *
+ * @see talloc()
+ * @see talloc_zero_array()
+ */
+void *talloc_array(const void *ctx, #type, unsigned count);
+#else
+#define talloc_array(ctx, type, count) (type *)_talloc_array(ctx, sizeof(type), count, #type)
+void *_talloc_array(const void *ctx, size_t el_size, unsigned count, const char *name);
+#endif
+
+#ifdef DOXYGEN
+/**
+ * @brief Allocate an array.
+ *
+ * @param[in]  ctx      The talloc context to hang the result off.
+ *
+ * @param[in]  size     The size of an array element.
+ *
+ * @param[in]  count    The number of elements you want to allocate.
+ *
+ * @return              The allocated result, NULL on error.
+ */
+void *talloc_array_size(const void *ctx, size_t size, unsigned count);
+#else
+#define talloc_array_size(ctx, size, count) _talloc_array(ctx, size, count, __location__)
+#endif
+
+#ifdef DOXYGEN
+/**
+ * @brief Allocate an array into a typed pointer.
+ *
+ * The macro should be used when you have a pointer to an array and want to
+ * allocate memory of an array to point at with this pointer. When compiling
+ * with gcc >= 3 it is typesafe. Note this is a wrapper of talloc_array_size()
+ * and talloc_get_name() will return the current location in the source file
+ * and not the type.
+ *
+ * @param[in]  ctx      The talloc context to hang the result off.
+ *
+ * @param[in]  ptr      The pointer you want to assign the result to.
+ *
+ * @param[in]  count    The number of elements you want to allocate.
+ *
+ * @return              The allocated memory chunk, properly casted. NULL on
+ *                      error.
+ */
+void *talloc_array_ptrtype(const void *ctx, const void *ptr, unsigned count);
+#else
+#define talloc_array_ptrtype(ctx, ptr, count) (_TALLOC_TYPEOF(ptr))talloc_array_size(ctx, sizeof(*(ptr)), count)
+#endif
+
+#ifdef DOXYGEN
+/**
+ * @brief Get the number of elements in a talloc'ed array.
+ *
+ * A talloc chunk carries its own size, so for talloc'ed arrays it is not
+ * necessary to store the number of elements explicitly.
+ *
+ * @param[in]  ctx      The allocated array.
+ *
+ * @return              The number of elements in ctx.
+ */
+size_t talloc_array_length(const void *ctx);
+#else
+#define talloc_array_length(ctx) (talloc_get_size(ctx)/sizeof(*ctx))
+#endif
+
+#ifdef DOXYGEN
+/**
+ * @brief Allocate a zero-initialized array
+ *
+ * @param[in]  ctx      The talloc context to hang the result off.
+ *
+ * @param[in]  type     The type that we want to allocate.
+ *
+ * @param[in]  count    The number of "type" elements you want to allocate.
+ *
+ * @return              The allocated result casted to "type *", NULL on error.
+ *
+ * The talloc_zero_array() macro is equivalent to:
+ *
+ * @code
+ *     ptr = talloc_array(ctx, type, count);
+ *     if (ptr) memset(ptr, sizeof(type) * count);
+ * @endcode
+ */
+void *talloc_zero_array(const void *ctx, #type, unsigned count);
+#else
+#define talloc_zero_array(ctx, type, count) (type *)_talloc_zero_array(ctx, sizeof(type), count, #type)
+void *_talloc_zero_array(const void *ctx,
+			 size_t el_size,
+			 unsigned count,
+			 const char *name);
+#endif
+
+#ifdef DOXYGEN
+/**
+ * @brief Change the size of a talloc array.
+ *
+ * The macro changes the size of a talloc pointer. The 'count' argument is the
+ * number of elements of type 'type' that you want the resulting pointer to
+ * hold.
+ *
+ * talloc_realloc() has the following equivalences:
+ *
+ * @code
+ *      talloc_realloc(ctx, NULL, type, 1) ==> talloc(ctx, type);
+ *      talloc_realloc(ctx, NULL, type, N) ==> talloc_array(ctx, type, N);
+ *      talloc_realloc(ctx, ptr, type, 0)  ==> talloc_free(ptr);
+ * @endcode
+ *
+ * The "context" argument is only used if "ptr" is NULL, otherwise it is
+ * ignored.
+ *
+ * @param[in]  ctx      The parent context used if ptr is NULL.
+ *
+ * @param[in]  ptr      The chunk to be resized.
+ *
+ * @param[in]  type     The type of the array element inside ptr.
+ *
+ * @param[in]  count    The intended number of array elements.
+ *
+ * @return              The new array, NULL on error. The call will fail either
+ *                      due to a lack of memory, or because the pointer has more
+ *                      than one parent (see talloc_reference()).
+ */
+void *talloc_realloc(const void *ctx, void *ptr, #type, size_t count);
+#else
+#define talloc_realloc(ctx, p, type, count) (type *)_talloc_realloc_array(ctx, p, sizeof(type), count, #type)
+void *_talloc_realloc_array(const void *ctx, void *ptr, size_t el_size, unsigned count, const char *name);
+#endif
+
+#ifdef DOXYGEN
+/**
+ * @brief Untyped realloc to change the size of a talloc array.
+ *
+ * The macro is useful when the type is not known so the typesafe
+ * talloc_realloc() cannot be used.
+ *
+ * @param[in]  ctx      The parent context used if 'ptr' is NULL.
+ *
+ * @param[in]  ptr      The chunk to be resized.
+ *
+ * @param[in]  size     The new chunk size.
+ *
+ * @return              The new array, NULL on error.
+ */
+void *talloc_realloc_size(const void *ctx, void *ptr, size_t size);
+#else
+#define talloc_realloc_size(ctx, ptr, size) _talloc_realloc(ctx, ptr, size, __location__)
+void *_talloc_realloc(const void *context, void *ptr, size_t size, const char *name);
+#endif
+
+/**
+ * @brief Provide a function version of talloc_realloc_size.
+ *
+ * This is a non-macro version of talloc_realloc(), which is useful as
+ * libraries sometimes want a ralloc function pointer. A realloc()
+ * implementation encapsulates the functionality of malloc(), free() and
+ * realloc() in one call, which is why it is useful to be able to pass around
+ * a single function pointer.
+ *
+ * @param[in]  context  The parent context used if ptr is NULL.
+ *
+ * @param[in]  ptr      The chunk to be resized.
+ *
+ * @param[in]  size     The new chunk size.
+ *
+ * @return              The new chunk, NULL on error.
+ */
+void *talloc_realloc_fn(const void *context, void *ptr, size_t size);
+
+/* @} ******************************************************************/
+
+/**
+ * @defgroup talloc_string The talloc string functions.
+ * @ingroup talloc
+ *
+ * talloc string allocation and manipulation functions.
+ * @{
+ */
+
+/**
+ * @brief Duplicate a string into a talloc chunk.
+ *
+ * This function is equivalent to:
+ *
+ * @code
+ *      ptr = talloc_size(ctx, strlen(p)+1);
+ *      if (ptr) memcpy(ptr, p, strlen(p)+1);
+ * @endcode
+ *
+ * This functions sets the name of the new pointer to the passed
+ * string. This is equivalent to:
+ *
+ * @code
+ *      talloc_set_name_const(ptr, ptr)
+ * @endcode
+ *
+ * @param[in]  t        The talloc context to hang the result off.
+ *
+ * @param[in]  p        The string you want to duplicate.
+ *
+ * @return              The duplicated string, NULL on error.
+ */
+char *talloc_strdup(const void *t, const char *p);
+
+/**
+ * @brief Append a string to given string.
+ *
+ * The destination string is reallocated to take
+ * <code>strlen(s) + strlen(a) + 1</code> characters.
+ *
+ * This functions sets the name of the new pointer to the new
+ * string. This is equivalent to:
+ *
+ * @code
+ *      talloc_set_name_const(ptr, ptr)
+ * @endcode
+ *
+ * If <code>s == NULL</code> then new context is created.
+ *
+ * @param[in]  s        The destination to append to.
+ *
+ * @param[in]  a        The string you want to append.
+ *
+ * @return              The concatenated strings, NULL on error.
+ *
+ * @see talloc_strdup()
+ * @see talloc_strdup_append_buffer()
+ */
+char *talloc_strdup_append(char *s, const char *a);
+
+/**
+ * @brief Append a string to a given buffer.
+ *
+ * This is a more efficient version of talloc_strdup_append(). It determines the
+ * length of the destination string by the size of the talloc context.
+ *
+ * Use this very carefully as it produces a different result than
+ * talloc_strdup_append() when a zero character is in the middle of the
+ * destination string.
+ *
+ * @code
+ *      char *str_a = talloc_strdup(NULL, "hello world");
+ *      char *str_b = talloc_strdup(NULL, "hello world");
+ *      str_a[5] = str_b[5] = '\0'
+ *
+ *      char *app = talloc_strdup_append(str_a, ", hello");
+ *      char *buf = talloc_strdup_append_buffer(str_b, ", hello");
+ *
+ *      printf("%s\n", app); // hello, hello (app = "hello, hello")
+ *      printf("%s\n", buf); // hello (buf = "hello\0world, hello")
+ * @endcode
+ *
+ * If <code>s == NULL</code> then new context is created.
+ *
+ * @param[in]  s        The destination buffer to append to.
+ *
+ * @param[in]  a        The string you want to append.
+ *
+ * @return              The concatenated strings, NULL on error.
+ *
+ * @see talloc_strdup()
+ * @see talloc_strdup_append()
+ * @see talloc_array_length()
+ */
+char *talloc_strdup_append_buffer(char *s, const char *a);
+
+/**
+ * @brief Duplicate a length-limited string into a talloc chunk.
+ *
+ * This function is the talloc equivalent of the C library function strndup(3).
+ *
+ * This functions sets the name of the new pointer to the passed string. This is
+ * equivalent to:
+ *
+ * @code
+ *      talloc_set_name_const(ptr, ptr)
+ * @endcode
+ *
+ * @param[in]  t        The talloc context to hang the result off.
+ *
+ * @param[in]  p        The string you want to duplicate.
+ *
+ * @param[in]  n        The maximum string length to duplicate.
+ *
+ * @return              The duplicated string, NULL on error.
+ */
+char *talloc_strndup(const void *t, const char *p, size_t n);
+
+/**
+ * @brief Append at most n characters of a string to given string.
+ *
+ * The destination string is reallocated to take
+ * <code>strlen(s) + strnlen(a, n) + 1</code> characters.
+ *
+ * This functions sets the name of the new pointer to the new
+ * string. This is equivalent to:
+ *
+ * @code
+ *      talloc_set_name_const(ptr, ptr)
+ * @endcode
+ *
+ * If <code>s == NULL</code> then new context is created.
+ *
+ * @param[in]  s        The destination string to append to.
+ *
+ * @param[in]  a        The source string you want to append.
+ *
+ * @param[in]  n        The number of characters you want to append from the
+ *                      string.
+ *
+ * @return              The concatenated strings, NULL on error.
+ *
+ * @see talloc_strndup()
+ * @see talloc_strndup_append_buffer()
+ */
+char *talloc_strndup_append(char *s, const char *a, size_t n);
+
+/**
+ * @brief Append at most n characters of a string to given buffer
+ *
+ * This is a more efficient version of talloc_strndup_append(). It determines
+ * the length of the destination string by the size of the talloc context.
+ *
+ * Use this very carefully as it produces a different result than
+ * talloc_strndup_append() when a zero character is in the middle of the
+ * destination string.
+ *
+ * @code
+ *      char *str_a = talloc_strdup(NULL, "hello world");
+ *      char *str_b = talloc_strdup(NULL, "hello world");
+ *      str_a[5] = str_b[5] = '\0'
+ *
+ *      char *app = talloc_strndup_append(str_a, ", hello", 7);
+ *      char *buf = talloc_strndup_append_buffer(str_b, ", hello", 7);
+ *
+ *      printf("%s\n", app); // hello, hello (app = "hello, hello")
+ *      printf("%s\n", buf); // hello (buf = "hello\0world, hello")
+ * @endcode
+ *
+ * If <code>s == NULL</code> then new context is created.
+ *
+ * @param[in]  s        The destination buffer to append to.
+ *
+ * @param[in]  a        The source string you want to append.
+ *
+ * @param[in]  n        The number of characters you want to append from the
+ *                      string.
+ *
+ * @return              The concatenated strings, NULL on error.
+ *
+ * @see talloc_strndup()
+ * @see talloc_strndup_append()
+ * @see talloc_array_length()
+ */
+char *talloc_strndup_append_buffer(char *s, const char *a, size_t n);
+
+/**
+ * @brief Format a string given a va_list.
+ *
+ * This function is the talloc equivalent of the C library function
+ * vasprintf(3).
+ *
+ * This functions sets the name of the new pointer to the new string. This is
+ * equivalent to:
+ *
+ * @code
+ *      talloc_set_name_const(ptr, ptr)
+ * @endcode
+ *
+ * @param[in]  t        The talloc context to hang the result off.
+ *
+ * @param[in]  fmt      The format string.
+ *
+ * @param[in]  ap       The parameters used to fill fmt.
+ *
+ * @return              The formatted string, NULL on error.
+ */
+char *talloc_vasprintf(const void *t, const char *fmt, va_list ap) PRINTF_ATTRIBUTE(2,0);
+
+/**
+ * @brief Format a string given a va_list and append it to the given destination
+ *        string.
+ *
+ * @param[in]  s        The destination string to append to.
+ *
+ * @param[in]  fmt      The format string.
+ *
+ * @param[in]  ap       The parameters used to fill fmt.
+ *
+ * @return              The formatted string, NULL on error.
+ *
+ * @see talloc_vasprintf()
+ */
+char *talloc_vasprintf_append(char *s, const char *fmt, va_list ap) PRINTF_ATTRIBUTE(2,0);
+
+/**
+ * @brief Format a string given a va_list and append it to the given destination
+ *        buffer.
+ *
+ * @param[in]  s        The destination buffer to append to.
+ *
+ * @param[in]  fmt      The format string.
+ *
+ * @param[in]  ap       The parameters used to fill fmt.
+ *
+ * @return              The formatted string, NULL on error.
+ *
+ * @see talloc_vasprintf()
+ */
+char *talloc_vasprintf_append_buffer(char *s, const char *fmt, va_list ap) PRINTF_ATTRIBUTE(2,0);
+
+/**
+ * @brief Format a string.
+ *
+ * This function is the talloc equivalent of the C library function asprintf(3).
+ *
+ * This functions sets the name of the new pointer to the new string. This is
+ * equivalent to:
+ *
+ * @code
+ *      talloc_set_name_const(ptr, ptr)
+ * @endcode
+ *
+ * @param[in]  t        The talloc context to hang the result off.
+ *
+ * @param[in]  fmt      The format string.
+ *
+ * @param[in]  ...      The parameters used to fill fmt.
+ *
+ * @return              The formatted string, NULL on error.
+ */
+char *talloc_asprintf(const void *t, const char *fmt, ...) PRINTF_ATTRIBUTE(2,3);
+
+/**
+ * @brief Append a formatted string to another string.
+ *
+ * This function appends the given formatted string to the given string. Use
+ * this variant when the string in the current talloc buffer may have been
+ * truncated in length.
+ *
+ * This functions sets the name of the new pointer to the new
+ * string. This is equivalent to:
+ *
+ * @code
+ *      talloc_set_name_const(ptr, ptr)
+ * @endcode
+ *
+ * If <code>s == NULL</code> then new context is created.
+ *
+ * @param[in]  s        The string to append to.
+ *
+ * @param[in]  fmt      The format string.
+ *
+ * @param[in]  ...      The parameters used to fill fmt.
+ *
+ * @return              The formatted string, NULL on error.
+ */
+char *talloc_asprintf_append(char *s, const char *fmt, ...) PRINTF_ATTRIBUTE(2,3);
+
+/**
+ * @brief Append a formatted string to another string.
+ *
+ * This is a more efficient version of talloc_asprintf_append(). It determines
+ * the length of the destination string by the size of the talloc context.
+ *
+ * Use this very carefully as it produces a different result than
+ * talloc_asprintf_append() when a zero character is in the middle of the
+ * destination string.
+ *
+ * @code
+ *      char *str_a = talloc_strdup(NULL, "hello world");
+ *      char *str_b = talloc_strdup(NULL, "hello world");
+ *      str_a[5] = str_b[5] = '\0'
+ *
+ *      char *app = talloc_asprintf_append(str_a, "%s", ", hello");
+ *      char *buf = talloc_strdup_append_buffer(str_b, "%s", ", hello");
+ *
+ *      printf("%s\n", app); // hello, hello (app = "hello, hello")
+ *      printf("%s\n", buf); // hello (buf = "hello\0world, hello")
+ * @endcode
+ *
+ * If <code>s == NULL</code> then new context is created.
+ *
+ * @param[in]  s        The string to append to
+ *
+ * @param[in]  fmt      The format string.
+ *
+ * @param[in]  ...      The parameters used to fill fmt.
+ *
+ * @return              The formatted string, NULL on error.
+ *
+ * @see talloc_asprintf()
+ * @see talloc_asprintf_append()
+ */
+char *talloc_asprintf_append_buffer(char *s, const char *fmt, ...) PRINTF_ATTRIBUTE(2,3);
+
+/* @} ******************************************************************/
+
+/**
+ * @defgroup talloc_debug The talloc debugging support functions
+ * @ingroup talloc
+ *
+ * To aid memory debugging, talloc contains routines to inspect the currently
+ * allocated memory hierarchy.
+ *
+ * @{
+ */
+
+/**
+ * @brief Walk a complete talloc hierarchy.
+ *
+ * This provides a more flexible reports than talloc_report(). It
+ * will recursively call the callback for the entire tree of memory
+ * referenced by the pointer. References in the tree are passed with
+ * is_ref = 1 and the pointer that is referenced.
+ *
+ * You can pass NULL for the pointer, in which case a report is
+ * printed for the top level memory context, but only if
+ * talloc_enable_leak_report() or talloc_enable_leak_report_full()
+ * has been called.
+ *
+ * The recursion is stopped when depth >= max_depth.
+ * max_depth = -1 means only stop at leaf nodes.
+ *
+ * @param[in]  ptr      The talloc chunk.
+ *
+ * @param[in]  depth    Internal parameter to control recursion. Call with 0.
+ *
+ * @param[in]  max_depth  Maximum recursion level.
+ *
+ * @param[in]  callback  Function to be called on every chunk.
+ *
+ * @param[in]  private_data  Private pointer passed to callback.
+ */
+void talloc_report_depth_cb(const void *ptr, int depth, int max_depth,
+			    void (*callback)(const void *ptr,
+					     int depth, int max_depth,
+					     int is_ref,
+					     void *private_data),
+			    void *private_data);
+
+/**
+ * @brief Print a talloc hierarchy.
+ *
+ * This provides a more flexible reports than talloc_report(). It
+ * will let you specify the depth and max_depth.
+ *
+ * @param[in]  ptr      The talloc chunk.
+ *
+ * @param[in]  depth    Internal parameter to control recursion. Call with 0.
+ *
+ * @param[in]  max_depth  Maximum recursion level.
+ *
+ * @param[in]  f        The file handle to print to.
+ */
+void talloc_report_depth_file(const void *ptr, int depth, int max_depth, FILE *f);
+
+/**
+ * @brief Print a summary report of all memory used by ptr.
+ *
+ * This provides a more detailed report than talloc_report(). It will
+ * recursively print the entire tree of memory referenced by the
+ * pointer. References in the tree are shown by giving the name of the
+ * pointer that is referenced.
+ *
+ * You can pass NULL for the pointer, in which case a report is printed
+ * for the top level memory context, but only if
+ * talloc_enable_leak_report() or talloc_enable_leak_report_full() has
+ * been called.
+ *
+ * @param[in]  ptr      The talloc chunk.
+ *
+ * @param[in]  f        The file handle to print to.
+ *
+ * Example:
+ * @code
+ *      unsigned int *a, *b;
+ *      a = talloc(NULL, unsigned int);
+ *      b = talloc(a, unsigned int);
+ *      fprintf(stderr, "Dumping memory tree for a:\n");
+ *      talloc_report_full(a, stderr);
+ * @endcode
+ *
+ * @see talloc_report()
+ */
+void talloc_report_full(const void *ptr, FILE *f);
+
+/**
+ * @brief Print a summary report of all memory used by ptr.
+ *
+ * This function prints a summary report of all memory used by ptr. One line of
+ * report is printed for each immediate child of ptr, showing the total memory
+ * and number of blocks used by that child.
+ *
+ * You can pass NULL for the pointer, in which case a report is printed
+ * for the top level memory context, but only if talloc_enable_leak_report()
+ * or talloc_enable_leak_report_full() has been called.
+ *
+ * @param[in]  ptr      The talloc chunk.
+ *
+ * @param[in]  f        The file handle to print to.
+ *
+ * Example:
+ * @code
+ *      unsigned int *a, *b;
+ *      a = talloc(NULL, unsigned int);
+ *      b = talloc(a, unsigned int);
+ *      fprintf(stderr, "Summary of memory tree for a:\n");
+ *      talloc_report(a, stderr);
+ * @endcode
+ *
+ * @see talloc_report_full()
+ */
+void talloc_report(const void *ptr, FILE *f);
+
+/**
+ * @brief Enable tracking the use of NULL memory contexts.
+ *
+ * This enables tracking of the NULL memory context without enabling leak
+ * reporting on exit. Useful for when you want to do your own leak
+ * reporting call via talloc_report_null_full();
+ */
+void talloc_enable_null_tracking(void);
+
+/**
+ * @brief Enable tracking the use of NULL memory contexts.
+ *
+ * This enables tracking of the NULL memory context without enabling leak
+ * reporting on exit. Useful for when you want to do your own leak
+ * reporting call via talloc_report_null_full();
+ */
+void talloc_enable_null_tracking_no_autofree(void);
+
+/**
+ * @brief Disable tracking of the NULL memory context.
+ *
+ * This disables tracking of the NULL memory context.
+ */
+void talloc_disable_null_tracking(void);
+
+/**
+ * @brief Enable leak report when a program exits.
+ *
+ * This enables calling of talloc_report(NULL, stderr) when the program
+ * exits. In Samba4 this is enabled by using the --leak-report command
+ * line option.
+ *
+ * For it to be useful, this function must be called before any other
+ * talloc function as it establishes a "null context" that acts as the
+ * top of the tree. If you don't call this function first then passing
+ * NULL to talloc_report() or talloc_report_full() won't give you the
+ * full tree printout.
+ *
+ * Here is a typical talloc report:
+ *
+ * @code
+ * talloc report on 'null_context' (total 267 bytes in 15 blocks)
+ *      libcli/auth/spnego_parse.c:55  contains     31 bytes in   2 blocks
+ *      libcli/auth/spnego_parse.c:55  contains     31 bytes in   2 blocks
+ *      iconv(UTF8,CP850)              contains     42 bytes in   2 blocks
+ *      libcli/auth/spnego_parse.c:55  contains     31 bytes in   2 blocks
+ *      iconv(CP850,UTF8)              contains     42 bytes in   2 blocks
+ *      iconv(UTF8,UTF-16LE)           contains     45 bytes in   2 blocks
+ *      iconv(UTF-16LE,UTF8)           contains     45 bytes in   2 blocks
+ * @endcode
+ */
+void talloc_enable_leak_report(void);
+
+/**
+ * @brief Enable full leak report when a program exits.
+ *
+ * This enables calling of talloc_report_full(NULL, stderr) when the
+ * program exits. In Samba4 this is enabled by using the
+ * --leak-report-full command line option.
+ *
+ * For it to be useful, this function must be called before any other
+ * talloc function as it establishes a "null context" that acts as the
+ * top of the tree. If you don't call this function first then passing
+ * NULL to talloc_report() or talloc_report_full() won't give you the
+ * full tree printout.
+ *
+ * Here is a typical full report:
+ *
+ * @code
+ * full talloc report on 'root' (total 18 bytes in 8 blocks)
+ *      p1                             contains     18 bytes in   7 blocks (ref 0)
+ *      r1                             contains     13 bytes in   2 blocks (ref 0)
+ *      reference to: p2
+ *      p2                             contains      1 bytes in   1 blocks (ref 1)
+ *      x3                             contains      1 bytes in   1 blocks (ref 0)
+ *      x2                             contains      1 bytes in   1 blocks (ref 0)
+ *      x1                             contains      1 bytes in   1 blocks (ref 0)
+ * @endcode
+ */
+void talloc_enable_leak_report_full(void);
+
+/**
+ * @brief Set a custom "abort" function that is called on serious error.
+ *
+ * The default "abort" function is <code>abort()</code>.
+ *
+ * The "abort" function is called when:
+ *
+ * <ul>
+ *  <li>talloc_get_type_abort() fails</li>
+ *  <li>the provided pointer is not a valid talloc context</li>
+ *  <li>when the context meta data are invalid</li>
+ *  <li>when access after free is detected</li>
+ * </ul>
+ *
+ * Example:
+ *
+ * @code
+ * void my_abort(const char *reason)
+ * {
+ *      fprintf(stderr, "talloc abort: %s\n", reason);
+ *      abort();
+ * }
+ *
+ *      talloc_set_abort_fn(my_abort);
+ * @endcode
+ *
+ * @param[in]  abort_fn      The new "abort" function.
+ *
+ * @see talloc_set_log_fn()
+ * @see talloc_get_type()
+ */
+void talloc_set_abort_fn(void (*abort_fn)(const char *reason));
+
+/**
+ * @brief Set a logging function.
+ *
+ * @param[in]  log_fn      The logging function.
+ *
+ * @see talloc_set_log_stderr()
+ * @see talloc_set_abort_fn()
+ */
+void talloc_set_log_fn(void (*log_fn)(const char *message));
+
+/**
+ * @brief Set stderr as the output for logs.
+ *
+ * @see talloc_set_log_fn()
+ * @see talloc_set_abort_fn()
+ */
+void talloc_set_log_stderr(void);
+
+/**
+ * @brief Set a max memory limit for the current context hierarchy
+ *	  This affects all children of this context and constrain any
+ *	  allocation in the hierarchy to never exceed the limit set.
+ *	  The limit can be removed by setting 0 (unlimited) as the
+ *	  max_size by calling the funciton again on the sam context.
+ *	  Memory limits can also be nested, meaning a hild can have
+ *	  a stricter memory limit than a parent.
+ *	  Memory limits are enforced only at memory allocation time.
+ *	  Stealing a context into a 'limited' hierarchy properly
+ *	  updates memory usage but does *not* cause failure if the
+ *	  move causes the new parent to exceed its limits. However
+ *	  any further allocation on that hierarchy will then fail.
+ *
+ * @param[in]	ctx		The talloc context to set the limit on
+ * @param[in]	max_size	The (new) max_size
+ */
+int talloc_set_memlimit(const void *ctx, size_t max_size);
+
+/* @} ******************************************************************/
+
+#if TALLOC_DEPRECATED
+#define talloc_zero_p(ctx, type) talloc_zero(ctx, type)
+#define talloc_p(ctx, type) talloc(ctx, type)
+#define talloc_array_p(ctx, type, count) talloc_array(ctx, type, count)
+#define talloc_realloc_p(ctx, p, type, count) talloc_realloc(ctx, p, type, count)
+#define talloc_destroy(ctx) talloc_free(ctx)
+#define talloc_append_string(c, s, a) (s?talloc_strdup_append(s,a):talloc_strdup(c, a))
+#endif
+
+#ifndef TALLOC_MAX_DEPTH
+#define TALLOC_MAX_DEPTH 10000
+#endif
+
+#ifdef __cplusplus
+} /* end of extern "C" */
+#endif
+
+#endif
diff --git a/ctdb/lib/talloc/talloc.i b/ctdb/lib/talloc/talloc.i
new file mode 100644
index 0000000000..a9afb97ed7
--- /dev/null
+++ b/ctdb/lib/talloc/talloc.i
@@ -0,0 +1,31 @@
+/* 
+   Unix SMB/CIFS implementation.
+   Copyright (C) Jelmer Vernooij <jelmer@samba.org> 2007
+   
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.
+*/
+
+/* Don't expose talloc contexts in Python code. Python does reference 
+   counting for us, so just create a new top-level talloc context.
+ */
+%typemap(in, numinputs=0, noblock=1) TALLOC_CTX * {
+    $1 = NULL;
+}
+
+%define %talloctype(TYPE)
+%nodefaultctor TYPE;
+%extend TYPE {
+    ~TYPE() { talloc_free($self); }
+}
+%enddef
diff --git a/ctdb/lib/talloc/talloc.pc.in b/ctdb/lib/talloc/talloc.pc.in
new file mode 100644
index 0000000000..437281a69a
--- /dev/null
+++ b/ctdb/lib/talloc/talloc.pc.in
@@ -0,0 +1,11 @@
+prefix=@prefix@
+exec_prefix=@exec_prefix@
+libdir=@libdir@
+includedir=@includedir@
+
+Name: talloc 
+Description: A hierarchical pool based memory system with destructors
+Version: @TALLOC_VERSION@
+Libs: @LIB_RPATH@ -L${libdir} -ltalloc
+Cflags: -I${includedir} 
+URL: http://talloc.samba.org/
diff --git a/ctdb/lib/talloc/talloc_guide.txt b/ctdb/lib/talloc/talloc_guide.txt
new file mode 100644
index 0000000000..16afc9b6b8
--- /dev/null
+++ b/ctdb/lib/talloc/talloc_guide.txt
@@ -0,0 +1,767 @@
+Using talloc in Samba4
+======================
+
+.. contents::
+
+Andrew Tridgell
+August 2009
+
+The most current version of this document is available at
+   http://samba.org/ftp/unpacked/talloc/talloc_guide.txt
+
+If you are used to the "old" talloc from Samba3 before 3.0.20 then please read
+this carefully, as talloc has changed a lot. With 3.0.20 (or 3.0.14?) the
+Samba4 talloc has been ported back to Samba3, so this guide applies to both.
+
+The new talloc is a hierarchical, reference counted memory pool system
+with destructors. Quite a mouthful really, but not too bad once you
+get used to it.
+
+Perhaps the biggest change from Samba3 is that there is no distinction
+between a "talloc context" and a "talloc pointer". Any pointer
+returned from talloc() is itself a valid talloc context. This means
+you can do this::
+
+  struct foo *X = talloc(mem_ctx, struct foo);
+  X->name = talloc_strdup(X, "foo");
+
+and the pointer X->name would be a "child" of the talloc context "X"
+which is itself a child of "mem_ctx". So if you do talloc_free(mem_ctx)
+then it is all destroyed, whereas if you do talloc_free(X) then just X
+and X->name are destroyed, and if you do talloc_free(X->name) then
+just the name element of X is destroyed.
+
+If you think about this, then what this effectively gives you is an
+n-ary tree, where you can free any part of the tree with
+talloc_free().
+
+If you find this confusing, then I suggest you run the testsuite to
+watch talloc in action. You may also like to add your own tests to
+testsuite.c to clarify how some particular situation is handled.
+
+
+Performance
+-----------
+
+All the additional features of talloc() over malloc() do come at a
+price. We have a simple performance test in Samba4 that measures
+talloc() versus malloc() performance, and it seems that talloc() is
+about 4% slower than malloc() on my x86 Debian Linux box. For Samba,
+the great reduction in code complexity that we get by using talloc
+makes this worthwhile, especially as the total overhead of
+talloc/malloc in Samba is already quite small.
+
+
+talloc API
+----------
+
+The following is a complete guide to the talloc API. Read it all at
+least twice.
+
+Multi-threading
+---------------
+
+talloc itself does not deal with threads. It is thread-safe (assuming  
+the underlying "malloc" is), as long as each thread uses different  
+memory contexts.
+If two threads use the same context then they need to synchronize in
+order to be safe. In particular:
+- when using talloc_enable_leak_report(), giving directly NULL as a  
+parent context implicitly refers to a hidden "null context" global  
+variable, so this should not be used in a multi-threaded environment  
+without proper synchronization ;
+- the context returned by talloc_autofree_context() is also global so  
+shouldn't be used by several threads simultaneously without  
+synchronization.
+
+talloc and shared objects
+-------------------------
+
+talloc can be used in shared objects. Special care needs to be taken
+to never use talloc_autofree_context() in code that might be loaded
+with dlopen() and unloaded with dlclose(), as talloc_autofree_context()
+internally uses atexit(3). Some platforms like modern Linux handles
+this fine, but for example FreeBSD does not deal well with dlopen()
+and atexit() used simultaneously: dlclose() does not clean up the list
+of atexit-handlers, so when the program exits the code that was
+registered from within talloc_autofree_context() is gone, the program
+crashes at exit.
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+(type *)talloc(const void *context, type);
+
+The talloc() macro is the core of the talloc library. It takes a
+memory context and a type, and returns a pointer to a new area of
+memory of the given type.
+
+The returned pointer is itself a talloc context, so you can use it as
+the context argument to more calls to talloc if you wish.
+
+The returned pointer is a "child" of the supplied context. This means
+that if you talloc_free() the context then the new child disappears as
+well. Alternatively you can free just the child.
+
+The context argument to talloc() can be NULL, in which case a new top
+level context is created. 
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void *talloc_size(const void *context, size_t size);
+
+The function talloc_size() should be used when you don't have a
+convenient type to pass to talloc(). Unlike talloc(), it is not type
+safe (as it returns a void *), so you are on your own for type checking.
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+(typeof(ptr)) talloc_ptrtype(const void *ctx, ptr);
+
+The talloc_ptrtype() macro should be used when you have a pointer and
+want to allocate memory to point at with this pointer. When compiling
+with gcc >= 3 it is typesafe. Note this is a wrapper of talloc_size()
+and talloc_get_name() will return the current location in the source file.
+and not the type.
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+int talloc_free(void *ptr);
+
+The talloc_free() function frees a piece of talloc memory, and all its
+children. You can call talloc_free() on any pointer returned by
+talloc().
+
+The return value of talloc_free() indicates success or failure, with 0
+returned for success and -1 for failure. A possible failure condition
+is if the pointer had a destructor attached to it and the destructor
+returned -1. See talloc_set_destructor() for details on
+destructors. Likewise, if "ptr" is NULL, then the function will make
+no modifications and returns -1.
+
+From version 2.0 and onwards, as a special case, talloc_free() is
+refused on pointers that have more than one parent associated, as talloc
+would have no way of knowing which parent should be removed. This is
+different from older versions in the sense that always the reference to
+the most recently established parent has been destroyed. Hence to free a
+pointer that has more than one parent please use talloc_unlink().
+
+To help you find problems in your code caused by this behaviour, if
+you do try and free a pointer with more than one parent then the
+talloc logging function will be called to give output like this:
+
+  ERROR: talloc_free with references at some_dir/source/foo.c:123
+  	reference at some_dir/source/other.c:325
+  	reference at some_dir/source/third.c:121
+
+Please see the documentation for talloc_set_log_fn() and
+talloc_set_log_stderr() for more information on talloc logging
+functions.
+
+talloc_free() operates recursively on its children.
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void talloc_free_children(void *ptr);
+
+The talloc_free_children() walks along the list of all children of a
+talloc context and talloc_free()s only the children, not the context
+itself.
+
+A NULL argument is handled as no-op.
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void *talloc_reference(const void *context, const void *ptr);
+
+The talloc_reference() function makes "context" an additional parent
+of "ptr".
+
+The return value of talloc_reference() is always the original pointer
+"ptr", unless talloc ran out of memory in creating the reference in
+which case it will return NULL (each additional reference consumes
+around 48 bytes of memory on intel x86 platforms).
+
+If "ptr" is NULL, then the function is a no-op, and simply returns NULL.
+
+After creating a reference you can free it in one of the following
+ways:
+
+  - you can talloc_free() any parent of the original pointer. That
+    will reduce the number of parents of this pointer by 1, and will
+    cause this pointer to be freed if it runs out of parents.
+
+  - you can talloc_free() the pointer itself if it has at maximum one
+    parent. This behaviour has been changed since the release of version
+    2.0. Further informations in the description of "talloc_free".
+
+For more control on which parent to remove, see talloc_unlink()
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+int talloc_unlink(const void *context, const void *ptr);
+
+The talloc_unlink() function removes a specific parent from ptr. The
+context passed must either be a context used in talloc_reference()
+with this pointer, or must be a direct parent of ptr. 
+
+Note that if the parent has already been removed using talloc_free()
+then this function will fail and will return -1.  Likewise, if "ptr"
+is NULL, then the function will make no modifications and return -1.
+
+You can just use talloc_free() instead of talloc_unlink() if there
+is at maximum one parent. This behaviour has been changed since the
+release of version 2.0. Further informations in the description of
+"talloc_free".
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void talloc_set_destructor(const void *ptr, int (*destructor)(void *));
+
+The function talloc_set_destructor() sets the "destructor" for the
+pointer "ptr". A destructor is a function that is called when the
+memory used by a pointer is about to be released. The destructor
+receives the pointer as an argument, and should return 0 for success
+and -1 for failure.
+
+The destructor can do anything it wants to, including freeing other
+pieces of memory. A common use for destructors is to clean up
+operating system resources (such as open file descriptors) contained
+in the structure the destructor is placed on.
+
+You can only place one destructor on a pointer. If you need more than
+one destructor then you can create a zero-length child of the pointer
+and place an additional destructor on that.
+
+To remove a destructor call talloc_set_destructor() with NULL for the
+destructor.
+
+If your destructor attempts to talloc_free() the pointer that it is
+the destructor for then talloc_free() will return -1 and the free will
+be ignored. This would be a pointless operation anyway, as the
+destructor is only called when the memory is just about to go away.
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+int talloc_increase_ref_count(const void *ptr);
+
+The talloc_increase_ref_count(ptr) function is exactly equivalent to:
+
+  talloc_reference(NULL, ptr);
+
+You can use either syntax, depending on which you think is clearer in
+your code.
+
+It returns 0 on success and -1 on failure.
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+size_t talloc_reference_count(const void *ptr);
+
+Return the number of references to the pointer.
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void talloc_set_name(const void *ptr, const char *fmt, ...);
+
+Each talloc pointer has a "name". The name is used principally for
+debugging purposes, although it is also possible to set and get the
+name on a pointer in as a way of "marking" pointers in your code.
+
+The main use for names on pointer is for "talloc reports". See
+talloc_report() and talloc_report_full() for details. Also see
+talloc_enable_leak_report() and talloc_enable_leak_report_full().
+
+The talloc_set_name() function allocates memory as a child of the
+pointer. It is logically equivalent to:
+  talloc_set_name_const(ptr, talloc_asprintf(ptr, fmt, ...));
+
+Note that multiple calls to talloc_set_name() will allocate more
+memory without releasing the name. All of the memory is released when
+the ptr is freed using talloc_free().
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void talloc_set_name_const(const void *ptr, const char *name);
+
+The function talloc_set_name_const() is just like talloc_set_name(),
+but it takes a string constant, and is much faster. It is extensively
+used by the "auto naming" macros, such as talloc_p().
+
+This function does not allocate any memory. It just copies the
+supplied pointer into the internal representation of the talloc
+ptr. This means you must not pass a name pointer to memory that will
+disappear before the ptr is freed with talloc_free().
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void *talloc_named(const void *context, size_t size, const char *fmt, ...);
+
+The talloc_named() function creates a named talloc pointer. It is
+equivalent to:
+
+   ptr = talloc_size(context, size);
+   talloc_set_name(ptr, fmt, ....);
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void *talloc_named_const(const void *context, size_t size, const char *name);
+
+This is equivalent to::
+
+   ptr = talloc_size(context, size);
+   talloc_set_name_const(ptr, name);
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+const char *talloc_get_name(const void *ptr);
+
+This returns the current name for the given talloc pointer. See
+talloc_set_name() for details.
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void *talloc_init(const char *fmt, ...);
+
+This function creates a zero length named talloc context as a top
+level context. It is equivalent to::
+
+  talloc_named(NULL, 0, fmt, ...);
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void *talloc_new(void *ctx);
+
+This is a utility macro that creates a new memory context hanging
+off an exiting context, automatically naming it "talloc_new: __location__"
+where __location__ is the source line it is called from. It is
+particularly useful for creating a new temporary working context.
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+(type *)talloc_realloc(const void *context, void *ptr, type, count);
+
+The talloc_realloc() macro changes the size of a talloc
+pointer. The "count" argument is the number of elements of type "type"
+that you want the resulting pointer to hold. 
+
+talloc_realloc() has the following equivalences::
+
+  talloc_realloc(context, NULL, type, 1) ==> talloc(context, type);
+  talloc_realloc(context, NULL, type, N) ==> talloc_array(context, type, N);
+  talloc_realloc(context, ptr, type, 0)  ==> talloc_free(ptr);
+
+The "context" argument is only used if "ptr" is NULL, otherwise it is
+ignored.
+
+talloc_realloc() returns the new pointer, or NULL on failure. The call
+will fail either due to a lack of memory, or because the pointer has
+more than one parent (see talloc_reference()).
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void *talloc_realloc_size(const void *context, void *ptr, size_t size);
+
+the talloc_realloc_size() function is useful when the type is not 
+known so the typesafe talloc_realloc() cannot be used.
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void *talloc_steal(const void *new_ctx, const void *ptr);
+
+The talloc_steal() function changes the parent context of a talloc
+pointer. It is typically used when the context that the pointer is
+currently a child of is going to be freed and you wish to keep the
+memory for a longer time. 
+
+The talloc_steal() function returns the pointer that you pass it. It
+does not have any failure modes.
+
+NOTE: It is possible to produce loops in the parent/child relationship
+if you are not careful with talloc_steal(). No guarantees are provided
+as to your sanity or the safety of your data if you do this.
+
+talloc_steal (new_ctx, NULL) will return NULL with no sideeffects.
+
+Note that if you try and call talloc_steal() on a pointer that has
+more than one parent then the result is ambiguous. Talloc will choose
+to remove the parent that is currently indicated by talloc_parent()
+and replace it with the chosen parent. You will also get a message
+like this via the talloc logging functions:
+
+  WARNING: talloc_steal with references at some_dir/source/foo.c:123
+  	reference at some_dir/source/other.c:325
+  	reference at some_dir/source/third.c:121
+
+To unambiguously change the parent of a pointer please see the
+function talloc_reparent(). See the talloc_set_log_fn() documentation
+for more information on talloc logging.
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void *talloc_reparent(const void *old_parent, const void *new_parent, const void *ptr);
+
+The talloc_reparent() function changes the parent context of a talloc
+pointer. It is typically used when the context that the pointer is
+currently a child of is going to be freed and you wish to keep the
+memory for a longer time.
+
+The talloc_reparent() function returns the pointer that you pass it. It
+does not have any failure modes.
+
+The difference between talloc_reparent() and talloc_steal() is that
+talloc_reparent() can specify which parent you wish to change. This is
+useful when a pointer has multiple parents via references.
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void *talloc_parent(const void *ptr);
+
+The talloc_parent() function returns the current talloc parent. This
+is usually the pointer under which this memory was originally created,
+but it may have changed due to a talloc_steal() or talloc_reparent()
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+size_t talloc_total_size(const void *ptr);
+
+The talloc_total_size() function returns the total size in bytes used
+by this pointer and all child pointers. Mostly useful for debugging.
+
+Passing NULL is allowed, but it will only give a meaningful result if
+talloc_enable_leak_report() or talloc_enable_leak_report_full() has
+been called.
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+size_t talloc_total_blocks(const void *ptr);
+
+The talloc_total_blocks() function returns the total memory block
+count used by this pointer and all child pointers. Mostly useful for
+debugging.
+
+Passing NULL is allowed, but it will only give a meaningful result if
+talloc_enable_leak_report() or talloc_enable_leak_report_full() has
+been called.
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void talloc_report_depth_cb(const void *ptr, int depth, int max_depth,
+			    void (*callback)(const void *ptr,
+			    		     int depth, int max_depth,
+					     int is_ref,
+					     void *priv),
+			    void *priv);
+
+This provides a more flexible reports than talloc_report(). It
+will recursively call the callback for the entire tree of memory
+referenced by the pointer. References in the tree are passed with
+is_ref = 1 and the pointer that is referenced.
+
+You can pass NULL for the pointer, in which case a report is
+printed for the top level memory context, but only if
+talloc_enable_leak_report() or talloc_enable_leak_report_full()
+has been called.
+
+The recursion is stopped when depth >= max_depth.
+max_depth = -1 means only stop at leaf nodes.
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void talloc_report_depth_file(const void *ptr, int depth, int max_depth, FILE *f);
+
+This provides a more flexible reports than talloc_report(). It
+will let you specify the depth and max_depth.
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void talloc_report(const void *ptr, FILE *f);
+
+The talloc_report() function prints a summary report of all memory
+used by ptr. One line of report is printed for each immediate child of
+ptr, showing the total memory and number of blocks used by that child.
+
+You can pass NULL for the pointer, in which case a report is printed
+for the top level memory context, but only if
+talloc_enable_leak_report() or talloc_enable_leak_report_full() has
+been called.
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void talloc_report_full(const void *ptr, FILE *f);
+
+This provides a more detailed report than talloc_report(). It will
+recursively print the entire tree of memory referenced by the
+pointer. References in the tree are shown by giving the name of the
+pointer that is referenced.
+
+You can pass NULL for the pointer, in which case a report is printed
+for the top level memory context, but only if
+talloc_enable_leak_report() or talloc_enable_leak_report_full() has
+been called.
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void talloc_enable_leak_report(void);
+
+This enables calling of talloc_report(NULL, stderr) when the program
+exits. In Samba4 this is enabled by using the --leak-report command
+line option.
+
+For it to be useful, this function must be called before any other
+talloc function as it establishes a "null context" that acts as the
+top of the tree. If you don't call this function first then passing
+NULL to talloc_report() or talloc_report_full() won't give you the
+full tree printout.
+
+Here is a typical talloc report:
+
+talloc report on 'null_context' (total 267 bytes in 15 blocks)
+        libcli/auth/spnego_parse.c:55  contains     31 bytes in   2 blocks
+        libcli/auth/spnego_parse.c:55  contains     31 bytes in   2 blocks
+        iconv(UTF8,CP850)              contains     42 bytes in   2 blocks
+        libcli/auth/spnego_parse.c:55  contains     31 bytes in   2 blocks
+        iconv(CP850,UTF8)              contains     42 bytes in   2 blocks
+        iconv(UTF8,UTF-16LE)           contains     45 bytes in   2 blocks
+        iconv(UTF-16LE,UTF8)           contains     45 bytes in   2 blocks
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void talloc_enable_leak_report_full(void);
+
+This enables calling of talloc_report_full(NULL, stderr) when the
+program exits. In Samba4 this is enabled by using the
+--leak-report-full command line option.
+
+For it to be useful, this function must be called before any other
+talloc function as it establishes a "null context" that acts as the
+top of the tree. If you don't call this function first then passing
+NULL to talloc_report() or talloc_report_full() won't give you the
+full tree printout.
+
+Here is a typical full report:
+
+full talloc report on 'root' (total 18 bytes in 8 blocks)
+    p1                             contains     18 bytes in   7 blocks (ref 0)
+        r1                             contains     13 bytes in   2 blocks (ref 0)
+            reference to: p2
+        p2                             contains      1 bytes in   1 blocks (ref 1)
+        x3                             contains      1 bytes in   1 blocks (ref 0)
+        x2                             contains      1 bytes in   1 blocks (ref 0)
+        x1                             contains      1 bytes in   1 blocks (ref 0)
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void talloc_enable_null_tracking(void);
+
+This enables tracking of the NULL memory context without enabling leak
+reporting on exit. Useful for when you want to do your own leak
+reporting call via talloc_report_null_full();
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void talloc_disable_null_tracking(void);
+
+This disables tracking of the NULL memory context.
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+(type *)talloc_zero(const void *ctx, type);
+
+The talloc_zero() macro is equivalent to::
+
+  ptr = talloc(ctx, type);
+  if (ptr) memset(ptr, 0, sizeof(type));
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void *talloc_zero_size(const void *ctx, size_t size)
+
+The talloc_zero_size() function is useful when you don't have a known type
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void *talloc_memdup(const void *ctx, const void *p, size_t size);
+
+The talloc_memdup() function is equivalent to::
+
+  ptr = talloc_size(ctx, size);
+  if (ptr) memcpy(ptr, p, size);
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+char *talloc_strdup(const void *ctx, const char *p);
+
+The talloc_strdup() function is equivalent to::
+
+  ptr = talloc_size(ctx, strlen(p)+1);
+  if (ptr) memcpy(ptr, p, strlen(p)+1);
+
+This functions sets the name of the new pointer to the passed
+string. This is equivalent to::
+
+   talloc_set_name_const(ptr, ptr)
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+char *talloc_strndup(const void *t, const char *p, size_t n);
+
+The talloc_strndup() function is the talloc equivalent of the C
+library function strndup()
+
+This functions sets the name of the new pointer to the passed
+string. This is equivalent to:
+   talloc_set_name_const(ptr, ptr)
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+char *talloc_append_string(const void *t, char *orig, const char *append);
+
+The talloc_append_string() function appends the given formatted
+string to the given string.
+
+This function sets the name of the new pointer to the new
+string. This is equivalent to::
+
+   talloc_set_name_const(ptr, ptr)
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+char *talloc_vasprintf(const void *t, const char *fmt, va_list ap);
+
+The talloc_vasprintf() function is the talloc equivalent of the C
+library function vasprintf()
+
+This functions sets the name of the new pointer to the new
+string. This is equivalent to::
+
+   talloc_set_name_const(ptr, ptr)
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+char *talloc_asprintf(const void *t, const char *fmt, ...);
+
+The talloc_asprintf() function is the talloc equivalent of the C
+library function asprintf()
+
+This functions sets the name of the new pointer to the new
+string. This is equivalent to::
+
+   talloc_set_name_const(ptr, ptr)
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+char *talloc_asprintf_append(char *s, const char *fmt, ...);
+
+The talloc_asprintf_append() function appends the given formatted
+string to the given string.
+Use this variant when the string in the current talloc buffer may
+have been truncated in length.
+
+This functions sets the name of the new pointer to the new
+string. This is equivalent to::
+
+   talloc_set_name_const(ptr, ptr)
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+char *talloc_asprintf_append_buffer(char *s, const char *fmt, ...);
+
+The talloc_asprintf_append() function appends the given formatted 
+string to the end of the currently allocated talloc buffer.
+Use this variant when the string in the current talloc buffer has
+not been changed.
+
+This functions sets the name of the new pointer to the new
+string. This is equivalent to::
+
+   talloc_set_name_const(ptr, ptr)
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+((type *)talloc_array(const void *ctx, type, unsigned int count);
+
+The talloc_array() macro is equivalent to::
+
+  (type *)talloc_size(ctx, sizeof(type) * count);
+
+except that it provides integer overflow protection for the multiply,
+returning NULL if the multiply overflows.
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void *talloc_array_size(const void *ctx, size_t size, unsigned int count);
+
+The talloc_array_size() function is useful when the type is not
+known. It operates in the same way as talloc_array(), but takes a size
+instead of a type.
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+(typeof(ptr)) talloc_array_ptrtype(const void *ctx, ptr, unsigned int count);
+
+The talloc_ptrtype() macro should be used when you have a pointer to an array
+and want to allocate memory of an array to point at with this pointer. When compiling
+with gcc >= 3 it is typesafe. Note this is a wrapper of talloc_array_size()
+and talloc_get_name() will return the current location in the source file.
+and not the type.
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void *talloc_realloc_fn(const void *ctx, void *ptr, size_t size);
+
+This is a non-macro version of talloc_realloc(), which is useful 
+as libraries sometimes want a ralloc function pointer. A realloc()
+implementation encapsulates the functionality of malloc(), free() and
+realloc() in one call, which is why it is useful to be able to pass
+around a single function pointer.
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void *talloc_autofree_context(void);
+
+This is a handy utility function that returns a talloc context
+which will be automatically freed on program exit. This can be used
+to reduce the noise in memory leak reports.
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void *talloc_check_name(const void *ptr, const char *name);
+
+This function checks if a pointer has the specified name. If it does
+then the pointer is returned. It it doesn't then NULL is returned.
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+(type *)talloc_get_type(const void *ptr, type);
+
+This macro allows you to do type checking on talloc pointers. It is
+particularly useful for void* private pointers. It is equivalent to
+this::
+
+   (type *)talloc_check_name(ptr, #type)
+
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+talloc_set_type(const void *ptr, type);
+
+This macro allows you to force the name of a pointer to be of a
+particular type. This can be used in conjunction with
+talloc_get_type() to do type checking on void* pointers.
+
+It is equivalent to this::
+
+   talloc_set_name_const(ptr, #type)
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+talloc_get_size(const void *ctx);
+
+This function lets you know the amount of memory allocated so far by
+this context. It does NOT account for subcontext memory.
+This can be used to calculate the size of an array.
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void *talloc_find_parent_byname(const void *ctx, const char *name);
+
+Find a parent memory context of the current context that has the given
+name. This can be very useful in complex programs where it may be
+difficult to pass all information down to the level you need, but you
+know the structure you want is a parent of another context.
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+(type *)talloc_find_parent_bytype(ctx, type);
+
+Like talloc_find_parent_byname() but takes a type, making it typesafe.
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void talloc_set_log_fn(void (*log_fn)(const char *message));
+
+This function sets a logging function that talloc will use for
+warnings and errors. By default talloc will not print any warnings or
+errors.
+
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+void talloc_set_log_stderr(void)
+
+This sets the talloc log function to write log messages to stderr.
diff --git a/ctdb/lib/talloc/talloc_testsuite.h b/ctdb/lib/talloc/talloc_testsuite.h
new file mode 100644
index 0000000000..acb9701041
--- /dev/null
+++ b/ctdb/lib/talloc/talloc_testsuite.h
@@ -0,0 +1,7 @@
+#ifndef __LIB_TALLOC_TALLOC_TESTSUITE_H__
+#define __LIB_TALLOC_TALLOC_TESTSUITE_H__
+
+struct torture_context;
+bool torture_local_talloc(struct torture_context *tctx);
+
+#endif
diff --git a/ctdb/lib/talloc/testsuite.c b/ctdb/lib/talloc/testsuite.c
new file mode 100644
index 0000000000..d456cbb0c2
--- /dev/null
+++ b/ctdb/lib/talloc/testsuite.c
@@ -0,0 +1,1602 @@
+/* 
+   Unix SMB/CIFS implementation.
+
+   local testing of talloc routines.
+
+   Copyright (C) Andrew Tridgell 2004
+   
+     ** NOTE! The following LGPL license applies to the talloc
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "replace.h"
+#include "system/time.h"
+#include <talloc.h>
+
+#include "talloc_testsuite.h"
+
+static struct timeval timeval_current(void)
+{
+	struct timeval tv;
+	gettimeofday(&tv, NULL);
+	return tv;
+}
+
+static double timeval_elapsed(struct timeval *tv)
+{
+	struct timeval tv2 = timeval_current();
+	return (tv2.tv_sec - tv->tv_sec) + 
+	       (tv2.tv_usec - tv->tv_usec)*1.0e-6;
+}
+
+#define torture_assert(test, expr, str) if (!(expr)) { \
+	printf("failure: %s [\n%s: Expression %s failed: %s\n]\n", \
+		test, __location__, #expr, str); \
+	return false; \
+}
+
+#define torture_assert_str_equal(test, arg1, arg2, desc) \
+	if (arg1 == NULL && arg2 == NULL) {				\
+	} else if (strcmp(arg1, arg2)) {			\
+		printf("failure: %s [\n%s: Expected %s, got %s: %s\n]\n", \
+		   test, __location__, arg1, arg2, desc); \
+		return false; \
+	}
+
+#define CHECK_SIZE(test, ptr, tsize) do { \
+	if (talloc_total_size(ptr) != (tsize)) { \
+		printf("failed: %s [\n%s: wrong '%s' tree size: got %u  expected %u\n]\n", \
+		       test, __location__, #ptr, \
+		       (unsigned)talloc_total_size(ptr), \
+		       (unsigned)tsize); \
+		talloc_report_full(ptr, stdout); \
+		return false; \
+	} \
+} while (0)
+
+#define CHECK_BLOCKS(test, ptr, tblocks) do { \
+	if (talloc_total_blocks(ptr) != (tblocks)) { \
+		printf("failed: %s [\n%s: wrong '%s' tree blocks: got %u  expected %u\n]\n", \
+		       test, __location__, #ptr, \
+		       (unsigned)talloc_total_blocks(ptr), \
+		       (unsigned)tblocks); \
+		talloc_report_full(ptr, stdout); \
+		return false; \
+	} \
+} while (0)
+
+#define CHECK_PARENT(test, ptr, parent) do { \
+	if (talloc_parent(ptr) != (parent)) { \
+		printf("failed: %s [\n%s: '%s' has wrong parent: got %p  expected %p\n]\n", \
+		       test, __location__, #ptr, \
+		       talloc_parent(ptr), \
+		       (parent)); \
+		talloc_report_full(ptr, stdout); \
+		talloc_report_full(parent, stdout); \
+		talloc_report_full(NULL, stdout); \
+		return false; \
+	} \
+} while (0)
+
+static unsigned int test_abort_count;
+
+#if 0
+static void test_abort_fn(const char *reason)
+{
+	printf("# test_abort_fn(%s)\n", reason);
+	test_abort_count++;
+}
+
+static void test_abort_start(void)
+{
+	test_abort_count = 0;
+	talloc_set_abort_fn(test_abort_fn);
+}
+#endif
+
+static void test_abort_stop(void)
+{
+	test_abort_count = 0;
+	talloc_set_abort_fn(NULL);
+}
+
+static void test_log_stdout(const char *message)
+{
+	fprintf(stdout, "%s", message);
+}
+
+/*
+  test references 
+*/
+static bool test_ref1(void)
+{
+	void *root, *p1, *p2, *ref, *r1;
+
+	printf("test: ref1\n# SINGLE REFERENCE FREE\n");
+
+	root = talloc_named_const(NULL, 0, "root");
+	p1 = talloc_named_const(root, 1, "p1");
+	p2 = talloc_named_const(p1, 1, "p2");
+	talloc_named_const(p1, 1, "x1");
+	talloc_named_const(p1, 2, "x2");
+	talloc_named_const(p1, 3, "x3");
+
+	r1 = talloc_named_const(root, 1, "r1");	
+	ref = talloc_reference(r1, p2);
+	talloc_report_full(root, stderr);
+
+	CHECK_BLOCKS("ref1", p1, 5);
+	CHECK_BLOCKS("ref1", p2, 1);
+	CHECK_BLOCKS("ref1", r1, 2);
+
+	fprintf(stderr, "Freeing p2\n");
+	talloc_unlink(r1, p2);
+	talloc_report_full(root, stderr);
+
+	CHECK_BLOCKS("ref1", p1, 5);
+	CHECK_BLOCKS("ref1", p2, 1);
+	CHECK_BLOCKS("ref1", r1, 1);
+
+	fprintf(stderr, "Freeing p1\n");
+	talloc_free(p1);
+	talloc_report_full(root, stderr);
+
+	CHECK_BLOCKS("ref1", r1, 1);
+
+	fprintf(stderr, "Freeing r1\n");
+	talloc_free(r1);
+	talloc_report_full(NULL, stderr);
+
+	fprintf(stderr, "Testing NULL\n");
+	if (talloc_reference(root, NULL)) {
+		return false;
+	}
+
+	CHECK_BLOCKS("ref1", root, 1);
+
+	CHECK_SIZE("ref1", root, 0);
+
+	talloc_free(root);
+	printf("success: ref1\n");
+	return true;
+}
+
+/*
+  test references 
+*/
+static bool test_ref2(void)
+{
+	void *root, *p1, *p2, *ref, *r1;
+
+	printf("test: ref2\n# DOUBLE REFERENCE FREE\n");
+	root = talloc_named_const(NULL, 0, "root");
+	p1 = talloc_named_const(root, 1, "p1");
+	talloc_named_const(p1, 1, "x1");
+	talloc_named_const(p1, 1, "x2");
+	talloc_named_const(p1, 1, "x3");
+	p2 = talloc_named_const(p1, 1, "p2");
+
+	r1 = talloc_named_const(root, 1, "r1");	
+	ref = talloc_reference(r1, p2);
+	talloc_report_full(root, stderr);
+
+	CHECK_BLOCKS("ref2", p1, 5);
+	CHECK_BLOCKS("ref2", p2, 1);
+	CHECK_BLOCKS("ref2", r1, 2);
+
+	fprintf(stderr, "Freeing ref\n");
+	talloc_unlink(r1, ref);
+	talloc_report_full(root, stderr);
+
+	CHECK_BLOCKS("ref2", p1, 5);
+	CHECK_BLOCKS("ref2", p2, 1);
+	CHECK_BLOCKS("ref2", r1, 1);
+
+	fprintf(stderr, "Freeing p2\n");
+	talloc_free(p2);
+	talloc_report_full(root, stderr);
+
+	CHECK_BLOCKS("ref2", p1, 4);
+	CHECK_BLOCKS("ref2", r1, 1);
+
+	fprintf(stderr, "Freeing p1\n");
+	talloc_free(p1);
+	talloc_report_full(root, stderr);
+
+	CHECK_BLOCKS("ref2", r1, 1);
+
+	fprintf(stderr, "Freeing r1\n");
+	talloc_free(r1);
+	talloc_report_full(root, stderr);
+
+	CHECK_SIZE("ref2", root, 0);
+
+	talloc_free(root);
+	printf("success: ref2\n");
+	return true;
+}
+
+/*
+  test references 
+*/
+static bool test_ref3(void)
+{
+	void *root, *p1, *p2, *ref, *r1;
+
+	printf("test: ref3\n# PARENT REFERENCE FREE\n");
+
+	root = talloc_named_const(NULL, 0, "root");
+	p1 = talloc_named_const(root, 1, "p1");
+	p2 = talloc_named_const(root, 1, "p2");
+	r1 = talloc_named_const(p1, 1, "r1");
+	ref = talloc_reference(p2, r1);
+	talloc_report_full(root, stderr);
+
+	CHECK_BLOCKS("ref3", p1, 2);
+	CHECK_BLOCKS("ref3", p2, 2);
+	CHECK_BLOCKS("ref3", r1, 1);
+
+	fprintf(stderr, "Freeing p1\n");
+	talloc_free(p1);
+	talloc_report_full(root, stderr);
+
+	CHECK_BLOCKS("ref3", p2, 2);
+	CHECK_BLOCKS("ref3", r1, 1);
+
+	fprintf(stderr, "Freeing p2\n");
+	talloc_free(p2);
+	talloc_report_full(root, stderr);
+
+	CHECK_SIZE("ref3", root, 0);
+
+	talloc_free(root);
+
+	printf("success: ref3\n");
+	return true;
+}
+
+/*
+  test references 
+*/
+static bool test_ref4(void)
+{
+	void *root, *p1, *p2, *ref, *r1;
+
+	printf("test: ref4\n# REFERRER REFERENCE FREE\n");
+
+	root = talloc_named_const(NULL, 0, "root");
+	p1 = talloc_named_const(root, 1, "p1");
+	talloc_named_const(p1, 1, "x1");
+	talloc_named_const(p1, 1, "x2");
+	talloc_named_const(p1, 1, "x3");
+	p2 = talloc_named_const(p1, 1, "p2");
+
+	r1 = talloc_named_const(root, 1, "r1");	
+	ref = talloc_reference(r1, p2);
+	talloc_report_full(root, stderr);
+
+	CHECK_BLOCKS("ref4", p1, 5);
+	CHECK_BLOCKS("ref4", p2, 1);
+	CHECK_BLOCKS("ref4", r1, 2);
+
+	fprintf(stderr, "Freeing r1\n");
+	talloc_free(r1);
+	talloc_report_full(root, stderr);
+
+	CHECK_BLOCKS("ref4", p1, 5);
+	CHECK_BLOCKS("ref4", p2, 1);
+
+	fprintf(stderr, "Freeing p2\n");
+	talloc_free(p2);
+	talloc_report_full(root, stderr);
+
+	CHECK_BLOCKS("ref4", p1, 4);
+
+	fprintf(stderr, "Freeing p1\n");
+	talloc_free(p1);
+	talloc_report_full(root, stderr);
+
+	CHECK_SIZE("ref4", root, 0);
+
+	talloc_free(root);
+
+	printf("success: ref4\n");
+	return true;
+}
+
+
+/*
+  test references 
+*/
+static bool test_unlink1(void)
+{
+	void *root, *p1, *p2, *ref, *r1;
+
+	printf("test: unlink\n# UNLINK\n");
+
+	root = talloc_named_const(NULL, 0, "root");
+	p1 = talloc_named_const(root, 1, "p1");
+	talloc_named_const(p1, 1, "x1");
+	talloc_named_const(p1, 1, "x2");
+	talloc_named_const(p1, 1, "x3");
+	p2 = talloc_named_const(p1, 1, "p2");
+
+	r1 = talloc_named_const(p1, 1, "r1");	
+	ref = talloc_reference(r1, p2);
+	talloc_report_full(root, stderr);
+
+	CHECK_BLOCKS("unlink", p1, 7);
+	CHECK_BLOCKS("unlink", p2, 1);
+	CHECK_BLOCKS("unlink", r1, 2);
+
+	fprintf(stderr, "Unreferencing r1\n");
+	talloc_unlink(r1, p2);
+	talloc_report_full(root, stderr);
+
+	CHECK_BLOCKS("unlink", p1, 6);
+	CHECK_BLOCKS("unlink", p2, 1);
+	CHECK_BLOCKS("unlink", r1, 1);
+
+	fprintf(stderr, "Freeing p1\n");
+	talloc_free(p1);
+	talloc_report_full(root, stderr);
+
+	CHECK_SIZE("unlink", root, 0);
+
+	talloc_free(root);
+
+	printf("success: unlink\n");
+	return true;
+}
+
+static int fail_destructor(void *ptr)
+{
+	return -1;
+}
+
+/*
+  miscellaneous tests to try to get a higher test coverage percentage
+*/
+static bool test_misc(void)
+{
+	void *root, *p1;
+	char *p2;
+	double *d;
+	const char *name;
+
+	printf("test: misc\n# MISCELLANEOUS\n");
+
+	root = talloc_new(NULL);
+
+	p1 = talloc_size(root, 0x7fffffff);
+	torture_assert("misc", !p1, "failed: large talloc allowed\n");
+
+	p1 = talloc_strdup(root, "foo");
+	talloc_increase_ref_count(p1);
+	talloc_increase_ref_count(p1);
+	talloc_increase_ref_count(p1);
+	CHECK_BLOCKS("misc", p1, 1);
+	CHECK_BLOCKS("misc", root, 2);
+	talloc_unlink(NULL, p1);
+	CHECK_BLOCKS("misc", p1, 1);
+	CHECK_BLOCKS("misc", root, 2);
+	talloc_unlink(NULL, p1);
+	CHECK_BLOCKS("misc", p1, 1);
+	CHECK_BLOCKS("misc", root, 2);
+	p2 = talloc_strdup(p1, "foo");
+	torture_assert("misc", talloc_unlink(root, p2) == -1,
+				   "failed: talloc_unlink() of non-reference context should return -1\n");
+	torture_assert("misc", talloc_unlink(p1, p2) == 0,
+		"failed: talloc_unlink() of parent should succeed\n");
+	talloc_unlink(NULL, p1);
+	CHECK_BLOCKS("misc", p1, 1);
+	CHECK_BLOCKS("misc", root, 2);
+
+	name = talloc_set_name(p1, "my name is %s", "foo");
+	torture_assert_str_equal("misc", talloc_get_name(p1), "my name is foo",
+		"failed: wrong name after talloc_set_name(my name is foo)");
+	CHECK_BLOCKS("misc", p1, 2);
+	CHECK_BLOCKS("misc", root, 3);
+
+	talloc_set_name_const(p1, NULL);
+	torture_assert_str_equal ("misc", talloc_get_name(p1), "UNNAMED",
+		"failed: wrong name after talloc_set_name(NULL)");
+	CHECK_BLOCKS("misc", p1, 2);
+	CHECK_BLOCKS("misc", root, 3);
+
+	torture_assert("misc", talloc_free(NULL) == -1, 
+				   "talloc_free(NULL) should give -1\n");
+
+	talloc_set_destructor(p1, fail_destructor);
+	torture_assert("misc", talloc_free(p1) == -1, 
+		"Failed destructor should cause talloc_free to fail\n");
+	talloc_set_destructor(p1, NULL);
+
+	talloc_report(root, stderr);
+
+
+	p2 = (char *)talloc_zero_size(p1, 20);
+	torture_assert("misc", p2[19] == 0, "Failed to give zero memory\n");
+	talloc_free(p2);
+
+	torture_assert("misc", talloc_strdup(root, NULL) == NULL,
+		"failed: strdup on NULL should give NULL\n");
+
+	p2 = talloc_strndup(p1, "foo", 2);
+	torture_assert("misc", strcmp("fo", p2) == 0, 
+				   "strndup doesn't work\n");
+	p2 = talloc_asprintf_append_buffer(p2, "o%c", 'd');
+	torture_assert("misc", strcmp("food", p2) == 0, 
+				   "talloc_asprintf_append_buffer doesn't work\n");
+	CHECK_BLOCKS("misc", p2, 1);
+	CHECK_BLOCKS("misc", p1, 3);
+
+	p2 = talloc_asprintf_append_buffer(NULL, "hello %s", "world");
+	torture_assert("misc", strcmp("hello world", p2) == 0,
+		"talloc_asprintf_append_buffer doesn't work\n");
+	CHECK_BLOCKS("misc", p2, 1);
+	CHECK_BLOCKS("misc", p1, 3);
+	talloc_free(p2);
+
+	d = talloc_array(p1, double, 0x20000000);
+	torture_assert("misc", !d, "failed: integer overflow not detected\n");
+
+	d = talloc_realloc(p1, d, double, 0x20000000);
+	torture_assert("misc", !d, "failed: integer overflow not detected\n");
+
+	talloc_free(p1);
+	CHECK_BLOCKS("misc", root, 1);
+
+	p1 = talloc_named(root, 100, "%d bytes", 100);
+	CHECK_BLOCKS("misc", p1, 2);
+	CHECK_BLOCKS("misc", root, 3);
+	talloc_unlink(root, p1);
+
+	p1 = talloc_init("%d bytes", 200);
+	p2 = talloc_asprintf(p1, "my test '%s'", "string");
+	torture_assert_str_equal("misc", p2, "my test 'string'",
+		"failed: talloc_asprintf(\"my test '%%s'\", \"string\") gave: \"%s\"");
+	CHECK_BLOCKS("misc", p1, 3);
+	CHECK_SIZE("misc", p2, 17);
+	CHECK_BLOCKS("misc", root, 1);
+	talloc_unlink(NULL, p1);
+
+	p1 = talloc_named_const(root, 10, "p1");
+	p2 = (char *)talloc_named_const(root, 20, "p2");
+	(void)talloc_reference(p1, p2);
+	talloc_report_full(root, stderr);
+	talloc_unlink(root, p2);
+	talloc_report_full(root, stderr);
+	CHECK_BLOCKS("misc", p2, 1);
+	CHECK_BLOCKS("misc", p1, 2);
+	CHECK_BLOCKS("misc", root, 3);
+	talloc_unlink(p1, p2);
+	talloc_unlink(root, p1);
+
+	p1 = talloc_named_const(root, 10, "p1");
+	p2 = (char *)talloc_named_const(root, 20, "p2");
+	(void)talloc_reference(NULL, p2);
+	talloc_report_full(root, stderr);
+	talloc_unlink(root, p2);
+	talloc_report_full(root, stderr);
+	CHECK_BLOCKS("misc", p2, 1);
+	CHECK_BLOCKS("misc", p1, 1);
+	CHECK_BLOCKS("misc", root, 2);
+	talloc_unlink(NULL, p2);
+	talloc_unlink(root, p1);
+
+	/* Test that talloc_unlink is a no-op */
+
+	torture_assert("misc", talloc_unlink(root, NULL) == -1,
+		"failed: talloc_unlink(root, NULL) == -1\n");
+
+	talloc_report(root, stderr);
+	talloc_report(NULL, stderr);
+
+	CHECK_SIZE("misc", root, 0);
+
+	talloc_free(root);
+
+	CHECK_SIZE("misc", NULL, 0);
+
+	talloc_enable_null_tracking_no_autofree();
+	talloc_enable_leak_report();
+	talloc_enable_leak_report_full();
+
+	printf("success: misc\n");
+
+	return true;
+}
+
+
+/*
+  test realloc
+*/
+static bool test_realloc(void)
+{
+	void *root, *p1, *p2;
+
+	printf("test: realloc\n# REALLOC\n");
+
+	root = talloc_new(NULL);
+
+	p1 = talloc_size(root, 10);
+	CHECK_SIZE("realloc", p1, 10);
+
+	p1 = talloc_realloc_size(NULL, p1, 20);
+	CHECK_SIZE("realloc", p1, 20);
+
+	talloc_new(p1);
+
+	p2 = talloc_realloc_size(p1, NULL, 30);
+
+	talloc_new(p1);
+
+	p2 = talloc_realloc_size(p1, p2, 40);
+
+	CHECK_SIZE("realloc", p2, 40);
+	CHECK_SIZE("realloc", root, 60);
+	CHECK_BLOCKS("realloc", p1, 4);
+
+	p1 = talloc_realloc_size(NULL, p1, 20);
+	CHECK_SIZE("realloc", p1, 60);
+
+	talloc_increase_ref_count(p2);
+	torture_assert("realloc", talloc_realloc_size(NULL, p2, 5) == NULL,
+		"failed: talloc_realloc() on a referenced pointer should fail\n");
+	CHECK_BLOCKS("realloc", p1, 4);
+
+	talloc_realloc_size(NULL, p2, 0);
+	talloc_realloc_size(NULL, p2, 0);
+	CHECK_BLOCKS("realloc", p1, 4);
+	talloc_realloc_size(p1, p2, 0);
+	CHECK_BLOCKS("realloc", p1, 3);
+
+	torture_assert("realloc", talloc_realloc_size(NULL, p1, 0x7fffffff) == NULL,
+		"failed: oversize talloc should fail\n");
+
+	talloc_realloc_size(NULL, p1, 0);
+	CHECK_BLOCKS("realloc", root, 4);
+	talloc_realloc_size(root, p1, 0);
+	CHECK_BLOCKS("realloc", root, 1);
+
+	CHECK_SIZE("realloc", root, 0);
+
+	talloc_free(root);
+
+	printf("success: realloc\n");
+
+	return true;
+}
+
+/*
+  test realloc with a child
+*/
+static bool test_realloc_child(void)
+{
+	void *root;
+	struct el2 {
+		const char *name;
+	} *el2;	
+	struct el1 {
+		int count;
+		struct el2 **list, **list2, **list3;
+	} *el1;
+
+	printf("test: REALLOC WITH CHILD\n");
+
+	root = talloc_new(NULL);
+
+	el1 = talloc(root, struct el1);
+	el1->list = talloc(el1, struct el2 *);
+	el1->list[0] = talloc(el1->list, struct el2);
+	el1->list[0]->name = talloc_strdup(el1->list[0], "testing");
+
+	el1->list2 = talloc(el1, struct el2 *);
+	el1->list2[0] = talloc(el1->list2, struct el2);
+	el1->list2[0]->name = talloc_strdup(el1->list2[0], "testing2");
+
+	el1->list3 = talloc(el1, struct el2 *);
+	el1->list3[0] = talloc(el1->list3, struct el2);
+	el1->list3[0]->name = talloc_strdup(el1->list3[0], "testing2");
+	
+	el2 = talloc(el1->list, struct el2);
+	el2 = talloc(el1->list2, struct el2);
+	el2 = talloc(el1->list3, struct el2);
+
+	el1->list = talloc_realloc(el1, el1->list, struct el2 *, 100);
+	el1->list2 = talloc_realloc(el1, el1->list2, struct el2 *, 200);
+	el1->list3 = talloc_realloc(el1, el1->list3, struct el2 *, 300);
+
+	talloc_free(root);
+
+	printf("success: REALLOC WITH CHILD\n");
+	return true;
+}
+
+/*
+  test type checking
+*/
+static bool test_type(void)
+{
+	void *root;
+	struct el1 {
+		int count;
+	};
+	struct el2 {
+		int count;
+	};
+	struct el1 *el1;
+
+	printf("test: type\n# talloc type checking\n");
+
+	root = talloc_new(NULL);
+
+	el1 = talloc(root, struct el1);
+
+	el1->count = 1;
+
+	torture_assert("type", talloc_get_type(el1, struct el1) == el1,
+		"type check failed on el1\n");
+	torture_assert("type", talloc_get_type(el1, struct el2) == NULL,
+		"type check failed on el1 with el2\n");
+	talloc_set_type(el1, struct el2);
+	torture_assert("type", talloc_get_type(el1, struct el2) == (struct el2 *)el1,
+		"type set failed on el1 with el2\n");
+
+	talloc_free(root);
+
+	printf("success: type\n");
+	return true;
+}
+
+/*
+  test steal
+*/
+static bool test_steal(void)
+{
+	void *root, *p1, *p2;
+
+	printf("test: steal\n# STEAL\n");
+
+	root = talloc_new(NULL);
+
+	p1 = talloc_array(root, char, 10);
+	CHECK_SIZE("steal", p1, 10);
+
+	p2 = talloc_realloc(root, NULL, char, 20);
+	CHECK_SIZE("steal", p1, 10);
+	CHECK_SIZE("steal", root, 30);
+
+	torture_assert("steal", talloc_steal(p1, NULL) == NULL,
+		"failed: stealing NULL should give NULL\n");
+
+	torture_assert("steal", talloc_steal(p1, p1) == p1,
+		"failed: stealing to ourselves is a nop\n");
+	CHECK_BLOCKS("steal", root, 3);
+	CHECK_SIZE("steal", root, 30);
+
+	talloc_steal(NULL, p1);
+	talloc_steal(NULL, p2);
+	CHECK_BLOCKS("steal", root, 1);
+	CHECK_SIZE("steal", root, 0);
+
+	talloc_free(p1);
+	talloc_steal(root, p2);
+	CHECK_BLOCKS("steal", root, 2);
+	CHECK_SIZE("steal", root, 20);
+	
+	talloc_free(p2);
+
+	CHECK_BLOCKS("steal", root, 1);
+	CHECK_SIZE("steal", root, 0);
+
+	talloc_free(root);
+
+	p1 = talloc_size(NULL, 3);
+	talloc_report_full(NULL, stderr);
+	CHECK_SIZE("steal", NULL, 3);
+	talloc_free(p1);
+
+	printf("success: steal\n");
+	return true;
+}
+
+/*
+  test move
+*/
+static bool test_move(void)
+{
+	void *root;
+	struct t_move {
+		char *p;
+		int *x;
+	} *t1, *t2;
+
+	printf("test: move\n# MOVE\n");
+
+	root = talloc_new(NULL);
+
+	t1 = talloc(root, struct t_move);
+	t2 = talloc(root, struct t_move);
+	t1->p = talloc_strdup(t1, "foo");
+	t1->x = talloc(t1, int);
+	*t1->x = 42;
+
+	t2->p = talloc_move(t2, &t1->p);
+	t2->x = talloc_move(t2, &t1->x);
+	torture_assert("move", t1->p == NULL && t1->x == NULL &&
+	    strcmp(t2->p, "foo") == 0 && *t2->x == 42,
+		"talloc move failed");
+
+	talloc_free(root);
+
+	printf("success: move\n");
+
+	return true;
+}
+
+/*
+  test talloc_realloc_fn
+*/
+static bool test_realloc_fn(void)
+{
+	void *root, *p1;
+
+	printf("test: realloc_fn\n# talloc_realloc_fn\n");
+
+	root = talloc_new(NULL);
+
+	p1 = talloc_realloc_fn(root, NULL, 10);
+	CHECK_BLOCKS("realloc_fn", root, 2);
+	CHECK_SIZE("realloc_fn", root, 10);
+	p1 = talloc_realloc_fn(root, p1, 20);
+	CHECK_BLOCKS("realloc_fn", root, 2);
+	CHECK_SIZE("realloc_fn", root, 20);
+	p1 = talloc_realloc_fn(root, p1, 0);
+	CHECK_BLOCKS("realloc_fn", root, 1);
+	CHECK_SIZE("realloc_fn", root, 0);
+
+	talloc_free(root);
+
+	printf("success: realloc_fn\n");
+	return true;
+}
+
+
+static bool test_unref_reparent(void)
+{
+	void *root, *p1, *p2, *c1;
+
+	printf("test: unref_reparent\n# UNREFERENCE AFTER PARENT FREED\n");
+
+	root = talloc_named_const(NULL, 0, "root");
+	p1 = talloc_named_const(root, 1, "orig parent");
+	p2 = talloc_named_const(root, 1, "parent by reference");
+
+	c1 = talloc_named_const(p1, 1, "child");
+	talloc_reference(p2, c1);
+
+	CHECK_PARENT("unref_reparent", c1, p1);
+
+	talloc_free(p1);
+
+	CHECK_PARENT("unref_reparent", c1, p2);
+
+	talloc_unlink(p2, c1);
+
+	CHECK_SIZE("unref_reparent", root, 1);
+
+	talloc_free(p2);
+	talloc_free(root);
+
+	printf("success: unref_reparent\n");
+	return true;
+}
+
+/*
+  measure the speed of talloc versus malloc
+*/
+static bool test_speed(void)
+{
+	void *ctx = talloc_new(NULL);
+	unsigned count;
+	const int loop = 1000;
+	int i;
+	struct timeval tv;
+
+	printf("test: speed\n# TALLOC VS MALLOC SPEED\n");
+
+	tv = timeval_current();
+	count = 0;
+	do {
+		void *p1, *p2, *p3;
+		for (i=0;i<loop;i++) {
+			p1 = talloc_size(ctx, loop % 100);
+			p2 = talloc_strdup(p1, "foo bar");
+			p3 = talloc_size(p1, 300);
+			talloc_free(p1);
+		}
+		count += 3 * loop;
+	} while (timeval_elapsed(&tv) < 5.0);
+
+	fprintf(stderr, "talloc: %.0f ops/sec\n", count/timeval_elapsed(&tv));
+
+	talloc_free(ctx);
+
+	ctx = talloc_pool(NULL, 1024);
+
+	tv = timeval_current();
+	count = 0;
+	do {
+		void *p1, *p2, *p3;
+		for (i=0;i<loop;i++) {
+			p1 = talloc_size(ctx, loop % 100);
+			p2 = talloc_strdup(p1, "foo bar");
+			p3 = talloc_size(p1, 300);
+			talloc_free(p1);
+		}
+		count += 3 * loop;
+	} while (timeval_elapsed(&tv) < 5.0);
+
+	talloc_free(ctx);
+
+	fprintf(stderr, "talloc_pool: %.0f ops/sec\n", count/timeval_elapsed(&tv));
+
+	tv = timeval_current();
+	count = 0;
+	do {
+		void *p1, *p2, *p3;
+		for (i=0;i<loop;i++) {
+			p1 = malloc(loop % 100);
+			p2 = strdup("foo bar");
+			p3 = malloc(300);
+			free(p1);
+			free(p2);
+			free(p3);
+		}
+		count += 3 * loop;
+	} while (timeval_elapsed(&tv) < 5.0);
+	fprintf(stderr, "malloc: %.0f ops/sec\n", count/timeval_elapsed(&tv));
+
+	printf("success: speed\n");
+
+	return true;
+}
+
+static bool test_lifeless(void)
+{
+	void *top = talloc_new(NULL);
+	char *parent, *child; 
+	void *child_owner = talloc_new(NULL);
+
+	printf("test: lifeless\n# TALLOC_UNLINK LOOP\n");
+
+	parent = talloc_strdup(top, "parent");
+	child = talloc_strdup(parent, "child");  
+	(void)talloc_reference(child, parent);
+	(void)talloc_reference(child_owner, child); 
+	talloc_report_full(top, stderr);
+	talloc_unlink(top, parent);
+	talloc_unlink(top, child);
+	talloc_report_full(top, stderr);
+	talloc_free(top);
+	talloc_free(child_owner);
+	talloc_free(child);
+
+	printf("success: lifeless\n");
+	return true;
+}
+
+static int loop_destructor_count;
+
+static int test_loop_destructor(char *ptr)
+{
+	loop_destructor_count++;
+	return 0;
+}
+
+static bool test_loop(void)
+{
+	void *top = talloc_new(NULL);
+	char *parent;
+	struct req1 {
+		char *req2, *req3;
+	} *req1;
+
+	printf("test: loop\n# TALLOC LOOP DESTRUCTION\n");
+
+	parent = talloc_strdup(top, "parent");
+	req1 = talloc(parent, struct req1);
+	req1->req2 = talloc_strdup(req1, "req2");  
+	talloc_set_destructor(req1->req2, test_loop_destructor);
+	req1->req3 = talloc_strdup(req1, "req3");
+	(void)talloc_reference(req1->req3, req1);
+	talloc_report_full(top, stderr);
+	talloc_free(parent);
+	talloc_report_full(top, stderr);
+	talloc_report_full(NULL, stderr);
+	talloc_free(top);
+
+	torture_assert("loop", loop_destructor_count == 1, 
+				   "FAILED TO FIRE LOOP DESTRUCTOR\n");
+	loop_destructor_count = 0;
+
+	printf("success: loop\n");
+	return true;
+}
+
+static int fail_destructor_str(char *ptr)
+{
+	return -1;
+}
+
+static bool test_free_parent_deny_child(void)
+{
+	void *top = talloc_new(NULL);
+	char *level1;
+	char *level2;
+	char *level3;
+
+	printf("test: free_parent_deny_child\n# TALLOC FREE PARENT DENY CHILD\n");
+
+	level1 = talloc_strdup(top, "level1");
+	level2 = talloc_strdup(level1, "level2");
+	level3 = talloc_strdup(level2, "level3");
+
+	talloc_set_destructor(level3, fail_destructor_str);
+	talloc_free(level1);
+	talloc_set_destructor(level3, NULL);
+
+	CHECK_PARENT("free_parent_deny_child", level3, top);
+
+	talloc_free(top);
+
+	printf("success: free_parent_deny_child\n");
+	return true;
+}
+
+static bool test_talloc_ptrtype(void)
+{
+	void *top = talloc_new(NULL);
+	struct struct1 {
+		int foo;
+		int bar;
+	} *s1, *s2, **s3, ***s4;
+	const char *location1;
+	const char *location2;
+	const char *location3;
+	const char *location4;
+
+	printf("test: ptrtype\n# TALLOC PTRTYPE\n");
+
+	s1 = talloc_ptrtype(top, s1);location1 = __location__;
+
+	if (talloc_get_size(s1) != sizeof(struct struct1)) {
+		printf("failure: ptrtype [\n"
+		  "talloc_ptrtype() allocated the wrong size %lu (should be %lu)\n"
+		  "]\n", (unsigned long)talloc_get_size(s1),
+		           (unsigned long)sizeof(struct struct1));
+		return false;
+	}
+
+	if (strcmp(location1, talloc_get_name(s1)) != 0) {
+		printf("failure: ptrtype [\n"
+		  "talloc_ptrtype() sets the wrong name '%s' (should be '%s')\n]\n",
+			talloc_get_name(s1), location1);
+		return false;
+	}
+
+	s2 = talloc_array_ptrtype(top, s2, 10);location2 = __location__;
+
+	if (talloc_get_size(s2) != (sizeof(struct struct1) * 10)) {
+		printf("failure: ptrtype [\n"
+			   "talloc_array_ptrtype() allocated the wrong size "
+		       "%lu (should be %lu)\n]\n",
+			(unsigned long)talloc_get_size(s2),
+		    (unsigned long)(sizeof(struct struct1)*10));
+		return false;
+	}
+
+	if (strcmp(location2, talloc_get_name(s2)) != 0) {
+		printf("failure: ptrtype [\n"
+		"talloc_array_ptrtype() sets the wrong name '%s' (should be '%s')\n]\n",
+			talloc_get_name(s2), location2);
+		return false;
+	}
+
+	s3 = talloc_array_ptrtype(top, s3, 10);location3 = __location__;
+
+	if (talloc_get_size(s3) != (sizeof(struct struct1 *) * 10)) {
+		printf("failure: ptrtype [\n"
+			   "talloc_array_ptrtype() allocated the wrong size "
+		       "%lu (should be %lu)\n]\n",
+			   (unsigned long)talloc_get_size(s3),
+		       (unsigned long)(sizeof(struct struct1 *)*10));
+		return false;
+	}
+
+	torture_assert_str_equal("ptrtype", location3, talloc_get_name(s3),
+		"talloc_array_ptrtype() sets the wrong name");
+
+	s4 = talloc_array_ptrtype(top, s4, 10);location4 = __location__;
+
+	if (talloc_get_size(s4) != (sizeof(struct struct1 **) * 10)) {
+		printf("failure: ptrtype [\n"
+		      "talloc_array_ptrtype() allocated the wrong size "
+		       "%lu (should be %lu)\n]\n",
+			   (unsigned long)talloc_get_size(s4),
+		       (unsigned long)(sizeof(struct struct1 **)*10));
+		return false;
+	}
+
+	torture_assert_str_equal("ptrtype", location4, talloc_get_name(s4),
+		"talloc_array_ptrtype() sets the wrong name");
+
+	talloc_free(top);
+
+	printf("success: ptrtype\n");
+	return true;
+}
+
+static int _test_talloc_free_in_destructor(void **ptr)
+{
+	talloc_free(*ptr);
+	return 0;
+}
+
+static bool test_talloc_free_in_destructor(void)
+{
+	void *level0;
+	void *level1;
+	void *level2;
+	void *level3;
+	void *level4;
+	void **level5;
+
+	printf("test: free_in_destructor\n# TALLOC FREE IN DESTRUCTOR\n");
+
+	level0 = talloc_new(NULL);
+	level1 = talloc_new(level0);
+	level2 = talloc_new(level1);
+	level3 = talloc_new(level2);
+	level4 = talloc_new(level3);
+	level5 = talloc(level4, void *);
+
+	*level5 = level3;
+	(void)talloc_reference(level0, level3);
+	(void)talloc_reference(level3, level3);
+	(void)talloc_reference(level5, level3);
+
+	talloc_set_destructor(level5, _test_talloc_free_in_destructor);
+
+	talloc_free(level1);
+
+	talloc_free(level0);
+
+	printf("success: free_in_destructor\n");
+	return true;
+}
+
+static bool test_autofree(void)
+{
+#if _SAMBA_BUILD_ < 4
+	/* autofree test would kill smbtorture */
+	void *p;
+	printf("test: autofree\n# TALLOC AUTOFREE CONTEXT\n");
+
+	p = talloc_autofree_context();
+	talloc_free(p);
+
+	p = talloc_autofree_context();
+	talloc_free(p);
+
+	printf("success: autofree\n");
+#endif
+	return true;
+}
+
+static bool test_pool(void)
+{
+	void *pool;
+	void *p1, *p2, *p3, *p4;
+	void *p2_2;
+
+	pool = talloc_pool(NULL, 1024);
+
+	p1 = talloc_size(pool, 80);
+	memset(p1, 0x11, talloc_get_size(p1));
+	p2 = talloc_size(pool, 20);
+	memset(p2, 0x11, talloc_get_size(p2));
+	p3 = talloc_size(p1, 50);
+	memset(p3, 0x11, talloc_get_size(p3));
+	p4 = talloc_size(p3, 1000);
+	memset(p4, 0x11, talloc_get_size(p4));
+
+#if 1 /* this relies on ALWAYS_REALLOC == 0 in talloc.c */
+	p2_2 = talloc_realloc_size(pool, p2, 20+1);
+	torture_assert("pool realloc 20+1", p2_2 == p2, "failed: pointer changed");
+	memset(p2, 0x11, talloc_get_size(p2));
+	p2_2 = talloc_realloc_size(pool, p2, 20-1);
+	torture_assert("pool realloc 20-1", p2_2 == p2, "failed: pointer changed");
+	memset(p2, 0x11, talloc_get_size(p2));
+	p2_2 = talloc_realloc_size(pool, p2, 20-1);
+	torture_assert("pool realloc 20-1", p2_2 == p2, "failed: pointer changed");
+	memset(p2, 0x11, talloc_get_size(p2));
+
+	talloc_free(p3);
+
+	/* this should reclaim the memory of p4 and p3 */
+	p2_2 = talloc_realloc_size(pool, p2, 400);
+	torture_assert("pool realloc 400", p2_2 == p2, "failed: pointer changed");
+	memset(p2, 0x11, talloc_get_size(p2));
+
+	talloc_free(p1);
+
+	/* this should reclaim the memory of p1 */
+	p2_2 = talloc_realloc_size(pool, p2, 800);
+	torture_assert("pool realloc 800", p2_2 == p1, "failed: pointer not changed");
+	p2 = p2_2;
+	memset(p2, 0x11, talloc_get_size(p2));
+
+	/* this should do a malloc */
+	p2_2 = talloc_realloc_size(pool, p2, 1800);
+	torture_assert("pool realloc 1800", p2_2 != p2, "failed: pointer not changed");
+	p2 = p2_2;
+	memset(p2, 0x11, talloc_get_size(p2));
+
+	/* this should reclaim the memory from the pool */
+	p3 = talloc_size(pool, 80);
+	torture_assert("pool alloc 80", p3 == p1, "failed: pointer changed");
+	memset(p3, 0x11, talloc_get_size(p3));
+
+	talloc_free(p2);
+	talloc_free(p3);
+
+	p1 = talloc_size(pool, 80);
+	memset(p1, 0x11, talloc_get_size(p1));
+	p2 = talloc_size(pool, 20);
+	memset(p2, 0x11, talloc_get_size(p2));
+
+	talloc_free(p1);
+
+	p2_2 = talloc_realloc_size(pool, p2, 20-1);
+	torture_assert("pool realloc 20-1", p2_2 == p2, "failed: pointer changed");
+	memset(p2, 0x11, talloc_get_size(p2));
+	p2_2 = talloc_realloc_size(pool, p2, 20-1);
+	torture_assert("pool realloc 20-1", p2_2 == p2, "failed: pointer changed");
+	memset(p2, 0x11, talloc_get_size(p2));
+
+	/* this should do a malloc */
+	p2_2 = talloc_realloc_size(pool, p2, 1800);
+	torture_assert("pool realloc 1800", p2_2 != p2, "failed: pointer not changed");
+	p2 = p2_2;
+	memset(p2, 0x11, talloc_get_size(p2));
+
+	/* this should reclaim the memory from the pool */
+	p3 = talloc_size(pool, 800);
+	torture_assert("pool alloc 800", p3 == p1, "failed: pointer changed");
+	memset(p3, 0x11, talloc_get_size(p3));
+
+#endif /* this relies on ALWAYS_REALLOC == 0 in talloc.c */
+
+	talloc_free(pool);
+
+	return true;
+}
+
+static bool test_pool_steal(void)
+{
+	void *root;
+	void *pool;
+	void *p1, *p2;
+	void *p1_2, *p2_2;
+	size_t hdr;
+	size_t ofs1, ofs2;
+
+	root = talloc_new(NULL);
+	pool = talloc_pool(root, 1024);
+
+	p1 = talloc_size(pool, 4 * 16);
+	torture_assert("pool allocate 4 * 16", p1 != NULL, "failed ");
+	memset(p1, 0x11, talloc_get_size(p1));
+	p2 = talloc_size(pool, 4 * 16);
+	torture_assert("pool allocate 4 * 16", p2 > p1, "failed: !(p2 > p1) ");
+	memset(p2, 0x11, talloc_get_size(p2));
+
+	ofs1 = PTR_DIFF(p2, p1);
+	hdr = ofs1 - talloc_get_size(p1);
+
+	talloc_steal(root, p1);
+	talloc_steal(root, p2);
+
+	talloc_free(pool);
+
+	p1_2 = p1;
+
+#if 1 /* this relies on ALWAYS_REALLOC == 0 in talloc.c */
+	p1_2 = talloc_realloc_size(root, p1, 5 * 16);
+	torture_assert("pool realloc 5 * 16", p1_2 > p2, "failed: pointer not changed");
+	memset(p1_2, 0x11, talloc_get_size(p1_2));
+	ofs1 = PTR_DIFF(p1_2, p2);
+	ofs2 = talloc_get_size(p2) + hdr;
+
+	torture_assert("pool realloc ", ofs1 == ofs2, "failed: pointer offset unexpected");
+
+	p2_2 = talloc_realloc_size(root, p2, 3 * 16);
+	torture_assert("pool realloc 5 * 16", p2_2 == p2, "failed: pointer changed");
+	memset(p2_2, 0x11, talloc_get_size(p2_2));
+#endif /* this relies on ALWAYS_REALLOC == 0 in talloc.c */
+
+	talloc_free(p1_2);
+
+	p2_2 = p2;
+
+#if 1 /* this relies on ALWAYS_REALLOC == 0 in talloc.c */
+	/* now we should reclaim the full pool */
+	p2_2 = talloc_realloc_size(root, p2, 8 * 16);
+	torture_assert("pool realloc 8 * 16", p2_2 == p1, "failed: pointer not expected");
+	p2 = p2_2;
+	memset(p2_2, 0x11, talloc_get_size(p2_2));
+
+	/* now we malloc and free the full pool space */
+	p2_2 = talloc_realloc_size(root, p2, 2 * 1024);
+	torture_assert("pool realloc 2 * 1024", p2_2 != p1, "failed: pointer not expected");
+	memset(p2_2, 0x11, talloc_get_size(p2_2));
+
+#endif /* this relies on ALWAYS_REALLOC == 0 in talloc.c */
+
+	talloc_free(p2_2);
+
+	talloc_free(root);
+
+	return true;
+}
+
+static bool test_free_ref_null_context(void)
+{
+	void *p1, *p2, *p3;
+	int ret;
+
+	talloc_disable_null_tracking();
+	p1 = talloc_new(NULL);
+	p2 = talloc_new(NULL);
+
+	p3 = talloc_reference(p2, p1);
+	torture_assert("reference", p3 == p1, "failed: reference on null");
+
+	ret = talloc_free(p1);
+	torture_assert("ref free with null parent", ret == 0, "failed: free with null parent");
+	talloc_free(p2);
+
+	talloc_enable_null_tracking_no_autofree();
+	p1 = talloc_new(NULL);
+	p2 = talloc_new(NULL);
+
+	p3 = talloc_reference(p2, p1);
+	torture_assert("reference", p3 == p1, "failed: reference on null");
+
+	ret = talloc_free(p1);
+	torture_assert("ref free with null tracked parent", ret == 0, "failed: free with null parent");
+	talloc_free(p2);
+
+	return true;
+}
+
+static bool test_rusty(void)
+{
+	void *root;
+	const char *p1;
+
+	talloc_enable_null_tracking();
+	root = talloc_new(NULL);
+	p1 = talloc_strdup(root, "foo");
+	talloc_increase_ref_count(p1);
+	talloc_report_full(root, stdout);
+	talloc_free(root);
+	CHECK_BLOCKS("null_context", NULL, 2);
+	return true;
+}
+
+static bool test_free_children(void)
+{
+	void *root;
+	char *p1, *p2;
+	const char *name, *name2;
+
+	talloc_enable_null_tracking();
+	root = talloc_new(NULL);
+	p1 = talloc_strdup(root, "foo1");
+	p2 = talloc_strdup(p1, "foo2");
+
+	talloc_set_name(p1, "%s", "testname");
+	talloc_free_children(p1);
+	/* check its still a valid talloc ptr */
+	talloc_get_size(talloc_get_name(p1));
+	if (strcmp(talloc_get_name(p1), "testname") != 0) {
+		return false;
+	}
+
+	talloc_set_name(p1, "%s", "testname");
+	name = talloc_get_name(p1);
+	talloc_free_children(p1);
+	/* check its still a valid talloc ptr */
+	talloc_get_size(talloc_get_name(p1));
+	torture_assert("name", name == talloc_get_name(p1), "name ptr changed");
+	torture_assert("namecheck", strcmp(talloc_get_name(p1), "testname") == 0,
+		       "wrong name");
+	CHECK_BLOCKS("name1", p1, 2);
+
+	/* note that this does not free the old child name */
+	talloc_set_name_const(p1, "testname2");
+	name2 = talloc_get_name(p1);
+	/* but this does */
+	talloc_free_children(p1);
+	torture_assert("namecheck", strcmp(talloc_get_name(p1), "testname2") == 0,
+		       "wrong name");
+	CHECK_BLOCKS("name1", p1, 1);
+
+	talloc_report_full(root, stdout);
+	talloc_free(root);
+	return true;
+}
+
+static bool test_memlimit(void)
+{
+	void *root;
+	char *l1, *l2, *l3, *l4, *l5, *t;
+
+	printf("test: memlimit\n# MEMORY LIMITS\n");
+
+	printf("==== talloc_new(NULL)\n");
+	root = talloc_new(NULL);
+
+	talloc_report_full(root, stdout);
+
+	printf("==== talloc_size(root, 2048)\n");
+	l1 = talloc_size(root, 2048);
+	torture_assert("memlimit", l1 != NULL,
+		"failed: alloc should not fail due to memory limit\n");
+
+	talloc_report_full(root, stdout);
+
+	printf("==== talloc_free(l1)\n");
+	talloc_free(l1);
+
+	talloc_report_full(root, stdout);
+
+	printf("==== talloc_strdup(root, level 1)\n");
+	l1 = talloc_strdup(root, "level 1");
+	torture_assert("memlimit", l1 != NULL,
+		"failed: alloc should not fail due to memory limit\n");
+
+	talloc_report_full(root, stdout);
+
+	printf("==== talloc_set_memlimit(l1, 2048)\n");
+	torture_assert("memlimit", talloc_set_memlimit(l1, 2048) == 0,
+		"failed: setting memlimit should never fail\n");
+
+	talloc_report_full(root, stdout);
+
+	printf("==== talloc_size(root, 2048)\n");
+	l2 = talloc_size(l1, 2048);
+	torture_assert("memlimit", l2 == NULL,
+		"failed: alloc should fail due to memory limit\n");
+
+	talloc_report_full(root, stdout);
+
+	printf("==== talloc_strdup(l1, level 2)\n");
+	l2 = talloc_strdup(l1, "level 2");
+	torture_assert("memlimit", l2 != NULL,
+		"failed: alloc should not fail due to memory limit\n");
+
+	talloc_report_full(root, stdout);
+
+	printf("==== talloc_free(l2)\n");
+	talloc_free(l2);
+
+	talloc_report_full(root, stdout);
+
+	printf("==== talloc_size(NULL, 2048)\n");
+	l2 = talloc_size(NULL, 2048);
+
+	talloc_report_full(root, stdout);
+
+	printf("==== talloc_steal(l1, l2)\n");
+	talloc_steal(l1, l2);
+
+	talloc_report_full(root, stdout);
+
+	printf("==== talloc_strdup(l2, level 3)\n");
+	l3 = talloc_strdup(l2, "level 3");
+	torture_assert("memlimit", l3 == NULL,
+		"failed: alloc should fail due to memory limit\n");
+
+	talloc_report_full(root, stdout);
+
+	printf("==== talloc_free(l2)\n");
+	talloc_free(l2);
+
+	talloc_report_full(root, stdout);
+
+	printf("==== talloc_strdup(NULL, level 2)\n");
+	l2 = talloc_strdup(NULL, "level 2");
+	talloc_steal(l1, l2);
+
+	talloc_report_full(root, stdout);
+
+	printf("==== talloc_strdup(l2, level 3)\n");
+	l3 = talloc_strdup(l2, "level 3");
+	torture_assert("memlimit", l3 != NULL,
+		"failed: alloc should not fail due to memory limit\n");
+
+	talloc_report_full(root, stdout);
+
+	printf("==== talloc_set_memlimit(l3, 1024)\n");
+	torture_assert("memlimit", talloc_set_memlimit(l3, 1024) == 0,
+		"failed: setting memlimit should never fail\n");
+
+	talloc_report_full(root, stdout);
+
+	printf("==== talloc_strdup(l3, level 4)\n");
+	l4 = talloc_strdup(l3, "level 4");
+	torture_assert("memlimit", l4 != NULL,
+		"failed: alloc should not fail due to memory limit\n");
+
+	talloc_report_full(root, stdout);
+
+	printf("==== talloc_set_memlimit(l4, 512)\n");
+	torture_assert("memlimit", talloc_set_memlimit(l4, 512) == 0,
+		"failed: setting memlimit should never fail\n");
+
+	talloc_report_full(root, stdout);
+
+	printf("==== talloc_strdup(l4, level 5)\n");
+	l5 = talloc_strdup(l4, "level 5");
+	torture_assert("memlimit", l5 != NULL,
+		"failed: alloc should not fail due to memory limit\n");
+
+	talloc_report_full(root, stdout);
+
+	printf("==== talloc_realloc(NULL, l5, char, 600)\n");
+	t = talloc_realloc(NULL, l5, char, 600);
+	torture_assert("memlimit", t == NULL,
+		"failed: alloc should fail due to memory limit\n");
+
+	talloc_report_full(root, stdout);
+
+	printf("==== talloc_realloc(NULL, l5, char, 5)\n");
+	l5 = talloc_realloc(NULL, l5, char, 5);
+	torture_assert("memlimit", l5 != NULL,
+		"failed: alloc should not fail due to memory limit\n");
+
+	talloc_report_full(root, stdout);
+
+	printf("==== talloc_strdup(l3, level 4)\n");
+	l4 = talloc_strdup(l3, "level 4");
+	torture_assert("memlimit", l4 != NULL,
+		"failed: alloc should not fail due to memory limit\n");
+
+	talloc_report_full(root, stdout);
+
+	printf("==== talloc_set_memlimit(l4, 512)\n");
+	torture_assert("memlimit", talloc_set_memlimit(l4, 512) == 0,
+		"failed: setting memlimit should never fail\n");
+
+	talloc_report_full(root, stdout);
+
+	printf("==== talloc_strdup(l4, level 5)\n");
+	l5 = talloc_strdup(l4, "level 5");
+	torture_assert("memlimit", l5 != NULL,
+		"failed: alloc should not fail due to memory limit\n");
+
+	talloc_report_full(root, stdout);
+
+	printf("==== Make new temp context and steal l5\n");
+	t = talloc_new(root);
+	talloc_steal(t, l5);
+
+	talloc_report_full(root, stdout);
+
+	printf("==== talloc_size(t, 2048)\n");
+	l1 = talloc_size(t, 2048);
+	torture_assert("memlimit", l1 != NULL,
+		"failed: alloc should not fail due to memory limit\n");
+
+	talloc_report_full(root, stdout);
+	talloc_free(root);
+
+	printf("success: memlimit\n");
+
+	return true;
+}
+
+static void test_reset(void)
+{
+	talloc_set_log_fn(test_log_stdout);
+	test_abort_stop();
+	talloc_disable_null_tracking();
+	talloc_enable_null_tracking_no_autofree();
+}
+
+bool torture_local_talloc(struct torture_context *tctx)
+{
+	bool ret = true;
+
+	setlinebuf(stdout);
+
+	test_reset();
+	ret &= test_ref1();
+	test_reset();
+	ret &= test_ref2();
+	test_reset();
+	ret &= test_ref3();
+	test_reset();
+	ret &= test_ref4();
+	test_reset();
+	ret &= test_unlink1(); 
+	test_reset();
+	ret &= test_misc();
+	test_reset();
+	ret &= test_realloc();
+	test_reset();
+	ret &= test_realloc_child(); 
+	test_reset();
+	ret &= test_steal(); 
+	test_reset();
+	ret &= test_move(); 
+	test_reset();
+	ret &= test_unref_reparent();
+	test_reset();
+	ret &= test_realloc_fn(); 
+	test_reset();
+	ret &= test_type();
+	test_reset();
+	ret &= test_lifeless(); 
+	test_reset();
+	ret &= test_loop();
+	test_reset();
+	ret &= test_free_parent_deny_child(); 
+	test_reset();
+	ret &= test_talloc_ptrtype();
+	test_reset();
+	ret &= test_talloc_free_in_destructor();
+	test_reset();
+	ret &= test_pool();
+	test_reset();
+	ret &= test_pool_steal();
+	test_reset();
+	ret &= test_free_ref_null_context();
+	test_reset();
+	ret &= test_rusty();
+	test_reset();
+	ret &= test_free_children();
+	test_reset();
+	ret &= test_memlimit();
+
+
+	if (ret) {
+		test_reset();
+		ret &= test_speed();
+	}
+	test_reset();
+	ret &= test_autofree();
+
+	test_reset();
+	talloc_disable_null_tracking();
+	return ret;
+}
diff --git a/ctdb/lib/talloc/testsuite_main.c b/ctdb/lib/talloc/testsuite_main.c
new file mode 100644
index 0000000000..50ce0f8e3b
--- /dev/null
+++ b/ctdb/lib/talloc/testsuite_main.c
@@ -0,0 +1,36 @@
+/*
+   Unix SMB/CIFS implementation.
+
+   local testing of talloc routines.
+
+   Copyright (C) Andrew Tridgell 2004
+
+     ** NOTE! The following LGPL license applies to the talloc
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "replace.h"
+
+#include "talloc_testsuite.h"
+
+int main(void)
+{
+	bool ret = torture_local_talloc(NULL);
+	if (!ret)
+		return -1;
+	return 0;
+}
diff --git a/ctdb/lib/talloc/web/index.html b/ctdb/lib/talloc/web/index.html
new file mode 100644
index 0000000000..388ec2cde2
--- /dev/null
+++ b/ctdb/lib/talloc/web/index.html
@@ -0,0 +1,51 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
+<HTML>
+<HEAD>
+<TITLE>talloc</TITLE>
+</HEAD>
+<BODY BGCOLOR="#ffffff" TEXT="#000000" VLINK="#292555" LINK="#292555" ALINK="#cc0033">
+
+<h1>talloc</h1>
+
+talloc is a hierarchical pool based memory allocator with
+destructors. It is the core memory allocator used in Samba, and has
+made a huge difference in many aspects of Samba4 development.<p>
+
+To get started with talloc, I would recommend you read the <a
+href="http://samba.org/ftp/unpacked/talloc/talloc_guide.txt">talloc guide</a>.
+
+<h2>Download</h2>
+You can download the latest releases of talloc from the <a
+href="http://samba.org/ftp/talloc">talloc directory</a> on the samba public
+source archive.
+
+<h2>Discussion and bug reports</h2>
+
+talloc does not currently have its own mailing list or bug tracking
+system. For now, please use the <a
+href="https://lists.samba.org/mailman/listinfo/samba-technical">samba-technical</a>
+mailing list, and the <a href="http://bugzilla.samba.org/">Samba
+bugzilla</a> bug tracking system.
+
+<h2>Development</h2>
+
+You can download the latest code either via git or rsync.<br>
+<br>
+To fetch via git see the following guide:<br>
+<a href="http://wiki.samba.org/index.php/Using_Git_for_Samba_Development">Using Git for Samba Development</a><br>
+Once you have cloned the tree switch to the master branch and cd into the lib/talloc directory.<br>
+<br>
+To fetch via rsync use this command:
+
+<pre>
+  rsync -Pavz samba.org::ftp/unpacked/standalone_projects/lib/talloc .
+</pre>
+
+<hr>
+<tiny>
+<a href="http://samba.org/~tridge/">Andrew Tridgell</a><br>
+talloc AT tridgell.net
+</tiny>
+
+</BODY>
+</HTML>
diff --git a/ctdb/lib/talloc/wscript b/ctdb/lib/talloc/wscript
new file mode 100644
index 0000000000..8d3246b26b
--- /dev/null
+++ b/ctdb/lib/talloc/wscript
@@ -0,0 +1,154 @@
+#!/usr/bin/env python
+
+APPNAME = 'talloc'
+VERSION = '2.0.8'
+
+
+blddir = 'bin'
+
+import Logs
+import os, sys
+
+# find the buildtools directory
+srcdir = '.'
+while not os.path.exists(srcdir+'/buildtools') and len(srcdir.split('/')) < 5:
+    srcdir = '../' + srcdir
+sys.path.insert(0, srcdir + '/buildtools/wafsamba')
+
+import sys
+sys.path.insert(0, srcdir+"/buildtools/wafsamba")
+import wafsamba, samba_dist, Options
+
+# setup what directories to put in a tarball
+samba_dist.DIST_DIRS('lib/talloc:. lib/replace:lib/replace buildtools:buildtools')
+
+
+def set_options(opt):
+    opt.BUILTIN_DEFAULT('replace')
+    opt.PRIVATE_EXTENSION_DEFAULT('talloc', noextension='talloc')
+    opt.RECURSE('lib/replace')
+    opt.add_option('--enable-talloc-compat1',
+                   help=("Build talloc 1.x.x compat library [False]"),
+                   action="store_true", dest='TALLOC_COMPAT1', default=False)
+    if opt.IN_LAUNCH_DIR():
+        opt.add_option('--disable-python',
+                       help=("disable the pytalloc module"),
+                       action="store_true", dest='disable_python', default=False)
+
+
+def configure(conf):
+    conf.RECURSE('lib/replace')
+
+    conf.env.standalone_talloc = conf.IN_LAUNCH_DIR()
+
+    conf.env.disable_python = getattr(Options.options, 'disable_python', False)
+
+    if not conf.env.standalone_talloc:
+        if conf.CHECK_BUNDLED_SYSTEM_PKG('talloc', minversion=VERSION,
+                                     implied_deps='replace'):
+            conf.define('USING_SYSTEM_TALLOC', 1)
+        if conf.CHECK_BUNDLED_SYSTEM_PKG('pytalloc-util', minversion=VERSION,
+                                     implied_deps='talloc replace'):
+            conf.define('USING_SYSTEM_PYTALLOC_UTIL', 1)
+
+    conf.env.TALLOC_COMPAT1 = Options.options.TALLOC_COMPAT1
+
+    conf.CHECK_XSLTPROC_MANPAGES()
+
+    if not conf.env.disable_python:
+        # also disable if we don't have the python libs installed
+        conf.find_program('python', var='PYTHON')
+        conf.check_tool('python')
+        conf.check_python_version((2,4,2))
+        conf.SAMBA_CHECK_PYTHON_HEADERS(mandatory=False)
+        if not conf.env.HAVE_PYTHON_H:
+            Logs.warn('Disabling pytalloc-util as python devel libs not found')
+            conf.env.disable_python = True
+
+    conf.SAMBA_CONFIG_H()
+
+    conf.SAMBA_CHECK_UNDEFINED_SYMBOL_FLAGS()
+
+
+def build(bld):
+    bld.RECURSE('lib/replace')
+
+    if bld.env.standalone_talloc:
+        bld.env.PKGCONFIGDIR = '${LIBDIR}/pkgconfig'
+        bld.env.TALLOC_VERSION = VERSION
+        private_library = False
+
+        # should we also install the symlink to libtalloc1.so here?
+        bld.SAMBA_LIBRARY('talloc-compat1-%s' % (VERSION),
+                          'compat/talloc_compat1.c',
+                          public_deps='talloc',
+                          soname='libtalloc.so.1',
+                          pc_files=[],
+                          public_headers=[],
+                          enabled=bld.env.TALLOC_COMPAT1)
+
+        bld.SAMBA_BINARY('talloc_testsuite',
+                         'testsuite_main.c testsuite.c',
+                         deps='talloc',
+                         install=False)
+
+    else:
+        private_library = True
+
+    if not bld.CONFIG_SET('USING_SYSTEM_TALLOC'):
+
+        bld.SAMBA_LIBRARY('talloc',
+                          'talloc.c',
+                          deps='replace',
+                          abi_directory='ABI',
+                          abi_match='talloc* _talloc*',
+                          hide_symbols=True,
+                          vnum=VERSION,
+                          public_headers='talloc.h',
+                          pc_files='talloc.pc',
+                          public_headers_install=not private_library,
+                          private_library=private_library,
+                          manpages='talloc.3')
+
+    if not bld.CONFIG_SET('USING_SYSTEM_PYTALLOC_UTIL') and not bld.env.disable_python:
+        bld.SAMBA_LIBRARY('pytalloc-util',
+            source='pytalloc_util.c',
+            public_deps='talloc',
+            pyembed=True,
+            vnum=VERSION,
+            hide_symbols=True,
+            abi_directory='ABI',
+            abi_match='pytalloc_*',
+            private_library=private_library,
+            public_headers='pytalloc.h',
+            pc_files='pytalloc-util.pc'
+            )
+        bld.SAMBA_PYTHON('pytalloc',
+                         'pytalloc.c',
+                         deps='talloc pytalloc-util',
+                         enabled=True,
+                         realname='talloc.so')
+
+def test(ctx):
+    '''run talloc testsuite'''
+    import Utils, samba_utils
+    cmd = os.path.join(Utils.g_module.blddir, 'talloc_testsuite')
+    ret = samba_utils.RUN_COMMAND(cmd)
+    print("testsuite returned %d" % ret)
+    sys.exit(ret)
+
+def dist():
+    '''makes a tarball for distribution'''
+    samba_dist.dist()
+
+def reconfigure(ctx):
+    '''reconfigure if config scripts have changed'''
+    import samba_utils
+    samba_utils.reconfigure(ctx)
+
+
+def pydoctor(ctx):
+    '''build python apidocs'''
+    cmd='PYTHONPATH=bin/python pydoctor --project-name=talloc --project-url=http://talloc.samba.org/ --make-html --docformat=restructuredtext --introspect-c-modules --add-module bin/python/talloc.*'
+    print("Running: %s" % cmd)
+    os.system(cmd)
diff --git a/ctdb/lib/tdb/ABI/tdb-1.2.1.sigs b/ctdb/lib/tdb/ABI/tdb-1.2.1.sigs
new file mode 100644
index 0000000000..84f200745e
--- /dev/null
+++ b/ctdb/lib/tdb/ABI/tdb-1.2.1.sigs
@@ -0,0 +1,95 @@
+tdb_add_flags: void (struct tdb_context *, unsigned int)
+tdb_alloc_read: unsigned char *(struct tdb_context *, tdb_off_t, tdb_len_t)
+tdb_allocate: tdb_off_t (struct tdb_context *, tdb_len_t, struct tdb_record *)
+tdb_allrecord_lock: int (struct tdb_context *, int, enum tdb_lock_flags, bool)
+tdb_allrecord_unlock: int (struct tdb_context *, int, bool)
+tdb_allrecord_upgrade: int (struct tdb_context *)
+tdb_append: int (struct tdb_context *, TDB_DATA, TDB_DATA)
+tdb_brlock: int (struct tdb_context *, int, tdb_off_t, size_t, enum tdb_lock_flags)
+tdb_brunlock: int (struct tdb_context *, int, tdb_off_t, size_t)
+tdb_chainlock: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_mark: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_nonblock: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_read: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_unmark: int (struct tdb_context *, TDB_DATA)
+tdb_chainunlock: int (struct tdb_context *, TDB_DATA)
+tdb_chainunlock_read: int (struct tdb_context *, TDB_DATA)
+tdb_check: int (struct tdb_context *, int (*)(TDB_DATA, TDB_DATA, void *), void *)
+tdb_close: int (struct tdb_context *)
+tdb_convert: void *(void *, uint32_t)
+tdb_delete: int (struct tdb_context *, TDB_DATA)
+tdb_do_delete: int (struct tdb_context *, tdb_off_t, struct tdb_record *)
+tdb_dump_all: void (struct tdb_context *)
+tdb_enable_seqnum: void (struct tdb_context *)
+tdb_error: enum TDB_ERROR (struct tdb_context *)
+tdb_errorstr: const char *(struct tdb_context *)
+tdb_exists: int (struct tdb_context *, TDB_DATA)
+tdb_expand: int (struct tdb_context *, tdb_off_t)
+tdb_fd: int (struct tdb_context *)
+tdb_fetch: TDB_DATA (struct tdb_context *, TDB_DATA)
+tdb_find_lock_hash: tdb_off_t (struct tdb_context *, TDB_DATA, uint32_t, int, struct tdb_record *)
+tdb_firstkey: TDB_DATA (struct tdb_context *)
+tdb_free: int (struct tdb_context *, tdb_off_t, struct tdb_record *)
+tdb_freelist_size: int (struct tdb_context *)
+tdb_get_flags: int (struct tdb_context *)
+tdb_get_logging_private: void *(struct tdb_context *)
+tdb_get_seqnum: int (struct tdb_context *)
+tdb_hash_size: int (struct tdb_context *)
+tdb_have_extra_locks: bool (struct tdb_context *)
+tdb_increment_seqnum_nonblock: void (struct tdb_context *)
+tdb_io_init: void (struct tdb_context *)
+tdb_lock: int (struct tdb_context *, int, int)
+tdb_lock_nonblock: int (struct tdb_context *, int, int)
+tdb_lock_record: int (struct tdb_context *, tdb_off_t)
+tdb_lockall: int (struct tdb_context *)
+tdb_lockall_mark: int (struct tdb_context *)
+tdb_lockall_nonblock: int (struct tdb_context *)
+tdb_lockall_read: int (struct tdb_context *)
+tdb_lockall_read_nonblock: int (struct tdb_context *)
+tdb_lockall_unmark: int (struct tdb_context *)
+tdb_log_fn: tdb_log_func (struct tdb_context *)
+tdb_map_size: size_t (struct tdb_context *)
+tdb_mmap: void (struct tdb_context *)
+tdb_munmap: int (struct tdb_context *)
+tdb_name: const char *(struct tdb_context *)
+tdb_needs_recovery: bool (struct tdb_context *)
+tdb_nest_lock: int (struct tdb_context *, uint32_t, int, enum tdb_lock_flags)
+tdb_nest_unlock: int (struct tdb_context *, uint32_t, int, bool)
+tdb_nextkey: TDB_DATA (struct tdb_context *, TDB_DATA)
+tdb_ofs_read: int (struct tdb_context *, tdb_off_t, tdb_off_t *)
+tdb_ofs_write: int (struct tdb_context *, tdb_off_t, tdb_off_t *)
+tdb_open: struct tdb_context *(const char *, int, int, int, mode_t)
+tdb_open_ex: struct tdb_context *(const char *, int, int, int, mode_t, const struct tdb_logging_context *, tdb_hash_func)
+tdb_parse_data: int (struct tdb_context *, TDB_DATA, tdb_off_t, tdb_len_t, int (*)(TDB_DATA, TDB_DATA, void *), void *)
+tdb_parse_record: int (struct tdb_context *, TDB_DATA, int (*)(TDB_DATA, TDB_DATA, void *), void *)
+tdb_printfreelist: int (struct tdb_context *)
+tdb_rec_free_read: int (struct tdb_context *, tdb_off_t, struct tdb_record *)
+tdb_rec_read: int (struct tdb_context *, tdb_off_t, struct tdb_record *)
+tdb_rec_write: int (struct tdb_context *, tdb_off_t, struct tdb_record *)
+tdb_release_transaction_locks: void (struct tdb_context *)
+tdb_remove_flags: void (struct tdb_context *, unsigned int)
+tdb_reopen: int (struct tdb_context *)
+tdb_reopen_all: int (int)
+tdb_repack: int (struct tdb_context *)
+tdb_set_logging_function: void (struct tdb_context *, const struct tdb_logging_context *)
+tdb_set_max_dead: void (struct tdb_context *, int)
+tdb_setalarm_sigptr: void (struct tdb_context *, volatile sig_atomic_t *)
+tdb_store: int (struct tdb_context *, TDB_DATA, TDB_DATA, int)
+tdb_transaction_cancel: int (struct tdb_context *)
+tdb_transaction_commit: int (struct tdb_context *)
+tdb_transaction_lock: int (struct tdb_context *, int, enum tdb_lock_flags)
+tdb_transaction_prepare_commit: int (struct tdb_context *)
+tdb_transaction_recover: int (struct tdb_context *)
+tdb_transaction_start: int (struct tdb_context *)
+tdb_transaction_start_nonblock: int (struct tdb_context *)
+tdb_transaction_unlock: int (struct tdb_context *, int)
+tdb_traverse: int (struct tdb_context *, tdb_traverse_func, void *)
+tdb_traverse_read: int (struct tdb_context *, tdb_traverse_func, void *)
+tdb_unlock: int (struct tdb_context *, int, int)
+tdb_unlock_record: int (struct tdb_context *, tdb_off_t)
+tdb_unlockall: int (struct tdb_context *)
+tdb_unlockall_read: int (struct tdb_context *)
+tdb_validate_freelist: int (struct tdb_context *, int *)
+tdb_wipe_all: int (struct tdb_context *)
+tdb_write_lock_record: int (struct tdb_context *, tdb_off_t)
+tdb_write_unlock_record: int (struct tdb_context *, tdb_off_t)
diff --git a/ctdb/lib/tdb/ABI/tdb-1.2.10.sigs b/ctdb/lib/tdb/ABI/tdb-1.2.10.sigs
new file mode 100644
index 0000000000..61f6c19ee3
--- /dev/null
+++ b/ctdb/lib/tdb/ABI/tdb-1.2.10.sigs
@@ -0,0 +1,66 @@
+tdb_add_flags: void (struct tdb_context *, unsigned int)
+tdb_append: int (struct tdb_context *, TDB_DATA, TDB_DATA)
+tdb_chainlock: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_mark: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_nonblock: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_read: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_unmark: int (struct tdb_context *, TDB_DATA)
+tdb_chainunlock: int (struct tdb_context *, TDB_DATA)
+tdb_chainunlock_read: int (struct tdb_context *, TDB_DATA)
+tdb_check: int (struct tdb_context *, int (*)(TDB_DATA, TDB_DATA, void *), void *)
+tdb_close: int (struct tdb_context *)
+tdb_delete: int (struct tdb_context *, TDB_DATA)
+tdb_dump_all: void (struct tdb_context *)
+tdb_enable_seqnum: void (struct tdb_context *)
+tdb_error: enum TDB_ERROR (struct tdb_context *)
+tdb_errorstr: const char *(struct tdb_context *)
+tdb_exists: int (struct tdb_context *, TDB_DATA)
+tdb_fd: int (struct tdb_context *)
+tdb_fetch: TDB_DATA (struct tdb_context *, TDB_DATA)
+tdb_firstkey: TDB_DATA (struct tdb_context *)
+tdb_freelist_size: int (struct tdb_context *)
+tdb_get_flags: int (struct tdb_context *)
+tdb_get_logging_private: void *(struct tdb_context *)
+tdb_get_seqnum: int (struct tdb_context *)
+tdb_hash_size: int (struct tdb_context *)
+tdb_increment_seqnum_nonblock: void (struct tdb_context *)
+tdb_jenkins_hash: unsigned int (TDB_DATA *)
+tdb_lock_nonblock: int (struct tdb_context *, int, int)
+tdb_lockall: int (struct tdb_context *)
+tdb_lockall_mark: int (struct tdb_context *)
+tdb_lockall_nonblock: int (struct tdb_context *)
+tdb_lockall_read: int (struct tdb_context *)
+tdb_lockall_read_nonblock: int (struct tdb_context *)
+tdb_lockall_unmark: int (struct tdb_context *)
+tdb_log_fn: tdb_log_func (struct tdb_context *)
+tdb_map_size: size_t (struct tdb_context *)
+tdb_name: const char *(struct tdb_context *)
+tdb_nextkey: TDB_DATA (struct tdb_context *, TDB_DATA)
+tdb_null: dptr = 0xXXXX, dsize = 0
+tdb_open: struct tdb_context *(const char *, int, int, int, mode_t)
+tdb_open_ex: struct tdb_context *(const char *, int, int, int, mode_t, const struct tdb_logging_context *, tdb_hash_func)
+tdb_parse_record: int (struct tdb_context *, TDB_DATA, int (*)(TDB_DATA, TDB_DATA, void *), void *)
+tdb_printfreelist: int (struct tdb_context *)
+tdb_remove_flags: void (struct tdb_context *, unsigned int)
+tdb_reopen: int (struct tdb_context *)
+tdb_reopen_all: int (int)
+tdb_repack: int (struct tdb_context *)
+tdb_set_logging_function: void (struct tdb_context *, const struct tdb_logging_context *)
+tdb_set_max_dead: void (struct tdb_context *, int)
+tdb_setalarm_sigptr: void (struct tdb_context *, volatile sig_atomic_t *)
+tdb_store: int (struct tdb_context *, TDB_DATA, TDB_DATA, int)
+tdb_summary: char *(struct tdb_context *)
+tdb_transaction_cancel: int (struct tdb_context *)
+tdb_transaction_commit: int (struct tdb_context *)
+tdb_transaction_prepare_commit: int (struct tdb_context *)
+tdb_transaction_start: int (struct tdb_context *)
+tdb_transaction_start_nonblock: int (struct tdb_context *)
+tdb_transaction_write_lock_mark: int (struct tdb_context *)
+tdb_transaction_write_lock_unmark: int (struct tdb_context *)
+tdb_traverse: int (struct tdb_context *, tdb_traverse_func, void *)
+tdb_traverse_read: int (struct tdb_context *, tdb_traverse_func, void *)
+tdb_unlock: int (struct tdb_context *, int, int)
+tdb_unlockall: int (struct tdb_context *)
+tdb_unlockall_read: int (struct tdb_context *)
+tdb_validate_freelist: int (struct tdb_context *, int *)
+tdb_wipe_all: int (struct tdb_context *)
diff --git a/ctdb/lib/tdb/ABI/tdb-1.2.11.sigs b/ctdb/lib/tdb/ABI/tdb-1.2.11.sigs
new file mode 100644
index 0000000000..d727f2163d
--- /dev/null
+++ b/ctdb/lib/tdb/ABI/tdb-1.2.11.sigs
@@ -0,0 +1,67 @@
+tdb_add_flags: void (struct tdb_context *, unsigned int)
+tdb_append: int (struct tdb_context *, TDB_DATA, TDB_DATA)
+tdb_chainlock: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_mark: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_nonblock: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_read: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_unmark: int (struct tdb_context *, TDB_DATA)
+tdb_chainunlock: int (struct tdb_context *, TDB_DATA)
+tdb_chainunlock_read: int (struct tdb_context *, TDB_DATA)
+tdb_check: int (struct tdb_context *, int (*)(TDB_DATA, TDB_DATA, void *), void *)
+tdb_close: int (struct tdb_context *)
+tdb_delete: int (struct tdb_context *, TDB_DATA)
+tdb_dump_all: void (struct tdb_context *)
+tdb_enable_seqnum: void (struct tdb_context *)
+tdb_error: enum TDB_ERROR (struct tdb_context *)
+tdb_errorstr: const char *(struct tdb_context *)
+tdb_exists: int (struct tdb_context *, TDB_DATA)
+tdb_fd: int (struct tdb_context *)
+tdb_fetch: TDB_DATA (struct tdb_context *, TDB_DATA)
+tdb_firstkey: TDB_DATA (struct tdb_context *)
+tdb_freelist_size: int (struct tdb_context *)
+tdb_get_flags: int (struct tdb_context *)
+tdb_get_logging_private: void *(struct tdb_context *)
+tdb_get_seqnum: int (struct tdb_context *)
+tdb_hash_size: int (struct tdb_context *)
+tdb_increment_seqnum_nonblock: void (struct tdb_context *)
+tdb_jenkins_hash: unsigned int (TDB_DATA *)
+tdb_lock_nonblock: int (struct tdb_context *, int, int)
+tdb_lockall: int (struct tdb_context *)
+tdb_lockall_mark: int (struct tdb_context *)
+tdb_lockall_nonblock: int (struct tdb_context *)
+tdb_lockall_read: int (struct tdb_context *)
+tdb_lockall_read_nonblock: int (struct tdb_context *)
+tdb_lockall_unmark: int (struct tdb_context *)
+tdb_log_fn: tdb_log_func (struct tdb_context *)
+tdb_map_size: size_t (struct tdb_context *)
+tdb_name: const char *(struct tdb_context *)
+tdb_nextkey: TDB_DATA (struct tdb_context *, TDB_DATA)
+tdb_null: dptr = 0xXXXX, dsize = 0
+tdb_open: struct tdb_context *(const char *, int, int, int, mode_t)
+tdb_open_ex: struct tdb_context *(const char *, int, int, int, mode_t, const struct tdb_logging_context *, tdb_hash_func)
+tdb_parse_record: int (struct tdb_context *, TDB_DATA, int (*)(TDB_DATA, TDB_DATA, void *), void *)
+tdb_printfreelist: int (struct tdb_context *)
+tdb_remove_flags: void (struct tdb_context *, unsigned int)
+tdb_reopen: int (struct tdb_context *)
+tdb_reopen_all: int (int)
+tdb_repack: int (struct tdb_context *)
+tdb_rescue: int (struct tdb_context *, void (*)(TDB_DATA, TDB_DATA, void *), void *)
+tdb_set_logging_function: void (struct tdb_context *, const struct tdb_logging_context *)
+tdb_set_max_dead: void (struct tdb_context *, int)
+tdb_setalarm_sigptr: void (struct tdb_context *, volatile sig_atomic_t *)
+tdb_store: int (struct tdb_context *, TDB_DATA, TDB_DATA, int)
+tdb_summary: char *(struct tdb_context *)
+tdb_transaction_cancel: int (struct tdb_context *)
+tdb_transaction_commit: int (struct tdb_context *)
+tdb_transaction_prepare_commit: int (struct tdb_context *)
+tdb_transaction_start: int (struct tdb_context *)
+tdb_transaction_start_nonblock: int (struct tdb_context *)
+tdb_transaction_write_lock_mark: int (struct tdb_context *)
+tdb_transaction_write_lock_unmark: int (struct tdb_context *)
+tdb_traverse: int (struct tdb_context *, tdb_traverse_func, void *)
+tdb_traverse_read: int (struct tdb_context *, tdb_traverse_func, void *)
+tdb_unlock: int (struct tdb_context *, int, int)
+tdb_unlockall: int (struct tdb_context *)
+tdb_unlockall_read: int (struct tdb_context *)
+tdb_validate_freelist: int (struct tdb_context *, int *)
+tdb_wipe_all: int (struct tdb_context *)
diff --git a/ctdb/lib/tdb/ABI/tdb-1.2.2.sigs b/ctdb/lib/tdb/ABI/tdb-1.2.2.sigs
new file mode 100644
index 0000000000..043790d27e
--- /dev/null
+++ b/ctdb/lib/tdb/ABI/tdb-1.2.2.sigs
@@ -0,0 +1,60 @@
+tdb_add_flags: void (struct tdb_context *, unsigned int)
+tdb_append: int (struct tdb_context *, TDB_DATA, TDB_DATA)
+tdb_chainlock: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_mark: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_nonblock: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_read: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_unmark: int (struct tdb_context *, TDB_DATA)
+tdb_chainunlock: int (struct tdb_context *, TDB_DATA)
+tdb_chainunlock_read: int (struct tdb_context *, TDB_DATA)
+tdb_check: int (struct tdb_context *, int (*)(TDB_DATA, TDB_DATA, void *), void *)
+tdb_close: int (struct tdb_context *)
+tdb_delete: int (struct tdb_context *, TDB_DATA)
+tdb_dump_all: void (struct tdb_context *)
+tdb_enable_seqnum: void (struct tdb_context *)
+tdb_error: enum TDB_ERROR (struct tdb_context *)
+tdb_errorstr: const char *(struct tdb_context *)
+tdb_exists: int (struct tdb_context *, TDB_DATA)
+tdb_fd: int (struct tdb_context *)
+tdb_fetch: TDB_DATA (struct tdb_context *, TDB_DATA)
+tdb_firstkey: TDB_DATA (struct tdb_context *)
+tdb_freelist_size: int (struct tdb_context *)
+tdb_get_flags: int (struct tdb_context *)
+tdb_get_logging_private: void *(struct tdb_context *)
+tdb_get_seqnum: int (struct tdb_context *)
+tdb_hash_size: int (struct tdb_context *)
+tdb_increment_seqnum_nonblock: void (struct tdb_context *)
+tdb_lockall: int (struct tdb_context *)
+tdb_lockall_mark: int (struct tdb_context *)
+tdb_lockall_nonblock: int (struct tdb_context *)
+tdb_lockall_read: int (struct tdb_context *)
+tdb_lockall_read_nonblock: int (struct tdb_context *)
+tdb_lockall_unmark: int (struct tdb_context *)
+tdb_log_fn: tdb_log_func (struct tdb_context *)
+tdb_map_size: size_t (struct tdb_context *)
+tdb_name: const char *(struct tdb_context *)
+tdb_nextkey: TDB_DATA (struct tdb_context *, TDB_DATA)
+tdb_null: dptr = 0xXXXX, dsize = 0
+tdb_open: struct tdb_context *(const char *, int, int, int, mode_t)
+tdb_open_ex: struct tdb_context *(const char *, int, int, int, mode_t, const struct tdb_logging_context *, tdb_hash_func)
+tdb_parse_record: int (struct tdb_context *, TDB_DATA, int (*)(TDB_DATA, TDB_DATA, void *), void *)
+tdb_printfreelist: int (struct tdb_context *)
+tdb_remove_flags: void (struct tdb_context *, unsigned int)
+tdb_reopen: int (struct tdb_context *)
+tdb_reopen_all: int (int)
+tdb_repack: int (struct tdb_context *)
+tdb_set_logging_function: void (struct tdb_context *, const struct tdb_logging_context *)
+tdb_set_max_dead: void (struct tdb_context *, int)
+tdb_setalarm_sigptr: void (struct tdb_context *, volatile sig_atomic_t *)
+tdb_store: int (struct tdb_context *, TDB_DATA, TDB_DATA, int)
+tdb_transaction_cancel: int (struct tdb_context *)
+tdb_transaction_commit: int (struct tdb_context *)
+tdb_transaction_prepare_commit: int (struct tdb_context *)
+tdb_transaction_start: int (struct tdb_context *)
+tdb_transaction_start_nonblock: int (struct tdb_context *)
+tdb_traverse: int (struct tdb_context *, tdb_traverse_func, void *)
+tdb_traverse_read: int (struct tdb_context *, tdb_traverse_func, void *)
+tdb_unlockall: int (struct tdb_context *)
+tdb_unlockall_read: int (struct tdb_context *)
+tdb_validate_freelist: int (struct tdb_context *, int *)
+tdb_wipe_all: int (struct tdb_context *)
diff --git a/ctdb/lib/tdb/ABI/tdb-1.2.3.sigs b/ctdb/lib/tdb/ABI/tdb-1.2.3.sigs
new file mode 100644
index 0000000000..043790d27e
--- /dev/null
+++ b/ctdb/lib/tdb/ABI/tdb-1.2.3.sigs
@@ -0,0 +1,60 @@
+tdb_add_flags: void (struct tdb_context *, unsigned int)
+tdb_append: int (struct tdb_context *, TDB_DATA, TDB_DATA)
+tdb_chainlock: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_mark: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_nonblock: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_read: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_unmark: int (struct tdb_context *, TDB_DATA)
+tdb_chainunlock: int (struct tdb_context *, TDB_DATA)
+tdb_chainunlock_read: int (struct tdb_context *, TDB_DATA)
+tdb_check: int (struct tdb_context *, int (*)(TDB_DATA, TDB_DATA, void *), void *)
+tdb_close: int (struct tdb_context *)
+tdb_delete: int (struct tdb_context *, TDB_DATA)
+tdb_dump_all: void (struct tdb_context *)
+tdb_enable_seqnum: void (struct tdb_context *)
+tdb_error: enum TDB_ERROR (struct tdb_context *)
+tdb_errorstr: const char *(struct tdb_context *)
+tdb_exists: int (struct tdb_context *, TDB_DATA)
+tdb_fd: int (struct tdb_context *)
+tdb_fetch: TDB_DATA (struct tdb_context *, TDB_DATA)
+tdb_firstkey: TDB_DATA (struct tdb_context *)
+tdb_freelist_size: int (struct tdb_context *)
+tdb_get_flags: int (struct tdb_context *)
+tdb_get_logging_private: void *(struct tdb_context *)
+tdb_get_seqnum: int (struct tdb_context *)
+tdb_hash_size: int (struct tdb_context *)
+tdb_increment_seqnum_nonblock: void (struct tdb_context *)
+tdb_lockall: int (struct tdb_context *)
+tdb_lockall_mark: int (struct tdb_context *)
+tdb_lockall_nonblock: int (struct tdb_context *)
+tdb_lockall_read: int (struct tdb_context *)
+tdb_lockall_read_nonblock: int (struct tdb_context *)
+tdb_lockall_unmark: int (struct tdb_context *)
+tdb_log_fn: tdb_log_func (struct tdb_context *)
+tdb_map_size: size_t (struct tdb_context *)
+tdb_name: const char *(struct tdb_context *)
+tdb_nextkey: TDB_DATA (struct tdb_context *, TDB_DATA)
+tdb_null: dptr = 0xXXXX, dsize = 0
+tdb_open: struct tdb_context *(const char *, int, int, int, mode_t)
+tdb_open_ex: struct tdb_context *(const char *, int, int, int, mode_t, const struct tdb_logging_context *, tdb_hash_func)
+tdb_parse_record: int (struct tdb_context *, TDB_DATA, int (*)(TDB_DATA, TDB_DATA, void *), void *)
+tdb_printfreelist: int (struct tdb_context *)
+tdb_remove_flags: void (struct tdb_context *, unsigned int)
+tdb_reopen: int (struct tdb_context *)
+tdb_reopen_all: int (int)
+tdb_repack: int (struct tdb_context *)
+tdb_set_logging_function: void (struct tdb_context *, const struct tdb_logging_context *)
+tdb_set_max_dead: void (struct tdb_context *, int)
+tdb_setalarm_sigptr: void (struct tdb_context *, volatile sig_atomic_t *)
+tdb_store: int (struct tdb_context *, TDB_DATA, TDB_DATA, int)
+tdb_transaction_cancel: int (struct tdb_context *)
+tdb_transaction_commit: int (struct tdb_context *)
+tdb_transaction_prepare_commit: int (struct tdb_context *)
+tdb_transaction_start: int (struct tdb_context *)
+tdb_transaction_start_nonblock: int (struct tdb_context *)
+tdb_traverse: int (struct tdb_context *, tdb_traverse_func, void *)
+tdb_traverse_read: int (struct tdb_context *, tdb_traverse_func, void *)
+tdb_unlockall: int (struct tdb_context *)
+tdb_unlockall_read: int (struct tdb_context *)
+tdb_validate_freelist: int (struct tdb_context *, int *)
+tdb_wipe_all: int (struct tdb_context *)
diff --git a/ctdb/lib/tdb/ABI/tdb-1.2.4.sigs b/ctdb/lib/tdb/ABI/tdb-1.2.4.sigs
new file mode 100644
index 0000000000..043790d27e
--- /dev/null
+++ b/ctdb/lib/tdb/ABI/tdb-1.2.4.sigs
@@ -0,0 +1,60 @@
+tdb_add_flags: void (struct tdb_context *, unsigned int)
+tdb_append: int (struct tdb_context *, TDB_DATA, TDB_DATA)
+tdb_chainlock: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_mark: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_nonblock: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_read: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_unmark: int (struct tdb_context *, TDB_DATA)
+tdb_chainunlock: int (struct tdb_context *, TDB_DATA)
+tdb_chainunlock_read: int (struct tdb_context *, TDB_DATA)
+tdb_check: int (struct tdb_context *, int (*)(TDB_DATA, TDB_DATA, void *), void *)
+tdb_close: int (struct tdb_context *)
+tdb_delete: int (struct tdb_context *, TDB_DATA)
+tdb_dump_all: void (struct tdb_context *)
+tdb_enable_seqnum: void (struct tdb_context *)
+tdb_error: enum TDB_ERROR (struct tdb_context *)
+tdb_errorstr: const char *(struct tdb_context *)
+tdb_exists: int (struct tdb_context *, TDB_DATA)
+tdb_fd: int (struct tdb_context *)
+tdb_fetch: TDB_DATA (struct tdb_context *, TDB_DATA)
+tdb_firstkey: TDB_DATA (struct tdb_context *)
+tdb_freelist_size: int (struct tdb_context *)
+tdb_get_flags: int (struct tdb_context *)
+tdb_get_logging_private: void *(struct tdb_context *)
+tdb_get_seqnum: int (struct tdb_context *)
+tdb_hash_size: int (struct tdb_context *)
+tdb_increment_seqnum_nonblock: void (struct tdb_context *)
+tdb_lockall: int (struct tdb_context *)
+tdb_lockall_mark: int (struct tdb_context *)
+tdb_lockall_nonblock: int (struct tdb_context *)
+tdb_lockall_read: int (struct tdb_context *)
+tdb_lockall_read_nonblock: int (struct tdb_context *)
+tdb_lockall_unmark: int (struct tdb_context *)
+tdb_log_fn: tdb_log_func (struct tdb_context *)
+tdb_map_size: size_t (struct tdb_context *)
+tdb_name: const char *(struct tdb_context *)
+tdb_nextkey: TDB_DATA (struct tdb_context *, TDB_DATA)
+tdb_null: dptr = 0xXXXX, dsize = 0
+tdb_open: struct tdb_context *(const char *, int, int, int, mode_t)
+tdb_open_ex: struct tdb_context *(const char *, int, int, int, mode_t, const struct tdb_logging_context *, tdb_hash_func)
+tdb_parse_record: int (struct tdb_context *, TDB_DATA, int (*)(TDB_DATA, TDB_DATA, void *), void *)
+tdb_printfreelist: int (struct tdb_context *)
+tdb_remove_flags: void (struct tdb_context *, unsigned int)
+tdb_reopen: int (struct tdb_context *)
+tdb_reopen_all: int (int)
+tdb_repack: int (struct tdb_context *)
+tdb_set_logging_function: void (struct tdb_context *, const struct tdb_logging_context *)
+tdb_set_max_dead: void (struct tdb_context *, int)
+tdb_setalarm_sigptr: void (struct tdb_context *, volatile sig_atomic_t *)
+tdb_store: int (struct tdb_context *, TDB_DATA, TDB_DATA, int)
+tdb_transaction_cancel: int (struct tdb_context *)
+tdb_transaction_commit: int (struct tdb_context *)
+tdb_transaction_prepare_commit: int (struct tdb_context *)
+tdb_transaction_start: int (struct tdb_context *)
+tdb_transaction_start_nonblock: int (struct tdb_context *)
+tdb_traverse: int (struct tdb_context *, tdb_traverse_func, void *)
+tdb_traverse_read: int (struct tdb_context *, tdb_traverse_func, void *)
+tdb_unlockall: int (struct tdb_context *)
+tdb_unlockall_read: int (struct tdb_context *)
+tdb_validate_freelist: int (struct tdb_context *, int *)
+tdb_wipe_all: int (struct tdb_context *)
diff --git a/ctdb/lib/tdb/ABI/tdb-1.2.5.sigs b/ctdb/lib/tdb/ABI/tdb-1.2.5.sigs
new file mode 100644
index 0000000000..1e01f3ba24
--- /dev/null
+++ b/ctdb/lib/tdb/ABI/tdb-1.2.5.sigs
@@ -0,0 +1,61 @@
+tdb_add_flags: void (struct tdb_context *, unsigned int)
+tdb_append: int (struct tdb_context *, TDB_DATA, TDB_DATA)
+tdb_chainlock: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_mark: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_nonblock: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_read: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_unmark: int (struct tdb_context *, TDB_DATA)
+tdb_chainunlock: int (struct tdb_context *, TDB_DATA)
+tdb_chainunlock_read: int (struct tdb_context *, TDB_DATA)
+tdb_check: int (struct tdb_context *, int (*)(TDB_DATA, TDB_DATA, void *), void *)
+tdb_close: int (struct tdb_context *)
+tdb_delete: int (struct tdb_context *, TDB_DATA)
+tdb_dump_all: void (struct tdb_context *)
+tdb_enable_seqnum: void (struct tdb_context *)
+tdb_error: enum TDB_ERROR (struct tdb_context *)
+tdb_errorstr: const char *(struct tdb_context *)
+tdb_exists: int (struct tdb_context *, TDB_DATA)
+tdb_fd: int (struct tdb_context *)
+tdb_fetch: TDB_DATA (struct tdb_context *, TDB_DATA)
+tdb_firstkey: TDB_DATA (struct tdb_context *)
+tdb_freelist_size: int (struct tdb_context *)
+tdb_get_flags: int (struct tdb_context *)
+tdb_get_logging_private: void *(struct tdb_context *)
+tdb_get_seqnum: int (struct tdb_context *)
+tdb_hash_size: int (struct tdb_context *)
+tdb_increment_seqnum_nonblock: void (struct tdb_context *)
+tdb_jenkins_hash: unsigned int (TDB_DATA *)
+tdb_lockall: int (struct tdb_context *)
+tdb_lockall_mark: int (struct tdb_context *)
+tdb_lockall_nonblock: int (struct tdb_context *)
+tdb_lockall_read: int (struct tdb_context *)
+tdb_lockall_read_nonblock: int (struct tdb_context *)
+tdb_lockall_unmark: int (struct tdb_context *)
+tdb_log_fn: tdb_log_func (struct tdb_context *)
+tdb_map_size: size_t (struct tdb_context *)
+tdb_name: const char *(struct tdb_context *)
+tdb_nextkey: TDB_DATA (struct tdb_context *, TDB_DATA)
+tdb_null: dptr = 0xXXXX, dsize = 0
+tdb_open: struct tdb_context *(const char *, int, int, int, mode_t)
+tdb_open_ex: struct tdb_context *(const char *, int, int, int, mode_t, const struct tdb_logging_context *, tdb_hash_func)
+tdb_parse_record: int (struct tdb_context *, TDB_DATA, int (*)(TDB_DATA, TDB_DATA, void *), void *)
+tdb_printfreelist: int (struct tdb_context *)
+tdb_remove_flags: void (struct tdb_context *, unsigned int)
+tdb_reopen: int (struct tdb_context *)
+tdb_reopen_all: int (int)
+tdb_repack: int (struct tdb_context *)
+tdb_set_logging_function: void (struct tdb_context *, const struct tdb_logging_context *)
+tdb_set_max_dead: void (struct tdb_context *, int)
+tdb_setalarm_sigptr: void (struct tdb_context *, volatile sig_atomic_t *)
+tdb_store: int (struct tdb_context *, TDB_DATA, TDB_DATA, int)
+tdb_transaction_cancel: int (struct tdb_context *)
+tdb_transaction_commit: int (struct tdb_context *)
+tdb_transaction_prepare_commit: int (struct tdb_context *)
+tdb_transaction_start: int (struct tdb_context *)
+tdb_transaction_start_nonblock: int (struct tdb_context *)
+tdb_traverse: int (struct tdb_context *, tdb_traverse_func, void *)
+tdb_traverse_read: int (struct tdb_context *, tdb_traverse_func, void *)
+tdb_unlockall: int (struct tdb_context *)
+tdb_unlockall_read: int (struct tdb_context *)
+tdb_validate_freelist: int (struct tdb_context *, int *)
+tdb_wipe_all: int (struct tdb_context *)
diff --git a/ctdb/lib/tdb/ABI/tdb-1.2.6.sigs b/ctdb/lib/tdb/ABI/tdb-1.2.6.sigs
new file mode 100644
index 0000000000..1e01f3ba24
--- /dev/null
+++ b/ctdb/lib/tdb/ABI/tdb-1.2.6.sigs
@@ -0,0 +1,61 @@
+tdb_add_flags: void (struct tdb_context *, unsigned int)
+tdb_append: int (struct tdb_context *, TDB_DATA, TDB_DATA)
+tdb_chainlock: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_mark: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_nonblock: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_read: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_unmark: int (struct tdb_context *, TDB_DATA)
+tdb_chainunlock: int (struct tdb_context *, TDB_DATA)
+tdb_chainunlock_read: int (struct tdb_context *, TDB_DATA)
+tdb_check: int (struct tdb_context *, int (*)(TDB_DATA, TDB_DATA, void *), void *)
+tdb_close: int (struct tdb_context *)
+tdb_delete: int (struct tdb_context *, TDB_DATA)
+tdb_dump_all: void (struct tdb_context *)
+tdb_enable_seqnum: void (struct tdb_context *)
+tdb_error: enum TDB_ERROR (struct tdb_context *)
+tdb_errorstr: const char *(struct tdb_context *)
+tdb_exists: int (struct tdb_context *, TDB_DATA)
+tdb_fd: int (struct tdb_context *)
+tdb_fetch: TDB_DATA (struct tdb_context *, TDB_DATA)
+tdb_firstkey: TDB_DATA (struct tdb_context *)
+tdb_freelist_size: int (struct tdb_context *)
+tdb_get_flags: int (struct tdb_context *)
+tdb_get_logging_private: void *(struct tdb_context *)
+tdb_get_seqnum: int (struct tdb_context *)
+tdb_hash_size: int (struct tdb_context *)
+tdb_increment_seqnum_nonblock: void (struct tdb_context *)
+tdb_jenkins_hash: unsigned int (TDB_DATA *)
+tdb_lockall: int (struct tdb_context *)
+tdb_lockall_mark: int (struct tdb_context *)
+tdb_lockall_nonblock: int (struct tdb_context *)
+tdb_lockall_read: int (struct tdb_context *)
+tdb_lockall_read_nonblock: int (struct tdb_context *)
+tdb_lockall_unmark: int (struct tdb_context *)
+tdb_log_fn: tdb_log_func (struct tdb_context *)
+tdb_map_size: size_t (struct tdb_context *)
+tdb_name: const char *(struct tdb_context *)
+tdb_nextkey: TDB_DATA (struct tdb_context *, TDB_DATA)
+tdb_null: dptr = 0xXXXX, dsize = 0
+tdb_open: struct tdb_context *(const char *, int, int, int, mode_t)
+tdb_open_ex: struct tdb_context *(const char *, int, int, int, mode_t, const struct tdb_logging_context *, tdb_hash_func)
+tdb_parse_record: int (struct tdb_context *, TDB_DATA, int (*)(TDB_DATA, TDB_DATA, void *), void *)
+tdb_printfreelist: int (struct tdb_context *)
+tdb_remove_flags: void (struct tdb_context *, unsigned int)
+tdb_reopen: int (struct tdb_context *)
+tdb_reopen_all: int (int)
+tdb_repack: int (struct tdb_context *)
+tdb_set_logging_function: void (struct tdb_context *, const struct tdb_logging_context *)
+tdb_set_max_dead: void (struct tdb_context *, int)
+tdb_setalarm_sigptr: void (struct tdb_context *, volatile sig_atomic_t *)
+tdb_store: int (struct tdb_context *, TDB_DATA, TDB_DATA, int)
+tdb_transaction_cancel: int (struct tdb_context *)
+tdb_transaction_commit: int (struct tdb_context *)
+tdb_transaction_prepare_commit: int (struct tdb_context *)
+tdb_transaction_start: int (struct tdb_context *)
+tdb_transaction_start_nonblock: int (struct tdb_context *)
+tdb_traverse: int (struct tdb_context *, tdb_traverse_func, void *)
+tdb_traverse_read: int (struct tdb_context *, tdb_traverse_func, void *)
+tdb_unlockall: int (struct tdb_context *)
+tdb_unlockall_read: int (struct tdb_context *)
+tdb_validate_freelist: int (struct tdb_context *, int *)
+tdb_wipe_all: int (struct tdb_context *)
diff --git a/ctdb/lib/tdb/ABI/tdb-1.2.7.sigs b/ctdb/lib/tdb/ABI/tdb-1.2.7.sigs
new file mode 100644
index 0000000000..1e01f3ba24
--- /dev/null
+++ b/ctdb/lib/tdb/ABI/tdb-1.2.7.sigs
@@ -0,0 +1,61 @@
+tdb_add_flags: void (struct tdb_context *, unsigned int)
+tdb_append: int (struct tdb_context *, TDB_DATA, TDB_DATA)
+tdb_chainlock: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_mark: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_nonblock: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_read: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_unmark: int (struct tdb_context *, TDB_DATA)
+tdb_chainunlock: int (struct tdb_context *, TDB_DATA)
+tdb_chainunlock_read: int (struct tdb_context *, TDB_DATA)
+tdb_check: int (struct tdb_context *, int (*)(TDB_DATA, TDB_DATA, void *), void *)
+tdb_close: int (struct tdb_context *)
+tdb_delete: int (struct tdb_context *, TDB_DATA)
+tdb_dump_all: void (struct tdb_context *)
+tdb_enable_seqnum: void (struct tdb_context *)
+tdb_error: enum TDB_ERROR (struct tdb_context *)
+tdb_errorstr: const char *(struct tdb_context *)
+tdb_exists: int (struct tdb_context *, TDB_DATA)
+tdb_fd: int (struct tdb_context *)
+tdb_fetch: TDB_DATA (struct tdb_context *, TDB_DATA)
+tdb_firstkey: TDB_DATA (struct tdb_context *)
+tdb_freelist_size: int (struct tdb_context *)
+tdb_get_flags: int (struct tdb_context *)
+tdb_get_logging_private: void *(struct tdb_context *)
+tdb_get_seqnum: int (struct tdb_context *)
+tdb_hash_size: int (struct tdb_context *)
+tdb_increment_seqnum_nonblock: void (struct tdb_context *)
+tdb_jenkins_hash: unsigned int (TDB_DATA *)
+tdb_lockall: int (struct tdb_context *)
+tdb_lockall_mark: int (struct tdb_context *)
+tdb_lockall_nonblock: int (struct tdb_context *)
+tdb_lockall_read: int (struct tdb_context *)
+tdb_lockall_read_nonblock: int (struct tdb_context *)
+tdb_lockall_unmark: int (struct tdb_context *)
+tdb_log_fn: tdb_log_func (struct tdb_context *)
+tdb_map_size: size_t (struct tdb_context *)
+tdb_name: const char *(struct tdb_context *)
+tdb_nextkey: TDB_DATA (struct tdb_context *, TDB_DATA)
+tdb_null: dptr = 0xXXXX, dsize = 0
+tdb_open: struct tdb_context *(const char *, int, int, int, mode_t)
+tdb_open_ex: struct tdb_context *(const char *, int, int, int, mode_t, const struct tdb_logging_context *, tdb_hash_func)
+tdb_parse_record: int (struct tdb_context *, TDB_DATA, int (*)(TDB_DATA, TDB_DATA, void *), void *)
+tdb_printfreelist: int (struct tdb_context *)
+tdb_remove_flags: void (struct tdb_context *, unsigned int)
+tdb_reopen: int (struct tdb_context *)
+tdb_reopen_all: int (int)
+tdb_repack: int (struct tdb_context *)
+tdb_set_logging_function: void (struct tdb_context *, const struct tdb_logging_context *)
+tdb_set_max_dead: void (struct tdb_context *, int)
+tdb_setalarm_sigptr: void (struct tdb_context *, volatile sig_atomic_t *)
+tdb_store: int (struct tdb_context *, TDB_DATA, TDB_DATA, int)
+tdb_transaction_cancel: int (struct tdb_context *)
+tdb_transaction_commit: int (struct tdb_context *)
+tdb_transaction_prepare_commit: int (struct tdb_context *)
+tdb_transaction_start: int (struct tdb_context *)
+tdb_transaction_start_nonblock: int (struct tdb_context *)
+tdb_traverse: int (struct tdb_context *, tdb_traverse_func, void *)
+tdb_traverse_read: int (struct tdb_context *, tdb_traverse_func, void *)
+tdb_unlockall: int (struct tdb_context *)
+tdb_unlockall_read: int (struct tdb_context *)
+tdb_validate_freelist: int (struct tdb_context *, int *)
+tdb_wipe_all: int (struct tdb_context *)
diff --git a/ctdb/lib/tdb/ABI/tdb-1.2.8.sigs b/ctdb/lib/tdb/ABI/tdb-1.2.8.sigs
new file mode 100644
index 0000000000..1e01f3ba24
--- /dev/null
+++ b/ctdb/lib/tdb/ABI/tdb-1.2.8.sigs
@@ -0,0 +1,61 @@
+tdb_add_flags: void (struct tdb_context *, unsigned int)
+tdb_append: int (struct tdb_context *, TDB_DATA, TDB_DATA)
+tdb_chainlock: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_mark: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_nonblock: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_read: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_unmark: int (struct tdb_context *, TDB_DATA)
+tdb_chainunlock: int (struct tdb_context *, TDB_DATA)
+tdb_chainunlock_read: int (struct tdb_context *, TDB_DATA)
+tdb_check: int (struct tdb_context *, int (*)(TDB_DATA, TDB_DATA, void *), void *)
+tdb_close: int (struct tdb_context *)
+tdb_delete: int (struct tdb_context *, TDB_DATA)
+tdb_dump_all: void (struct tdb_context *)
+tdb_enable_seqnum: void (struct tdb_context *)
+tdb_error: enum TDB_ERROR (struct tdb_context *)
+tdb_errorstr: const char *(struct tdb_context *)
+tdb_exists: int (struct tdb_context *, TDB_DATA)
+tdb_fd: int (struct tdb_context *)
+tdb_fetch: TDB_DATA (struct tdb_context *, TDB_DATA)
+tdb_firstkey: TDB_DATA (struct tdb_context *)
+tdb_freelist_size: int (struct tdb_context *)
+tdb_get_flags: int (struct tdb_context *)
+tdb_get_logging_private: void *(struct tdb_context *)
+tdb_get_seqnum: int (struct tdb_context *)
+tdb_hash_size: int (struct tdb_context *)
+tdb_increment_seqnum_nonblock: void (struct tdb_context *)
+tdb_jenkins_hash: unsigned int (TDB_DATA *)
+tdb_lockall: int (struct tdb_context *)
+tdb_lockall_mark: int (struct tdb_context *)
+tdb_lockall_nonblock: int (struct tdb_context *)
+tdb_lockall_read: int (struct tdb_context *)
+tdb_lockall_read_nonblock: int (struct tdb_context *)
+tdb_lockall_unmark: int (struct tdb_context *)
+tdb_log_fn: tdb_log_func (struct tdb_context *)
+tdb_map_size: size_t (struct tdb_context *)
+tdb_name: const char *(struct tdb_context *)
+tdb_nextkey: TDB_DATA (struct tdb_context *, TDB_DATA)
+tdb_null: dptr = 0xXXXX, dsize = 0
+tdb_open: struct tdb_context *(const char *, int, int, int, mode_t)
+tdb_open_ex: struct tdb_context *(const char *, int, int, int, mode_t, const struct tdb_logging_context *, tdb_hash_func)
+tdb_parse_record: int (struct tdb_context *, TDB_DATA, int (*)(TDB_DATA, TDB_DATA, void *), void *)
+tdb_printfreelist: int (struct tdb_context *)
+tdb_remove_flags: void (struct tdb_context *, unsigned int)
+tdb_reopen: int (struct tdb_context *)
+tdb_reopen_all: int (int)
+tdb_repack: int (struct tdb_context *)
+tdb_set_logging_function: void (struct tdb_context *, const struct tdb_logging_context *)
+tdb_set_max_dead: void (struct tdb_context *, int)
+tdb_setalarm_sigptr: void (struct tdb_context *, volatile sig_atomic_t *)
+tdb_store: int (struct tdb_context *, TDB_DATA, TDB_DATA, int)
+tdb_transaction_cancel: int (struct tdb_context *)
+tdb_transaction_commit: int (struct tdb_context *)
+tdb_transaction_prepare_commit: int (struct tdb_context *)
+tdb_transaction_start: int (struct tdb_context *)
+tdb_transaction_start_nonblock: int (struct tdb_context *)
+tdb_traverse: int (struct tdb_context *, tdb_traverse_func, void *)
+tdb_traverse_read: int (struct tdb_context *, tdb_traverse_func, void *)
+tdb_unlockall: int (struct tdb_context *)
+tdb_unlockall_read: int (struct tdb_context *)
+tdb_validate_freelist: int (struct tdb_context *, int *)
+tdb_wipe_all: int (struct tdb_context *)
diff --git a/ctdb/lib/tdb/ABI/tdb-1.2.9.sigs b/ctdb/lib/tdb/ABI/tdb-1.2.9.sigs
new file mode 100644
index 0000000000..9e4149b4e5
--- /dev/null
+++ b/ctdb/lib/tdb/ABI/tdb-1.2.9.sigs
@@ -0,0 +1,62 @@
+tdb_add_flags: void (struct tdb_context *, unsigned int)
+tdb_append: int (struct tdb_context *, TDB_DATA, TDB_DATA)
+tdb_chainlock: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_mark: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_nonblock: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_read: int (struct tdb_context *, TDB_DATA)
+tdb_chainlock_unmark: int (struct tdb_context *, TDB_DATA)
+tdb_chainunlock: int (struct tdb_context *, TDB_DATA)
+tdb_chainunlock_read: int (struct tdb_context *, TDB_DATA)
+tdb_check: int (struct tdb_context *, int (*)(TDB_DATA, TDB_DATA, void *), void *)
+tdb_close: int (struct tdb_context *)
+tdb_delete: int (struct tdb_context *, TDB_DATA)
+tdb_dump_all: void (struct tdb_context *)
+tdb_enable_seqnum: void (struct tdb_context *)
+tdb_error: enum TDB_ERROR (struct tdb_context *)
+tdb_errorstr: const char *(struct tdb_context *)
+tdb_exists: int (struct tdb_context *, TDB_DATA)
+tdb_fd: int (struct tdb_context *)
+tdb_fetch: TDB_DATA (struct tdb_context *, TDB_DATA)
+tdb_firstkey: TDB_DATA (struct tdb_context *)
+tdb_freelist_size: int (struct tdb_context *)
+tdb_get_flags: int (struct tdb_context *)
+tdb_get_logging_private: void *(struct tdb_context *)
+tdb_get_seqnum: int (struct tdb_context *)
+tdb_hash_size: int (struct tdb_context *)
+tdb_increment_seqnum_nonblock: void (struct tdb_context *)
+tdb_jenkins_hash: unsigned int (TDB_DATA *)
+tdb_lockall: int (struct tdb_context *)
+tdb_lockall_mark: int (struct tdb_context *)
+tdb_lockall_nonblock: int (struct tdb_context *)
+tdb_lockall_read: int (struct tdb_context *)
+tdb_lockall_read_nonblock: int (struct tdb_context *)
+tdb_lockall_unmark: int (struct tdb_context *)
+tdb_log_fn: tdb_log_func (struct tdb_context *)
+tdb_map_size: size_t (struct tdb_context *)
+tdb_name: const char *(struct tdb_context *)
+tdb_nextkey: TDB_DATA (struct tdb_context *, TDB_DATA)
+tdb_null: dptr = 0xXXXX, dsize = 0
+tdb_open: struct tdb_context *(const char *, int, int, int, mode_t)
+tdb_open_ex: struct tdb_context *(const char *, int, int, int, mode_t, const struct tdb_logging_context *, tdb_hash_func)
+tdb_parse_record: int (struct tdb_context *, TDB_DATA, int (*)(TDB_DATA, TDB_DATA, void *), void *)
+tdb_printfreelist: int (struct tdb_context *)
+tdb_remove_flags: void (struct tdb_context *, unsigned int)
+tdb_reopen: int (struct tdb_context *)
+tdb_reopen_all: int (int)
+tdb_repack: int (struct tdb_context *)
+tdb_set_logging_function: void (struct tdb_context *, const struct tdb_logging_context *)
+tdb_set_max_dead: void (struct tdb_context *, int)
+tdb_setalarm_sigptr: void (struct tdb_context *, volatile sig_atomic_t *)
+tdb_store: int (struct tdb_context *, TDB_DATA, TDB_DATA, int)
+tdb_summary: char *(struct tdb_context *)
+tdb_transaction_cancel: int (struct tdb_context *)
+tdb_transaction_commit: int (struct tdb_context *)
+tdb_transaction_prepare_commit: int (struct tdb_context *)
+tdb_transaction_start: int (struct tdb_context *)
+tdb_transaction_start_nonblock: int (struct tdb_context *)
+tdb_traverse: int (struct tdb_context *, tdb_traverse_func, void *)
+tdb_traverse_read: int (struct tdb_context *, tdb_traverse_func, void *)
+tdb_unlockall: int (struct tdb_context *)
+tdb_unlockall_read: int (struct tdb_context *)
+tdb_validate_freelist: int (struct tdb_context *, int *)
+tdb_wipe_all: int (struct tdb_context *)
diff --git a/ctdb/lib/tdb/common/check.c b/ctdb/lib/tdb/common/check.c
new file mode 100644
index 0000000000..313f55cbb0
--- /dev/null
+++ b/ctdb/lib/tdb/common/check.c
@@ -0,0 +1,472 @@
+ /*
+   Unix SMB/CIFS implementation.
+
+   trivial database library
+
+   Copyright (C) Rusty Russell		   2009
+
+     ** NOTE! The following LGPL license applies to the tdb
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+#include "tdb_private.h"
+
+/* Since we opened it, these shouldn't fail unless it's recent corruption. */
+static bool tdb_check_header(struct tdb_context *tdb, tdb_off_t *recovery)
+{
+	struct tdb_header hdr;
+	uint32_t h1, h2;
+
+	if (tdb->methods->tdb_read(tdb, 0, &hdr, sizeof(hdr), 0) == -1)
+		return false;
+	if (strcmp(hdr.magic_food, TDB_MAGIC_FOOD) != 0)
+		goto corrupt;
+
+	CONVERT(hdr);
+	if (hdr.version != TDB_VERSION)
+		goto corrupt;
+
+	if (hdr.rwlocks != 0 && hdr.rwlocks != TDB_HASH_RWLOCK_MAGIC)
+		goto corrupt;
+
+	tdb_header_hash(tdb, &h1, &h2);
+	if (hdr.magic1_hash && hdr.magic2_hash &&
+	    (hdr.magic1_hash != h1 || hdr.magic2_hash != h2))
+		goto corrupt;
+
+	if (hdr.hash_size == 0)
+		goto corrupt;
+
+	if (hdr.hash_size != tdb->header.hash_size)
+		goto corrupt;
+
+	if (hdr.recovery_start != 0 &&
+	    hdr.recovery_start < TDB_DATA_START(tdb->header.hash_size))
+		goto corrupt;
+
+	*recovery = hdr.recovery_start;
+	return true;
+
+corrupt:
+	tdb->ecode = TDB_ERR_CORRUPT;
+	TDB_LOG((tdb, TDB_DEBUG_ERROR, "Header is corrupt\n"));
+	return false;
+}
+
+/* Generic record header check. */
+static bool tdb_check_record(struct tdb_context *tdb,
+			     tdb_off_t off,
+			     const struct tdb_record *rec)
+{
+	tdb_off_t tailer;
+
+	/* Check rec->next: 0 or points to record offset, aligned. */
+	if (rec->next > 0 && rec->next < TDB_DATA_START(tdb->header.hash_size)){
+		TDB_LOG((tdb, TDB_DEBUG_ERROR,
+			 "Record offset %d too small next %d\n",
+			 off, rec->next));
+		goto corrupt;
+	}
+	if (rec->next + sizeof(*rec) < rec->next) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR,
+			 "Record offset %d too large next %d\n",
+			 off, rec->next));
+		goto corrupt;
+	}
+	if ((rec->next % TDB_ALIGNMENT) != 0) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR,
+			 "Record offset %d misaligned next %d\n",
+			 off, rec->next));
+		goto corrupt;
+	}
+	if (tdb->methods->tdb_oob(tdb, rec->next, sizeof(*rec), 0))
+		goto corrupt;
+
+	/* Check rec_len: similar to rec->next, implies next record. */
+	if ((rec->rec_len % TDB_ALIGNMENT) != 0) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR,
+			 "Record offset %d misaligned length %d\n",
+			 off, rec->rec_len));
+		goto corrupt;
+	}
+	/* Must fit tailer. */
+	if (rec->rec_len < sizeof(tailer)) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR,
+			 "Record offset %d too short length %d\n",
+			 off, rec->rec_len));
+		goto corrupt;
+	}
+	/* OOB allows "right at the end" access, so this works for last rec. */
+	if (tdb->methods->tdb_oob(tdb, off, sizeof(*rec)+rec->rec_len, 0))
+		goto corrupt;
+
+	/* Check tailer. */
+	if (tdb_ofs_read(tdb, off+sizeof(*rec)+rec->rec_len-sizeof(tailer),
+			 &tailer) == -1)
+		goto corrupt;
+	if (tailer != sizeof(*rec) + rec->rec_len) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR,
+			 "Record offset %d invalid tailer\n", off));
+		goto corrupt;
+	}
+
+	return true;
+
+corrupt:
+	tdb->ecode = TDB_ERR_CORRUPT;
+	return false;
+}
+
+/* Grab some bytes: may copy if can't use mmap.
+   Caller has already done bounds check. */
+static TDB_DATA get_bytes(struct tdb_context *tdb,
+			  tdb_off_t off, tdb_len_t len)
+{
+	TDB_DATA d;
+
+	d.dsize = len;
+
+	if (tdb->transaction == NULL && tdb->map_ptr != NULL)
+		d.dptr = (unsigned char *)tdb->map_ptr + off;
+	else
+		d.dptr = tdb_alloc_read(tdb, off, d.dsize);
+	return d;
+}
+
+/* Frees data if we're not able to simply use mmap. */
+static void put_bytes(struct tdb_context *tdb, TDB_DATA d)
+{
+	if (tdb->transaction == NULL && tdb->map_ptr != NULL)
+		return;
+	free(d.dptr);
+}
+
+/* We use the excellent Jenkins lookup3 hash; this is based on hash_word2.
+ * See: http://burtleburtle.net/bob/c/lookup3.c
+ */
+#define rot(x,k) (((x)<<(k)) | ((x)>>(32-(k))))
+static void hash(uint32_t key, uint32_t *pc, uint32_t *pb)
+{
+	uint32_t a,b,c;
+
+	/* Set up the internal state */
+	a = b = c = 0xdeadbeef + *pc;
+	c += *pb;
+	a += key;
+	c ^= b; c -= rot(b,14);
+	a ^= c; a -= rot(c,11);
+	b ^= a; b -= rot(a,25);
+	c ^= b; c -= rot(b,16);
+	a ^= c; a -= rot(c,4);
+	b ^= a; b -= rot(a,14);
+	c ^= b; c -= rot(b,24);
+	*pc=c; *pb=b;
+}
+
+/*
+  We want to check that all free records are in the free list
+  (only once), and all free list entries are free records.  Similarly
+  for each hash chain of used records.
+
+  Doing that naively (without walking hash chains, since we want to be
+  linear) means keeping a list of records which have been seen in each
+  hash chain, and another of records pointed to (ie. next pointers
+  from records and the initial hash chain heads).  These two lists
+  should be equal.  This will take 8 bytes per record, and require
+  sorting at the end.
+
+  So instead, we record each offset in a bitmap such a way that
+  recording it twice will cancel out.  Since each offset should appear
+  exactly twice, the bitmap should be zero at the end.
+
+  The approach was inspired by Bloom Filters (see Wikipedia).  For
+  each value, we flip K bits in a bitmap of size N.  The number of
+  distinct arrangements is:
+
+	N! / (K! * (N-K)!)
+
+  Of course, not all arrangements are actually distinct, but testing
+  shows this formula to be close enough.
+
+  So, if K == 8 and N == 256, the probability of two things flipping the same
+  bits is 1 in 409,663,695,276,000.
+
+  Given that ldb uses a hash size of 10000, using 32 bytes per hash chain
+  (320k) seems reasonable.
+*/
+#define NUM_HASHES 8
+#define BITMAP_BITS 256
+
+static void bit_flip(unsigned char bits[], unsigned int idx)
+{
+	bits[idx / CHAR_BIT] ^= (1 << (idx % CHAR_BIT));
+}
+
+/* We record offsets in a bitmap for the particular chain it should be in.  */
+static void record_offset(unsigned char bits[], tdb_off_t off)
+{
+	uint32_t h1 = off, h2 = 0;
+	unsigned int i;
+
+	/* We get two good hash values out of jhash2, so we use both.  Then
+	 * we keep going to produce further hash values. */
+	for (i = 0; i < NUM_HASHES / 2; i++) {
+		hash(off, &h1, &h2);
+		bit_flip(bits, h1 % BITMAP_BITS);
+		bit_flip(bits, h2 % BITMAP_BITS);
+		h2++;
+	}
+}
+
+/* Check that an in-use record is valid. */
+static bool tdb_check_used_record(struct tdb_context *tdb,
+				  tdb_off_t off,
+				  const struct tdb_record *rec,
+				  unsigned char **hashes,
+				  int (*check)(TDB_DATA, TDB_DATA, void *),
+				  void *private_data)
+{
+	TDB_DATA key, data;
+
+	if (!tdb_check_record(tdb, off, rec))
+		return false;
+
+	/* key + data + tailer must fit in record */
+	if (rec->key_len + rec->data_len + sizeof(tdb_off_t) > rec->rec_len) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR,
+			 "Record offset %d too short for contents\n", off));
+		return false;
+	}
+
+	key = get_bytes(tdb, off + sizeof(*rec), rec->key_len);
+	if (!key.dptr)
+		return false;
+
+	if (tdb->hash_fn(&key) != rec->full_hash) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR,
+			 "Record offset %d has incorrect hash\n", off));
+		goto fail_put_key;
+	}
+
+	/* Mark this offset as a known value for this hash bucket. */
+	record_offset(hashes[BUCKET(rec->full_hash)+1], off);
+	/* And similarly if the next pointer is valid. */
+	if (rec->next)
+		record_offset(hashes[BUCKET(rec->full_hash)+1], rec->next);
+
+	/* If they supply a check function and this record isn't dead,
+	   get data and feed it. */
+	if (check && rec->magic != TDB_DEAD_MAGIC) {
+		data = get_bytes(tdb, off + sizeof(*rec) + rec->key_len,
+				 rec->data_len);
+		if (!data.dptr)
+			goto fail_put_key;
+
+		if (check(key, data, private_data) == -1)
+			goto fail_put_data;
+		put_bytes(tdb, data);
+	}
+
+	put_bytes(tdb, key);
+	return true;
+
+fail_put_data:
+	put_bytes(tdb, data);
+fail_put_key:
+	put_bytes(tdb, key);
+	return false;
+}
+
+/* Check that an unused record is valid. */
+static bool tdb_check_free_record(struct tdb_context *tdb,
+				  tdb_off_t off,
+				  const struct tdb_record *rec,
+				  unsigned char **hashes)
+{
+	if (!tdb_check_record(tdb, off, rec))
+		return false;
+
+	/* Mark this offset as a known value for the free list. */
+	record_offset(hashes[0], off);
+	/* And similarly if the next pointer is valid. */
+	if (rec->next)
+		record_offset(hashes[0], rec->next);
+	return true;
+}
+
+/* Slow, but should be very rare. */
+size_t tdb_dead_space(struct tdb_context *tdb, tdb_off_t off)
+{
+	size_t len;
+
+	for (len = 0; off + len < tdb->map_size; len++) {
+		char c;
+		if (tdb->methods->tdb_read(tdb, off, &c, 1, 0))
+			return 0;
+		if (c != 0 && c != 0x42)
+			break;
+	}
+	return len;
+}
+
+_PUBLIC_ int tdb_check(struct tdb_context *tdb,
+	      int (*check)(TDB_DATA key, TDB_DATA data, void *private_data),
+	      void *private_data)
+{
+	unsigned int h;
+	unsigned char **hashes;
+	tdb_off_t off, recovery_start;
+	struct tdb_record rec;
+	bool found_recovery = false;
+	tdb_len_t dead;
+	bool locked;
+
+	/* Read-only databases use no locking at all: it's best-effort.
+	 * We may have a write lock already, so skip that case too. */
+	if (tdb->read_only || tdb->allrecord_lock.count != 0) {
+		locked = false;
+	} else {
+		if (tdb_lockall_read(tdb) == -1)
+			return -1;
+		locked = true;
+	}
+
+	/* Make sure we know true size of the underlying file. */
+	tdb->methods->tdb_oob(tdb, tdb->map_size, 1, 1);
+
+	/* Header must be OK: also gets us the recovery ptr, if any. */
+	if (!tdb_check_header(tdb, &recovery_start))
+		goto unlock;
+
+	/* We should have the whole header, too. */
+	if (tdb->map_size < TDB_DATA_START(tdb->header.hash_size)) {
+		tdb->ecode = TDB_ERR_CORRUPT;
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "File too short for hashes\n"));
+		goto unlock;
+	}
+
+	/* One big malloc: pointers then bit arrays. */
+	hashes = (unsigned char **)calloc(
+			1, sizeof(hashes[0]) * (1+tdb->header.hash_size)
+			+ BITMAP_BITS / CHAR_BIT * (1+tdb->header.hash_size));
+	if (!hashes) {
+		tdb->ecode = TDB_ERR_OOM;
+		goto unlock;
+	}
+
+	/* Initialize pointers */
+	hashes[0] = (unsigned char *)(&hashes[1+tdb->header.hash_size]);
+	for (h = 1; h < 1+tdb->header.hash_size; h++)
+		hashes[h] = hashes[h-1] + BITMAP_BITS / CHAR_BIT;
+
+	/* Freelist and hash headers are all in a row: read them. */
+	for (h = 0; h < 1+tdb->header.hash_size; h++) {
+		if (tdb_ofs_read(tdb, FREELIST_TOP + h*sizeof(tdb_off_t),
+				 &off) == -1)
+			goto free;
+		if (off)
+			record_offset(hashes[h], off);
+	}
+
+	/* For each record, read it in and check it's ok. */
+	for (off = TDB_DATA_START(tdb->header.hash_size);
+	     off < tdb->map_size;
+	     off += sizeof(rec) + rec.rec_len) {
+		if (tdb->methods->tdb_read(tdb, off, &rec, sizeof(rec),
+					   DOCONV()) == -1)
+			goto free;
+		switch (rec.magic) {
+		case TDB_MAGIC:
+		case TDB_DEAD_MAGIC:
+			if (!tdb_check_used_record(tdb, off, &rec, hashes,
+						   check, private_data))
+				goto free;
+			break;
+		case TDB_FREE_MAGIC:
+			if (!tdb_check_free_record(tdb, off, &rec, hashes))
+				goto free;
+			break;
+		/* If we crash after ftruncate, we can get zeroes or fill. */
+		case TDB_RECOVERY_INVALID_MAGIC:
+		case 0x42424242:
+			if (recovery_start == off) {
+				found_recovery = true;
+				break;
+			}
+			dead = tdb_dead_space(tdb, off);
+			if (dead < sizeof(rec))
+				goto corrupt;
+
+			TDB_LOG((tdb, TDB_DEBUG_ERROR,
+				 "Dead space at %d-%d (of %u)\n",
+				 off, off + dead, tdb->map_size));
+			rec.rec_len = dead - sizeof(rec);
+			break;
+		case TDB_RECOVERY_MAGIC:
+			if (recovery_start != off) {
+				TDB_LOG((tdb, TDB_DEBUG_ERROR,
+					 "Unexpected recovery record at offset %d\n",
+					 off));
+				goto free;
+			}
+			found_recovery = true;
+			break;
+		default: ;
+		corrupt:
+			tdb->ecode = TDB_ERR_CORRUPT;
+			TDB_LOG((tdb, TDB_DEBUG_ERROR,
+				 "Bad magic 0x%x at offset %d\n",
+				 rec.magic, off));
+			goto free;
+		}
+	}
+
+	/* Now, hashes should all be empty: each record exists and is referred
+	 * to by one other. */
+	for (h = 0; h < 1+tdb->header.hash_size; h++) {
+		unsigned int i;
+		for (i = 0; i < BITMAP_BITS / CHAR_BIT; i++) {
+			if (hashes[h][i] != 0) {
+				tdb->ecode = TDB_ERR_CORRUPT;
+				TDB_LOG((tdb, TDB_DEBUG_ERROR,
+					 "Hashes do not match records\n"));
+				goto free;
+			}
+		}
+	}
+
+	/* We must have found recovery area if there was one. */
+	if (recovery_start != 0 && !found_recovery) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR,
+			 "Expected a recovery area at %u\n",
+			 recovery_start));
+		goto free;
+	}
+
+	free(hashes);
+	if (locked) {
+		tdb_unlockall_read(tdb);
+	}
+	return 0;
+
+free:
+	free(hashes);
+unlock:
+	if (locked) {
+		tdb_unlockall_read(tdb);
+	}
+	return -1;
+}
diff --git a/ctdb/lib/tdb/common/dump.c b/ctdb/lib/tdb/common/dump.c
new file mode 100644
index 0000000000..67de04e37c
--- /dev/null
+++ b/ctdb/lib/tdb/common/dump.c
@@ -0,0 +1,137 @@
+ /* 
+   Unix SMB/CIFS implementation.
+
+   trivial database library
+
+   Copyright (C) Andrew Tridgell              1999-2005
+   Copyright (C) Paul `Rusty' Russell		   2000
+   Copyright (C) Jeremy Allison			   2000-2003
+
+     ** NOTE! The following LGPL license applies to the tdb
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "tdb_private.h"
+
+static tdb_off_t tdb_dump_record(struct tdb_context *tdb, int hash,
+				 tdb_off_t offset)
+{
+	struct tdb_record rec;
+	tdb_off_t tailer_ofs, tailer;
+
+	if (tdb->methods->tdb_read(tdb, offset, (char *)&rec, 
+				   sizeof(rec), DOCONV()) == -1) {
+		printf("ERROR: failed to read record at %u\n", offset);
+		return 0;
+	}
+
+	printf(" rec: hash=%d offset=0x%08x next=0x%08x rec_len=%d "
+	       "key_len=%d data_len=%d full_hash=0x%x magic=0x%x\n",
+	       hash, offset, rec.next, rec.rec_len, rec.key_len, rec.data_len,
+	       rec.full_hash, rec.magic);
+
+	tailer_ofs = offset + sizeof(rec) + rec.rec_len - sizeof(tdb_off_t);
+
+	if (tdb_ofs_read(tdb, tailer_ofs, &tailer) == -1) {
+		printf("ERROR: failed to read tailer at %u\n", tailer_ofs);
+		return rec.next;
+	}
+
+	if (tailer != rec.rec_len + sizeof(rec)) {
+		printf("ERROR: tailer does not match record! tailer=%u totalsize=%u\n",
+				(unsigned int)tailer, (unsigned int)(rec.rec_len + sizeof(rec)));
+	}
+	return rec.next;
+}
+
+static int tdb_dump_chain(struct tdb_context *tdb, int i)
+{
+	tdb_off_t rec_ptr, top;
+
+	top = TDB_HASH_TOP(i);
+
+	if (tdb_lock(tdb, i, F_WRLCK) != 0)
+		return -1;
+
+	if (tdb_ofs_read(tdb, top, &rec_ptr) == -1)
+		return tdb_unlock(tdb, i, F_WRLCK);
+
+	if (rec_ptr)
+		printf("hash=%d\n", i);
+
+	while (rec_ptr) {
+		rec_ptr = tdb_dump_record(tdb, i, rec_ptr);
+	}
+
+	return tdb_unlock(tdb, i, F_WRLCK);
+}
+
+_PUBLIC_ void tdb_dump_all(struct tdb_context *tdb)
+{
+	int i;
+	for (i=0;i<tdb->header.hash_size;i++) {
+		tdb_dump_chain(tdb, i);
+	}
+	printf("freelist:\n");
+	tdb_dump_chain(tdb, -1);
+}
+
+_PUBLIC_ int tdb_printfreelist(struct tdb_context *tdb)
+{
+	int ret;
+	long total_free = 0;
+	tdb_off_t offset, rec_ptr;
+	struct tdb_record rec;
+
+	if ((ret = tdb_lock(tdb, -1, F_WRLCK)) != 0)
+		return ret;
+
+	offset = FREELIST_TOP;
+
+	/* read in the freelist top */
+	if (tdb_ofs_read(tdb, offset, &rec_ptr) == -1) {
+		tdb_unlock(tdb, -1, F_WRLCK);
+		return 0;
+	}
+
+	printf("freelist top=[0x%08x]\n", rec_ptr );
+	while (rec_ptr) {
+		if (tdb->methods->tdb_read(tdb, rec_ptr, (char *)&rec, 
+					   sizeof(rec), DOCONV()) == -1) {
+			tdb_unlock(tdb, -1, F_WRLCK);
+			return -1;
+		}
+
+		if (rec.magic != TDB_FREE_MAGIC) {
+			printf("bad magic 0x%08x in free list\n", rec.magic);
+			tdb_unlock(tdb, -1, F_WRLCK);
+			return -1;
+		}
+
+		printf("entry offset=[0x%08x], rec.rec_len = [0x%08x (%d)] (end = 0x%08x)\n", 
+		       rec_ptr, rec.rec_len, rec.rec_len, rec_ptr + rec.rec_len);
+		total_free += rec.rec_len;
+
+		/* move to the next record */
+		rec_ptr = rec.next;
+	}
+	printf("total rec_len = [0x%08x (%d)]\n", (int)total_free, 
+               (int)total_free);
+
+	return tdb_unlock(tdb, -1, F_WRLCK);
+}
+
diff --git a/ctdb/lib/tdb/common/error.c b/ctdb/lib/tdb/common/error.c
new file mode 100644
index 0000000000..2aaaa8134e
--- /dev/null
+++ b/ctdb/lib/tdb/common/error.c
@@ -0,0 +1,57 @@
+ /* 
+   Unix SMB/CIFS implementation.
+
+   trivial database library
+
+   Copyright (C) Andrew Tridgell              1999-2005
+   Copyright (C) Paul `Rusty' Russell		   2000
+   Copyright (C) Jeremy Allison			   2000-2003
+
+     ** NOTE! The following LGPL license applies to the tdb
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "tdb_private.h"
+
+_PUBLIC_ enum TDB_ERROR tdb_error(struct tdb_context *tdb)
+{
+	return tdb->ecode;
+}
+
+static struct tdb_errname {
+	enum TDB_ERROR ecode; const char *estring;
+} emap[] = { {TDB_SUCCESS, "Success"},
+	     {TDB_ERR_CORRUPT, "Corrupt database"},
+	     {TDB_ERR_IO, "IO Error"},
+	     {TDB_ERR_LOCK, "Locking error"},
+	     {TDB_ERR_OOM, "Out of memory"},
+	     {TDB_ERR_EXISTS, "Record exists"},
+	     {TDB_ERR_NOLOCK, "Lock exists on other keys"},
+	     {TDB_ERR_EINVAL, "Invalid parameter"},
+	     {TDB_ERR_NOEXIST, "Record does not exist"},
+	     {TDB_ERR_RDONLY, "write not permitted"} };
+
+/* Error string for the last tdb error */
+_PUBLIC_ const char *tdb_errorstr(struct tdb_context *tdb)
+{
+	uint32_t i;
+	for (i = 0; i < sizeof(emap) / sizeof(struct tdb_errname); i++)
+		if (tdb->ecode == emap[i].ecode)
+			return emap[i].estring;
+	return "Invalid error code";
+}
+
diff --git a/ctdb/lib/tdb/common/freelist.c b/ctdb/lib/tdb/common/freelist.c
new file mode 100644
index 0000000000..6358f64a04
--- /dev/null
+++ b/ctdb/lib/tdb/common/freelist.c
@@ -0,0 +1,386 @@
+ /* 
+   Unix SMB/CIFS implementation.
+
+   trivial database library
+
+   Copyright (C) Andrew Tridgell              1999-2005
+   Copyright (C) Paul `Rusty' Russell		   2000
+   Copyright (C) Jeremy Allison			   2000-2003
+
+     ** NOTE! The following LGPL license applies to the tdb
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "tdb_private.h"
+
+/* 'right' merges can involve O(n^2) cost when combined with a
+   traverse, so they are disabled until we find a way to do them in 
+   O(1) time
+*/
+#define USE_RIGHT_MERGES 0
+
+/* read a freelist record and check for simple errors */
+int tdb_rec_free_read(struct tdb_context *tdb, tdb_off_t off, struct tdb_record *rec)
+{
+	if (tdb->methods->tdb_read(tdb, off, rec, sizeof(*rec),DOCONV()) == -1)
+		return -1;
+
+	if (rec->magic == TDB_MAGIC) {
+		/* this happens when a app is showdown while deleting a record - we should
+		   not completely fail when this happens */
+		TDB_LOG((tdb, TDB_DEBUG_WARNING, "tdb_rec_free_read non-free magic 0x%x at offset=%d - fixing\n", 
+			 rec->magic, off));
+		rec->magic = TDB_FREE_MAGIC;
+		if (tdb->methods->tdb_write(tdb, off, rec, sizeof(*rec)) == -1)
+			return -1;
+	}
+
+	if (rec->magic != TDB_FREE_MAGIC) {
+		/* Ensure ecode is set for log fn. */
+		tdb->ecode = TDB_ERR_CORRUPT;
+		TDB_LOG((tdb, TDB_DEBUG_WARNING, "tdb_rec_free_read bad magic 0x%x at offset=%d\n", 
+			   rec->magic, off));
+		return -1;
+	}
+	if (tdb->methods->tdb_oob(tdb, rec->next, sizeof(*rec), 0) != 0)
+		return -1;
+	return 0;
+}
+
+
+#if USE_RIGHT_MERGES
+/* Remove an element from the freelist.  Must have alloc lock. */
+static int remove_from_freelist(struct tdb_context *tdb, tdb_off_t off, tdb_off_t next)
+{
+	tdb_off_t last_ptr, i;
+
+	/* read in the freelist top */
+	last_ptr = FREELIST_TOP;
+	while (tdb_ofs_read(tdb, last_ptr, &i) != -1 && i != 0) {
+		if (i == off) {
+			/* We've found it! */
+			return tdb_ofs_write(tdb, last_ptr, &next);
+		}
+		/* Follow chain (next offset is at start of record) */
+		last_ptr = i;
+	}
+	tdb->ecode = TDB_ERR_CORRUPT;
+	TDB_LOG((tdb, TDB_DEBUG_FATAL,"remove_from_freelist: not on list at off=%d\n", off));
+	return -1;
+}
+#endif
+
+
+/* update a record tailer (must hold allocation lock) */
+static int update_tailer(struct tdb_context *tdb, tdb_off_t offset,
+			 const struct tdb_record *rec)
+{
+	tdb_off_t totalsize;
+
+	/* Offset of tailer from record header */
+	totalsize = sizeof(*rec) + rec->rec_len;
+	return tdb_ofs_write(tdb, offset + totalsize - sizeof(tdb_off_t),
+			 &totalsize);
+}
+
+/* Add an element into the freelist. Merge adjacent records if
+   necessary. */
+int tdb_free(struct tdb_context *tdb, tdb_off_t offset, struct tdb_record *rec)
+{
+	/* Allocation and tailer lock */
+	if (tdb_lock(tdb, -1, F_WRLCK) != 0)
+		return -1;
+
+	/* set an initial tailer, so if we fail we don't leave a bogus record */
+	if (update_tailer(tdb, offset, rec) != 0) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_free: update_tailer failed!\n"));
+		goto fail;
+	}
+
+#if USE_RIGHT_MERGES
+	/* Look right first (I'm an Australian, dammit) */
+	if (offset + sizeof(*rec) + rec->rec_len + sizeof(*rec) <= tdb->map_size) {
+		tdb_off_t right = offset + sizeof(*rec) + rec->rec_len;
+		struct tdb_record r;
+
+		if (tdb->methods->tdb_read(tdb, right, &r, sizeof(r), DOCONV()) == -1) {
+			TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_free: right read failed at %u\n", right));
+			goto left;
+		}
+
+		/* If it's free, expand to include it. */
+		if (r.magic == TDB_FREE_MAGIC) {
+			if (remove_from_freelist(tdb, right, r.next) == -1) {
+				TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_free: right free failed at %u\n", right));
+				goto left;
+			}
+			rec->rec_len += sizeof(r) + r.rec_len;
+			if (update_tailer(tdb, offset, rec) == -1) {
+				TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_free: update_tailer failed at %u\n", offset));
+				goto fail;
+			}
+		}
+	}
+left:
+#endif
+
+	/* Look left */
+	if (offset - sizeof(tdb_off_t) > TDB_DATA_START(tdb->header.hash_size)) {
+		tdb_off_t left = offset - sizeof(tdb_off_t);
+		struct tdb_record l;
+		tdb_off_t leftsize;
+
+		/* Read in tailer and jump back to header */
+		if (tdb_ofs_read(tdb, left, &leftsize) == -1) {
+			TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_free: left offset read failed at %u\n", left));
+			goto update;
+		}
+
+		/* it could be uninitialised data */
+		if (leftsize == 0 || leftsize == TDB_PAD_U32) {
+			goto update;
+		}
+
+		left = offset - leftsize;
+
+		if (leftsize > offset ||
+		    left < TDB_DATA_START(tdb->header.hash_size)) {
+			goto update;
+		}
+
+		/* Now read in the left record */
+		if (tdb->methods->tdb_read(tdb, left, &l, sizeof(l), DOCONV()) == -1) {
+			TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_free: left read failed at %u (%u)\n", left, leftsize));
+			goto update;
+		}
+
+		/* If it's free, expand to include it. */
+		if (l.magic == TDB_FREE_MAGIC) {
+			/* we now merge the new record into the left record, rather than the other 
+			   way around. This makes the operation O(1) instead of O(n). This change
+			   prevents traverse from being O(n^2) after a lot of deletes */
+			l.rec_len += sizeof(*rec) + rec->rec_len;
+			if (tdb_rec_write(tdb, left, &l) == -1) {
+				TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_free: update_left failed at %u\n", left));
+				goto fail;
+			}
+			if (update_tailer(tdb, left, &l) == -1) {
+				TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_free: update_tailer failed at %u\n", offset));
+				goto fail;
+			}
+			tdb_unlock(tdb, -1, F_WRLCK);
+			return 0;
+		}
+	}
+
+update:
+
+	/* Now, prepend to free list */
+	rec->magic = TDB_FREE_MAGIC;
+
+	if (tdb_ofs_read(tdb, FREELIST_TOP, &rec->next) == -1 ||
+	    tdb_rec_write(tdb, offset, rec) == -1 ||
+	    tdb_ofs_write(tdb, FREELIST_TOP, &offset) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_free record write failed at offset=%d\n", offset));
+		goto fail;
+	}
+
+	/* And we're done. */
+	tdb_unlock(tdb, -1, F_WRLCK);
+	return 0;
+
+ fail:
+	tdb_unlock(tdb, -1, F_WRLCK);
+	return -1;
+}
+
+
+
+/* 
+   the core of tdb_allocate - called when we have decided which
+   free list entry to use
+
+   Note that we try to allocate by grabbing data from the end of an existing record,
+   not the beginning. This is so the left merge in a free is more likely to be
+   able to free up the record without fragmentation
+ */
+static tdb_off_t tdb_allocate_ofs(struct tdb_context *tdb, 
+				  tdb_len_t length, tdb_off_t rec_ptr,
+				  struct tdb_record *rec, tdb_off_t last_ptr)
+{
+#define MIN_REC_SIZE (sizeof(struct tdb_record) + sizeof(tdb_off_t) + 8)
+
+	if (rec->rec_len < length + MIN_REC_SIZE) {
+		/* we have to grab the whole record */
+
+		/* unlink it from the previous record */
+		if (tdb_ofs_write(tdb, last_ptr, &rec->next) == -1) {
+			return 0;
+		}
+
+		/* mark it not free */
+		rec->magic = TDB_MAGIC;
+		if (tdb_rec_write(tdb, rec_ptr, rec) == -1) {
+			return 0;
+		}
+		return rec_ptr;
+	}
+
+	/* we're going to just shorten the existing record */
+	rec->rec_len -= (length + sizeof(*rec));
+	if (tdb_rec_write(tdb, rec_ptr, rec) == -1) {
+		return 0;
+	}
+	if (update_tailer(tdb, rec_ptr, rec) == -1) {
+		return 0;
+	}
+
+	/* and setup the new record */
+	rec_ptr += sizeof(*rec) + rec->rec_len;	
+
+	memset(rec, '\0', sizeof(*rec));
+	rec->rec_len = length;
+	rec->magic = TDB_MAGIC;
+
+	if (tdb_rec_write(tdb, rec_ptr, rec) == -1) {
+		return 0;
+	}
+
+	if (update_tailer(tdb, rec_ptr, rec) == -1) {
+		return 0;
+	}
+
+	return rec_ptr;
+}
+
+/* allocate some space from the free list. The offset returned points
+   to a unconnected tdb_record within the database with room for at
+   least length bytes of total data
+
+   0 is returned if the space could not be allocated
+ */
+tdb_off_t tdb_allocate(struct tdb_context *tdb, tdb_len_t length, struct tdb_record *rec)
+{
+	tdb_off_t rec_ptr, last_ptr, newrec_ptr;
+	struct {
+		tdb_off_t rec_ptr, last_ptr;
+		tdb_len_t rec_len;
+	} bestfit;
+	float multiplier = 1.0;
+
+	if (tdb_lock(tdb, -1, F_WRLCK) == -1)
+		return 0;
+
+	/* over-allocate to reduce fragmentation */
+	length *= 1.25;
+
+	/* Extra bytes required for tailer */
+	length += sizeof(tdb_off_t);
+	length = TDB_ALIGN(length, TDB_ALIGNMENT);
+
+ again:
+	last_ptr = FREELIST_TOP;
+
+	/* read in the freelist top */
+	if (tdb_ofs_read(tdb, FREELIST_TOP, &rec_ptr) == -1)
+		goto fail;
+
+	bestfit.rec_ptr = 0;
+	bestfit.last_ptr = 0;
+	bestfit.rec_len = 0;
+
+	/* 
+	   this is a best fit allocation strategy. Originally we used
+	   a first fit strategy, but it suffered from massive fragmentation
+	   issues when faced with a slowly increasing record size.
+	 */
+	while (rec_ptr) {
+		if (tdb_rec_free_read(tdb, rec_ptr, rec) == -1) {
+			goto fail;
+		}
+
+		if (rec->rec_len >= length) {
+			if (bestfit.rec_ptr == 0 ||
+			    rec->rec_len < bestfit.rec_len) {
+				bestfit.rec_len = rec->rec_len;
+				bestfit.rec_ptr = rec_ptr;
+				bestfit.last_ptr = last_ptr;
+			}
+		}
+
+		/* move to the next record */
+		last_ptr = rec_ptr;
+		rec_ptr = rec->next;
+
+		/* if we've found a record that is big enough, then
+		   stop searching if its also not too big. The
+		   definition of 'too big' changes as we scan
+		   through */
+		if (bestfit.rec_len > 0 &&
+		    bestfit.rec_len < length * multiplier) {
+			break;
+		}
+
+		/* this multiplier means we only extremely rarely
+		   search more than 50 or so records. At 50 records we
+		   accept records up to 11 times larger than what we
+		   want */
+		multiplier *= 1.05;
+	}
+
+	if (bestfit.rec_ptr != 0) {
+		if (tdb_rec_free_read(tdb, bestfit.rec_ptr, rec) == -1) {
+			goto fail;
+		}
+
+		newrec_ptr = tdb_allocate_ofs(tdb, length, bestfit.rec_ptr, 
+					      rec, bestfit.last_ptr);
+		tdb_unlock(tdb, -1, F_WRLCK);
+		return newrec_ptr;
+	}
+
+	/* we didn't find enough space. See if we can expand the
+	   database and if we can then try again */
+	if (tdb_expand(tdb, length + sizeof(*rec)) == 0)
+		goto again;
+ fail:
+	tdb_unlock(tdb, -1, F_WRLCK);
+	return 0;
+}
+
+
+
+/* 
+   return the size of the freelist - used to decide if we should repack 
+*/
+_PUBLIC_ int tdb_freelist_size(struct tdb_context *tdb)
+{
+	tdb_off_t ptr;
+	int count=0;
+
+	if (tdb_lock(tdb, -1, F_RDLCK) == -1) {
+		return -1;
+	}
+
+	ptr = FREELIST_TOP;
+	while (tdb_ofs_read(tdb, ptr, &ptr) == 0 && ptr != 0) {
+		count++;
+	}
+
+	tdb_unlock(tdb, -1, F_RDLCK);
+	return count;
+}
diff --git a/ctdb/lib/tdb/common/freelistcheck.c b/ctdb/lib/tdb/common/freelistcheck.c
new file mode 100644
index 0000000000..ab6e78f02d
--- /dev/null
+++ b/ctdb/lib/tdb/common/freelistcheck.c
@@ -0,0 +1,109 @@
+/*
+   Unix SMB/CIFS implementation.
+
+   trivial database library
+
+   Copyright (C) Jeremy Allison                    2006
+
+     ** NOTE! The following LGPL license applies to the tdb
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "tdb_private.h"
+
+/* Check the freelist is good and contains no loops.
+   Very memory intensive - only do this as a consistency
+   checker. Heh heh - uses an in memory tdb as the storage
+   for the "seen" record list. For some reason this strikes
+   me as extremely clever as I don't have to write another tree
+   data structure implementation :-).
+ */
+
+static int seen_insert(struct tdb_context *mem_tdb, tdb_off_t rec_ptr)
+{
+	TDB_DATA key, data;
+
+	memset(&data, '\0', sizeof(data));
+	key.dptr = (unsigned char *)&rec_ptr;
+	key.dsize = sizeof(rec_ptr);
+	return tdb_store(mem_tdb, key, data, TDB_INSERT);
+}
+
+_PUBLIC_ int tdb_validate_freelist(struct tdb_context *tdb, int *pnum_entries)
+{
+	struct tdb_context *mem_tdb = NULL;
+	struct tdb_record rec;
+	tdb_off_t rec_ptr, last_ptr;
+	int ret = -1;
+
+	*pnum_entries = 0;
+
+	mem_tdb = tdb_open("flval", tdb->header.hash_size,
+				TDB_INTERNAL, O_RDWR, 0600);
+	if (!mem_tdb) {
+		return -1;
+	}
+
+	if (tdb_lock(tdb, -1, F_WRLCK) == -1) {
+		tdb_close(mem_tdb);
+		return 0;
+	}
+
+	last_ptr = FREELIST_TOP;
+
+	/* Store the FREELIST_TOP record. */
+	if (seen_insert(mem_tdb, last_ptr) == -1) {
+		tdb->ecode = TDB_ERR_CORRUPT;
+		ret = -1;
+		goto fail;
+	}
+
+	/* read in the freelist top */
+	if (tdb_ofs_read(tdb, FREELIST_TOP, &rec_ptr) == -1) {
+		goto fail;
+	}
+
+	while (rec_ptr) {
+
+		/* If we can't store this record (we've seen it
+		   before) then the free list has a loop and must
+		   be corrupt. */
+
+		if (seen_insert(mem_tdb, rec_ptr)) {
+			tdb->ecode = TDB_ERR_CORRUPT;
+			ret = -1;
+			goto fail;
+		}
+
+		if (tdb_rec_free_read(tdb, rec_ptr, &rec) == -1) {
+			goto fail;
+		}
+
+		/* move to the next record */
+		last_ptr = rec_ptr;
+		rec_ptr = rec.next;
+		*pnum_entries += 1;
+	}
+
+	ret = 0;
+
+  fail:
+
+	tdb_close(mem_tdb);
+	tdb_unlock(tdb, -1, F_WRLCK);
+	return ret;
+}
diff --git a/ctdb/lib/tdb/common/hash.c b/ctdb/lib/tdb/common/hash.c
new file mode 100644
index 0000000000..1eed7221d2
--- /dev/null
+++ b/ctdb/lib/tdb/common/hash.c
@@ -0,0 +1,345 @@
+ /*
+   Unix SMB/CIFS implementation.
+
+   trivial database library
+
+   Copyright (C) Rusty Russell		   2010
+
+     ** NOTE! The following LGPL license applies to the tdb
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+#include "tdb_private.h"
+
+/* This is based on the hash algorithm from gdbm */
+unsigned int tdb_old_hash(TDB_DATA *key)
+{
+	uint32_t value;	/* Used to compute the hash value.  */
+	uint32_t   i;	/* Used to cycle through random values. */
+
+	/* Set the initial value from the key size. */
+	for (value = 0x238F13AF * key->dsize, i=0; i < key->dsize; i++)
+		value = (value + (key->dptr[i] << (i*5 % 24)));
+
+	return (1103515243 * value + 12345);
+}
+
+#ifndef WORDS_BIGENDIAN
+# define HASH_LITTLE_ENDIAN 1
+# define HASH_BIG_ENDIAN 0
+#else
+# define HASH_LITTLE_ENDIAN 0
+# define HASH_BIG_ENDIAN 1
+#endif
+
+/*
+-------------------------------------------------------------------------------
+lookup3.c, by Bob Jenkins, May 2006, Public Domain.
+
+These are functions for producing 32-bit hashes for hash table lookup.
+hash_word(), hashlittle(), hashlittle2(), hashbig(), mix(), and final()
+are externally useful functions.  Routines to test the hash are included
+if SELF_TEST is defined.  You can use this free for any purpose.  It's in
+the public domain.  It has no warranty.
+
+You probably want to use hashlittle().  hashlittle() and hashbig()
+hash byte arrays.  hashlittle() is is faster than hashbig() on
+little-endian machines.  Intel and AMD are little-endian machines.
+On second thought, you probably want hashlittle2(), which is identical to
+hashlittle() except it returns two 32-bit hashes for the price of one.
+You could implement hashbig2() if you wanted but I haven't bothered here.
+
+If you want to find a hash of, say, exactly 7 integers, do
+  a = i1;  b = i2;  c = i3;
+  mix(a,b,c);
+  a += i4; b += i5; c += i6;
+  mix(a,b,c);
+  a += i7;
+  final(a,b,c);
+then use c as the hash value.  If you have a variable length array of
+4-byte integers to hash, use hash_word().  If you have a byte array (like
+a character string), use hashlittle().  If you have several byte arrays, or
+a mix of things, see the comments above hashlittle().
+
+Why is this so big?  I read 12 bytes at a time into 3 4-byte integers,
+then mix those integers.  This is fast (you can do a lot more thorough
+mixing with 12*3 instructions on 3 integers than you can with 3 instructions
+on 1 byte), but shoehorning those bytes into integers efficiently is messy.
+*/
+
+#define hashsize(n) ((uint32_t)1<<(n))
+#define hashmask(n) (hashsize(n)-1)
+#define rot(x,k) (((x)<<(k)) | ((x)>>(32-(k))))
+
+/*
+-------------------------------------------------------------------------------
+mix -- mix 3 32-bit values reversibly.
+
+This is reversible, so any information in (a,b,c) before mix() is
+still in (a,b,c) after mix().
+
+If four pairs of (a,b,c) inputs are run through mix(), or through
+mix() in reverse, there are at least 32 bits of the output that
+are sometimes the same for one pair and different for another pair.
+This was tested for:
+* pairs that differed by one bit, by two bits, in any combination
+  of top bits of (a,b,c), or in any combination of bottom bits of
+  (a,b,c).
+* "differ" is defined as +, -, ^, or ~^.  For + and -, I transformed
+  the output delta to a Gray code (a^(a>>1)) so a string of 1's (as
+  is commonly produced by subtraction) look like a single 1-bit
+  difference.
+* the base values were pseudorandom, all zero but one bit set, or
+  all zero plus a counter that starts at zero.
+
+Some k values for my "a-=c; a^=rot(c,k); c+=b;" arrangement that
+satisfy this are
+    4  6  8 16 19  4
+    9 15  3 18 27 15
+   14  9  3  7 17  3
+Well, "9 15 3 18 27 15" didn't quite get 32 bits diffing
+for "differ" defined as + with a one-bit base and a two-bit delta.  I
+used http://burtleburtle.net/bob/hash/avalanche.html to choose
+the operations, constants, and arrangements of the variables.
+
+This does not achieve avalanche.  There are input bits of (a,b,c)
+that fail to affect some output bits of (a,b,c), especially of a.  The
+most thoroughly mixed value is c, but it doesn't really even achieve
+avalanche in c.
+
+This allows some parallelism.  Read-after-writes are good at doubling
+the number of bits affected, so the goal of mixing pulls in the opposite
+direction as the goal of parallelism.  I did what I could.  Rotates
+seem to cost as much as shifts on every machine I could lay my hands
+on, and rotates are much kinder to the top and bottom bits, so I used
+rotates.
+-------------------------------------------------------------------------------
+*/
+#define mix(a,b,c) \
+{ \
+  a -= c;  a ^= rot(c, 4);  c += b; \
+  b -= a;  b ^= rot(a, 6);  a += c; \
+  c -= b;  c ^= rot(b, 8);  b += a; \
+  a -= c;  a ^= rot(c,16);  c += b; \
+  b -= a;  b ^= rot(a,19);  a += c; \
+  c -= b;  c ^= rot(b, 4);  b += a; \
+}
+
+/*
+-------------------------------------------------------------------------------
+final -- final mixing of 3 32-bit values (a,b,c) into c
+
+Pairs of (a,b,c) values differing in only a few bits will usually
+produce values of c that look totally different.  This was tested for
+* pairs that differed by one bit, by two bits, in any combination
+  of top bits of (a,b,c), or in any combination of bottom bits of
+  (a,b,c).
+* "differ" is defined as +, -, ^, or ~^.  For + and -, I transformed
+  the output delta to a Gray code (a^(a>>1)) so a string of 1's (as
+  is commonly produced by subtraction) look like a single 1-bit
+  difference.
+* the base values were pseudorandom, all zero but one bit set, or
+  all zero plus a counter that starts at zero.
+
+These constants passed:
+ 14 11 25 16 4 14 24
+ 12 14 25 16 4 14 24
+and these came close:
+  4  8 15 26 3 22 24
+ 10  8 15 26 3 22 24
+ 11  8 15 26 3 22 24
+-------------------------------------------------------------------------------
+*/
+#define final(a,b,c) \
+{ \
+  c ^= b; c -= rot(b,14); \
+  a ^= c; a -= rot(c,11); \
+  b ^= a; b -= rot(a,25); \
+  c ^= b; c -= rot(b,16); \
+  a ^= c; a -= rot(c,4);  \
+  b ^= a; b -= rot(a,14); \
+  c ^= b; c -= rot(b,24); \
+}
+
+
+/*
+-------------------------------------------------------------------------------
+hashlittle() -- hash a variable-length key into a 32-bit value
+  k       : the key (the unaligned variable-length array of bytes)
+  length  : the length of the key, counting by bytes
+  val2    : IN: can be any 4-byte value OUT: second 32 bit hash.
+Returns a 32-bit value.  Every bit of the key affects every bit of
+the return value.  Two keys differing by one or two bits will have
+totally different hash values.  Note that the return value is better
+mixed than val2, so use that first.
+
+The best hash table sizes are powers of 2.  There is no need to do
+mod a prime (mod is sooo slow!).  If you need less than 32 bits,
+use a bitmask.  For example, if you need only 10 bits, do
+  h = (h & hashmask(10));
+In which case, the hash table should have hashsize(10) elements.
+
+If you are hashing n strings (uint8_t **)k, do it like this:
+  for (i=0, h=0; i<n; ++i) h = hashlittle( k[i], len[i], h);
+
+By Bob Jenkins, 2006.  bob_jenkins@burtleburtle.net.  You may use this
+code any way you wish, private, educational, or commercial.  It's free.
+
+Use for hash table lookup, or anything where one collision in 2^^32 is
+acceptable.  Do NOT use for cryptographic purposes.
+-------------------------------------------------------------------------------
+*/
+
+static uint32_t hashlittle( const void *key, size_t length )
+{
+  uint32_t a,b,c;                                          /* internal state */
+  union { const void *ptr; size_t i; } u;     /* needed for Mac Powerbook G4 */
+
+  /* Set up the internal state */
+  a = b = c = 0xdeadbeef + ((uint32_t)length);
+
+  u.ptr = key;
+  if (HASH_LITTLE_ENDIAN && ((u.i & 0x3) == 0)) {
+    const uint32_t *k = (const uint32_t *)key;         /* read 32-bit chunks */
+    const uint8_t  *k8;
+
+    /*------ all but last block: aligned reads and affect 32 bits of (a,b,c) */
+    while (length > 12)
+    {
+      a += k[0];
+      b += k[1];
+      c += k[2];
+      mix(a,b,c);
+      length -= 12;
+      k += 3;
+    }
+
+    /*----------------------------- handle the last (probably partial) block */
+    k8 = (const uint8_t *)k;
+    switch(length)
+    {
+    case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
+    case 11: c+=((uint32_t)k8[10])<<16;  /* fall through */
+    case 10: c+=((uint32_t)k8[9])<<8;    /* fall through */
+    case 9 : c+=k8[8];                   /* fall through */
+    case 8 : b+=k[1]; a+=k[0]; break;
+    case 7 : b+=((uint32_t)k8[6])<<16;   /* fall through */
+    case 6 : b+=((uint32_t)k8[5])<<8;    /* fall through */
+    case 5 : b+=k8[4];                   /* fall through */
+    case 4 : a+=k[0]; break;
+    case 3 : a+=((uint32_t)k8[2])<<16;   /* fall through */
+    case 2 : a+=((uint32_t)k8[1])<<8;    /* fall through */
+    case 1 : a+=k8[0]; break;
+    case 0 : return c;
+    }
+  } else if (HASH_LITTLE_ENDIAN && ((u.i & 0x1) == 0)) {
+    const uint16_t *k = (const uint16_t *)key;         /* read 16-bit chunks */
+    const uint8_t  *k8;
+
+    /*--------------- all but last block: aligned reads and different mixing */
+    while (length > 12)
+    {
+      a += k[0] + (((uint32_t)k[1])<<16);
+      b += k[2] + (((uint32_t)k[3])<<16);
+      c += k[4] + (((uint32_t)k[5])<<16);
+      mix(a,b,c);
+      length -= 12;
+      k += 6;
+    }
+
+    /*----------------------------- handle the last (probably partial) block */
+    k8 = (const uint8_t *)k;
+    switch(length)
+    {
+    case 12: c+=k[4]+(((uint32_t)k[5])<<16);
+             b+=k[2]+(((uint32_t)k[3])<<16);
+             a+=k[0]+(((uint32_t)k[1])<<16);
+             break;
+    case 11: c+=((uint32_t)k8[10])<<16;     /* fall through */
+    case 10: c+=k[4];
+             b+=k[2]+(((uint32_t)k[3])<<16);
+             a+=k[0]+(((uint32_t)k[1])<<16);
+             break;
+    case 9 : c+=k8[8];                      /* fall through */
+    case 8 : b+=k[2]+(((uint32_t)k[3])<<16);
+             a+=k[0]+(((uint32_t)k[1])<<16);
+             break;
+    case 7 : b+=((uint32_t)k8[6])<<16;      /* fall through */
+    case 6 : b+=k[2];
+             a+=k[0]+(((uint32_t)k[1])<<16);
+             break;
+    case 5 : b+=k8[4];                      /* fall through */
+    case 4 : a+=k[0]+(((uint32_t)k[1])<<16);
+             break;
+    case 3 : a+=((uint32_t)k8[2])<<16;      /* fall through */
+    case 2 : a+=k[0];
+             break;
+    case 1 : a+=k8[0];
+             break;
+    case 0 : return c;                     /* zero length requires no mixing */
+    }
+
+  } else {                        /* need to read the key one byte at a time */
+    const uint8_t *k = (const uint8_t *)key;
+
+    /*--------------- all but the last block: affect some 32 bits of (a,b,c) */
+    while (length > 12)
+    {
+      a += k[0];
+      a += ((uint32_t)k[1])<<8;
+      a += ((uint32_t)k[2])<<16;
+      a += ((uint32_t)k[3])<<24;
+      b += k[4];
+      b += ((uint32_t)k[5])<<8;
+      b += ((uint32_t)k[6])<<16;
+      b += ((uint32_t)k[7])<<24;
+      c += k[8];
+      c += ((uint32_t)k[9])<<8;
+      c += ((uint32_t)k[10])<<16;
+      c += ((uint32_t)k[11])<<24;
+      mix(a,b,c);
+      length -= 12;
+      k += 12;
+    }
+
+    /*-------------------------------- last block: affect all 32 bits of (c) */
+    switch(length)                   /* all the case statements fall through */
+    {
+    case 12: c+=((uint32_t)k[11])<<24;
+    case 11: c+=((uint32_t)k[10])<<16;
+    case 10: c+=((uint32_t)k[9])<<8;
+    case 9 : c+=k[8];
+    case 8 : b+=((uint32_t)k[7])<<24;
+    case 7 : b+=((uint32_t)k[6])<<16;
+    case 6 : b+=((uint32_t)k[5])<<8;
+    case 5 : b+=k[4];
+    case 4 : a+=((uint32_t)k[3])<<24;
+    case 3 : a+=((uint32_t)k[2])<<16;
+    case 2 : a+=((uint32_t)k[1])<<8;
+    case 1 : a+=k[0];
+             break;
+    case 0 : return c;
+    }
+  }
+
+  final(a,b,c);
+  return c;
+}
+
+_PUBLIC_ unsigned int tdb_jenkins_hash(TDB_DATA *key)
+{
+	return hashlittle(key->dptr, key->dsize);
+}
diff --git a/ctdb/lib/tdb/common/io.c b/ctdb/lib/tdb/common/io.c
new file mode 100644
index 0000000000..25968bfef2
--- /dev/null
+++ b/ctdb/lib/tdb/common/io.c
@@ -0,0 +1,534 @@
+ /* 
+   Unix SMB/CIFS implementation.
+
+   trivial database library
+
+   Copyright (C) Andrew Tridgell              1999-2005
+   Copyright (C) Paul `Rusty' Russell		   2000
+   Copyright (C) Jeremy Allison			   2000-2003
+
+     ** NOTE! The following LGPL license applies to the tdb
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+
+#include "tdb_private.h"
+
+/* check for an out of bounds access - if it is out of bounds then
+   see if the database has been expanded by someone else and expand
+   if necessary 
+*/
+static int tdb_oob(struct tdb_context *tdb, tdb_off_t off, tdb_len_t len,
+		   int probe)
+{
+	struct stat st;
+	if (len + off < len) {
+		if (!probe) {
+			/* Ensure ecode is set for log fn. */
+			tdb->ecode = TDB_ERR_IO;
+			TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_oob off %d len %d wrap\n",
+				 (int)off, (int)len));
+		}
+		return -1;
+	}
+
+	if (off + len <= tdb->map_size)
+		return 0;
+	if (tdb->flags & TDB_INTERNAL) {
+		if (!probe) {
+			/* Ensure ecode is set for log fn. */
+			tdb->ecode = TDB_ERR_IO;
+			TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_oob len %u beyond internal malloc size %u\n",
+				 (int)(off + len), (int)tdb->map_size));
+		}
+		return -1;
+	}
+
+	if (fstat(tdb->fd, &st) == -1) {
+		tdb->ecode = TDB_ERR_IO;
+		return -1;
+	}
+
+	/* Beware >4G files! */
+	if ((tdb_off_t)st.st_size != st.st_size) {
+		/* Ensure ecode is set for log fn. */
+		tdb->ecode = TDB_ERR_IO;
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_oob len %llu too large!\n",
+			 (long long)st.st_size));
+		return -1;
+	}
+
+	/* Unmap, update size, remap.  We do this unconditionally, to handle
+	 * the unusual case where the db is truncated.
+	 *
+	 * This can happen to a child using tdb_reopen_all(true) on a
+	 * TDB_CLEAR_IF_FIRST tdb whose parent crashes: the next
+	 * opener will truncate the database. */
+	if (tdb_munmap(tdb) == -1) {
+		tdb->ecode = TDB_ERR_IO;
+		return -1;
+	}
+	tdb->map_size = st.st_size;
+	if (tdb_mmap(tdb) != 0) {
+		return - 1;
+	}
+
+	if (st.st_size < (size_t)off + len) {
+		if (!probe) {
+			/* Ensure ecode is set for log fn. */
+			tdb->ecode = TDB_ERR_IO;
+			TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_oob len %u beyond eof at %u\n",
+				 (int)(off + len), (int)st.st_size));
+		}
+		return -1;
+	}
+	return 0;
+}
+
+/* write a lump of data at a specified offset */
+static int tdb_write(struct tdb_context *tdb, tdb_off_t off, 
+		     const void *buf, tdb_len_t len)
+{
+	if (len == 0) {
+		return 0;
+	}
+
+	if (tdb->read_only || tdb->traverse_read) {
+		tdb->ecode = TDB_ERR_RDONLY;
+		return -1;
+	}
+
+	if (tdb->methods->tdb_oob(tdb, off, len, 0) != 0)
+		return -1;
+
+	if (tdb->map_ptr) {
+		memcpy(off + (char *)tdb->map_ptr, buf, len);
+	} else {
+#ifdef HAVE_INCOHERENT_MMAP
+		tdb->ecode = TDB_ERR_IO;
+		return -1;
+#else
+		ssize_t written = pwrite(tdb->fd, buf, len, off);
+		if ((written != (ssize_t)len) && (written != -1)) {
+			/* try once more */
+			tdb->ecode = TDB_ERR_IO;
+			TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_write: wrote only "
+				 "%d of %d bytes at %d, trying once more\n",
+				 (int)written, len, off));
+			written = pwrite(tdb->fd, (const char *)buf+written,
+					 len-written,
+					 off+written);
+		}
+		if (written == -1) {
+			/* Ensure ecode is set for log fn. */
+			tdb->ecode = TDB_ERR_IO;
+			TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_write failed at %d "
+				 "len=%d (%s)\n", off, len, strerror(errno)));
+			return -1;
+		} else if (written != (ssize_t)len) {
+			tdb->ecode = TDB_ERR_IO;
+			TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_write: failed to "
+				 "write %d bytes at %d in two attempts\n",
+				 len, off));
+			return -1;
+		}
+#endif
+	}
+	return 0;
+}
+
+/* Endian conversion: we only ever deal with 4 byte quantities */
+void *tdb_convert(void *buf, uint32_t size)
+{
+	uint32_t i, *p = (uint32_t *)buf;
+	for (i = 0; i < size / 4; i++)
+		p[i] = TDB_BYTEREV(p[i]);
+	return buf;
+}
+
+
+/* read a lump of data at a specified offset, maybe convert */
+static int tdb_read(struct tdb_context *tdb, tdb_off_t off, void *buf, 
+		    tdb_len_t len, int cv)
+{
+	if (tdb->methods->tdb_oob(tdb, off, len, 0) != 0) {
+		return -1;
+	}
+
+	if (tdb->map_ptr) {
+		memcpy(buf, off + (char *)tdb->map_ptr, len);
+	} else {
+#ifdef HAVE_INCOHERENT_MMAP
+		tdb->ecode = TDB_ERR_IO;
+		return -1;
+#else
+		ssize_t ret = pread(tdb->fd, buf, len, off);
+		if (ret != (ssize_t)len) {
+			/* Ensure ecode is set for log fn. */
+			tdb->ecode = TDB_ERR_IO;
+			TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_read failed at %d "
+				 "len=%d ret=%d (%s) map_size=%d\n",
+				 (int)off, (int)len, (int)ret, strerror(errno),
+				 (int)tdb->map_size));
+			return -1;
+		}
+#endif
+	}
+	if (cv) {
+		tdb_convert(buf, len);
+	}
+	return 0;
+}
+
+
+
+/*
+  do an unlocked scan of the hash table heads to find the next non-zero head. The value
+  will then be confirmed with the lock held
+*/		
+static void tdb_next_hash_chain(struct tdb_context *tdb, uint32_t *chain)
+{
+	uint32_t h = *chain;
+	if (tdb->map_ptr) {
+		for (;h < tdb->header.hash_size;h++) {
+			if (0 != *(uint32_t *)(TDB_HASH_TOP(h) + (unsigned char *)tdb->map_ptr)) {
+				break;
+			}
+		}
+	} else {
+		uint32_t off=0;
+		for (;h < tdb->header.hash_size;h++) {
+			if (tdb_ofs_read(tdb, TDB_HASH_TOP(h), &off) != 0 || off != 0) {
+				break;
+			}
+		}
+	}
+	(*chain) = h;
+}
+
+
+int tdb_munmap(struct tdb_context *tdb)
+{
+	if (tdb->flags & TDB_INTERNAL)
+		return 0;
+
+#ifdef HAVE_MMAP
+	if (tdb->map_ptr) {
+		int ret;
+
+		ret = munmap(tdb->map_ptr, tdb->map_size);
+		if (ret != 0)
+			return ret;
+	}
+#endif
+	tdb->map_ptr = NULL;
+	return 0;
+}
+
+/* If mmap isn't coherent, *everyone* must always mmap. */
+static bool should_mmap(const struct tdb_context *tdb)
+{
+#ifdef HAVE_INCOHERENT_MMAP
+	return true;
+#else
+	return !(tdb->flags & TDB_NOMMAP);
+#endif
+}
+
+int tdb_mmap(struct tdb_context *tdb)
+{
+	if (tdb->flags & TDB_INTERNAL)
+		return 0;
+
+#ifdef HAVE_MMAP
+	if (should_mmap(tdb)) {
+		tdb->map_ptr = mmap(NULL, tdb->map_size, 
+				    PROT_READ|(tdb->read_only? 0:PROT_WRITE), 
+				    MAP_SHARED|MAP_FILE, tdb->fd, 0);
+
+		/*
+		 * NB. When mmap fails it returns MAP_FAILED *NOT* NULL !!!!
+		 */
+
+		if (tdb->map_ptr == MAP_FAILED) {
+			tdb->map_ptr = NULL;
+			TDB_LOG((tdb, TDB_DEBUG_WARNING, "tdb_mmap failed for size %d (%s)\n", 
+				 tdb->map_size, strerror(errno)));
+#ifdef HAVE_INCOHERENT_MMAP
+			tdb->ecode = TDB_ERR_IO;
+			return -1;
+#endif
+		}
+	} else {
+		tdb->map_ptr = NULL;
+	}
+#else
+	tdb->map_ptr = NULL;
+#endif
+	return 0;
+}
+
+/* expand a file.  we prefer to use ftruncate, as that is what posix
+  says to use for mmap expansion */
+static int tdb_expand_file(struct tdb_context *tdb, tdb_off_t size, tdb_off_t addition)
+{
+	char buf[8192];
+
+	if (tdb->read_only || tdb->traverse_read) {
+		tdb->ecode = TDB_ERR_RDONLY;
+		return -1;
+	}
+
+	if (ftruncate(tdb->fd, size+addition) == -1) {
+		char b = 0;
+		ssize_t written = pwrite(tdb->fd,  &b, 1, (size+addition) - 1);
+		if (written == 0) {
+			/* try once more, potentially revealing errno */
+			written = pwrite(tdb->fd,  &b, 1, (size+addition) - 1);
+		}
+		if (written == 0) {
+			/* again - give up, guessing errno */
+			errno = ENOSPC;
+		}
+		if (written != 1) {
+			TDB_LOG((tdb, TDB_DEBUG_FATAL, "expand_file to %d failed (%s)\n", 
+				 size+addition, strerror(errno)));
+			return -1;
+		}
+	}
+
+	/* now fill the file with something. This ensures that the
+	   file isn't sparse, which would be very bad if we ran out of
+	   disk. This must be done with write, not via mmap */
+	memset(buf, TDB_PAD_BYTE, sizeof(buf));
+	while (addition) {
+		size_t n = addition>sizeof(buf)?sizeof(buf):addition;
+		ssize_t written = pwrite(tdb->fd, buf, n, size);
+		if (written == 0) {
+			/* prevent infinite loops: try _once_ more */
+			written = pwrite(tdb->fd, buf, n, size);
+		}
+		if (written == 0) {
+			/* give up, trying to provide a useful errno */
+			TDB_LOG((tdb, TDB_DEBUG_FATAL, "expand_file write "
+				"returned 0 twice: giving up!\n"));
+			errno = ENOSPC;
+			return -1;
+		} else if (written == -1) {
+			TDB_LOG((tdb, TDB_DEBUG_FATAL, "expand_file write of "
+				 "%d bytes failed (%s)\n", (int)n,
+				 strerror(errno)));
+			return -1;
+		} else if (written != n) {
+			TDB_LOG((tdb, TDB_DEBUG_WARNING, "expand_file: wrote "
+				 "only %d of %d bytes - retrying\n", (int)written,
+				 (int)n));
+		}
+		addition -= written;
+		size += written;
+	}
+	return 0;
+}
+
+
+/* You need 'size', this tells you how much you should expand by. */
+tdb_off_t tdb_expand_adjust(tdb_off_t map_size, tdb_off_t size, int page_size)
+{
+	tdb_off_t new_size, top_size;
+
+	/* limit size in order to avoid using up huge amounts of memory for
+	 * in memory tdbs if an oddball huge record creeps in */
+	if (size > 100 * 1024) {
+		top_size = map_size + size * 2;
+	} else {
+		top_size = map_size + size * 100;
+	}
+
+	/* always make room for at least top_size more records, and at
+	   least 25% more space. if the DB is smaller than 100MiB,
+	   otherwise grow it by 10% only. */
+	if (map_size > 100 * 1024 * 1024) {
+		new_size = map_size * 1.10;
+	} else {
+		new_size = map_size * 1.25;
+	}
+
+	/* Round the database up to a multiple of the page size */
+	new_size = MAX(top_size, new_size);
+	return TDB_ALIGN(new_size, page_size) - map_size;
+}
+
+/* expand the database at least size bytes by expanding the underlying
+   file and doing the mmap again if necessary */
+int tdb_expand(struct tdb_context *tdb, tdb_off_t size)
+{
+	struct tdb_record rec;
+	tdb_off_t offset;
+
+	if (tdb_lock(tdb, -1, F_WRLCK) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "lock failed in tdb_expand\n"));
+		return -1;
+	}
+
+	/* must know about any previous expansions by another process */
+	tdb->methods->tdb_oob(tdb, tdb->map_size, 1, 1);
+
+	size = tdb_expand_adjust(tdb->map_size, size, tdb->page_size);
+
+	/* expand the file itself */
+	if (!(tdb->flags & TDB_INTERNAL)) {
+		if (tdb->methods->tdb_expand_file(tdb, tdb->map_size, size) != 0)
+			goto fail;
+	}
+
+	/* form a new freelist record */
+	offset = tdb->map_size;
+	memset(&rec,'\0',sizeof(rec));
+	rec.rec_len = size - sizeof(rec);
+
+	if (tdb->flags & TDB_INTERNAL) {
+		char *new_map_ptr = (char *)realloc(tdb->map_ptr,
+						    tdb->map_size + size);
+		if (!new_map_ptr) {
+			goto fail;
+		}
+		tdb->map_ptr = new_map_ptr;
+		tdb->map_size += size;
+	} else {
+		/* Explicitly remap: if we're in a transaction, this won't
+		 * happen automatically! */
+		tdb_munmap(tdb);
+		tdb->map_size += size;
+		if (tdb_mmap(tdb) != 0) {
+			goto fail;
+		}
+	}
+
+	/* link it into the free list */
+	if (tdb_free(tdb, offset, &rec) == -1)
+		goto fail;
+
+	tdb_unlock(tdb, -1, F_WRLCK);
+	return 0;
+ fail:
+	tdb_unlock(tdb, -1, F_WRLCK);
+	return -1;
+}
+
+/* read/write a tdb_off_t */
+int tdb_ofs_read(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d)
+{
+	return tdb->methods->tdb_read(tdb, offset, (char*)d, sizeof(*d), DOCONV());
+}
+
+int tdb_ofs_write(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d)
+{
+	tdb_off_t off = *d;
+	return tdb->methods->tdb_write(tdb, offset, CONVERT(off), sizeof(*d));
+}
+
+
+/* read a lump of data, allocating the space for it */
+unsigned char *tdb_alloc_read(struct tdb_context *tdb, tdb_off_t offset, tdb_len_t len)
+{
+	unsigned char *buf;
+
+	/* some systems don't like zero length malloc */
+
+	if (!(buf = (unsigned char *)malloc(len ? len : 1))) {
+		/* Ensure ecode is set for log fn. */
+		tdb->ecode = TDB_ERR_OOM;
+		TDB_LOG((tdb, TDB_DEBUG_ERROR,"tdb_alloc_read malloc failed len=%d (%s)\n",
+			   len, strerror(errno)));
+		return NULL;
+	}
+	if (tdb->methods->tdb_read(tdb, offset, buf, len, 0) == -1) {
+		SAFE_FREE(buf);
+		return NULL;
+	}
+	return buf;
+}
+
+/* Give a piece of tdb data to a parser */
+
+int tdb_parse_data(struct tdb_context *tdb, TDB_DATA key,
+		   tdb_off_t offset, tdb_len_t len,
+		   int (*parser)(TDB_DATA key, TDB_DATA data,
+				 void *private_data),
+		   void *private_data)
+{
+	TDB_DATA data;
+	int result;
+
+	data.dsize = len;
+
+	if ((tdb->transaction == NULL) && (tdb->map_ptr != NULL)) {
+		/*
+		 * Optimize by avoiding the malloc/memcpy/free, point the
+		 * parser directly at the mmap area.
+		 */
+		if (tdb->methods->tdb_oob(tdb, offset, len, 0) != 0) {
+			return -1;
+		}
+		data.dptr = offset + (unsigned char *)tdb->map_ptr;
+		return parser(key, data, private_data);
+	}
+
+	if (!(data.dptr = tdb_alloc_read(tdb, offset, len))) {
+		return -1;
+	}
+
+	result = parser(key, data, private_data);
+	free(data.dptr);
+	return result;
+}
+
+/* read/write a record */
+int tdb_rec_read(struct tdb_context *tdb, tdb_off_t offset, struct tdb_record *rec)
+{
+	if (tdb->methods->tdb_read(tdb, offset, rec, sizeof(*rec),DOCONV()) == -1)
+		return -1;
+	if (TDB_BAD_MAGIC(rec)) {
+		/* Ensure ecode is set for log fn. */
+		tdb->ecode = TDB_ERR_CORRUPT;
+		TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_rec_read bad magic 0x%x at offset=%d\n", rec->magic, offset));
+		return -1;
+	}
+	return tdb->methods->tdb_oob(tdb, rec->next, sizeof(*rec), 0);
+}
+
+int tdb_rec_write(struct tdb_context *tdb, tdb_off_t offset, struct tdb_record *rec)
+{
+	struct tdb_record r = *rec;
+	return tdb->methods->tdb_write(tdb, offset, CONVERT(r), sizeof(r));
+}
+
+static const struct tdb_methods io_methods = {
+	tdb_read,
+	tdb_write,
+	tdb_next_hash_chain,
+	tdb_oob,
+	tdb_expand_file,
+};
+
+/*
+  initialise the default methods table
+*/
+void tdb_io_init(struct tdb_context *tdb)
+{
+	tdb->methods = &io_methods;
+}
diff --git a/ctdb/lib/tdb/common/lock.c b/ctdb/lib/tdb/common/lock.c
new file mode 100644
index 0000000000..260fab6681
--- /dev/null
+++ b/ctdb/lib/tdb/common/lock.c
@@ -0,0 +1,875 @@
+ /* 
+   Unix SMB/CIFS implementation.
+
+   trivial database library
+
+   Copyright (C) Andrew Tridgell              1999-2005
+   Copyright (C) Paul `Rusty' Russell		   2000
+   Copyright (C) Jeremy Allison			   2000-2003
+
+     ** NOTE! The following LGPL license applies to the tdb
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "tdb_private.h"
+
+_PUBLIC_ void tdb_setalarm_sigptr(struct tdb_context *tdb, volatile sig_atomic_t *ptr)
+{
+	tdb->interrupt_sig_ptr = ptr;
+}
+
+static int fcntl_lock(struct tdb_context *tdb,
+		      int rw, off_t off, off_t len, bool waitflag)
+{
+	struct flock fl;
+
+	fl.l_type = rw;
+	fl.l_whence = SEEK_SET;
+	fl.l_start = off;
+	fl.l_len = len;
+	fl.l_pid = 0;
+
+	if (waitflag)
+		return fcntl(tdb->fd, F_SETLKW, &fl);
+	else
+		return fcntl(tdb->fd, F_SETLK, &fl);
+}
+
+static int fcntl_unlock(struct tdb_context *tdb, int rw, off_t off, off_t len)
+{
+	struct flock fl;
+#if 0 /* Check they matched up locks and unlocks correctly. */
+	char line[80];
+	FILE *locks;
+	bool found = false;
+
+	locks = fopen("/proc/locks", "r");
+
+	while (fgets(line, 80, locks)) {
+		char *p;
+		int type, start, l;
+
+		/* eg. 1: FLOCK  ADVISORY  WRITE 2440 08:01:2180826 0 EOF */
+		p = strchr(line, ':') + 1;
+		if (strncmp(p, " POSIX  ADVISORY  ", strlen(" POSIX  ADVISORY  ")))
+			continue;
+		p += strlen(" FLOCK  ADVISORY  ");
+		if (strncmp(p, "READ  ", strlen("READ  ")) == 0)
+			type = F_RDLCK;
+		else if (strncmp(p, "WRITE ", strlen("WRITE ")) == 0)
+			type = F_WRLCK;
+		else
+			abort();
+		p += 6;
+		if (atoi(p) != getpid())
+			continue;
+		p = strchr(strchr(p, ' ') + 1, ' ') + 1;
+		start = atoi(p);
+		p = strchr(p, ' ') + 1;
+		if (strncmp(p, "EOF", 3) == 0)
+			l = 0;
+		else
+			l = atoi(p) - start + 1;
+
+		if (off == start) {
+			if (len != l) {
+				fprintf(stderr, "Len %u should be %u: %s",
+					(int)len, l, line);
+				abort();
+			}
+			if (type != rw) {
+				fprintf(stderr, "Type %s wrong: %s",
+					rw == F_RDLCK ? "READ" : "WRITE", line);
+				abort();
+			}
+			found = true;
+			break;
+		}
+	}
+
+	if (!found) {
+		fprintf(stderr, "Unlock on %u@%u not found!\n",
+			(int)off, (int)len);
+		abort();
+	}
+
+	fclose(locks);
+#endif
+
+	fl.l_type = F_UNLCK;
+	fl.l_whence = SEEK_SET;
+	fl.l_start = off;
+	fl.l_len = len;
+	fl.l_pid = 0;
+
+	return fcntl(tdb->fd, F_SETLKW, &fl);
+}
+
+/* list -1 is the alloc list, otherwise a hash chain. */
+static tdb_off_t lock_offset(int list)
+{
+	return FREELIST_TOP + 4*list;
+}
+
+/* a byte range locking function - return 0 on success
+   this functions locks/unlocks 1 byte at the specified offset.
+
+   On error, errno is also set so that errors are passed back properly
+   through tdb_open(). 
+
+   note that a len of zero means lock to end of file
+*/
+int tdb_brlock(struct tdb_context *tdb,
+	       int rw_type, tdb_off_t offset, size_t len,
+	       enum tdb_lock_flags flags)
+{
+	int ret;
+
+	if (tdb->flags & TDB_NOLOCK) {
+		return 0;
+	}
+
+	if (flags & TDB_LOCK_MARK_ONLY) {
+		return 0;
+	}
+
+	if ((rw_type == F_WRLCK) && (tdb->read_only || tdb->traverse_read)) {
+		tdb->ecode = TDB_ERR_RDONLY;
+		return -1;
+	}
+
+	do {
+		ret = fcntl_lock(tdb, rw_type, offset, len,
+				 flags & TDB_LOCK_WAIT);
+		/* Check for a sigalarm break. */
+		if (ret == -1 && errno == EINTR &&
+				tdb->interrupt_sig_ptr &&
+				*tdb->interrupt_sig_ptr) {
+			break;
+		}
+	} while (ret == -1 && errno == EINTR);
+
+	if (ret == -1) {
+		tdb->ecode = TDB_ERR_LOCK;
+		/* Generic lock error. errno set by fcntl.
+		 * EAGAIN is an expected return from non-blocking
+		 * locks. */
+		if (!(flags & TDB_LOCK_PROBE) && errno != EAGAIN) {
+			TDB_LOG((tdb, TDB_DEBUG_TRACE,"tdb_brlock failed (fd=%d) at offset %d rw_type=%d flags=%d len=%d\n",
+				 tdb->fd, offset, rw_type, flags, (int)len));
+		}
+		return -1;
+	}
+	return 0;
+}
+
+int tdb_brunlock(struct tdb_context *tdb,
+		 int rw_type, tdb_off_t offset, size_t len)
+{
+	int ret;
+
+	if (tdb->flags & TDB_NOLOCK) {
+		return 0;
+	}
+
+	do {
+		ret = fcntl_unlock(tdb, rw_type, offset, len);
+	} while (ret == -1 && errno == EINTR);
+
+	if (ret == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_TRACE,"tdb_brunlock failed (fd=%d) at offset %d rw_type=%d len=%d\n",
+			 tdb->fd, offset, rw_type, (int)len));
+	}
+	return ret;
+}
+
+/*
+  upgrade a read lock to a write lock. This needs to be handled in a
+  special way as some OSes (such as solaris) have too conservative
+  deadlock detection and claim a deadlock when progress can be
+  made. For those OSes we may loop for a while.  
+*/
+int tdb_allrecord_upgrade(struct tdb_context *tdb)
+{
+	int count = 1000;
+
+	if (tdb->allrecord_lock.count != 1) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR,
+			 "tdb_allrecord_upgrade failed: count %u too high\n",
+			 tdb->allrecord_lock.count));
+		return -1;
+	}
+
+	if (tdb->allrecord_lock.off != 1) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR,
+			 "tdb_allrecord_upgrade failed: already upgraded?\n"));
+		return -1;
+	}
+
+	while (count--) {
+		struct timeval tv;
+		if (tdb_brlock(tdb, F_WRLCK, FREELIST_TOP, 0,
+			       TDB_LOCK_WAIT|TDB_LOCK_PROBE) == 0) {
+			tdb->allrecord_lock.ltype = F_WRLCK;
+			tdb->allrecord_lock.off = 0;
+			return 0;
+		}
+		if (errno != EDEADLK) {
+			break;
+		}
+		/* sleep for as short a time as we can - more portable than usleep() */
+		tv.tv_sec = 0;
+		tv.tv_usec = 1;
+		select(0, NULL, NULL, NULL, &tv);
+	}
+	TDB_LOG((tdb, TDB_DEBUG_TRACE,"tdb_allrecord_upgrade failed\n"));
+	return -1;
+}
+
+static struct tdb_lock_type *find_nestlock(struct tdb_context *tdb,
+					   tdb_off_t offset)
+{
+	unsigned int i;
+
+	for (i=0; i<tdb->num_lockrecs; i++) {
+		if (tdb->lockrecs[i].off == offset) {
+			return &tdb->lockrecs[i];
+		}
+	}
+	return NULL;
+}
+
+/* lock an offset in the database. */
+int tdb_nest_lock(struct tdb_context *tdb, uint32_t offset, int ltype,
+		  enum tdb_lock_flags flags)
+{
+	struct tdb_lock_type *new_lck;
+
+	if (offset >= lock_offset(tdb->header.hash_size)) {
+		tdb->ecode = TDB_ERR_LOCK;
+		TDB_LOG((tdb, TDB_DEBUG_ERROR,"tdb_lock: invalid offset %u for ltype=%d\n",
+			 offset, ltype));
+		return -1;
+	}
+	if (tdb->flags & TDB_NOLOCK)
+		return 0;
+
+	new_lck = find_nestlock(tdb, offset);
+	if (new_lck) {
+		/*
+		 * Just increment the in-memory struct, posix locks
+		 * don't stack.
+		 */
+		new_lck->count++;
+		return 0;
+	}
+
+	new_lck = (struct tdb_lock_type *)realloc(
+		tdb->lockrecs,
+		sizeof(*tdb->lockrecs) * (tdb->num_lockrecs+1));
+	if (new_lck == NULL) {
+		errno = ENOMEM;
+		return -1;
+	}
+	tdb->lockrecs = new_lck;
+
+	/* Since fcntl locks don't nest, we do a lock for the first one,
+	   and simply bump the count for future ones */
+	if (tdb_brlock(tdb, ltype, offset, 1, flags)) {
+		return -1;
+	}
+
+	tdb->lockrecs[tdb->num_lockrecs].off = offset;
+	tdb->lockrecs[tdb->num_lockrecs].count = 1;
+	tdb->lockrecs[tdb->num_lockrecs].ltype = ltype;
+	tdb->num_lockrecs++;
+
+	return 0;
+}
+
+static int tdb_lock_and_recover(struct tdb_context *tdb)
+{
+	int ret;
+
+	/* We need to match locking order in transaction commit. */
+	if (tdb_brlock(tdb, F_WRLCK, FREELIST_TOP, 0, TDB_LOCK_WAIT)) {
+		return -1;
+	}
+
+	if (tdb_brlock(tdb, F_WRLCK, OPEN_LOCK, 1, TDB_LOCK_WAIT)) {
+		tdb_brunlock(tdb, F_WRLCK, FREELIST_TOP, 0);
+		return -1;
+	}
+
+	ret = tdb_transaction_recover(tdb);
+
+	tdb_brunlock(tdb, F_WRLCK, OPEN_LOCK, 1);
+	tdb_brunlock(tdb, F_WRLCK, FREELIST_TOP, 0);
+
+	return ret;
+}
+
+static bool have_data_locks(const struct tdb_context *tdb)
+{
+	unsigned int i;
+
+	for (i = 0; i < tdb->num_lockrecs; i++) {
+		if (tdb->lockrecs[i].off >= lock_offset(-1))
+			return true;
+	}
+	return false;
+}
+
+static int tdb_lock_list(struct tdb_context *tdb, int list, int ltype,
+			 enum tdb_lock_flags waitflag)
+{
+	int ret;
+	bool check = false;
+
+	/* a allrecord lock allows us to avoid per chain locks */
+	if (tdb->allrecord_lock.count &&
+	    (ltype == tdb->allrecord_lock.ltype || ltype == F_RDLCK)) {
+		return 0;
+	}
+
+	if (tdb->allrecord_lock.count) {
+		tdb->ecode = TDB_ERR_LOCK;
+		ret = -1;
+	} else {
+		/* Only check when we grab first data lock. */
+		check = !have_data_locks(tdb);
+		ret = tdb_nest_lock(tdb, lock_offset(list), ltype, waitflag);
+
+		if (ret == 0 && check && tdb_needs_recovery(tdb)) {
+			tdb_nest_unlock(tdb, lock_offset(list), ltype, false);
+
+			if (tdb_lock_and_recover(tdb) == -1) {
+				return -1;
+			}
+			return tdb_lock_list(tdb, list, ltype, waitflag);
+		}
+	}
+	return ret;
+}
+
+/* lock a list in the database. list -1 is the alloc list */
+int tdb_lock(struct tdb_context *tdb, int list, int ltype)
+{
+	int ret;
+
+	ret = tdb_lock_list(tdb, list, ltype, TDB_LOCK_WAIT);
+	if (ret) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_lock failed on list %d "
+			 "ltype=%d (%s)\n",  list, ltype, strerror(errno)));
+	}
+	return ret;
+}
+
+/* lock a list in the database. list -1 is the alloc list. non-blocking lock */
+_PUBLIC_ int tdb_lock_nonblock(struct tdb_context *tdb, int list, int ltype)
+{
+	return tdb_lock_list(tdb, list, ltype, TDB_LOCK_NOWAIT);
+}
+
+
+int tdb_nest_unlock(struct tdb_context *tdb, uint32_t offset, int ltype,
+		    bool mark_lock)
+{
+	int ret = -1;
+	struct tdb_lock_type *lck;
+
+	if (tdb->flags & TDB_NOLOCK)
+		return 0;
+
+	/* Sanity checks */
+	if (offset >= lock_offset(tdb->header.hash_size)) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_unlock: offset %u invalid (%d)\n", offset, tdb->header.hash_size));
+		return ret;
+	}
+
+	lck = find_nestlock(tdb, offset);
+	if ((lck == NULL) || (lck->count == 0)) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_unlock: count is 0\n"));
+		return -1;
+	}
+
+	if (lck->count > 1) {
+		lck->count--;
+		return 0;
+	}
+
+	/*
+	 * This lock has count==1 left, so we need to unlock it in the
+	 * kernel. We don't bother with decrementing the in-memory array
+	 * element, we're about to overwrite it with the last array element
+	 * anyway.
+	 */
+
+	if (mark_lock) {
+		ret = 0;
+	} else {
+		ret = tdb_brunlock(tdb, ltype, offset, 1);
+	}
+
+	/*
+	 * Shrink the array by overwriting the element just unlocked with the
+	 * last array element.
+	 */
+	*lck = tdb->lockrecs[--tdb->num_lockrecs];
+
+	/*
+	 * We don't bother with realloc when the array shrinks, but if we have
+	 * a completely idle tdb we should get rid of the locked array.
+	 */
+
+	if (tdb->num_lockrecs == 0) {
+		SAFE_FREE(tdb->lockrecs);
+	}
+
+	if (ret)
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_unlock: An error occurred unlocking!\n")); 
+	return ret;
+}
+
+_PUBLIC_ int tdb_unlock(struct tdb_context *tdb, int list, int ltype)
+{
+	/* a global lock allows us to avoid per chain locks */
+	if (tdb->allrecord_lock.count &&
+	    (ltype == tdb->allrecord_lock.ltype || ltype == F_RDLCK)) {
+		return 0;
+	}
+
+	if (tdb->allrecord_lock.count) {
+		tdb->ecode = TDB_ERR_LOCK;
+		return -1;
+	}
+
+	return tdb_nest_unlock(tdb, lock_offset(list), ltype, false);
+}
+
+/*
+  get the transaction lock
+ */
+int tdb_transaction_lock(struct tdb_context *tdb, int ltype,
+			 enum tdb_lock_flags lockflags)
+{
+	return tdb_nest_lock(tdb, TRANSACTION_LOCK, ltype, lockflags);
+}
+
+/*
+  release the transaction lock
+ */
+int tdb_transaction_unlock(struct tdb_context *tdb, int ltype)
+{
+	return tdb_nest_unlock(tdb, TRANSACTION_LOCK, ltype, false);
+}
+
+/* Returns 0 if all done, -1 if error, 1 if ok. */
+static int tdb_allrecord_check(struct tdb_context *tdb, int ltype,
+			       enum tdb_lock_flags flags, bool upgradable)
+{
+	/* There are no locks on read-only dbs */
+	if (tdb->read_only || tdb->traverse_read) {
+		tdb->ecode = TDB_ERR_LOCK;
+		return -1;
+	}
+
+	if (tdb->allrecord_lock.count && tdb->allrecord_lock.ltype == ltype) {
+		tdb->allrecord_lock.count++;
+		return 0;
+	}
+
+	if (tdb->allrecord_lock.count) {
+		/* a global lock of a different type exists */
+		tdb->ecode = TDB_ERR_LOCK;
+		return -1;
+	}
+
+	if (tdb_have_extra_locks(tdb)) {
+		/* can't combine global and chain locks */
+		tdb->ecode = TDB_ERR_LOCK;
+		return -1;
+	}
+
+	if (upgradable && ltype != F_RDLCK) {
+		/* tdb error: you can't upgrade a write lock! */
+		tdb->ecode = TDB_ERR_LOCK;
+		return -1;
+	}
+	return 1;
+}
+
+/* We only need to lock individual bytes, but Linux merges consecutive locks
+ * so we lock in contiguous ranges. */
+static int tdb_chainlock_gradual(struct tdb_context *tdb,
+				 int ltype, enum tdb_lock_flags flags,
+				 size_t off, size_t len)
+{
+	int ret;
+	enum tdb_lock_flags nb_flags = (flags & ~TDB_LOCK_WAIT);
+
+	if (len <= 4) {
+		/* Single record.  Just do blocking lock. */
+		return tdb_brlock(tdb, ltype, off, len, flags);
+	}
+
+	/* First we try non-blocking. */
+	ret = tdb_brlock(tdb, ltype, off, len, nb_flags);
+	if (ret == 0) {
+		return 0;
+	}
+
+	/* Try locking first half, then second. */
+	ret = tdb_chainlock_gradual(tdb, ltype, flags, off, len / 2);
+	if (ret == -1)
+		return -1;
+
+	ret = tdb_chainlock_gradual(tdb, ltype, flags,
+				    off + len / 2, len - len / 2);
+	if (ret == -1) {
+		tdb_brunlock(tdb, ltype, off, len / 2);
+		return -1;
+	}
+	return 0;
+}
+
+/* lock/unlock entire database.  It can only be upgradable if you have some
+ * other way of guaranteeing exclusivity (ie. transaction write lock).
+ * We do the locking gradually to avoid being starved by smaller locks. */
+int tdb_allrecord_lock(struct tdb_context *tdb, int ltype,
+		       enum tdb_lock_flags flags, bool upgradable)
+{
+	switch (tdb_allrecord_check(tdb, ltype, flags, upgradable)) {
+	case -1:
+		return -1;
+	case 0:
+		return 0;
+	}
+
+	/* We cover two kinds of locks:
+	 * 1) Normal chain locks.  Taken for almost all operations.
+	 * 2) Individual records locks.  Taken after normal or free
+	 *    chain locks.
+	 *
+	 * It is (1) which cause the starvation problem, so we're only
+	 * gradual for that. */
+	if (tdb_chainlock_gradual(tdb, ltype, flags, FREELIST_TOP,
+				  tdb->header.hash_size * 4) == -1) {
+		return -1;
+	}
+
+	/* Grab individual record locks. */
+	if (tdb_brlock(tdb, ltype, lock_offset(tdb->header.hash_size), 0,
+		       flags) == -1) {
+		tdb_brunlock(tdb, ltype, FREELIST_TOP,
+			     tdb->header.hash_size * 4);
+		return -1;
+	}
+
+	tdb->allrecord_lock.count = 1;
+	/* If it's upgradable, it's actually exclusive so we can treat
+	 * it as a write lock. */
+	tdb->allrecord_lock.ltype = upgradable ? F_WRLCK : ltype;
+	tdb->allrecord_lock.off = upgradable;
+
+	if (tdb_needs_recovery(tdb)) {
+		bool mark = flags & TDB_LOCK_MARK_ONLY;
+		tdb_allrecord_unlock(tdb, ltype, mark);
+		if (mark) {
+			tdb->ecode = TDB_ERR_LOCK;
+			TDB_LOG((tdb, TDB_DEBUG_ERROR,
+				 "tdb_lockall_mark cannot do recovery\n"));
+			return -1;
+		}
+		if (tdb_lock_and_recover(tdb) == -1) {
+			return -1;
+		}
+		return tdb_allrecord_lock(tdb, ltype, flags, upgradable);
+	}
+
+	return 0;
+}
+
+
+
+/* unlock entire db */
+int tdb_allrecord_unlock(struct tdb_context *tdb, int ltype, bool mark_lock)
+{
+	/* There are no locks on read-only dbs */
+	if (tdb->read_only || tdb->traverse_read) {
+		tdb->ecode = TDB_ERR_LOCK;
+		return -1;
+	}
+
+	if (tdb->allrecord_lock.count == 0) {
+		tdb->ecode = TDB_ERR_LOCK;
+		return -1;
+	}
+
+	/* Upgradable locks are marked as write locks. */
+	if (tdb->allrecord_lock.ltype != ltype
+	    && (!tdb->allrecord_lock.off || ltype != F_RDLCK)) {
+		tdb->ecode = TDB_ERR_LOCK;
+		return -1;
+	}
+
+	if (tdb->allrecord_lock.count > 1) {
+		tdb->allrecord_lock.count--;
+		return 0;
+	}
+
+	if (!mark_lock && tdb_brunlock(tdb, ltype, FREELIST_TOP, 0)) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_unlockall failed (%s)\n", strerror(errno)));
+		return -1;
+	}
+
+	tdb->allrecord_lock.count = 0;
+	tdb->allrecord_lock.ltype = 0;
+
+	return 0;
+}
+
+/* lock entire database with write lock */
+_PUBLIC_ int tdb_lockall(struct tdb_context *tdb)
+{
+	tdb_trace(tdb, "tdb_lockall");
+	return tdb_allrecord_lock(tdb, F_WRLCK, TDB_LOCK_WAIT, false);
+}
+
+/* lock entire database with write lock - mark only */
+_PUBLIC_ int tdb_lockall_mark(struct tdb_context *tdb)
+{
+	tdb_trace(tdb, "tdb_lockall_mark");
+	return tdb_allrecord_lock(tdb, F_WRLCK, TDB_LOCK_MARK_ONLY, false);
+}
+
+/* unlock entire database with write lock - unmark only */
+_PUBLIC_ int tdb_lockall_unmark(struct tdb_context *tdb)
+{
+	tdb_trace(tdb, "tdb_lockall_unmark");
+	return tdb_allrecord_unlock(tdb, F_WRLCK, true);
+}
+
+/* lock entire database with write lock - nonblocking varient */
+_PUBLIC_ int tdb_lockall_nonblock(struct tdb_context *tdb)
+{
+	int ret = tdb_allrecord_lock(tdb, F_WRLCK, TDB_LOCK_NOWAIT, false);
+	tdb_trace_ret(tdb, "tdb_lockall_nonblock", ret);
+	return ret;
+}
+
+/* unlock entire database with write lock */
+_PUBLIC_ int tdb_unlockall(struct tdb_context *tdb)
+{
+	tdb_trace(tdb, "tdb_unlockall");
+	return tdb_allrecord_unlock(tdb, F_WRLCK, false);
+}
+
+/* lock entire database with read lock */
+_PUBLIC_ int tdb_lockall_read(struct tdb_context *tdb)
+{
+	tdb_trace(tdb, "tdb_lockall_read");
+	return tdb_allrecord_lock(tdb, F_RDLCK, TDB_LOCK_WAIT, false);
+}
+
+/* lock entire database with read lock - nonblock varient */
+_PUBLIC_ int tdb_lockall_read_nonblock(struct tdb_context *tdb)
+{
+	int ret = tdb_allrecord_lock(tdb, F_RDLCK, TDB_LOCK_NOWAIT, false);
+	tdb_trace_ret(tdb, "tdb_lockall_read_nonblock", ret);
+	return ret;
+}
+
+/* unlock entire database with read lock */
+_PUBLIC_ int tdb_unlockall_read(struct tdb_context *tdb)
+{
+	tdb_trace(tdb, "tdb_unlockall_read");
+	return tdb_allrecord_unlock(tdb, F_RDLCK, false);
+}
+
+/* lock/unlock one hash chain. This is meant to be used to reduce
+   contention - it cannot guarantee how many records will be locked */
+_PUBLIC_ int tdb_chainlock(struct tdb_context *tdb, TDB_DATA key)
+{
+	int ret = tdb_lock(tdb, BUCKET(tdb->hash_fn(&key)), F_WRLCK);
+	tdb_trace_1rec(tdb, "tdb_chainlock", key);
+	return ret;
+}
+
+/* lock/unlock one hash chain, non-blocking. This is meant to be used
+   to reduce contention - it cannot guarantee how many records will be
+   locked */
+_PUBLIC_ int tdb_chainlock_nonblock(struct tdb_context *tdb, TDB_DATA key)
+{
+	int ret = tdb_lock_nonblock(tdb, BUCKET(tdb->hash_fn(&key)), F_WRLCK);
+	tdb_trace_1rec_ret(tdb, "tdb_chainlock_nonblock", key, ret);
+	return ret;
+}
+
+/* mark a chain as locked without actually locking it. Warning! use with great caution! */
+_PUBLIC_ int tdb_chainlock_mark(struct tdb_context *tdb, TDB_DATA key)
+{
+	int ret = tdb_nest_lock(tdb, lock_offset(BUCKET(tdb->hash_fn(&key))),
+				F_WRLCK, TDB_LOCK_MARK_ONLY);
+	tdb_trace_1rec(tdb, "tdb_chainlock_mark", key);
+	return ret;
+}
+
+/* unmark a chain as locked without actually locking it. Warning! use with great caution! */
+_PUBLIC_ int tdb_chainlock_unmark(struct tdb_context *tdb, TDB_DATA key)
+{
+	tdb_trace_1rec(tdb, "tdb_chainlock_unmark", key);
+	return tdb_nest_unlock(tdb, lock_offset(BUCKET(tdb->hash_fn(&key))),
+			       F_WRLCK, true);
+}
+
+_PUBLIC_ int tdb_chainunlock(struct tdb_context *tdb, TDB_DATA key)
+{
+	tdb_trace_1rec(tdb, "tdb_chainunlock", key);
+	return tdb_unlock(tdb, BUCKET(tdb->hash_fn(&key)), F_WRLCK);
+}
+
+_PUBLIC_ int tdb_chainlock_read(struct tdb_context *tdb, TDB_DATA key)
+{
+	int ret;
+	ret = tdb_lock(tdb, BUCKET(tdb->hash_fn(&key)), F_RDLCK);
+	tdb_trace_1rec(tdb, "tdb_chainlock_read", key);
+	return ret;
+}
+
+_PUBLIC_ int tdb_chainunlock_read(struct tdb_context *tdb, TDB_DATA key)
+{
+	tdb_trace_1rec(tdb, "tdb_chainunlock_read", key);
+	return tdb_unlock(tdb, BUCKET(tdb->hash_fn(&key)), F_RDLCK);
+}
+
+/* record lock stops delete underneath */
+int tdb_lock_record(struct tdb_context *tdb, tdb_off_t off)
+{
+	if (tdb->allrecord_lock.count) {
+		return 0;
+	}
+	return off ? tdb_brlock(tdb, F_RDLCK, off, 1, TDB_LOCK_WAIT) : 0;
+}
+
+/*
+  Write locks override our own fcntl readlocks, so check it here.
+  Note this is meant to be F_SETLK, *not* F_SETLKW, as it's not
+  an error to fail to get the lock here.
+*/
+int tdb_write_lock_record(struct tdb_context *tdb, tdb_off_t off)
+{
+	struct tdb_traverse_lock *i;
+	for (i = &tdb->travlocks; i; i = i->next)
+		if (i->off == off)
+			return -1;
+	if (tdb->allrecord_lock.count) {
+		if (tdb->allrecord_lock.ltype == F_WRLCK) {
+			return 0;
+		}
+		return -1;
+	}
+	return tdb_brlock(tdb, F_WRLCK, off, 1, TDB_LOCK_NOWAIT|TDB_LOCK_PROBE);
+}
+
+int tdb_write_unlock_record(struct tdb_context *tdb, tdb_off_t off)
+{
+	if (tdb->allrecord_lock.count) {
+		return 0;
+	}
+	return tdb_brunlock(tdb, F_WRLCK, off, 1);
+}
+
+/* fcntl locks don't stack: avoid unlocking someone else's */
+int tdb_unlock_record(struct tdb_context *tdb, tdb_off_t off)
+{
+	struct tdb_traverse_lock *i;
+	uint32_t count = 0;
+
+	if (tdb->allrecord_lock.count) {
+		return 0;
+	}
+
+	if (off == 0)
+		return 0;
+	for (i = &tdb->travlocks; i; i = i->next)
+		if (i->off == off)
+			count++;
+	return (count == 1 ? tdb_brunlock(tdb, F_RDLCK, off, 1) : 0);
+}
+
+bool tdb_have_extra_locks(struct tdb_context *tdb)
+{
+	unsigned int extra = tdb->num_lockrecs;
+
+	/* A transaction holds the lock for all records. */
+	if (!tdb->transaction && tdb->allrecord_lock.count) {
+		return true;
+	}
+
+	/* We always hold the active lock if CLEAR_IF_FIRST. */
+	if (find_nestlock(tdb, ACTIVE_LOCK)) {
+		extra--;
+	}
+
+	/* In a transaction, we expect to hold the transaction lock */
+	if (tdb->transaction && find_nestlock(tdb, TRANSACTION_LOCK)) {
+		extra--;
+	}
+
+	return extra;
+}
+
+/* The transaction code uses this to remove all locks. */
+void tdb_release_transaction_locks(struct tdb_context *tdb)
+{
+	unsigned int i, active = 0;
+
+	if (tdb->allrecord_lock.count != 0) {
+		tdb_brunlock(tdb, tdb->allrecord_lock.ltype, FREELIST_TOP, 0);
+		tdb->allrecord_lock.count = 0;
+	}
+
+	for (i=0;i<tdb->num_lockrecs;i++) {
+		struct tdb_lock_type *lck = &tdb->lockrecs[i];
+
+		/* Don't release the active lock!  Copy it to first entry. */
+		if (lck->off == ACTIVE_LOCK) {
+			tdb->lockrecs[active++] = *lck;
+		} else {
+			tdb_brunlock(tdb, lck->ltype, lck->off, 1);
+		}
+	}
+	tdb->num_lockrecs = active;
+	if (tdb->num_lockrecs == 0) {
+		SAFE_FREE(tdb->lockrecs);
+	}
+}
+
+/* Following functions are added specifically to support CTDB. */
+
+/* Don't do actual fcntl locking, just mark tdb locked */
+_PUBLIC_ int tdb_transaction_write_lock_mark(struct tdb_context *tdb)
+{
+	return tdb_transaction_lock(tdb, F_WRLCK, TDB_LOCK_MARK_ONLY);
+}
+
+/* Don't do actual fcntl unlocking, just mark tdb unlocked */
+_PUBLIC_ int tdb_transaction_write_lock_unmark(struct tdb_context *tdb)
+{
+	return tdb_nest_unlock(tdb, TRANSACTION_LOCK, F_WRLCK, true);
+}
diff --git a/ctdb/lib/tdb/common/open.c b/ctdb/lib/tdb/common/open.c
new file mode 100644
index 0000000000..d9f76f0835
--- /dev/null
+++ b/ctdb/lib/tdb/common/open.c
@@ -0,0 +1,671 @@
+ /* 
+   Unix SMB/CIFS implementation.
+
+   trivial database library
+
+   Copyright (C) Andrew Tridgell              1999-2005
+   Copyright (C) Paul `Rusty' Russell		   2000
+   Copyright (C) Jeremy Allison			   2000-2003
+
+     ** NOTE! The following LGPL license applies to the tdb
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "tdb_private.h"
+
+/* all contexts, to ensure no double-opens (fcntl locks don't nest!) */
+static struct tdb_context *tdbs = NULL;
+
+/* We use two hashes to double-check they're using the right hash function. */
+void tdb_header_hash(struct tdb_context *tdb,
+		     uint32_t *magic1_hash, uint32_t *magic2_hash)
+{
+	TDB_DATA hash_key;
+	uint32_t tdb_magic = TDB_MAGIC;
+
+	hash_key.dptr = discard_const_p(unsigned char, TDB_MAGIC_FOOD);
+	hash_key.dsize = sizeof(TDB_MAGIC_FOOD);
+	*magic1_hash = tdb->hash_fn(&hash_key);
+
+	hash_key.dptr = (unsigned char *)CONVERT(tdb_magic);
+	hash_key.dsize = sizeof(tdb_magic);
+	*magic2_hash = tdb->hash_fn(&hash_key);
+
+	/* Make sure at least one hash is non-zero! */
+	if (*magic1_hash == 0 && *magic2_hash == 0)
+		*magic1_hash = 1;
+}
+
+/* initialise a new database with a specified hash size */
+static int tdb_new_database(struct tdb_context *tdb, int hash_size)
+{
+	struct tdb_header *newdb;
+	size_t size;
+	int ret = -1;
+
+	/* We make it up in memory, then write it out if not internal */
+	size = sizeof(struct tdb_header) + (hash_size+1)*sizeof(tdb_off_t);
+	if (!(newdb = (struct tdb_header *)calloc(size, 1))) {
+		tdb->ecode = TDB_ERR_OOM;
+		return -1;
+	}
+
+	/* Fill in the header */
+	newdb->version = TDB_VERSION;
+	newdb->hash_size = hash_size;
+
+	tdb_header_hash(tdb, &newdb->magic1_hash, &newdb->magic2_hash);
+
+	/* Make sure older tdbs (which don't check the magic hash fields)
+	 * will refuse to open this TDB. */
+	if (tdb->flags & TDB_INCOMPATIBLE_HASH)
+		newdb->rwlocks = TDB_HASH_RWLOCK_MAGIC;
+
+	if (tdb->flags & TDB_INTERNAL) {
+		tdb->map_size = size;
+		tdb->map_ptr = (char *)newdb;
+		memcpy(&tdb->header, newdb, sizeof(tdb->header));
+		/* Convert the `ondisk' version if asked. */
+		CONVERT(*newdb);
+		return 0;
+	}
+	if (lseek(tdb->fd, 0, SEEK_SET) == -1)
+		goto fail;
+
+	if (ftruncate(tdb->fd, 0) == -1)
+		goto fail;
+
+	/* This creates an endian-converted header, as if read from disk */
+	CONVERT(*newdb);
+	memcpy(&tdb->header, newdb, sizeof(tdb->header));
+	/* Don't endian-convert the magic food! */
+	memcpy(newdb->magic_food, TDB_MAGIC_FOOD, strlen(TDB_MAGIC_FOOD)+1);
+	/* we still have "ret == -1" here */
+	if (tdb_write_all(tdb->fd, newdb, size))
+		ret = 0;
+
+  fail:
+	SAFE_FREE(newdb);
+	return ret;
+}
+
+
+
+static int tdb_already_open(dev_t device,
+			    ino_t ino)
+{
+	struct tdb_context *i;
+
+	for (i = tdbs; i; i = i->next) {
+		if (i->device == device && i->inode == ino) {
+			return 1;
+		}
+	}
+
+	return 0;
+}
+
+/* open the database, creating it if necessary 
+
+   The open_flags and mode are passed straight to the open call on the
+   database file. A flags value of O_WRONLY is invalid. The hash size
+   is advisory, use zero for a default value.
+
+   Return is NULL on error, in which case errno is also set.  Don't 
+   try to call tdb_error or tdb_errname, just do strerror(errno).
+
+   @param name may be NULL for internal databases. */
+_PUBLIC_ struct tdb_context *tdb_open(const char *name, int hash_size, int tdb_flags,
+		      int open_flags, mode_t mode)
+{
+	return tdb_open_ex(name, hash_size, tdb_flags, open_flags, mode, NULL, NULL);
+}
+
+/* a default logging function */
+static void null_log_fn(struct tdb_context *tdb, enum tdb_debug_level level, const char *fmt, ...) PRINTF_ATTRIBUTE(3, 4);
+static void null_log_fn(struct tdb_context *tdb, enum tdb_debug_level level, const char *fmt, ...)
+{
+}
+
+static bool check_header_hash(struct tdb_context *tdb,
+			      bool default_hash, uint32_t *m1, uint32_t *m2)
+{
+	tdb_header_hash(tdb, m1, m2);
+	if (tdb->header.magic1_hash == *m1 &&
+	    tdb->header.magic2_hash == *m2) {
+		return true;
+	}
+
+	/* If they explicitly set a hash, always respect it. */
+	if (!default_hash)
+		return false;
+
+	/* Otherwise, try the other inbuilt hash. */
+	if (tdb->hash_fn == tdb_old_hash)
+		tdb->hash_fn = tdb_jenkins_hash;
+	else
+		tdb->hash_fn = tdb_old_hash;
+	return check_header_hash(tdb, false, m1, m2);
+}
+
+_PUBLIC_ struct tdb_context *tdb_open_ex(const char *name, int hash_size, int tdb_flags,
+				int open_flags, mode_t mode,
+				const struct tdb_logging_context *log_ctx,
+				tdb_hash_func hash_fn)
+{
+	struct tdb_context *tdb;
+	struct stat st;
+	int rev = 0, locked = 0;
+	unsigned char *vp;
+	uint32_t vertest;
+	unsigned v;
+	const char *hash_alg;
+	uint32_t magic1, magic2;
+
+	if (!(tdb = (struct tdb_context *)calloc(1, sizeof *tdb))) {
+		/* Can't log this */
+		errno = ENOMEM;
+		goto fail;
+	}
+	tdb_io_init(tdb);
+	tdb->fd = -1;
+#ifdef TDB_TRACE
+	tdb->tracefd = -1;
+#endif
+	tdb->name = NULL;
+	tdb->map_ptr = NULL;
+	tdb->flags = tdb_flags;
+	tdb->open_flags = open_flags;
+	if (log_ctx) {
+		tdb->log = *log_ctx;
+	} else {
+		tdb->log.log_fn = null_log_fn;
+		tdb->log.log_private = NULL;
+	}
+
+	if (name == NULL && (tdb_flags & TDB_INTERNAL)) {
+		name = "__TDB_INTERNAL__";
+	}
+
+	if (name == NULL) {
+		tdb->name = discard_const_p(char, "__NULL__");
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_open_ex: called with name == NULL\n"));
+		tdb->name = NULL;
+		errno = EINVAL;
+		goto fail;
+	}
+
+	/* now make a copy of the name, as the caller memory might went away */
+	if (!(tdb->name = (char *)strdup(name))) {
+		/*
+		 * set the name as the given string, so that tdb_name() will
+		 * work in case of an error.
+		 */
+		tdb->name = discard_const_p(char, name);
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_open_ex: can't strdup(%s)\n",
+			 name));
+		tdb->name = NULL;
+		errno = ENOMEM;
+		goto fail;
+	}
+
+	if (hash_fn) {
+		tdb->hash_fn = hash_fn;
+		hash_alg = "the user defined";
+	} else {
+		/* This controls what we use when creating a tdb. */
+		if (tdb->flags & TDB_INCOMPATIBLE_HASH) {
+			tdb->hash_fn = tdb_jenkins_hash;
+		} else {
+			tdb->hash_fn = tdb_old_hash;
+		}
+		hash_alg = "either default";
+	}
+
+	/* cache the page size */
+	tdb->page_size = getpagesize();
+	if (tdb->page_size <= 0) {
+		tdb->page_size = 0x2000;
+	}
+
+	tdb->max_dead_records = (tdb_flags & TDB_VOLATILE) ? 5 : 0;
+
+	if ((open_flags & O_ACCMODE) == O_WRONLY) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_open_ex: can't open tdb %s write-only\n",
+			 name));
+		errno = EINVAL;
+		goto fail;
+	}
+
+	if (hash_size == 0)
+		hash_size = DEFAULT_HASH_SIZE;
+	if ((open_flags & O_ACCMODE) == O_RDONLY) {
+		tdb->read_only = 1;
+		/* read only databases don't do locking or clear if first */
+		tdb->flags |= TDB_NOLOCK;
+		tdb->flags &= ~TDB_CLEAR_IF_FIRST;
+	}
+
+	if ((tdb->flags & TDB_ALLOW_NESTING) &&
+	    (tdb->flags & TDB_DISALLOW_NESTING)) {
+		tdb->ecode = TDB_ERR_NESTING;
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_open_ex: "
+			"allow_nesting and disallow_nesting are not allowed together!"));
+		errno = EINVAL;
+		goto fail;
+	}
+
+	if (getenv("TDB_NO_FSYNC")) {
+		tdb->flags |= TDB_NOSYNC;
+	}
+
+	/*
+	 * TDB_ALLOW_NESTING is the default behavior.
+	 * Note: this may change in future versions!
+	 */
+	if (!(tdb->flags & TDB_DISALLOW_NESTING)) {
+		tdb->flags |= TDB_ALLOW_NESTING;
+	}
+
+	/* internal databases don't mmap or lock, and start off cleared */
+	if (tdb->flags & TDB_INTERNAL) {
+		tdb->flags |= (TDB_NOLOCK | TDB_NOMMAP);
+		tdb->flags &= ~TDB_CLEAR_IF_FIRST;
+		if (tdb_new_database(tdb, hash_size) != 0) {
+			TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_open_ex: tdb_new_database failed!"));
+			goto fail;
+		}
+		goto internal;
+	}
+
+	if ((tdb->fd = open(name, open_flags, mode)) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_WARNING, "tdb_open_ex: could not open file %s: %s\n",
+			 name, strerror(errno)));
+		goto fail;	/* errno set by open(2) */
+	}
+
+	/* on exec, don't inherit the fd */
+	v = fcntl(tdb->fd, F_GETFD, 0);
+        fcntl(tdb->fd, F_SETFD, v | FD_CLOEXEC);
+
+	/* ensure there is only one process initialising at once */
+	if (tdb_nest_lock(tdb, OPEN_LOCK, F_WRLCK, TDB_LOCK_WAIT) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_open_ex: failed to get open lock on %s: %s\n",
+			 name, strerror(errno)));
+		goto fail;	/* errno set by tdb_brlock */
+	}
+
+	/* we need to zero database if we are the only one with it open */
+	if ((tdb_flags & TDB_CLEAR_IF_FIRST) &&
+	    (!tdb->read_only) &&
+	    (locked = (tdb_nest_lock(tdb, ACTIVE_LOCK, F_WRLCK, TDB_LOCK_NOWAIT|TDB_LOCK_PROBE) == 0))) {
+		int ret;
+		ret = tdb_brlock(tdb, F_WRLCK, FREELIST_TOP, 0,
+				 TDB_LOCK_WAIT);
+		if (ret == -1) {
+			TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_open_ex: "
+				 "tdb_brlock failed for %s: %s\n",
+				 name, strerror(errno)));
+			goto fail;
+		}
+		ret = tdb_new_database(tdb, hash_size);
+		if (ret == -1) {
+			TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_open_ex: "
+				 "tdb_new_database failed for %s: %s\n",
+				 name, strerror(errno)));
+			tdb_unlockall(tdb);
+			goto fail;
+		}
+		ret = tdb_brunlock(tdb, F_WRLCK, FREELIST_TOP, 0);
+		if (ret == -1) {
+			TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_open_ex: "
+				 "tdb_unlockall failed for %s: %s\n",
+				 name, strerror(errno)));
+			goto fail;
+		}
+		ret = lseek(tdb->fd, 0, SEEK_SET);
+		if (ret == -1) {
+			TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_open_ex: "
+				 "lseek failed for %s: %s\n",
+				 name, strerror(errno)));
+			goto fail;
+		}
+	}
+
+	errno = 0;
+	if (read(tdb->fd, &tdb->header, sizeof(tdb->header)) != sizeof(tdb->header)
+	    || strcmp(tdb->header.magic_food, TDB_MAGIC_FOOD) != 0) {
+		if (!(open_flags & O_CREAT) || tdb_new_database(tdb, hash_size) == -1) {
+			if (errno == 0) {
+				errno = EIO; /* ie bad format or something */
+			}
+			goto fail;
+		}
+		rev = (tdb->flags & TDB_CONVERT);
+	} else if (tdb->header.version != TDB_VERSION
+		   && !(rev = (tdb->header.version==TDB_BYTEREV(TDB_VERSION)))) {
+		/* wrong version */
+		errno = EIO;
+		goto fail;
+	}
+	vp = (unsigned char *)&tdb->header.version;
+	vertest = (((uint32_t)vp[0]) << 24) | (((uint32_t)vp[1]) << 16) |
+		  (((uint32_t)vp[2]) << 8) | (uint32_t)vp[3];
+	tdb->flags |= (vertest==TDB_VERSION) ? TDB_BIGENDIAN : 0;
+	if (!rev)
+		tdb->flags &= ~TDB_CONVERT;
+	else {
+		tdb->flags |= TDB_CONVERT;
+		tdb_convert(&tdb->header, sizeof(tdb->header));
+	}
+	if (fstat(tdb->fd, &st) == -1)
+		goto fail;
+
+	if (tdb->header.rwlocks != 0 &&
+	    tdb->header.rwlocks != TDB_HASH_RWLOCK_MAGIC) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_open_ex: spinlocks no longer supported\n"));
+		goto fail;
+	}
+
+	if ((tdb->header.magic1_hash == 0) && (tdb->header.magic2_hash == 0)) {
+		/* older TDB without magic hash references */
+		tdb->hash_fn = tdb_old_hash;
+	} else if (!check_header_hash(tdb, !hash_fn, &magic1, &magic2)) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_open_ex: "
+			 "%s was not created with %s hash function we are using\n"
+			 "magic1_hash[0x%08X %s 0x%08X] "
+			 "magic2_hash[0x%08X %s 0x%08X]\n",
+			 name, hash_alg,
+			 tdb->header.magic1_hash,
+			 (tdb->header.magic1_hash == magic1) ? "==" : "!=",
+			 magic1,
+			 tdb->header.magic2_hash,
+			 (tdb->header.magic2_hash == magic2) ? "==" : "!=",
+			 magic2));
+		errno = EINVAL;
+		goto fail;
+	}
+
+	/* Is it already in the open list?  If so, fail. */
+	if (tdb_already_open(st.st_dev, st.st_ino)) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_open_ex: "
+			 "%s (%d,%d) is already open in this process\n",
+			 name, (int)st.st_dev, (int)st.st_ino));
+		errno = EBUSY;
+		goto fail;
+	}
+
+	/* Beware truncation! */
+	tdb->map_size = st.st_size;
+	if (tdb->map_size != st.st_size) {
+		/* Ensure ecode is set for log fn. */
+		tdb->ecode = TDB_ERR_IO;
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_open_ex: "
+			 "len %llu too large!\n", (long long)st.st_size));
+		errno = EIO;
+		goto fail;
+	}
+
+	tdb->device = st.st_dev;
+	tdb->inode = st.st_ino;
+	tdb_mmap(tdb);
+	if (locked) {
+		if (tdb_nest_unlock(tdb, ACTIVE_LOCK, F_WRLCK, false) == -1) {
+			TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_open_ex: "
+				 "failed to release ACTIVE_LOCK on %s: %s\n",
+				 name, strerror(errno)));
+			goto fail;
+		}
+
+	}
+
+	/* We always need to do this if the CLEAR_IF_FIRST flag is set, even if
+	   we didn't get the initial exclusive lock as we need to let all other
+	   users know we're using it. */
+
+	if (tdb_flags & TDB_CLEAR_IF_FIRST) {
+		/* leave this lock in place to indicate it's in use */
+		if (tdb_nest_lock(tdb, ACTIVE_LOCK, F_RDLCK, TDB_LOCK_WAIT) == -1) {
+			goto fail;
+		}
+	}
+
+	/* if needed, run recovery */
+	if (tdb_transaction_recover(tdb) == -1) {
+		goto fail;
+	}
+
+#ifdef TDB_TRACE
+	{
+		char tracefile[strlen(name) + 32];
+
+		snprintf(tracefile, sizeof(tracefile),
+			 "%s.trace.%li", name, (long)getpid());
+		tdb->tracefd = open(tracefile, O_WRONLY|O_CREAT|O_EXCL, 0600);
+		if (tdb->tracefd >= 0) {
+			tdb_enable_seqnum(tdb);
+			tdb_trace_open(tdb, "tdb_open", hash_size, tdb_flags,
+				       open_flags);
+		} else
+			TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_open_ex: failed to open trace file %s!\n", tracefile));
+	}
+#endif
+
+ internal:
+	/* Internal (memory-only) databases skip all the code above to
+	 * do with disk files, and resume here by releasing their
+	 * open lock and hooking into the active list. */
+	if (tdb_nest_unlock(tdb, OPEN_LOCK, F_WRLCK, false) == -1) {
+		goto fail;
+	}
+	tdb->next = tdbs;
+	tdbs = tdb;
+	return tdb;
+
+ fail:
+	{ int save_errno = errno;
+
+	if (!tdb)
+		return NULL;
+
+#ifdef TDB_TRACE
+	close(tdb->tracefd);
+#endif
+	if (tdb->map_ptr) {
+		if (tdb->flags & TDB_INTERNAL)
+			SAFE_FREE(tdb->map_ptr);
+		else
+			tdb_munmap(tdb);
+	}
+	if (tdb->fd != -1)
+		if (close(tdb->fd) != 0)
+			TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_open_ex: failed to close tdb->fd on error!\n"));
+	SAFE_FREE(tdb->lockrecs);
+	SAFE_FREE(tdb->name);
+	SAFE_FREE(tdb);
+	errno = save_errno;
+	return NULL;
+	}
+}
+
+/*
+ * Set the maximum number of dead records per hash chain
+ */
+
+_PUBLIC_ void tdb_set_max_dead(struct tdb_context *tdb, int max_dead)
+{
+	tdb->max_dead_records = max_dead;
+}
+
+/**
+ * Close a database.
+ *
+ * @returns -1 for error; 0 for success.
+ **/
+_PUBLIC_ int tdb_close(struct tdb_context *tdb)
+{
+	struct tdb_context **i;
+	int ret = 0;
+
+	if (tdb->transaction) {
+		tdb_transaction_cancel(tdb);
+	}
+	tdb_trace(tdb, "tdb_close");
+
+	if (tdb->map_ptr) {
+		if (tdb->flags & TDB_INTERNAL)
+			SAFE_FREE(tdb->map_ptr);
+		else
+			tdb_munmap(tdb);
+	}
+	SAFE_FREE(tdb->name);
+	if (tdb->fd != -1) {
+		ret = close(tdb->fd);
+		tdb->fd = -1;
+	}
+	SAFE_FREE(tdb->lockrecs);
+
+	/* Remove from contexts list */
+	for (i = &tdbs; *i; i = &(*i)->next) {
+		if (*i == tdb) {
+			*i = tdb->next;
+			break;
+		}
+	}
+
+#ifdef TDB_TRACE
+	close(tdb->tracefd);
+#endif
+	memset(tdb, 0, sizeof(*tdb));
+	SAFE_FREE(tdb);
+
+	return ret;
+}
+
+/* register a loging function */
+_PUBLIC_ void tdb_set_logging_function(struct tdb_context *tdb,
+                                       const struct tdb_logging_context *log_ctx)
+{
+        tdb->log = *log_ctx;
+}
+
+_PUBLIC_ void *tdb_get_logging_private(struct tdb_context *tdb)
+{
+	return tdb->log.log_private;
+}
+
+static int tdb_reopen_internal(struct tdb_context *tdb, bool active_lock)
+{
+#if !defined(LIBREPLACE_PREAD_NOT_REPLACED) || \
+	!defined(LIBREPLACE_PWRITE_NOT_REPLACED)
+	struct stat st;
+#endif
+
+	if (tdb->flags & TDB_INTERNAL) {
+		return 0; /* Nothing to do. */
+	}
+
+	if (tdb_have_extra_locks(tdb)) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_reopen: reopen not allowed with locks held\n"));
+		goto fail;
+	}
+
+	if (tdb->transaction != 0) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_reopen: reopen not allowed inside a transaction\n"));
+		goto fail;
+	}
+
+/* If we have real pread & pwrite, we can skip reopen. */
+#if !defined(LIBREPLACE_PREAD_NOT_REPLACED) || \
+	!defined(LIBREPLACE_PWRITE_NOT_REPLACED)
+	if (tdb_munmap(tdb) != 0) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_reopen: munmap failed (%s)\n", strerror(errno)));
+		goto fail;
+	}
+	if (close(tdb->fd) != 0)
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_reopen: WARNING closing tdb->fd failed!\n"));
+	tdb->fd = open(tdb->name, tdb->open_flags & ~(O_CREAT|O_TRUNC), 0);
+	if (tdb->fd == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_reopen: open failed (%s)\n", strerror(errno)));
+		goto fail;
+	}
+	if (fstat(tdb->fd, &st) != 0) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_reopen: fstat failed (%s)\n", strerror(errno)));
+		goto fail;
+	}
+	if (st.st_ino != tdb->inode || st.st_dev != tdb->device) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_reopen: file dev/inode has changed!\n"));
+		goto fail;
+	}
+	if (tdb_mmap(tdb) != 0) {
+		goto fail;
+	}
+#endif /* fake pread or pwrite */
+
+	/* We may still think we hold the active lock. */
+	tdb->num_lockrecs = 0;
+	SAFE_FREE(tdb->lockrecs);
+
+	if (active_lock && tdb_nest_lock(tdb, ACTIVE_LOCK, F_RDLCK, TDB_LOCK_WAIT) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_reopen: failed to obtain active lock\n"));
+		goto fail;
+	}
+
+	return 0;
+
+fail:
+	tdb_close(tdb);
+	return -1;
+}
+
+/* reopen a tdb - this can be used after a fork to ensure that we have an independent
+   seek pointer from our parent and to re-establish locks */
+_PUBLIC_ int tdb_reopen(struct tdb_context *tdb)
+{
+	return tdb_reopen_internal(tdb, tdb->flags & TDB_CLEAR_IF_FIRST);
+}
+
+/* reopen all tdb's */
+_PUBLIC_ int tdb_reopen_all(int parent_longlived)
+{
+	struct tdb_context *tdb;
+
+	for (tdb=tdbs; tdb; tdb = tdb->next) {
+		bool active_lock = (tdb->flags & TDB_CLEAR_IF_FIRST);
+
+		/*
+		 * If the parent is longlived (ie. a
+		 * parent daemon architecture), we know
+		 * it will keep it's active lock on a
+		 * tdb opened with CLEAR_IF_FIRST. Thus
+		 * for child processes we don't have to
+		 * add an active lock. This is essential
+		 * to improve performance on systems that
+		 * keep POSIX locks as a non-scalable data
+		 * structure in the kernel.
+		 */
+		if (parent_longlived) {
+			/* Ensure no clear-if-first. */
+			active_lock = false;
+		}
+
+		if (tdb_reopen_internal(tdb, active_lock) != 0)
+			return -1;
+	}
+
+	return 0;
+}
diff --git a/ctdb/lib/tdb/common/rescue.c b/ctdb/lib/tdb/common/rescue.c
new file mode 100644
index 0000000000..03ae8d6ae5
--- /dev/null
+++ b/ctdb/lib/tdb/common/rescue.c
@@ -0,0 +1,349 @@
+ /*
+   Unix SMB/CIFS implementation.
+
+   trivial database library, rescue attempt code.
+
+   Copyright (C) Rusty Russell		   2012
+
+     ** NOTE! The following LGPL license applies to the tdb
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+#include "tdb_private.h"
+#include <assert.h>
+
+
+struct found {
+	tdb_off_t head; /* 0 -> invalid. */
+	struct tdb_record rec;
+	TDB_DATA key;
+	bool in_hash;
+	bool in_free;
+};
+
+struct found_table {
+	/* As an ordered array (by head offset). */
+	struct found *arr;
+	unsigned int num, max;
+};
+
+static bool looks_like_valid_record(struct tdb_context *tdb,
+				    tdb_off_t off,
+				    const struct tdb_record *rec,
+				    TDB_DATA *key)
+{
+	unsigned int hval;
+
+	if (rec->magic != TDB_MAGIC)
+		return false;
+
+	if (rec->key_len + rec->data_len > rec->rec_len)
+		return false;
+
+	if (rec->rec_len % TDB_ALIGNMENT)
+		return false;
+
+	/* Next pointer must make some sense. */
+	if (rec->next > 0 && rec->next < TDB_DATA_START(tdb->header.hash_size))
+		return false;
+
+	if (tdb->methods->tdb_oob(tdb, rec->next, sizeof(*rec), 1))
+		return false;
+
+	key->dsize = rec->key_len;
+	key->dptr = tdb_alloc_read(tdb, off + sizeof(*rec), key->dsize);
+	if (!key->dptr)
+		return false;
+
+	hval = tdb->hash_fn(key);
+	if (hval != rec->full_hash) {
+		free(key->dptr);
+		return false;
+	}
+
+	/* Caller frees up key->dptr */
+	return true;
+}
+
+static bool add_to_table(struct found_table *found,
+			 tdb_off_t off,
+			 struct tdb_record *rec,
+			 TDB_DATA key)
+{
+	if (found->num + 1 > found->max) {
+		struct found *new;
+		found->max = (found->max ? found->max * 2 : 128);
+		new = realloc(found->arr, found->max * sizeof(found->arr[0]));
+		if (!new)
+			return false;
+		found->arr = new;
+	}
+
+	found->arr[found->num].head = off;
+	found->arr[found->num].rec = *rec;
+	found->arr[found->num].key = key;
+	found->arr[found->num].in_hash = false;
+	found->arr[found->num].in_free = false;
+
+	found->num++;
+	return true;
+}
+
+static bool walk_record(struct tdb_context *tdb,
+			const struct found *f,
+			void (*walk)(TDB_DATA, TDB_DATA, void *private_data),
+			void *private_data)
+{
+	TDB_DATA data;
+
+	data.dsize = f->rec.data_len;
+	data.dptr = tdb_alloc_read(tdb,
+				   f->head + sizeof(f->rec) + f->rec.key_len,
+				   data.dsize);
+	if (!data.dptr) {
+		if (tdb->ecode == TDB_ERR_OOM)
+			return false;
+		/* I/O errors are expected. */
+		return true;
+	}
+
+	walk(f->key, data, private_data);
+	free(data.dptr);
+	return true;
+}
+
+/* First entry which has offset >= this one. */
+static unsigned int find_entry(struct found_table *found, tdb_off_t off)
+{
+	unsigned int start = 0, end = found->num;
+
+	while (start < end) {
+		/* We can't overflow here. */
+		unsigned int mid = (start + end) / 2;
+
+		if (off < found->arr[mid].head) {
+			end = mid;
+		} else if (off > found->arr[mid].head) {
+			start = mid + 1;
+		} else {
+			return mid;
+		}
+	}
+
+	assert(start == end);
+	return end;
+}
+
+static void found_in_hashchain(struct found_table *found, tdb_off_t head)
+{
+	unsigned int match;
+
+	match = find_entry(found, head);
+	if (match < found->num && found->arr[match].head == head) {
+		found->arr[match].in_hash = true;
+	}
+}
+
+static void mark_free_area(struct found_table *found, tdb_off_t head,
+			   tdb_len_t len)
+{
+	unsigned int match;
+
+	match = find_entry(found, head);
+	/* Mark everything within this free entry. */
+	while (match < found->num) {
+		if (found->arr[match].head >= head + len) {
+			break;
+		}
+		found->arr[match].in_free = true;
+		match++;
+	}
+}
+
+static int cmp_key(const void *a, const void *b)
+{
+	const struct found *fa = a, *fb = b;
+
+	if (fa->key.dsize < fb->key.dsize) {
+		return -1;
+	} else if (fa->key.dsize > fb->key.dsize) {
+		return 1;
+	}
+	return memcmp(fa->key.dptr, fb->key.dptr, fa->key.dsize);
+}
+
+static bool key_eq(TDB_DATA a, TDB_DATA b)
+{
+	return a.dsize == b.dsize
+		&& memcmp(a.dptr, b.dptr, a.dsize) == 0;
+}
+
+static void free_table(struct found_table *found)
+{
+	unsigned int i;
+
+	for (i = 0; i < found->num; i++) {
+		free(found->arr[i].key.dptr);
+	}
+	free(found->arr);
+}
+
+static void logging_suppressed(struct tdb_context *tdb,
+			       enum tdb_debug_level level, const char *fmt, ...)
+{
+}
+
+_PUBLIC_ int tdb_rescue(struct tdb_context *tdb,
+			void (*walk)(TDB_DATA, TDB_DATA, void *private_data),
+			void *private_data)
+{
+	struct found_table found = { NULL, 0, 0 };
+	tdb_off_t h, off, i;
+	tdb_log_func oldlog = tdb->log.log_fn;
+	struct tdb_record rec;
+	TDB_DATA key;
+	bool locked;
+
+	/* Read-only databases use no locking at all: it's best-effort.
+	 * We may have a write lock already, so skip that case too. */
+	if (tdb->read_only || tdb->allrecord_lock.count != 0) {
+		locked = false;
+	} else {
+		if (tdb_lockall_read(tdb) == -1)
+			return -1;
+		locked = true;
+	}
+
+	/* Make sure we know true size of the underlying file. */
+	tdb->methods->tdb_oob(tdb, tdb->map_size, 1, 1);
+
+	/* Suppress logging, since we anticipate errors. */
+	tdb->log.log_fn = logging_suppressed;
+
+	/* Now walk entire db looking for records. */
+	for (off = TDB_DATA_START(tdb->header.hash_size);
+	     off < tdb->map_size;
+	     off += TDB_ALIGNMENT) {
+		if (tdb->methods->tdb_read(tdb, off, &rec, sizeof(rec),
+					   DOCONV()) == -1)
+			continue;
+
+		if (looks_like_valid_record(tdb, off, &rec, &key)) {
+			if (!add_to_table(&found, off, &rec, key)) {
+				goto oom;
+			}
+		}
+	}
+
+	/* Walk hash chains to positive vet. */
+	for (h = 0; h < 1+tdb->header.hash_size; h++) {
+		bool slow_chase = false;
+		tdb_off_t slow_off = FREELIST_TOP + h*sizeof(tdb_off_t);
+
+		if (tdb_ofs_read(tdb, FREELIST_TOP + h*sizeof(tdb_off_t),
+				 &off) == -1)
+			continue;
+
+		while (off && off != slow_off) {
+			if (tdb->methods->tdb_read(tdb, off, &rec, sizeof(rec),
+						   DOCONV()) != 0) {
+				break;
+			}
+
+			/* 0 is the free list, rest are hash chains. */
+			if (h == 0) {
+				/* Don't mark garbage as free. */
+				if (rec.magic != TDB_FREE_MAGIC) {
+					break;
+				}
+				mark_free_area(&found, off,
+					       sizeof(rec) + rec.rec_len);
+			} else {
+				found_in_hashchain(&found, off);
+			}
+
+			off = rec.next;
+
+			/* Loop detection using second pointer at half-speed */
+			if (slow_chase) {
+				/* First entry happens to be next ptr */
+				tdb_ofs_read(tdb, slow_off, &slow_off);
+			}
+			slow_chase = !slow_chase;
+		}
+	}
+
+	/* Recovery area: must be marked as free, since it often has old
+	 * records in there! */
+	if (tdb_ofs_read(tdb, TDB_RECOVERY_HEAD, &off) == 0 && off != 0) {
+		if (tdb->methods->tdb_read(tdb, off, &rec, sizeof(rec),
+					   DOCONV()) == 0) {
+			mark_free_area(&found, off, sizeof(rec) + rec.rec_len);
+		}
+	}
+
+	/* Now sort by key! */
+	qsort(found.arr, found.num, sizeof(found.arr[0]), cmp_key);
+
+	for (i = 0; i < found.num; ) {
+		unsigned int num, num_in_hash = 0;
+
+		/* How many are identical? */
+		for (num = 0; num < found.num - i; num++) {
+			if (!key_eq(found.arr[i].key, found.arr[i+num].key)) {
+				break;
+			}
+			if (found.arr[i+num].in_hash) {
+				if (!walk_record(tdb, &found.arr[i+num],
+						 walk, private_data))
+					goto oom;
+				num_in_hash++;
+			}
+		}
+		assert(num);
+
+		/* If none were in the hash, we print any not in free list. */
+		if (num_in_hash == 0) {
+			unsigned int j;
+
+			for (j = i; j < i + num; j++) {
+				if (!found.arr[j].in_free) {
+					if (!walk_record(tdb, &found.arr[j],
+							 walk, private_data))
+						goto oom;
+				}
+			}
+		}
+
+		i += num;
+	}
+
+	tdb->log.log_fn = oldlog;
+	if (locked) {
+		tdb_unlockall_read(tdb);
+	}
+	return 0;
+
+oom:
+	tdb->log.log_fn = oldlog;
+	tdb->ecode = TDB_ERR_OOM;
+	TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_rescue: failed allocating\n"));
+	free_table(&found);
+	if (locked) {
+		tdb_unlockall_read(tdb);
+	}
+	return -1;
+}
diff --git a/ctdb/lib/tdb/common/summary.c b/ctdb/lib/tdb/common/summary.c
new file mode 100644
index 0000000000..171a1a2055
--- /dev/null
+++ b/ctdb/lib/tdb/common/summary.c
@@ -0,0 +1,201 @@
+ /* 
+   Trivial Database: human-readable summary code
+   Copyright (C) Rusty Russell 2010
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+#include "tdb_private.h"
+
+#define SUMMARY_FORMAT \
+	"Size of file/data: %u/%zu\n" \
+	"Number of records: %zu\n" \
+	"Smallest/average/largest keys: %zu/%zu/%zu\n" \
+	"Smallest/average/largest data: %zu/%zu/%zu\n" \
+	"Smallest/average/largest padding: %zu/%zu/%zu\n" \
+	"Number of dead records: %zu\n" \
+	"Smallest/average/largest dead records: %zu/%zu/%zu\n" \
+	"Number of free records: %zu\n" \
+	"Smallest/average/largest free records: %zu/%zu/%zu\n" \
+	"Number of hash chains: %zu\n" \
+	"Smallest/average/largest hash chains: %zu/%zu/%zu\n" \
+	"Number of uncoalesced records: %zu\n" \
+	"Smallest/average/largest uncoalesced runs: %zu/%zu/%zu\n" \
+	"Percentage keys/data/padding/free/dead/rechdrs&tailers/hashes: %.0f/%.0f/%.0f/%.0f/%.0f/%.0f/%.0f\n"
+
+/* We don't use tally module, to keep upstream happy. */
+struct tally {
+	size_t min, max, total;
+	size_t num;
+};
+
+static void tally_init(struct tally *tally)
+{
+	tally->total = 0;
+	tally->num = 0;
+	tally->min = tally->max = 0;
+}
+
+static void tally_add(struct tally *tally, size_t len)
+{
+	if (tally->num == 0)
+		tally->max = tally->min = len;
+	else if (len > tally->max)
+		tally->max = len;
+	else if (len < tally->min)
+		tally->min = len;
+	tally->num++;
+	tally->total += len;
+}
+
+static size_t tally_mean(const struct tally *tally)
+{
+	if (!tally->num)
+		return 0;
+	return tally->total / tally->num;
+}
+
+static size_t get_hash_length(struct tdb_context *tdb, unsigned int i)
+{
+	tdb_off_t rec_ptr;
+	size_t count = 0;
+
+	if (tdb_ofs_read(tdb, TDB_HASH_TOP(i), &rec_ptr) == -1)
+		return 0;
+
+	/* keep looking until we find the right record */
+	while (rec_ptr) {
+		struct tdb_record r;
+		++count;
+		if (tdb_rec_read(tdb, rec_ptr, &r) == -1)
+			return 0;
+		rec_ptr = r.next;
+	}
+	return count;
+}
+
+_PUBLIC_ char *tdb_summary(struct tdb_context *tdb)
+{
+	tdb_off_t off, rec_off;
+	struct tally freet, keys, data, dead, extra, hash, uncoal;
+	struct tdb_record rec;
+	char *ret = NULL;
+	bool locked;
+	size_t len, unc = 0;
+	struct tdb_record recovery;
+
+	/* Read-only databases use no locking at all: it's best-effort.
+	 * We may have a write lock already, so skip that case too. */
+	if (tdb->read_only || tdb->allrecord_lock.count != 0) {
+		locked = false;
+	} else {
+		if (tdb_lockall_read(tdb) == -1)
+			return NULL;
+		locked = true;
+	}
+
+	if (tdb_recovery_area(tdb, tdb->methods, &rec_off, &recovery) != 0) {
+		goto unlock;
+	}
+
+	tally_init(&freet);
+	tally_init(&keys);
+	tally_init(&data);
+	tally_init(&dead);
+	tally_init(&extra);
+	tally_init(&hash);
+	tally_init(&uncoal);
+
+	for (off = TDB_DATA_START(tdb->header.hash_size);
+	     off < tdb->map_size - 1;
+	     off += sizeof(rec) + rec.rec_len) {
+		if (tdb->methods->tdb_read(tdb, off, &rec, sizeof(rec),
+					   DOCONV()) == -1)
+			goto unlock;
+		switch (rec.magic) {
+		case TDB_MAGIC:
+			tally_add(&keys, rec.key_len);
+			tally_add(&data, rec.data_len);
+			tally_add(&extra, rec.rec_len - (rec.key_len
+							 + rec.data_len));
+			if (unc > 1)
+				tally_add(&uncoal, unc - 1);
+			unc = 0;
+			break;
+		case TDB_FREE_MAGIC:
+			tally_add(&freet, rec.rec_len);
+			unc++;
+			break;
+		/* If we crash after ftruncate, we can get zeroes or fill. */
+		case TDB_RECOVERY_INVALID_MAGIC:
+		case 0x42424242:
+			unc++;
+			/* If it's a valid recovery, we can trust rec_len. */
+			if (off != rec_off) {
+				rec.rec_len = tdb_dead_space(tdb, off)
+					- sizeof(rec);
+			}
+			/* Fall through */
+		case TDB_DEAD_MAGIC:
+			tally_add(&dead, rec.rec_len);
+			break;
+		default:
+			TDB_LOG((tdb, TDB_DEBUG_ERROR,
+				 "Unexpected record magic 0x%x at offset %d\n",
+				 rec.magic, off));
+			goto unlock;
+		}
+	}
+	if (unc > 1)
+		tally_add(&uncoal, unc - 1);
+
+	for (off = 0; off < tdb->header.hash_size; off++)
+		tally_add(&hash, get_hash_length(tdb, off));
+
+	/* 20 is max length of a %zu. */
+	len = strlen(SUMMARY_FORMAT) + 35*20 + 1;
+	ret = (char *)malloc(len);
+	if (!ret)
+		goto unlock;
+
+	snprintf(ret, len, SUMMARY_FORMAT,
+		 tdb->map_size, keys.total+data.total,
+		 keys.num,
+		 keys.min, tally_mean(&keys), keys.max,
+		 data.min, tally_mean(&data), data.max,
+		 extra.min, tally_mean(&extra), extra.max,
+		 dead.num,
+		 dead.min, tally_mean(&dead), dead.max,
+		 freet.num,
+		 freet.min, tally_mean(&freet), freet.max,
+		 hash.num,
+		 hash.min, tally_mean(&hash), hash.max,
+		 uncoal.total,
+		 uncoal.min, tally_mean(&uncoal), uncoal.max,
+		 keys.total * 100.0 / tdb->map_size,
+		 data.total * 100.0 / tdb->map_size,
+		 extra.total * 100.0 / tdb->map_size,
+		 freet.total * 100.0 / tdb->map_size,
+		 dead.total * 100.0 / tdb->map_size,
+		 (keys.num + freet.num + dead.num)
+		 * (sizeof(struct tdb_record) + sizeof(uint32_t))
+		 * 100.0 / tdb->map_size,
+		 tdb->header.hash_size * sizeof(tdb_off_t)
+		 * 100.0 / tdb->map_size);
+
+unlock:
+	if (locked) {
+		tdb_unlockall_read(tdb);
+	}
+	return ret;
+}
diff --git a/ctdb/lib/tdb/common/tdb.c b/ctdb/lib/tdb/common/tdb.c
new file mode 100644
index 0000000000..fc1f5608fa
--- /dev/null
+++ b/ctdb/lib/tdb/common/tdb.c
@@ -0,0 +1,1154 @@
+ /* 
+   Unix SMB/CIFS implementation.
+
+   trivial database library
+
+   Copyright (C) Andrew Tridgell              1999-2005
+   Copyright (C) Paul `Rusty' Russell		   2000
+   Copyright (C) Jeremy Allison			   2000-2003
+
+     ** NOTE! The following LGPL license applies to the tdb
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "tdb_private.h"
+
+_PUBLIC_ TDB_DATA tdb_null;
+
+/*
+  non-blocking increment of the tdb sequence number if the tdb has been opened using
+  the TDB_SEQNUM flag
+*/
+_PUBLIC_ void tdb_increment_seqnum_nonblock(struct tdb_context *tdb)
+{
+	tdb_off_t seqnum=0;
+
+	if (!(tdb->flags & TDB_SEQNUM)) {
+		return;
+	}
+
+	/* we ignore errors from this, as we have no sane way of
+	   dealing with them.
+	*/
+	tdb_ofs_read(tdb, TDB_SEQNUM_OFS, &seqnum);
+	seqnum++;
+	tdb_ofs_write(tdb, TDB_SEQNUM_OFS, &seqnum);
+}
+
+/*
+  increment the tdb sequence number if the tdb has been opened using
+  the TDB_SEQNUM flag
+*/
+static void tdb_increment_seqnum(struct tdb_context *tdb)
+{
+	if (!(tdb->flags & TDB_SEQNUM)) {
+		return;
+	}
+
+	if (tdb_nest_lock(tdb, TDB_SEQNUM_OFS, F_WRLCK,
+			  TDB_LOCK_WAIT|TDB_LOCK_PROBE) != 0) {
+		return;
+	}
+
+	tdb_increment_seqnum_nonblock(tdb);
+
+	tdb_nest_unlock(tdb, TDB_SEQNUM_OFS, F_WRLCK, false);
+}
+
+static int tdb_key_compare(TDB_DATA key, TDB_DATA data, void *private_data)
+{
+	return memcmp(data.dptr, key.dptr, data.dsize);
+}
+
+/* Returns 0 on fail.  On success, return offset of record, and fills
+   in rec */
+static tdb_off_t tdb_find(struct tdb_context *tdb, TDB_DATA key, uint32_t hash,
+			struct tdb_record *r)
+{
+	tdb_off_t rec_ptr;
+
+	/* read in the hash top */
+	if (tdb_ofs_read(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1)
+		return 0;
+
+	/* keep looking until we find the right record */
+	while (rec_ptr) {
+		if (tdb_rec_read(tdb, rec_ptr, r) == -1)
+			return 0;
+
+		if (!TDB_DEAD(r) && hash==r->full_hash
+		    && key.dsize==r->key_len
+		    && tdb_parse_data(tdb, key, rec_ptr + sizeof(*r),
+				      r->key_len, tdb_key_compare,
+				      NULL) == 0) {
+			return rec_ptr;
+		}
+		/* detect tight infinite loop */
+		if (rec_ptr == r->next) {
+			tdb->ecode = TDB_ERR_CORRUPT;
+			TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_find: loop detected.\n"));
+			return 0;
+		}
+		rec_ptr = r->next;
+	}
+	tdb->ecode = TDB_ERR_NOEXIST;
+	return 0;
+}
+
+/* As tdb_find, but if you succeed, keep the lock */
+tdb_off_t tdb_find_lock_hash(struct tdb_context *tdb, TDB_DATA key, uint32_t hash, int locktype,
+			   struct tdb_record *rec)
+{
+	uint32_t rec_ptr;
+
+	if (tdb_lock(tdb, BUCKET(hash), locktype) == -1)
+		return 0;
+	if (!(rec_ptr = tdb_find(tdb, key, hash, rec)))
+		tdb_unlock(tdb, BUCKET(hash), locktype);
+	return rec_ptr;
+}
+
+static TDB_DATA _tdb_fetch(struct tdb_context *tdb, TDB_DATA key);
+
+static int tdb_update_hash_cmp(TDB_DATA key, TDB_DATA data, void *private_data)
+{
+	TDB_DATA *dbuf = (TDB_DATA *)private_data;
+
+	if (dbuf->dsize != data.dsize) {
+		return -1;
+	}
+	if (memcmp(dbuf->dptr, data.dptr, data.dsize) != 0) {
+		return -1;
+	}
+	return 0;
+}
+
+/* update an entry in place - this only works if the new data size
+   is <= the old data size and the key exists.
+   on failure return -1.
+*/
+static int tdb_update_hash(struct tdb_context *tdb, TDB_DATA key, uint32_t hash, TDB_DATA dbuf)
+{
+	struct tdb_record rec;
+	tdb_off_t rec_ptr;
+
+	/* find entry */
+	if (!(rec_ptr = tdb_find(tdb, key, hash, &rec)))
+		return -1;
+
+	/* it could be an exact duplicate of what is there - this is
+	 * surprisingly common (eg. with a ldb re-index). */
+	if (rec.key_len == key.dsize && 
+	    rec.data_len == dbuf.dsize &&
+	    rec.full_hash == hash &&
+	    tdb_parse_record(tdb, key, tdb_update_hash_cmp, &dbuf) == 0) {
+		return 0;
+	}
+
+	/* must be long enough key, data and tailer */
+	if (rec.rec_len < key.dsize + dbuf.dsize + sizeof(tdb_off_t)) {
+		tdb->ecode = TDB_SUCCESS; /* Not really an error */
+		return -1;
+	}
+
+	if (tdb->methods->tdb_write(tdb, rec_ptr + sizeof(rec) + rec.key_len,
+		      dbuf.dptr, dbuf.dsize) == -1)
+		return -1;
+
+	if (dbuf.dsize != rec.data_len) {
+		/* update size */
+		rec.data_len = dbuf.dsize;
+		return tdb_rec_write(tdb, rec_ptr, &rec);
+	}
+
+	return 0;
+}
+
+/* find an entry in the database given a key */
+/* If an entry doesn't exist tdb_err will be set to
+ * TDB_ERR_NOEXIST. If a key has no data attached
+ * then the TDB_DATA will have zero length but
+ * a non-zero pointer
+ */
+static TDB_DATA _tdb_fetch(struct tdb_context *tdb, TDB_DATA key)
+{
+	tdb_off_t rec_ptr;
+	struct tdb_record rec;
+	TDB_DATA ret;
+	uint32_t hash;
+
+	/* find which hash bucket it is in */
+	hash = tdb->hash_fn(&key);
+	if (!(rec_ptr = tdb_find_lock_hash(tdb,key,hash,F_RDLCK,&rec)))
+		return tdb_null;
+
+	ret.dptr = tdb_alloc_read(tdb, rec_ptr + sizeof(rec) + rec.key_len,
+				  rec.data_len);
+	ret.dsize = rec.data_len;
+	tdb_unlock(tdb, BUCKET(rec.full_hash), F_RDLCK);
+	return ret;
+}
+
+_PUBLIC_ TDB_DATA tdb_fetch(struct tdb_context *tdb, TDB_DATA key)
+{
+	TDB_DATA ret = _tdb_fetch(tdb, key);
+
+	tdb_trace_1rec_retrec(tdb, "tdb_fetch", key, ret);
+	return ret;
+}
+
+/*
+ * Find an entry in the database and hand the record's data to a parsing
+ * function. The parsing function is executed under the chain read lock, so it
+ * should be fast and should not block on other syscalls.
+ *
+ * DON'T CALL OTHER TDB CALLS FROM THE PARSER, THIS MIGHT LEAD TO SEGFAULTS.
+ *
+ * For mmapped tdb's that do not have a transaction open it points the parsing
+ * function directly at the mmap area, it avoids the malloc/memcpy in this
+ * case. If a transaction is open or no mmap is available, it has to do
+ * malloc/read/parse/free.
+ *
+ * This is interesting for all readers of potentially large data structures in
+ * the tdb records, ldb indexes being one example.
+ *
+ * Return -1 if the record was not found.
+ */
+
+_PUBLIC_ int tdb_parse_record(struct tdb_context *tdb, TDB_DATA key,
+		     int (*parser)(TDB_DATA key, TDB_DATA data,
+				   void *private_data),
+		     void *private_data)
+{
+	tdb_off_t rec_ptr;
+	struct tdb_record rec;
+	int ret;
+	uint32_t hash;
+
+	/* find which hash bucket it is in */
+	hash = tdb->hash_fn(&key);
+
+	if (!(rec_ptr = tdb_find_lock_hash(tdb,key,hash,F_RDLCK,&rec))) {
+		/* record not found */
+		tdb_trace_1rec_ret(tdb, "tdb_parse_record", key, -1);
+		tdb->ecode = TDB_ERR_NOEXIST;
+		return -1;
+	}
+	tdb_trace_1rec_ret(tdb, "tdb_parse_record", key, 0);
+
+	ret = tdb_parse_data(tdb, key, rec_ptr + sizeof(rec) + rec.key_len,
+			     rec.data_len, parser, private_data);
+
+	tdb_unlock(tdb, BUCKET(rec.full_hash), F_RDLCK);
+
+	return ret;
+}
+
+/* check if an entry in the database exists 
+
+   note that 1 is returned if the key is found and 0 is returned if not found
+   this doesn't match the conventions in the rest of this module, but is
+   compatible with gdbm
+*/
+static int tdb_exists_hash(struct tdb_context *tdb, TDB_DATA key, uint32_t hash)
+{
+	struct tdb_record rec;
+
+	if (tdb_find_lock_hash(tdb, key, hash, F_RDLCK, &rec) == 0)
+		return 0;
+	tdb_unlock(tdb, BUCKET(rec.full_hash), F_RDLCK);
+	return 1;
+}
+
+_PUBLIC_ int tdb_exists(struct tdb_context *tdb, TDB_DATA key)
+{
+	uint32_t hash = tdb->hash_fn(&key);
+	int ret;
+
+	ret = tdb_exists_hash(tdb, key, hash);
+	tdb_trace_1rec_ret(tdb, "tdb_exists", key, ret);
+	return ret;
+}
+
+/* actually delete an entry in the database given the offset */
+int tdb_do_delete(struct tdb_context *tdb, tdb_off_t rec_ptr, struct tdb_record *rec)
+{
+	tdb_off_t last_ptr, i;
+	struct tdb_record lastrec;
+
+	if (tdb->read_only || tdb->traverse_read) return -1;
+
+	if (((tdb->traverse_write != 0) && (!TDB_DEAD(rec))) ||
+	    tdb_write_lock_record(tdb, rec_ptr) == -1) {
+		/* Someone traversing here: mark it as dead */
+		rec->magic = TDB_DEAD_MAGIC;
+		return tdb_rec_write(tdb, rec_ptr, rec);
+	}
+	if (tdb_write_unlock_record(tdb, rec_ptr) != 0)
+		return -1;
+
+	/* find previous record in hash chain */
+	if (tdb_ofs_read(tdb, TDB_HASH_TOP(rec->full_hash), &i) == -1)
+		return -1;
+	for (last_ptr = 0; i != rec_ptr; last_ptr = i, i = lastrec.next)
+		if (tdb_rec_read(tdb, i, &lastrec) == -1)
+			return -1;
+
+	/* unlink it: next ptr is at start of record. */
+	if (last_ptr == 0)
+		last_ptr = TDB_HASH_TOP(rec->full_hash);
+	if (tdb_ofs_write(tdb, last_ptr, &rec->next) == -1)
+		return -1;
+
+	/* recover the space */
+	if (tdb_free(tdb, rec_ptr, rec) == -1)
+		return -1;
+	return 0;
+}
+
+static int tdb_count_dead(struct tdb_context *tdb, uint32_t hash)
+{
+	int res = 0;
+	tdb_off_t rec_ptr;
+	struct tdb_record rec;
+
+	/* read in the hash top */
+	if (tdb_ofs_read(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1)
+		return 0;
+
+	while (rec_ptr) {
+		if (tdb_rec_read(tdb, rec_ptr, &rec) == -1)
+			return 0;
+
+		if (rec.magic == TDB_DEAD_MAGIC) {
+			res += 1;
+		}
+		rec_ptr = rec.next;
+	}
+	return res;
+}
+
+/*
+ * Purge all DEAD records from a hash chain
+ */
+static int tdb_purge_dead(struct tdb_context *tdb, uint32_t hash)
+{
+	int res = -1;
+	struct tdb_record rec;
+	tdb_off_t rec_ptr;
+
+	if (tdb_lock(tdb, -1, F_WRLCK) == -1) {
+		return -1;
+	}
+
+	/* read in the hash top */
+	if (tdb_ofs_read(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1)
+		goto fail;
+
+	while (rec_ptr) {
+		tdb_off_t next;
+
+		if (tdb_rec_read(tdb, rec_ptr, &rec) == -1) {
+			goto fail;
+		}
+
+		next = rec.next;
+
+		if (rec.magic == TDB_DEAD_MAGIC
+		    && tdb_do_delete(tdb, rec_ptr, &rec) == -1) {
+			goto fail;
+		}
+		rec_ptr = next;
+	}
+	res = 0;
+ fail:
+	tdb_unlock(tdb, -1, F_WRLCK);
+	return res;
+}
+
+/* delete an entry in the database given a key */
+static int tdb_delete_hash(struct tdb_context *tdb, TDB_DATA key, uint32_t hash)
+{
+	tdb_off_t rec_ptr;
+	struct tdb_record rec;
+	int ret;
+
+	if (tdb->max_dead_records != 0) {
+
+		/*
+		 * Allow for some dead records per hash chain, mainly for
+		 * tdb's with a very high create/delete rate like locking.tdb.
+		 */
+
+		if (tdb_lock(tdb, BUCKET(hash), F_WRLCK) == -1)
+			return -1;
+
+		if (tdb_count_dead(tdb, hash) >= tdb->max_dead_records) {
+			/*
+			 * Don't let the per-chain freelist grow too large,
+			 * delete all existing dead records
+			 */
+			tdb_purge_dead(tdb, hash);
+		}
+
+		if (!(rec_ptr = tdb_find(tdb, key, hash, &rec))) {
+			tdb_unlock(tdb, BUCKET(hash), F_WRLCK);
+			return -1;
+		}
+
+		/*
+		 * Just mark the record as dead.
+		 */
+		rec.magic = TDB_DEAD_MAGIC;
+		ret = tdb_rec_write(tdb, rec_ptr, &rec);
+	}
+	else {
+		if (!(rec_ptr = tdb_find_lock_hash(tdb, key, hash, F_WRLCK,
+						   &rec)))
+			return -1;
+
+		ret = tdb_do_delete(tdb, rec_ptr, &rec);
+	}
+
+	if (ret == 0) {
+		tdb_increment_seqnum(tdb);
+	}
+
+	if (tdb_unlock(tdb, BUCKET(rec.full_hash), F_WRLCK) != 0)
+		TDB_LOG((tdb, TDB_DEBUG_WARNING, "tdb_delete: WARNING tdb_unlock failed!\n"));
+	return ret;
+}
+
+_PUBLIC_ int tdb_delete(struct tdb_context *tdb, TDB_DATA key)
+{
+	uint32_t hash = tdb->hash_fn(&key);
+	int ret;
+
+	ret = tdb_delete_hash(tdb, key, hash);
+	tdb_trace_1rec_ret(tdb, "tdb_delete", key, ret);
+	return ret;
+}
+
+/*
+ * See if we have a dead record around with enough space
+ */
+static tdb_off_t tdb_find_dead(struct tdb_context *tdb, uint32_t hash,
+			       struct tdb_record *r, tdb_len_t length)
+{
+	tdb_off_t rec_ptr;
+
+	/* read in the hash top */
+	if (tdb_ofs_read(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1)
+		return 0;
+
+	/* keep looking until we find the right record */
+	while (rec_ptr) {
+		if (tdb_rec_read(tdb, rec_ptr, r) == -1)
+			return 0;
+
+		if (TDB_DEAD(r) && r->rec_len >= length) {
+			/*
+			 * First fit for simple coding, TODO: change to best
+			 * fit
+			 */
+			return rec_ptr;
+		}
+		rec_ptr = r->next;
+	}
+	return 0;
+}
+
+static int _tdb_store(struct tdb_context *tdb, TDB_DATA key,
+		       TDB_DATA dbuf, int flag, uint32_t hash)
+{
+	struct tdb_record rec;
+	tdb_off_t rec_ptr;
+	int ret = -1;
+
+	/* check for it existing, on insert. */
+	if (flag == TDB_INSERT) {
+		if (tdb_exists_hash(tdb, key, hash)) {
+			tdb->ecode = TDB_ERR_EXISTS;
+			goto fail;
+		}
+	} else {
+		/* first try in-place update, on modify or replace. */
+		if (tdb_update_hash(tdb, key, hash, dbuf) == 0) {
+			goto done;
+		}
+		if (tdb->ecode == TDB_ERR_NOEXIST &&
+		    flag == TDB_MODIFY) {
+			/* if the record doesn't exist and we are in TDB_MODIFY mode then
+			 we should fail the store */
+			goto fail;
+		}
+	}
+	/* reset the error code potentially set by the tdb_update() */
+	tdb->ecode = TDB_SUCCESS;
+
+	/* delete any existing record - if it doesn't exist we don't
+           care.  Doing this first reduces fragmentation, and avoids
+           coalescing with `allocated' block before it's updated. */
+	if (flag != TDB_INSERT)
+		tdb_delete_hash(tdb, key, hash);
+
+	if (tdb->max_dead_records != 0) {
+		/*
+		 * Allow for some dead records per hash chain, look if we can
+		 * find one that can hold the new record. We need enough space
+		 * for key, data and tailer. If we find one, we don't have to
+		 * consult the central freelist.
+		 */
+		rec_ptr = tdb_find_dead(
+			tdb, hash, &rec,
+			key.dsize + dbuf.dsize + sizeof(tdb_off_t));
+
+		if (rec_ptr != 0) {
+			rec.key_len = key.dsize;
+			rec.data_len = dbuf.dsize;
+			rec.full_hash = hash;
+			rec.magic = TDB_MAGIC;
+			if (tdb_rec_write(tdb, rec_ptr, &rec) == -1
+			    || tdb->methods->tdb_write(
+				    tdb, rec_ptr + sizeof(rec),
+				    key.dptr, key.dsize) == -1
+			    || tdb->methods->tdb_write(
+				    tdb, rec_ptr + sizeof(rec) + key.dsize,
+				    dbuf.dptr, dbuf.dsize) == -1) {
+				goto fail;
+			}
+			goto done;
+		}
+	}
+
+	/*
+	 * We have to allocate some space from the freelist, so this means we
+	 * have to lock it. Use the chance to purge all the DEAD records from
+	 * the hash chain under the freelist lock.
+	 */
+
+	if (tdb_lock(tdb, -1, F_WRLCK) == -1) {
+		goto fail;
+	}
+
+	if ((tdb->max_dead_records != 0)
+	    && (tdb_purge_dead(tdb, hash) == -1)) {
+		tdb_unlock(tdb, -1, F_WRLCK);
+		goto fail;
+	}
+
+	/* we have to allocate some space */
+	rec_ptr = tdb_allocate(tdb, key.dsize + dbuf.dsize, &rec);
+
+	tdb_unlock(tdb, -1, F_WRLCK);
+
+	if (rec_ptr == 0) {
+		goto fail;
+	}
+
+	/* Read hash top into next ptr */
+	if (tdb_ofs_read(tdb, TDB_HASH_TOP(hash), &rec.next) == -1)
+		goto fail;
+
+	rec.key_len = key.dsize;
+	rec.data_len = dbuf.dsize;
+	rec.full_hash = hash;
+	rec.magic = TDB_MAGIC;
+
+	/* write out and point the top of the hash chain at it */
+	if (tdb_rec_write(tdb, rec_ptr, &rec) == -1
+	    || tdb->methods->tdb_write(tdb, rec_ptr+sizeof(rec),
+				       key.dptr, key.dsize) == -1
+	    || tdb->methods->tdb_write(tdb, rec_ptr+sizeof(rec)+key.dsize,
+				       dbuf.dptr, dbuf.dsize) == -1
+	    || tdb_ofs_write(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1) {
+		/* Need to tdb_unallocate() here */
+		goto fail;
+	}
+
+ done:
+	ret = 0;
+ fail:
+	if (ret == 0) {
+		tdb_increment_seqnum(tdb);
+	}
+	return ret;
+}
+
+/* store an element in the database, replacing any existing element
+   with the same key
+
+   return 0 on success, -1 on failure
+*/
+_PUBLIC_ int tdb_store(struct tdb_context *tdb, TDB_DATA key, TDB_DATA dbuf, int flag)
+{
+	uint32_t hash;
+	int ret;
+
+	if (tdb->read_only || tdb->traverse_read) {
+		tdb->ecode = TDB_ERR_RDONLY;
+		tdb_trace_2rec_flag_ret(tdb, "tdb_store", key, dbuf, flag, -1);
+		return -1;
+	}
+
+	/* find which hash bucket it is in */
+	hash = tdb->hash_fn(&key);
+	if (tdb_lock(tdb, BUCKET(hash), F_WRLCK) == -1)
+		return -1;
+
+	ret = _tdb_store(tdb, key, dbuf, flag, hash);
+	tdb_trace_2rec_flag_ret(tdb, "tdb_store", key, dbuf, flag, ret);
+	tdb_unlock(tdb, BUCKET(hash), F_WRLCK);
+	return ret;
+}
+
+/* Append to an entry. Create if not exist. */
+_PUBLIC_ int tdb_append(struct tdb_context *tdb, TDB_DATA key, TDB_DATA new_dbuf)
+{
+	uint32_t hash;
+	TDB_DATA dbuf;
+	int ret = -1;
+
+	/* find which hash bucket it is in */
+	hash = tdb->hash_fn(&key);
+	if (tdb_lock(tdb, BUCKET(hash), F_WRLCK) == -1)
+		return -1;
+
+	dbuf = _tdb_fetch(tdb, key);
+
+	if (dbuf.dptr == NULL) {
+		dbuf.dptr = (unsigned char *)malloc(new_dbuf.dsize);
+	} else {
+		unsigned int new_len = dbuf.dsize + new_dbuf.dsize;
+		unsigned char *new_dptr;
+
+		/* realloc '0' is special: don't do that. */
+		if (new_len == 0)
+			new_len = 1;
+		new_dptr = (unsigned char *)realloc(dbuf.dptr, new_len);
+		if (new_dptr == NULL) {
+			free(dbuf.dptr);
+		}
+		dbuf.dptr = new_dptr;
+	}
+
+	if (dbuf.dptr == NULL) {
+		tdb->ecode = TDB_ERR_OOM;
+		goto failed;
+	}
+
+	memcpy(dbuf.dptr + dbuf.dsize, new_dbuf.dptr, new_dbuf.dsize);
+	dbuf.dsize += new_dbuf.dsize;
+
+	ret = _tdb_store(tdb, key, dbuf, 0, hash);
+	tdb_trace_2rec_retrec(tdb, "tdb_append", key, new_dbuf, dbuf);
+
+failed:
+	tdb_unlock(tdb, BUCKET(hash), F_WRLCK);
+	SAFE_FREE(dbuf.dptr);
+	return ret;
+}
+
+
+/*
+  return the name of the current tdb file
+  useful for external logging functions
+*/
+_PUBLIC_ const char *tdb_name(struct tdb_context *tdb)
+{
+	return tdb->name;
+}
+
+/*
+  return the underlying file descriptor being used by tdb, or -1
+  useful for external routines that want to check the device/inode
+  of the fd
+*/
+_PUBLIC_ int tdb_fd(struct tdb_context *tdb)
+{
+	return tdb->fd;
+}
+
+/*
+  return the current logging function
+  useful for external tdb routines that wish to log tdb errors
+*/
+_PUBLIC_ tdb_log_func tdb_log_fn(struct tdb_context *tdb)
+{
+	return tdb->log.log_fn;
+}
+
+
+/*
+  get the tdb sequence number. Only makes sense if the writers opened
+  with TDB_SEQNUM set. Note that this sequence number will wrap quite
+  quickly, so it should only be used for a 'has something changed'
+  test, not for code that relies on the count of the number of changes
+  made. If you want a counter then use a tdb record.
+
+  The aim of this sequence number is to allow for a very lightweight
+  test of a possible tdb change.
+*/
+_PUBLIC_ int tdb_get_seqnum(struct tdb_context *tdb)
+{
+	tdb_off_t seqnum=0;
+
+	tdb_ofs_read(tdb, TDB_SEQNUM_OFS, &seqnum);
+	return seqnum;
+}
+
+_PUBLIC_ int tdb_hash_size(struct tdb_context *tdb)
+{
+	return tdb->header.hash_size;
+}
+
+_PUBLIC_ size_t tdb_map_size(struct tdb_context *tdb)
+{
+	return tdb->map_size;
+}
+
+_PUBLIC_ int tdb_get_flags(struct tdb_context *tdb)
+{
+	return tdb->flags;
+}
+
+_PUBLIC_ void tdb_add_flags(struct tdb_context *tdb, unsigned flags)
+{
+	if ((flags & TDB_ALLOW_NESTING) &&
+	    (flags & TDB_DISALLOW_NESTING)) {
+		tdb->ecode = TDB_ERR_NESTING;
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_add_flags: "
+			"allow_nesting and disallow_nesting are not allowed together!"));
+		return;
+	}
+
+	if (flags & TDB_ALLOW_NESTING) {
+		tdb->flags &= ~TDB_DISALLOW_NESTING;
+	}
+	if (flags & TDB_DISALLOW_NESTING) {
+		tdb->flags &= ~TDB_ALLOW_NESTING;
+	}
+
+	tdb->flags |= flags;
+}
+
+_PUBLIC_ void tdb_remove_flags(struct tdb_context *tdb, unsigned flags)
+{
+	if ((flags & TDB_ALLOW_NESTING) &&
+	    (flags & TDB_DISALLOW_NESTING)) {
+		tdb->ecode = TDB_ERR_NESTING;
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_remove_flags: "
+			"allow_nesting and disallow_nesting are not allowed together!"));
+		return;
+	}
+
+	if (flags & TDB_ALLOW_NESTING) {
+		tdb->flags |= TDB_DISALLOW_NESTING;
+	}
+	if (flags & TDB_DISALLOW_NESTING) {
+		tdb->flags |= TDB_ALLOW_NESTING;
+	}
+
+	tdb->flags &= ~flags;
+}
+
+
+/*
+  enable sequence number handling on an open tdb
+*/
+_PUBLIC_ void tdb_enable_seqnum(struct tdb_context *tdb)
+{
+	tdb->flags |= TDB_SEQNUM;
+}
+
+
+/*
+  add a region of the file to the freelist. Length is the size of the region in bytes, 
+  which includes the free list header that needs to be added
+ */
+static int tdb_free_region(struct tdb_context *tdb, tdb_off_t offset, ssize_t length)
+{
+	struct tdb_record rec;
+	if (length <= sizeof(rec)) {
+		/* the region is not worth adding */
+		return 0;
+	}
+	if (length + offset > tdb->map_size) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_free_region: adding region beyond end of file\n"));
+		return -1;		
+	}
+	memset(&rec,'\0',sizeof(rec));
+	rec.rec_len = length - sizeof(rec);
+	if (tdb_free(tdb, offset, &rec) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_free_region: failed to add free record\n"));
+		return -1;
+	}
+	return 0;
+}
+
+/*
+  wipe the entire database, deleting all records. This can be done
+  very fast by using a allrecord lock. The entire data portion of the
+  file becomes a single entry in the freelist.
+
+  This code carefully steps around the recovery area, leaving it alone
+ */
+_PUBLIC_ int tdb_wipe_all(struct tdb_context *tdb)
+{
+	int i;
+	tdb_off_t offset = 0;
+	ssize_t data_len;
+	tdb_off_t recovery_head;
+	tdb_len_t recovery_size = 0;
+
+	if (tdb_lockall(tdb) != 0) {
+		return -1;
+	}
+
+	tdb_trace(tdb, "tdb_wipe_all");
+
+	/* see if the tdb has a recovery area, and remember its size
+	   if so. We don't want to lose this as otherwise each
+	   tdb_wipe_all() in a transaction will increase the size of
+	   the tdb by the size of the recovery area */
+	if (tdb_ofs_read(tdb, TDB_RECOVERY_HEAD, &recovery_head) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_wipe_all: failed to read recovery head\n"));
+		goto failed;
+	}
+
+	if (recovery_head != 0) {
+		struct tdb_record rec;
+		if (tdb->methods->tdb_read(tdb, recovery_head, &rec, sizeof(rec), DOCONV()) == -1) {
+			TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_wipe_all: failed to read recovery record\n"));
+			return -1;
+		}	
+		recovery_size = rec.rec_len + sizeof(rec);
+	}
+
+	/* wipe the hashes */
+	for (i=0;i<tdb->header.hash_size;i++) {
+		if (tdb_ofs_write(tdb, TDB_HASH_TOP(i), &offset) == -1) {
+			TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_wipe_all: failed to write hash %d\n", i));
+			goto failed;
+		}
+	}
+
+	/* wipe the freelist */
+	if (tdb_ofs_write(tdb, FREELIST_TOP, &offset) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_wipe_all: failed to write freelist\n"));
+		goto failed;
+	}
+
+	/* add all the rest of the file to the freelist, possibly leaving a gap 
+	   for the recovery area */
+	if (recovery_size == 0) {
+		/* the simple case - the whole file can be used as a freelist */
+		data_len = (tdb->map_size - TDB_DATA_START(tdb->header.hash_size));
+		if (tdb_free_region(tdb, TDB_DATA_START(tdb->header.hash_size), data_len) != 0) {
+			goto failed;
+		}
+	} else {
+		/* we need to add two freelist entries - one on either
+		   side of the recovery area 
+
+		   Note that we cannot shift the recovery area during
+		   this operation. Only the transaction.c code may
+		   move the recovery area or we risk subtle data
+		   corruption
+		*/
+		data_len = (recovery_head - TDB_DATA_START(tdb->header.hash_size));
+		if (tdb_free_region(tdb, TDB_DATA_START(tdb->header.hash_size), data_len) != 0) {
+			goto failed;
+		}
+		/* and the 2nd free list entry after the recovery area - if any */
+		data_len = tdb->map_size - (recovery_head+recovery_size);
+		if (tdb_free_region(tdb, recovery_head+recovery_size, data_len) != 0) {
+			goto failed;
+		}
+	}
+
+	tdb_increment_seqnum_nonblock(tdb);
+
+	if (tdb_unlockall(tdb) != 0) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_wipe_all: failed to unlock\n"));
+		goto failed;
+	}
+
+	return 0;
+
+failed:
+	tdb_unlockall(tdb);
+	return -1;
+}
+
+struct traverse_state {
+	bool error;
+	struct tdb_context *dest_db;
+};
+
+/*
+  traverse function for repacking
+ */
+static int repack_traverse(struct tdb_context *tdb, TDB_DATA key, TDB_DATA data, void *private_data)
+{
+	struct traverse_state *state = (struct traverse_state *)private_data;
+	if (tdb_store(state->dest_db, key, data, TDB_INSERT) != 0) {
+		state->error = true;
+		return -1;
+	}
+	return 0;
+}
+
+/*
+  repack a tdb
+ */
+_PUBLIC_ int tdb_repack(struct tdb_context *tdb)
+{
+	struct tdb_context *tmp_db;
+	struct traverse_state state;
+
+	tdb_trace(tdb, "tdb_repack");
+
+	if (tdb_transaction_start(tdb) != 0) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, __location__ " Failed to start transaction\n"));
+		return -1;
+	}
+
+	tmp_db = tdb_open("tmpdb", tdb_hash_size(tdb), TDB_INTERNAL, O_RDWR|O_CREAT, 0);
+	if (tmp_db == NULL) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, __location__ " Failed to create tmp_db\n"));
+		tdb_transaction_cancel(tdb);
+		return -1;
+	}
+
+	state.error = false;
+	state.dest_db = tmp_db;
+
+	if (tdb_traverse_read(tdb, repack_traverse, &state) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, __location__ " Failed to traverse copying out\n"));
+		tdb_transaction_cancel(tdb);
+		tdb_close(tmp_db);
+		return -1;		
+	}
+
+	if (state.error) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, __location__ " Error during traversal\n"));
+		tdb_transaction_cancel(tdb);
+		tdb_close(tmp_db);
+		return -1;
+	}
+
+	if (tdb_wipe_all(tdb) != 0) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, __location__ " Failed to wipe database\n"));
+		tdb_transaction_cancel(tdb);
+		tdb_close(tmp_db);
+		return -1;
+	}
+
+	state.error = false;
+	state.dest_db = tdb;
+
+	if (tdb_traverse_read(tmp_db, repack_traverse, &state) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, __location__ " Failed to traverse copying back\n"));
+		tdb_transaction_cancel(tdb);
+		tdb_close(tmp_db);
+		return -1;		
+	}
+
+	if (state.error) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, __location__ " Error during second traversal\n"));
+		tdb_transaction_cancel(tdb);
+		tdb_close(tmp_db);
+		return -1;
+	}
+
+	tdb_close(tmp_db);
+
+	if (tdb_transaction_commit(tdb) != 0) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, __location__ " Failed to commit\n"));
+		return -1;
+	}
+
+	return 0;
+}
+
+/* Even on files, we can get partial writes due to signals. */
+bool tdb_write_all(int fd, const void *buf, size_t count)
+{
+	while (count) {
+		ssize_t ret;
+		ret = write(fd, buf, count);
+		if (ret < 0)
+			return false;
+		buf = (const char *)buf + ret;
+		count -= ret;
+	}
+	return true;
+}
+
+#ifdef TDB_TRACE
+static void tdb_trace_write(struct tdb_context *tdb, const char *str)
+{
+	if (!tdb_write_all(tdb->tracefd, str, strlen(str))) {
+		close(tdb->tracefd);
+		tdb->tracefd = -1;
+	}
+}
+
+static void tdb_trace_start(struct tdb_context *tdb)
+{
+	tdb_off_t seqnum=0;
+	char msg[sizeof(tdb_off_t) * 4 + 1];
+
+	tdb_ofs_read(tdb, TDB_SEQNUM_OFS, &seqnum);
+	snprintf(msg, sizeof(msg), "%u ", seqnum);
+	tdb_trace_write(tdb, msg);
+}
+
+static void tdb_trace_end(struct tdb_context *tdb)
+{
+	tdb_trace_write(tdb, "\n");
+}
+
+static void tdb_trace_end_ret(struct tdb_context *tdb, int ret)
+{
+	char msg[sizeof(ret) * 4 + 4];
+	snprintf(msg, sizeof(msg), " = %i\n", ret);
+	tdb_trace_write(tdb, msg);
+}
+
+static void tdb_trace_record(struct tdb_context *tdb, TDB_DATA rec)
+{
+	char msg[20 + rec.dsize*2], *p;
+	unsigned int i;
+
+	/* We differentiate zero-length records from non-existent ones. */
+	if (rec.dptr == NULL) {
+		tdb_trace_write(tdb, " NULL");
+		return;
+	}
+
+	/* snprintf here is purely cargo-cult programming. */
+	p = msg;
+	p += snprintf(p, sizeof(msg), " %zu:", rec.dsize);
+	for (i = 0; i < rec.dsize; i++)
+		p += snprintf(p, 2, "%02x", rec.dptr[i]);
+
+	tdb_trace_write(tdb, msg);
+}
+
+void tdb_trace(struct tdb_context *tdb, const char *op)
+{
+	tdb_trace_start(tdb);
+	tdb_trace_write(tdb, op);
+	tdb_trace_end(tdb);
+}
+
+void tdb_trace_seqnum(struct tdb_context *tdb, uint32_t seqnum, const char *op)
+{
+	char msg[sizeof(tdb_off_t) * 4 + 1];
+
+	snprintf(msg, sizeof(msg), "%u ", seqnum);
+	tdb_trace_write(tdb, msg);
+	tdb_trace_write(tdb, op);
+	tdb_trace_end(tdb);
+}
+
+void tdb_trace_open(struct tdb_context *tdb, const char *op,
+		    unsigned hash_size, unsigned tdb_flags, unsigned open_flags)
+{
+	char msg[128];
+
+	snprintf(msg, sizeof(msg),
+		 "%s %u 0x%x 0x%x", op, hash_size, tdb_flags, open_flags);
+	tdb_trace_start(tdb);
+	tdb_trace_write(tdb, msg);
+	tdb_trace_end(tdb);
+}
+
+void tdb_trace_ret(struct tdb_context *tdb, const char *op, int ret)
+{
+	tdb_trace_start(tdb);
+	tdb_trace_write(tdb, op);
+	tdb_trace_end_ret(tdb, ret);
+}
+
+void tdb_trace_retrec(struct tdb_context *tdb, const char *op, TDB_DATA ret)
+{
+	tdb_trace_start(tdb);
+	tdb_trace_write(tdb, op);
+	tdb_trace_write(tdb, " =");
+	tdb_trace_record(tdb, ret);
+	tdb_trace_end(tdb);
+}
+
+void tdb_trace_1rec(struct tdb_context *tdb, const char *op,
+		    TDB_DATA rec)
+{
+	tdb_trace_start(tdb);
+	tdb_trace_write(tdb, op);
+	tdb_trace_record(tdb, rec);
+	tdb_trace_end(tdb);
+}
+
+void tdb_trace_1rec_ret(struct tdb_context *tdb, const char *op,
+			TDB_DATA rec, int ret)
+{
+	tdb_trace_start(tdb);
+	tdb_trace_write(tdb, op);
+	tdb_trace_record(tdb, rec);
+	tdb_trace_end_ret(tdb, ret);
+}
+
+void tdb_trace_1rec_retrec(struct tdb_context *tdb, const char *op,
+			   TDB_DATA rec, TDB_DATA ret)
+{
+	tdb_trace_start(tdb);
+	tdb_trace_write(tdb, op);
+	tdb_trace_record(tdb, rec);
+	tdb_trace_write(tdb, " =");
+	tdb_trace_record(tdb, ret);
+	tdb_trace_end(tdb);
+}
+
+void tdb_trace_2rec_flag_ret(struct tdb_context *tdb, const char *op,
+			     TDB_DATA rec1, TDB_DATA rec2, unsigned flag,
+			     int ret)
+{
+	char msg[1 + sizeof(ret) * 4];
+
+	snprintf(msg, sizeof(msg), " %#x", flag);
+	tdb_trace_start(tdb);
+	tdb_trace_write(tdb, op);
+	tdb_trace_record(tdb, rec1);
+	tdb_trace_record(tdb, rec2);
+	tdb_trace_write(tdb, msg);
+	tdb_trace_end_ret(tdb, ret);
+}
+
+void tdb_trace_2rec_retrec(struct tdb_context *tdb, const char *op,
+			   TDB_DATA rec1, TDB_DATA rec2, TDB_DATA ret)
+{
+	tdb_trace_start(tdb);
+	tdb_trace_write(tdb, op);
+	tdb_trace_record(tdb, rec1);
+	tdb_trace_record(tdb, rec2);
+	tdb_trace_write(tdb, " =");
+	tdb_trace_record(tdb, ret);
+	tdb_trace_end(tdb);
+}
+#endif
diff --git a/ctdb/lib/tdb/common/tdb_private.h b/ctdb/lib/tdb/common/tdb_private.h
new file mode 100644
index 0000000000..0441fb287f
--- /dev/null
+++ b/ctdb/lib/tdb/common/tdb_private.h
@@ -0,0 +1,285 @@
+#ifndef TDB_PRIVATE_H
+#define TDB_PRIVATE_H
+ /* 
+   Unix SMB/CIFS implementation.
+
+   trivial database library - private includes
+
+   Copyright (C) Andrew Tridgell              2005
+
+     ** NOTE! The following LGPL license applies to the tdb
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "replace.h"
+#include "system/filesys.h"
+#include "system/time.h"
+#include "system/shmem.h"
+#include "system/select.h"
+#include "system/wait.h"
+#include "tdb.h"
+
+/* #define TDB_TRACE 1 */
+#ifndef HAVE_GETPAGESIZE
+#define getpagesize() 0x2000
+#endif
+
+typedef uint32_t tdb_len_t;
+typedef uint32_t tdb_off_t;
+
+#ifndef offsetof
+#define offsetof(t,f) ((unsigned int)&((t *)0)->f)
+#endif
+
+#define TDB_MAGIC_FOOD "TDB file\n"
+#define TDB_VERSION (0x26011967 + 6)
+#define TDB_MAGIC (0x26011999U)
+#define TDB_FREE_MAGIC (~TDB_MAGIC)
+#define TDB_DEAD_MAGIC (0xFEE1DEAD)
+#define TDB_RECOVERY_MAGIC (0xf53bc0e7U)
+#define TDB_RECOVERY_INVALID_MAGIC (0x0)
+#define TDB_HASH_RWLOCK_MAGIC (0xbad1a51U)
+#define TDB_ALIGNMENT 4
+#define DEFAULT_HASH_SIZE 131
+#define FREELIST_TOP (sizeof(struct tdb_header))
+#define TDB_ALIGN(x,a) (((x) + (a)-1) & ~((a)-1))
+#define TDB_BYTEREV(x) (((((x)&0xff)<<24)|((x)&0xFF00)<<8)|(((x)>>8)&0xFF00)|((x)>>24))
+#define TDB_DEAD(r) ((r)->magic == TDB_DEAD_MAGIC)
+#define TDB_BAD_MAGIC(r) ((r)->magic != TDB_MAGIC && !TDB_DEAD(r))
+#define TDB_HASH_TOP(hash) (FREELIST_TOP + (BUCKET(hash)+1)*sizeof(tdb_off_t))
+#define TDB_HASHTABLE_SIZE(tdb) ((tdb->header.hash_size+1)*sizeof(tdb_off_t))
+#define TDB_DATA_START(hash_size) (TDB_HASH_TOP(hash_size-1) + sizeof(tdb_off_t))
+#define TDB_RECOVERY_HEAD offsetof(struct tdb_header, recovery_start)
+#define TDB_SEQNUM_OFS    offsetof(struct tdb_header, sequence_number)
+#define TDB_PAD_BYTE 0x42
+#define TDB_PAD_U32  0x42424242
+
+/* NB assumes there is a local variable called "tdb" that is the
+ * current context, also takes doubly-parenthesized print-style
+ * argument. */
+#define TDB_LOG(x) tdb->log.log_fn x
+
+#ifdef TDB_TRACE
+void tdb_trace(struct tdb_context *tdb, const char *op);
+void tdb_trace_seqnum(struct tdb_context *tdb, uint32_t seqnum, const char *op);
+void tdb_trace_open(struct tdb_context *tdb, const char *op,
+		    unsigned hash_size, unsigned tdb_flags, unsigned open_flags);
+void tdb_trace_ret(struct tdb_context *tdb, const char *op, int ret);
+void tdb_trace_retrec(struct tdb_context *tdb, const char *op, TDB_DATA ret);
+void tdb_trace_1rec(struct tdb_context *tdb, const char *op,
+		    TDB_DATA rec);
+void tdb_trace_1rec_ret(struct tdb_context *tdb, const char *op,
+			TDB_DATA rec, int ret);
+void tdb_trace_1rec_retrec(struct tdb_context *tdb, const char *op,
+			   TDB_DATA rec, TDB_DATA ret);
+void tdb_trace_2rec_flag_ret(struct tdb_context *tdb, const char *op,
+			     TDB_DATA rec1, TDB_DATA rec2, unsigned flag,
+			     int ret);
+void tdb_trace_2rec_retrec(struct tdb_context *tdb, const char *op,
+			   TDB_DATA rec1, TDB_DATA rec2, TDB_DATA ret);
+#else
+#define tdb_trace(tdb, op)
+#define tdb_trace_seqnum(tdb, seqnum, op)
+#define tdb_trace_open(tdb, op, hash_size, tdb_flags, open_flags)
+#define tdb_trace_ret(tdb, op, ret)
+#define tdb_trace_retrec(tdb, op, ret)
+#define tdb_trace_1rec(tdb, op, rec)
+#define tdb_trace_1rec_ret(tdb, op, rec, ret)
+#define tdb_trace_1rec_retrec(tdb, op, rec, ret)
+#define tdb_trace_2rec_flag_ret(tdb, op, rec1, rec2, flag, ret)
+#define tdb_trace_2rec_retrec(tdb, op, rec1, rec2, ret)
+#endif /* !TDB_TRACE */
+
+/* lock offsets */
+#define OPEN_LOCK        0
+#define ACTIVE_LOCK      4
+#define TRANSACTION_LOCK 8
+
+/* free memory if the pointer is valid and zero the pointer */
+#ifndef SAFE_FREE
+#define SAFE_FREE(x) do { if ((x) != NULL) {free(x); (x)=NULL;} } while(0)
+#endif
+
+#define BUCKET(hash) ((hash) % tdb->header.hash_size)
+
+#define DOCONV() (tdb->flags & TDB_CONVERT)
+#define CONVERT(x) (DOCONV() ? tdb_convert(&x, sizeof(x)) : &x)
+
+
+/* the body of the database is made of one tdb_record for the free space
+   plus a separate data list for each hash value */
+struct tdb_record {
+	tdb_off_t next; /* offset of the next record in the list */
+	tdb_len_t rec_len; /* total byte length of record */
+	tdb_len_t key_len; /* byte length of key */
+	tdb_len_t data_len; /* byte length of data */
+	uint32_t full_hash; /* the full 32 bit hash of the key */
+	uint32_t magic;   /* try to catch errors */
+	/* the following union is implied:
+		union {
+			char record[rec_len];
+			struct {
+				char key[key_len];
+				char data[data_len];
+			}
+			uint32_t totalsize; (tailer)
+		}
+	*/
+};
+
+
+/* this is stored at the front of every database */
+struct tdb_header {
+	char magic_food[32]; /* for /etc/magic */
+	uint32_t version; /* version of the code */
+	uint32_t hash_size; /* number of hash entries */
+	tdb_off_t rwlocks; /* obsolete - kept to detect old formats */
+	tdb_off_t recovery_start; /* offset of transaction recovery region */
+	tdb_off_t sequence_number; /* used when TDB_SEQNUM is set */
+	uint32_t magic1_hash; /* hash of TDB_MAGIC_FOOD. */
+	uint32_t magic2_hash; /* hash of TDB_MAGIC. */
+	tdb_off_t reserved[27];
+};
+
+struct tdb_lock_type {
+	uint32_t off;
+	uint32_t count;
+	uint32_t ltype;
+};
+
+struct tdb_traverse_lock {
+	struct tdb_traverse_lock *next;
+	uint32_t off;
+	uint32_t hash;
+	int lock_rw;
+};
+
+enum tdb_lock_flags {
+	/* WAIT == F_SETLKW, NOWAIT == F_SETLK */
+	TDB_LOCK_NOWAIT = 0,
+	TDB_LOCK_WAIT = 1,
+	/* If set, don't log an error on failure. */
+	TDB_LOCK_PROBE = 2,
+	/* If set, don't actually lock at all. */
+	TDB_LOCK_MARK_ONLY = 4,
+};
+
+struct tdb_methods {
+	int (*tdb_read)(struct tdb_context *, tdb_off_t , void *, tdb_len_t , int );
+	int (*tdb_write)(struct tdb_context *, tdb_off_t, const void *, tdb_len_t);
+	void (*next_hash_chain)(struct tdb_context *, uint32_t *);
+	int (*tdb_oob)(struct tdb_context *, tdb_off_t , tdb_len_t, int );
+	int (*tdb_expand_file)(struct tdb_context *, tdb_off_t , tdb_off_t );
+};
+
+struct tdb_context {
+	char *name; /* the name of the database */
+	void *map_ptr; /* where it is currently mapped */
+	int fd; /* open file descriptor for the database */
+	tdb_len_t map_size; /* how much space has been mapped */
+	int read_only; /* opened read-only */
+	int traverse_read; /* read-only traversal */
+	int traverse_write; /* read-write traversal */
+	struct tdb_lock_type allrecord_lock; /* .offset == upgradable */
+	int num_lockrecs;
+	struct tdb_lock_type *lockrecs; /* only real locks, all with count>0 */
+	enum TDB_ERROR ecode; /* error code for last tdb error */
+	struct tdb_header header; /* a cached copy of the header */
+	uint32_t flags; /* the flags passed to tdb_open */
+	struct tdb_traverse_lock travlocks; /* current traversal locks */
+	struct tdb_context *next; /* all tdbs to avoid multiple opens */
+	dev_t device;	/* uniquely identifies this tdb */
+	ino_t inode;	/* uniquely identifies this tdb */
+	struct tdb_logging_context log;
+	unsigned int (*hash_fn)(TDB_DATA *key);
+	int open_flags; /* flags used in the open - needed by reopen */
+	const struct tdb_methods *methods;
+	struct tdb_transaction *transaction;
+	int page_size;
+	int max_dead_records;
+#ifdef TDB_TRACE
+	int tracefd;
+#endif
+	volatile sig_atomic_t *interrupt_sig_ptr;
+};
+
+
+/*
+  internal prototypes
+*/
+int tdb_munmap(struct tdb_context *tdb);
+int tdb_mmap(struct tdb_context *tdb);
+int tdb_lock(struct tdb_context *tdb, int list, int ltype);
+int tdb_lock_nonblock(struct tdb_context *tdb, int list, int ltype);
+int tdb_nest_lock(struct tdb_context *tdb, uint32_t offset, int ltype,
+		  enum tdb_lock_flags flags);
+int tdb_nest_unlock(struct tdb_context *tdb, uint32_t offset, int ltype,
+		    bool mark_lock);
+int tdb_unlock(struct tdb_context *tdb, int list, int ltype);
+int tdb_brlock(struct tdb_context *tdb,
+	       int rw_type, tdb_off_t offset, size_t len,
+	       enum tdb_lock_flags flags);
+int tdb_brunlock(struct tdb_context *tdb,
+		 int rw_type, tdb_off_t offset, size_t len);
+bool tdb_have_extra_locks(struct tdb_context *tdb);
+void tdb_release_transaction_locks(struct tdb_context *tdb);
+int tdb_transaction_lock(struct tdb_context *tdb, int ltype,
+			 enum tdb_lock_flags lockflags);
+int tdb_transaction_unlock(struct tdb_context *tdb, int ltype);
+int tdb_recovery_area(struct tdb_context *tdb,
+		      const struct tdb_methods *methods,
+		      tdb_off_t *recovery_offset,
+		      struct tdb_record *rec);
+int tdb_allrecord_lock(struct tdb_context *tdb, int ltype,
+		       enum tdb_lock_flags flags, bool upgradable);
+int tdb_allrecord_unlock(struct tdb_context *tdb, int ltype, bool mark_lock);
+int tdb_allrecord_upgrade(struct tdb_context *tdb);
+int tdb_write_lock_record(struct tdb_context *tdb, tdb_off_t off);
+int tdb_write_unlock_record(struct tdb_context *tdb, tdb_off_t off);
+int tdb_ofs_read(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d);
+int tdb_ofs_write(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d);
+void *tdb_convert(void *buf, uint32_t size);
+int tdb_free(struct tdb_context *tdb, tdb_off_t offset, struct tdb_record *rec);
+tdb_off_t tdb_allocate(struct tdb_context *tdb, tdb_len_t length, struct tdb_record *rec);
+int tdb_ofs_read(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d);
+int tdb_ofs_write(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d);
+int tdb_lock_record(struct tdb_context *tdb, tdb_off_t off);
+int tdb_unlock_record(struct tdb_context *tdb, tdb_off_t off);
+bool tdb_needs_recovery(struct tdb_context *tdb);
+int tdb_rec_read(struct tdb_context *tdb, tdb_off_t offset, struct tdb_record *rec);
+int tdb_rec_write(struct tdb_context *tdb, tdb_off_t offset, struct tdb_record *rec);
+int tdb_do_delete(struct tdb_context *tdb, tdb_off_t rec_ptr, struct tdb_record *rec);
+unsigned char *tdb_alloc_read(struct tdb_context *tdb, tdb_off_t offset, tdb_len_t len);
+int tdb_parse_data(struct tdb_context *tdb, TDB_DATA key,
+		   tdb_off_t offset, tdb_len_t len,
+		   int (*parser)(TDB_DATA key, TDB_DATA data,
+				 void *private_data),
+		   void *private_data);
+tdb_off_t tdb_find_lock_hash(struct tdb_context *tdb, TDB_DATA key, uint32_t hash, int locktype,
+			   struct tdb_record *rec);
+void tdb_io_init(struct tdb_context *tdb);
+int tdb_expand(struct tdb_context *tdb, tdb_off_t size);
+tdb_off_t tdb_expand_adjust(tdb_off_t map_size, tdb_off_t size, int page_size);
+int tdb_rec_free_read(struct tdb_context *tdb, tdb_off_t off,
+		      struct tdb_record *rec);
+bool tdb_write_all(int fd, const void *buf, size_t count);
+int tdb_transaction_recover(struct tdb_context *tdb);
+void tdb_header_hash(struct tdb_context *tdb,
+		     uint32_t *magic1_hash, uint32_t *magic2_hash);
+unsigned int tdb_old_hash(TDB_DATA *key);
+size_t tdb_dead_space(struct tdb_context *tdb, tdb_off_t off);
+#endif /* TDB_PRIVATE_H */
diff --git a/ctdb/lib/tdb/common/transaction.c b/ctdb/lib/tdb/common/transaction.c
new file mode 100644
index 0000000000..f18b4c229e
--- /dev/null
+++ b/ctdb/lib/tdb/common/transaction.c
@@ -0,0 +1,1294 @@
+ /* 
+   Unix SMB/CIFS implementation.
+
+   trivial database library
+
+   Copyright (C) Andrew Tridgell              2005
+
+     ** NOTE! The following LGPL license applies to the tdb
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "tdb_private.h"
+
+/*
+  transaction design:
+
+  - only allow a single transaction at a time per database. This makes
+    using the transaction API simpler, as otherwise the caller would
+    have to cope with temporary failures in transactions that conflict
+    with other current transactions
+
+  - keep the transaction recovery information in the same file as the
+    database, using a special 'transaction recovery' record pointed at
+    by the header. This removes the need for extra journal files as
+    used by some other databases
+
+  - dynamically allocated the transaction recover record, re-using it
+    for subsequent transactions. If a larger record is needed then
+    tdb_free() the old record to place it on the normal tdb freelist
+    before allocating the new record
+
+  - during transactions, keep a linked list of writes all that have
+    been performed by intercepting all tdb_write() calls. The hooked
+    transaction versions of tdb_read() and tdb_write() check this
+    linked list and try to use the elements of the list in preference
+    to the real database.
+
+  - don't allow any locks to be held when a transaction starts,
+    otherwise we can end up with deadlock (plus lack of lock nesting
+    in posix locks would mean the lock is lost)
+
+  - if the caller gains a lock during the transaction but doesn't
+    release it then fail the commit
+
+  - allow for nested calls to tdb_transaction_start(), re-using the
+    existing transaction record. If the inner transaction is cancelled
+    then a subsequent commit will fail
+
+  - keep a mirrored copy of the tdb hash chain heads to allow for the
+    fast hash heads scan on traverse, updating the mirrored copy in
+    the transaction version of tdb_write
+
+  - allow callers to mix transaction and non-transaction use of tdb,
+    although once a transaction is started then an exclusive lock is
+    gained until the transaction is committed or cancelled
+
+  - the commit stategy involves first saving away all modified data
+    into a linearised buffer in the transaction recovery area, then
+    marking the transaction recovery area with a magic value to
+    indicate a valid recovery record. In total 4 fsync/msync calls are
+    needed per commit to prevent race conditions. It might be possible
+    to reduce this to 3 or even 2 with some more work.
+
+  - check for a valid recovery record on open of the tdb, while the
+    open lock is held. Automatically recover from the transaction
+    recovery area if needed, then continue with the open as
+    usual. This allows for smooth crash recovery with no administrator
+    intervention.
+
+  - if TDB_NOSYNC is passed to flags in tdb_open then transactions are
+    still available, but no fsync/msync calls are made.  This means we
+    are still proof against a process dying during transaction commit,
+    but not against machine reboot.
+
+  - if TDB_ALLOW_NESTING is passed to flags in tdb open, or added using
+    tdb_add_flags() transaction nesting is enabled.
+    It resets the TDB_DISALLOW_NESTING flag, as both cannot be used together.
+    The default is that transaction nesting is allowed.
+    Note: this default may change in future versions of tdb.
+
+    Beware. when transactions are nested a transaction successfully
+    completed with tdb_transaction_commit() can be silently unrolled later.
+
+  - if TDB_DISALLOW_NESTING is passed to flags in tdb open, or added using
+    tdb_add_flags() transaction nesting is disabled.
+    It resets the TDB_ALLOW_NESTING flag, as both cannot be used together.
+    An attempt create a nested transaction will fail with TDB_ERR_NESTING.
+    The default is that transaction nesting is allowed.
+    Note: this default may change in future versions of tdb.
+*/
+
+
+/*
+  hold the context of any current transaction
+*/
+struct tdb_transaction {
+	/* we keep a mirrored copy of the tdb hash heads here so
+	   tdb_next_hash_chain() can operate efficiently */
+	uint32_t *hash_heads;
+
+	/* the original io methods - used to do IOs to the real db */
+	const struct tdb_methods *io_methods;
+
+	/* the list of transaction blocks. When a block is first
+	   written to, it gets created in this list */
+	uint8_t **blocks;
+	uint32_t num_blocks;
+	uint32_t block_size;      /* bytes in each block */
+	uint32_t last_block_size; /* number of valid bytes in the last block */
+
+	/* non-zero when an internal transaction error has
+	   occurred. All write operations will then fail until the
+	   transaction is ended */
+	int transaction_error;
+
+	/* when inside a transaction we need to keep track of any
+	   nested tdb_transaction_start() calls, as these are allowed,
+	   but don't create a new transaction */
+	int nesting;
+
+	/* set when a prepare has already occurred */
+	bool prepared;
+	tdb_off_t magic_offset;
+
+	/* old file size before transaction */
+	tdb_len_t old_map_size;
+
+	/* did we expand in this transaction */
+	bool expanded;
+};
+
+
+/*
+  read while in a transaction. We need to check first if the data is in our list
+  of transaction elements, then if not do a real read
+*/
+static int transaction_read(struct tdb_context *tdb, tdb_off_t off, void *buf, 
+			    tdb_len_t len, int cv)
+{
+	uint32_t blk;
+
+	/* break it down into block sized ops */
+	while (len + (off % tdb->transaction->block_size) > tdb->transaction->block_size) {
+		tdb_len_t len2 = tdb->transaction->block_size - (off % tdb->transaction->block_size);
+		if (transaction_read(tdb, off, buf, len2, cv) != 0) {
+			return -1;
+		}
+		len -= len2;
+		off += len2;
+		buf = (void *)(len2 + (char *)buf);
+	}
+
+	if (len == 0) {
+		return 0;
+	}
+
+	blk = off / tdb->transaction->block_size;
+
+	/* see if we have it in the block list */
+	if (tdb->transaction->num_blocks <= blk ||
+	    tdb->transaction->blocks[blk] == NULL) {
+		/* nope, do a real read */
+		if (tdb->transaction->io_methods->tdb_read(tdb, off, buf, len, cv) != 0) {
+			goto fail;
+		}
+		return 0;
+	}
+
+	/* it is in the block list. Now check for the last block */
+	if (blk == tdb->transaction->num_blocks-1) {
+		if (len > tdb->transaction->last_block_size) {
+			goto fail;
+		}
+	}
+
+	/* now copy it out of this block */
+	memcpy(buf, tdb->transaction->blocks[blk] + (off % tdb->transaction->block_size), len);
+	if (cv) {
+		tdb_convert(buf, len);
+	}
+	return 0;
+
+fail:
+	TDB_LOG((tdb, TDB_DEBUG_FATAL, "transaction_read: failed at off=%d len=%d\n", off, len));
+	tdb->ecode = TDB_ERR_IO;
+	tdb->transaction->transaction_error = 1;
+	return -1;
+}
+
+
+/*
+  write while in a transaction
+*/
+static int transaction_write(struct tdb_context *tdb, tdb_off_t off, 
+			     const void *buf, tdb_len_t len)
+{
+	uint32_t blk;
+
+	/* Only a commit is allowed on a prepared transaction */
+	if (tdb->transaction->prepared) {
+		tdb->ecode = TDB_ERR_EINVAL;
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "transaction_write: transaction already prepared, write not allowed\n"));
+		tdb->transaction->transaction_error = 1;
+		return -1;
+	}
+
+	/* if the write is to a hash head, then update the transaction
+	   hash heads */
+	if (len == sizeof(tdb_off_t) && off >= FREELIST_TOP &&
+	    off < FREELIST_TOP+TDB_HASHTABLE_SIZE(tdb)) {
+		uint32_t chain = (off-FREELIST_TOP) / sizeof(tdb_off_t);
+		memcpy(&tdb->transaction->hash_heads[chain], buf, len);
+	}
+
+	/* break it up into block sized chunks */
+	while (len + (off % tdb->transaction->block_size) > tdb->transaction->block_size) {
+		tdb_len_t len2 = tdb->transaction->block_size - (off % tdb->transaction->block_size);
+		if (transaction_write(tdb, off, buf, len2) != 0) {
+			return -1;
+		}
+		len -= len2;
+		off += len2;
+		if (buf != NULL) {
+			buf = (const void *)(len2 + (const char *)buf);
+		}
+	}
+
+	if (len == 0) {
+		return 0;
+	}
+
+	blk = off / tdb->transaction->block_size;
+	off = off % tdb->transaction->block_size;
+
+	if (tdb->transaction->num_blocks <= blk) {
+		uint8_t **new_blocks;
+		/* expand the blocks array */
+		if (tdb->transaction->blocks == NULL) {
+			new_blocks = (uint8_t **)malloc(
+				(blk+1)*sizeof(uint8_t *));
+		} else {
+			new_blocks = (uint8_t **)realloc(
+				tdb->transaction->blocks,
+				(blk+1)*sizeof(uint8_t *));
+		}
+		if (new_blocks == NULL) {
+			tdb->ecode = TDB_ERR_OOM;
+			goto fail;
+		}
+		memset(&new_blocks[tdb->transaction->num_blocks], 0, 
+		       (1+(blk - tdb->transaction->num_blocks))*sizeof(uint8_t *));
+		tdb->transaction->blocks = new_blocks;
+		tdb->transaction->num_blocks = blk+1;
+		tdb->transaction->last_block_size = 0;
+	}
+
+	/* allocate and fill a block? */
+	if (tdb->transaction->blocks[blk] == NULL) {
+		tdb->transaction->blocks[blk] = (uint8_t *)calloc(tdb->transaction->block_size, 1);
+		if (tdb->transaction->blocks[blk] == NULL) {
+			tdb->ecode = TDB_ERR_OOM;
+			tdb->transaction->transaction_error = 1;
+			return -1;			
+		}
+		if (tdb->transaction->old_map_size > blk * tdb->transaction->block_size) {
+			tdb_len_t len2 = tdb->transaction->block_size;
+			if (len2 + (blk * tdb->transaction->block_size) > tdb->transaction->old_map_size) {
+				len2 = tdb->transaction->old_map_size - (blk * tdb->transaction->block_size);
+			}
+			if (tdb->transaction->io_methods->tdb_read(tdb, blk * tdb->transaction->block_size, 
+								   tdb->transaction->blocks[blk], 
+								   len2, 0) != 0) {
+				SAFE_FREE(tdb->transaction->blocks[blk]);				
+				tdb->ecode = TDB_ERR_IO;
+				goto fail;
+			}
+			if (blk == tdb->transaction->num_blocks-1) {
+				tdb->transaction->last_block_size = len2;
+			}			
+		}
+	}
+
+	/* overwrite part of an existing block */
+	if (buf == NULL) {
+		memset(tdb->transaction->blocks[blk] + off, 0, len);
+	} else {
+		memcpy(tdb->transaction->blocks[blk] + off, buf, len);
+	}
+	if (blk == tdb->transaction->num_blocks-1) {
+		if (len + off > tdb->transaction->last_block_size) {
+			tdb->transaction->last_block_size = len + off;
+		}
+	}
+
+	return 0;
+
+fail:
+	TDB_LOG((tdb, TDB_DEBUG_FATAL, "transaction_write: failed at off=%d len=%d\n", 
+		 (blk*tdb->transaction->block_size) + off, len));
+	tdb->transaction->transaction_error = 1;
+	return -1;
+}
+
+
+/*
+  write while in a transaction - this varient never expands the transaction blocks, it only
+  updates existing blocks. This means it cannot change the recovery size
+*/
+static int transaction_write_existing(struct tdb_context *tdb, tdb_off_t off, 
+				      const void *buf, tdb_len_t len)
+{
+	uint32_t blk;
+
+	/* break it up into block sized chunks */
+	while (len + (off % tdb->transaction->block_size) > tdb->transaction->block_size) {
+		tdb_len_t len2 = tdb->transaction->block_size - (off % tdb->transaction->block_size);
+		if (transaction_write_existing(tdb, off, buf, len2) != 0) {
+			return -1;
+		}
+		len -= len2;
+		off += len2;
+		if (buf != NULL) {
+			buf = (const void *)(len2 + (const char *)buf);
+		}
+	}
+
+	if (len == 0) {
+		return 0;
+	}
+
+	blk = off / tdb->transaction->block_size;
+	off = off % tdb->transaction->block_size;
+
+	if (tdb->transaction->num_blocks <= blk ||
+	    tdb->transaction->blocks[blk] == NULL) {
+		return 0;
+	}
+
+	if (blk == tdb->transaction->num_blocks-1 &&
+	    off + len > tdb->transaction->last_block_size) {
+		if (off >= tdb->transaction->last_block_size) {
+			return 0;
+		}
+		len = tdb->transaction->last_block_size - off;
+	}
+
+	/* overwrite part of an existing block */
+	memcpy(tdb->transaction->blocks[blk] + off, buf, len);
+
+	return 0;
+}
+
+
+/*
+  accelerated hash chain head search, using the cached hash heads
+*/
+static void transaction_next_hash_chain(struct tdb_context *tdb, uint32_t *chain)
+{
+	uint32_t h = *chain;
+	for (;h < tdb->header.hash_size;h++) {
+		/* the +1 takes account of the freelist */
+		if (0 != tdb->transaction->hash_heads[h+1]) {
+			break;
+		}
+	}
+	(*chain) = h;
+}
+
+/*
+  out of bounds check during a transaction
+*/
+static int transaction_oob(struct tdb_context *tdb, tdb_off_t off,
+			   tdb_len_t len, int probe)
+{
+	if (off + len >= off && off + len <= tdb->map_size) {
+		return 0;
+	}
+	tdb->ecode = TDB_ERR_IO;
+	return -1;
+}
+
+/*
+  transaction version of tdb_expand().
+*/
+static int transaction_expand_file(struct tdb_context *tdb, tdb_off_t size, 
+				   tdb_off_t addition)
+{
+	/* add a write to the transaction elements, so subsequent
+	   reads see the zero data */
+	if (transaction_write(tdb, size, NULL, addition) != 0) {
+		return -1;
+	}
+
+	tdb->transaction->expanded = true;
+
+	return 0;
+}
+
+static const struct tdb_methods transaction_methods = {
+	transaction_read,
+	transaction_write,
+	transaction_next_hash_chain,
+	transaction_oob,
+	transaction_expand_file,
+};
+
+
+/*
+  start a tdb transaction. No token is returned, as only a single
+  transaction is allowed to be pending per tdb_context
+*/
+static int _tdb_transaction_start(struct tdb_context *tdb,
+				  enum tdb_lock_flags lockflags)
+{
+	/* some sanity checks */
+	if (tdb->read_only || (tdb->flags & TDB_INTERNAL) || tdb->traverse_read) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_transaction_start: cannot start a transaction on a read-only or internal db\n"));
+		tdb->ecode = TDB_ERR_EINVAL;
+		return -1;
+	}
+
+	/* cope with nested tdb_transaction_start() calls */
+	if (tdb->transaction != NULL) {
+		if (!(tdb->flags & TDB_ALLOW_NESTING)) {
+			tdb->ecode = TDB_ERR_NESTING;
+			return -1;
+		}
+		tdb->transaction->nesting++;
+		TDB_LOG((tdb, TDB_DEBUG_TRACE, "tdb_transaction_start: nesting %d\n", 
+			 tdb->transaction->nesting));
+		return 0;
+	}
+
+	if (tdb_have_extra_locks(tdb)) {
+		/* the caller must not have any locks when starting a
+		   transaction as otherwise we'll be screwed by lack
+		   of nested locks in posix */
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_transaction_start: cannot start a transaction with locks held\n"));
+		tdb->ecode = TDB_ERR_LOCK;
+		return -1;
+	}
+
+	if (tdb->travlocks.next != NULL) {
+		/* you cannot use transactions inside a traverse (although you can use
+		   traverse inside a transaction) as otherwise you can end up with
+		   deadlock */
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_transaction_start: cannot start a transaction within a traverse\n"));
+		tdb->ecode = TDB_ERR_LOCK;
+		return -1;
+	}
+
+	tdb->transaction = (struct tdb_transaction *)
+		calloc(sizeof(struct tdb_transaction), 1);
+	if (tdb->transaction == NULL) {
+		tdb->ecode = TDB_ERR_OOM;
+		return -1;
+	}
+
+	/* a page at a time seems like a reasonable compromise between compactness and efficiency */
+	tdb->transaction->block_size = tdb->page_size;
+
+	/* get the transaction write lock. This is a blocking lock. As
+	   discussed with Volker, there are a number of ways we could
+	   make this async, which we will probably do in the future */
+	if (tdb_transaction_lock(tdb, F_WRLCK, lockflags) == -1) {
+		SAFE_FREE(tdb->transaction->blocks);
+		SAFE_FREE(tdb->transaction);
+		if ((lockflags & TDB_LOCK_WAIT) == 0) {
+			tdb->ecode = TDB_ERR_NOLOCK;
+		}
+		return -1;
+	}
+
+	/* get a read lock from the freelist to the end of file. This
+	   is upgraded to a write lock during the commit */
+	if (tdb_allrecord_lock(tdb, F_RDLCK, TDB_LOCK_WAIT, true) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_transaction_start: failed to get hash locks\n"));
+		goto fail_allrecord_lock;
+	}
+
+	/* setup a copy of the hash table heads so the hash scan in
+	   traverse can be fast */
+	tdb->transaction->hash_heads = (uint32_t *)
+		calloc(tdb->header.hash_size+1, sizeof(uint32_t));
+	if (tdb->transaction->hash_heads == NULL) {
+		tdb->ecode = TDB_ERR_OOM;
+		goto fail;
+	}
+	if (tdb->methods->tdb_read(tdb, FREELIST_TOP, tdb->transaction->hash_heads,
+				   TDB_HASHTABLE_SIZE(tdb), 0) != 0) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_start: failed to read hash heads\n"));
+		tdb->ecode = TDB_ERR_IO;
+		goto fail;
+	}
+
+	/* make sure we know about any file expansions already done by
+	   anyone else */
+	tdb->methods->tdb_oob(tdb, tdb->map_size, 1, 1);
+	tdb->transaction->old_map_size = tdb->map_size;
+
+	/* finally hook the io methods, replacing them with
+	   transaction specific methods */
+	tdb->transaction->io_methods = tdb->methods;
+	tdb->methods = &transaction_methods;
+
+	/* Trace at the end, so we get sequence number correct. */
+	tdb_trace(tdb, "tdb_transaction_start");
+	return 0;
+
+fail:
+	tdb_allrecord_unlock(tdb, F_RDLCK, false);
+fail_allrecord_lock:
+	tdb_transaction_unlock(tdb, F_WRLCK);
+	SAFE_FREE(tdb->transaction->blocks);
+	SAFE_FREE(tdb->transaction->hash_heads);
+	SAFE_FREE(tdb->transaction);
+	return -1;
+}
+
+_PUBLIC_ int tdb_transaction_start(struct tdb_context *tdb)
+{
+	return _tdb_transaction_start(tdb, TDB_LOCK_WAIT);
+}
+
+_PUBLIC_ int tdb_transaction_start_nonblock(struct tdb_context *tdb)
+{
+	return _tdb_transaction_start(tdb, TDB_LOCK_NOWAIT|TDB_LOCK_PROBE);
+}
+
+/*
+  sync to disk
+*/
+static int transaction_sync(struct tdb_context *tdb, tdb_off_t offset, tdb_len_t length)
+{	
+	if (tdb->flags & TDB_NOSYNC) {
+		return 0;
+	}
+
+#ifdef HAVE_FDATASYNC
+	if (fdatasync(tdb->fd) != 0) {
+#else
+	if (fsync(tdb->fd) != 0) {
+#endif
+		tdb->ecode = TDB_ERR_IO;
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction: fsync failed\n"));
+		return -1;
+	}
+#ifdef HAVE_MMAP
+	if (tdb->map_ptr) {
+		tdb_off_t moffset = offset & ~(tdb->page_size-1);
+		if (msync(moffset + (char *)tdb->map_ptr, 
+			  length + (offset - moffset), MS_SYNC) != 0) {
+			tdb->ecode = TDB_ERR_IO;
+			TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction: msync failed - %s\n",
+				 strerror(errno)));
+			return -1;
+		}
+	}
+#endif
+	return 0;
+}
+
+
+static int _tdb_transaction_cancel(struct tdb_context *tdb)
+{	
+	int i, ret = 0;
+
+	if (tdb->transaction == NULL) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_transaction_cancel: no transaction\n"));
+		return -1;
+	}
+
+	if (tdb->transaction->nesting != 0) {
+		tdb->transaction->transaction_error = 1;
+		tdb->transaction->nesting--;
+		return 0;
+	}		
+
+	tdb->map_size = tdb->transaction->old_map_size;
+
+	/* free all the transaction blocks */
+	for (i=0;i<tdb->transaction->num_blocks;i++) {
+		if (tdb->transaction->blocks[i] != NULL) {
+			free(tdb->transaction->blocks[i]);
+		}
+	}
+	SAFE_FREE(tdb->transaction->blocks);
+
+	if (tdb->transaction->magic_offset) {
+		const struct tdb_methods *methods = tdb->transaction->io_methods;
+		const uint32_t invalid = TDB_RECOVERY_INVALID_MAGIC;
+
+		/* remove the recovery marker */
+		if (methods->tdb_write(tdb, tdb->transaction->magic_offset, &invalid, 4) == -1 ||
+		transaction_sync(tdb, tdb->transaction->magic_offset, 4) == -1) {
+			TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_cancel: failed to remove recovery magic\n"));
+			ret = -1;
+		}
+	}
+
+	/* This also removes the OPEN_LOCK, if we have it. */
+	tdb_release_transaction_locks(tdb);
+
+	/* restore the normal io methods */
+	tdb->methods = tdb->transaction->io_methods;
+
+	SAFE_FREE(tdb->transaction->hash_heads);
+	SAFE_FREE(tdb->transaction);
+
+	return ret;
+}
+
+/*
+  cancel the current transaction
+*/
+_PUBLIC_ int tdb_transaction_cancel(struct tdb_context *tdb)
+{
+	tdb_trace(tdb, "tdb_transaction_cancel");
+	return _tdb_transaction_cancel(tdb);
+}
+
+/*
+  work out how much space the linearised recovery data will consume
+*/
+static tdb_len_t tdb_recovery_size(struct tdb_context *tdb)
+{
+	tdb_len_t recovery_size = 0;
+	int i;
+
+	recovery_size = sizeof(uint32_t);
+	for (i=0;i<tdb->transaction->num_blocks;i++) {
+		if (i * tdb->transaction->block_size >= tdb->transaction->old_map_size) {
+			break;
+		}
+		if (tdb->transaction->blocks[i] == NULL) {
+			continue;
+		}
+		recovery_size += 2*sizeof(tdb_off_t);
+		if (i == tdb->transaction->num_blocks-1) {
+			recovery_size += tdb->transaction->last_block_size;
+		} else {
+			recovery_size += tdb->transaction->block_size;
+		}
+	}	
+
+	return recovery_size;
+}
+
+int tdb_recovery_area(struct tdb_context *tdb,
+		      const struct tdb_methods *methods,
+		      tdb_off_t *recovery_offset,
+		      struct tdb_record *rec)
+{
+	if (tdb_ofs_read(tdb, TDB_RECOVERY_HEAD, recovery_offset) == -1) {
+		return -1;
+	}
+
+	if (*recovery_offset == 0) {
+		rec->rec_len = 0;
+		return 0;
+	}
+
+	if (methods->tdb_read(tdb, *recovery_offset, rec, sizeof(*rec),
+			      DOCONV()) == -1) {
+		return -1;
+	}
+
+	/* ignore invalid recovery regions: can happen in crash */
+	if (rec->magic != TDB_RECOVERY_MAGIC &&
+	    rec->magic != TDB_RECOVERY_INVALID_MAGIC) {
+		*recovery_offset = 0;
+		rec->rec_len = 0;
+	}
+	return 0;
+}
+
+/*
+  allocate the recovery area, or use an existing recovery area if it is
+  large enough
+*/
+static int tdb_recovery_allocate(struct tdb_context *tdb, 
+				 tdb_len_t *recovery_size,
+				 tdb_off_t *recovery_offset,
+				 tdb_len_t *recovery_max_size)
+{
+	struct tdb_record rec;
+	const struct tdb_methods *methods = tdb->transaction->io_methods;
+	tdb_off_t recovery_head, new_end;
+
+	if (tdb_recovery_area(tdb, methods, &recovery_head, &rec) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_recovery_allocate: failed to read recovery head\n"));
+		return -1;
+	}
+
+	*recovery_size = tdb_recovery_size(tdb);
+
+	/* Existing recovery area? */
+	if (recovery_head != 0 && *recovery_size <= rec.rec_len) {
+		/* it fits in the existing area */
+		*recovery_max_size = rec.rec_len;
+		*recovery_offset = recovery_head;
+		return 0;
+	}
+
+	/* If recovery area in middle of file, we need a new one. */
+	if (recovery_head == 0
+	    || recovery_head + sizeof(rec) + rec.rec_len != tdb->map_size) {
+		/* we need to free up the old recovery area, then allocate a
+		   new one at the end of the file. Note that we cannot use
+		   tdb_allocate() to allocate the new one as that might return
+		   us an area that is being currently used (as of the start of
+		   the transaction) */
+		if (recovery_head) {
+			if (tdb_free(tdb, recovery_head, &rec) == -1) {
+				TDB_LOG((tdb, TDB_DEBUG_FATAL,
+					 "tdb_recovery_allocate: failed to"
+					 " free previous recovery area\n"));
+				return -1;
+			}
+
+			/* the tdb_free() call might have increased
+			 * the recovery size */
+			*recovery_size = tdb_recovery_size(tdb);
+		}
+
+		/* New head will be at end of file. */
+		recovery_head = tdb->map_size;
+	}
+
+	/* Now we know where it will be. */
+	*recovery_offset = recovery_head;
+
+	/* Expand by more than we need, so we don't do it often. */
+	*recovery_max_size = tdb_expand_adjust(tdb->map_size,
+					       *recovery_size,
+					       tdb->page_size)
+		- sizeof(rec);
+
+	new_end = recovery_head + sizeof(rec) + *recovery_max_size;
+
+	if (methods->tdb_expand_file(tdb, tdb->transaction->old_map_size, 
+				     new_end - tdb->transaction->old_map_size)
+	    == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_recovery_allocate: failed to create recovery area\n"));
+		return -1;
+	}
+
+	/* remap the file (if using mmap) */
+	methods->tdb_oob(tdb, tdb->map_size, 1, 1);
+
+	/* we have to reset the old map size so that we don't try to expand the file
+	   again in the transaction commit, which would destroy the recovery area */
+	tdb->transaction->old_map_size = tdb->map_size;
+
+	/* write the recovery header offset and sync - we can sync without a race here
+	   as the magic ptr in the recovery record has not been set */
+	CONVERT(recovery_head);
+	if (methods->tdb_write(tdb, TDB_RECOVERY_HEAD, 
+			       &recovery_head, sizeof(tdb_off_t)) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_recovery_allocate: failed to write recovery head\n"));
+		return -1;
+	}
+	if (transaction_write_existing(tdb, TDB_RECOVERY_HEAD, &recovery_head, sizeof(tdb_off_t)) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_recovery_allocate: failed to write recovery head\n"));
+		return -1;
+	}
+
+	return 0;
+}
+
+
+/*
+  setup the recovery data that will be used on a crash during commit
+*/
+static int transaction_setup_recovery(struct tdb_context *tdb, 
+				      tdb_off_t *magic_offset)
+{
+	tdb_len_t recovery_size;
+	unsigned char *data, *p;
+	const struct tdb_methods *methods = tdb->transaction->io_methods;
+	struct tdb_record *rec;
+	tdb_off_t recovery_offset, recovery_max_size;
+	tdb_off_t old_map_size = tdb->transaction->old_map_size;
+	uint32_t magic, tailer;
+	int i;
+
+	/*
+	  check that the recovery area has enough space
+	*/
+	if (tdb_recovery_allocate(tdb, &recovery_size, 
+				  &recovery_offset, &recovery_max_size) == -1) {
+		return -1;
+	}
+
+	data = (unsigned char *)malloc(recovery_size + sizeof(*rec));
+	if (data == NULL) {
+		tdb->ecode = TDB_ERR_OOM;
+		return -1;
+	}
+
+	rec = (struct tdb_record *)data;
+	memset(rec, 0, sizeof(*rec));
+
+	rec->magic    = TDB_RECOVERY_INVALID_MAGIC;
+	rec->data_len = recovery_size;
+	rec->rec_len  = recovery_max_size;
+	rec->key_len  = old_map_size;
+	CONVERT(*rec);
+
+	/* build the recovery data into a single blob to allow us to do a single
+	   large write, which should be more efficient */
+	p = data + sizeof(*rec);
+	for (i=0;i<tdb->transaction->num_blocks;i++) {
+		tdb_off_t offset;
+		tdb_len_t length;
+
+		if (tdb->transaction->blocks[i] == NULL) {
+			continue;
+		}
+
+		offset = i * tdb->transaction->block_size;
+		length = tdb->transaction->block_size;
+		if (i == tdb->transaction->num_blocks-1) {
+			length = tdb->transaction->last_block_size;
+		}
+
+		if (offset >= old_map_size) {
+			continue;
+		}
+		if (offset + length > tdb->transaction->old_map_size) {
+			TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_setup_recovery: transaction data over new region boundary\n"));
+			free(data);
+			tdb->ecode = TDB_ERR_CORRUPT;
+			return -1;
+		}
+		memcpy(p, &offset, 4);
+		memcpy(p+4, &length, 4);
+		if (DOCONV()) {
+			tdb_convert(p, 8);
+		}
+		/* the recovery area contains the old data, not the
+		   new data, so we have to call the original tdb_read
+		   method to get it */
+		if (methods->tdb_read(tdb, offset, p + 8, length, 0) != 0) {
+			free(data);
+			tdb->ecode = TDB_ERR_IO;
+			return -1;
+		}
+		p += 8 + length;
+	}
+
+	/* and the tailer */
+	tailer = sizeof(*rec) + recovery_max_size;
+	memcpy(p, &tailer, 4);
+	if (DOCONV()) {
+		tdb_convert(p, 4);
+	}
+
+	/* write the recovery data to the recovery area */
+	if (methods->tdb_write(tdb, recovery_offset, data, sizeof(*rec) + recovery_size) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_setup_recovery: failed to write recovery data\n"));
+		free(data);
+		tdb->ecode = TDB_ERR_IO;
+		return -1;
+	}
+	if (transaction_write_existing(tdb, recovery_offset, data, sizeof(*rec) + recovery_size) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_setup_recovery: failed to write secondary recovery data\n"));
+		free(data);
+		tdb->ecode = TDB_ERR_IO;
+		return -1;
+	}
+
+	/* as we don't have ordered writes, we have to sync the recovery
+	   data before we update the magic to indicate that the recovery
+	   data is present */
+	if (transaction_sync(tdb, recovery_offset, sizeof(*rec) + recovery_size) == -1) {
+		free(data);
+		return -1;
+	}
+
+	free(data);
+
+	magic = TDB_RECOVERY_MAGIC;
+	CONVERT(magic);
+
+	*magic_offset = recovery_offset + offsetof(struct tdb_record, magic);
+
+	if (methods->tdb_write(tdb, *magic_offset, &magic, sizeof(magic)) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_setup_recovery: failed to write recovery magic\n"));
+		tdb->ecode = TDB_ERR_IO;
+		return -1;
+	}
+	if (transaction_write_existing(tdb, *magic_offset, &magic, sizeof(magic)) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_setup_recovery: failed to write secondary recovery magic\n"));
+		tdb->ecode = TDB_ERR_IO;
+		return -1;
+	}
+
+	/* ensure the recovery magic marker is on disk */
+	if (transaction_sync(tdb, *magic_offset, sizeof(magic)) == -1) {
+		return -1;
+	}
+
+	return 0;
+}
+
+static int _tdb_transaction_prepare_commit(struct tdb_context *tdb)
+{	
+	const struct tdb_methods *methods;
+
+	if (tdb->transaction == NULL) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_transaction_prepare_commit: no transaction\n"));
+		return -1;
+	}
+
+	if (tdb->transaction->prepared) {
+		tdb->ecode = TDB_ERR_EINVAL;
+		_tdb_transaction_cancel(tdb);
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_transaction_prepare_commit: transaction already prepared\n"));
+		return -1;
+	}
+
+	if (tdb->transaction->transaction_error) {
+		tdb->ecode = TDB_ERR_IO;
+		_tdb_transaction_cancel(tdb);
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_transaction_prepare_commit: transaction error pending\n"));
+		return -1;
+	}
+
+
+	if (tdb->transaction->nesting != 0) {
+		return 0;
+	}		
+
+	/* check for a null transaction */
+	if (tdb->transaction->blocks == NULL) {
+		return 0;
+	}
+
+	methods = tdb->transaction->io_methods;
+
+	/* if there are any locks pending then the caller has not
+	   nested their locks properly, so fail the transaction */
+	if (tdb_have_extra_locks(tdb)) {
+		tdb->ecode = TDB_ERR_LOCK;
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_transaction_prepare_commit: locks pending on commit\n"));
+		_tdb_transaction_cancel(tdb);
+		return -1;
+	}
+
+	/* upgrade the main transaction lock region to a write lock */
+	if (tdb_allrecord_upgrade(tdb) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_transaction_prepare_commit: failed to upgrade hash locks\n"));
+		_tdb_transaction_cancel(tdb);
+		return -1;
+	}
+
+	/* get the open lock - this prevents new users attaching to the database
+	   during the commit */
+	if (tdb_nest_lock(tdb, OPEN_LOCK, F_WRLCK, TDB_LOCK_WAIT) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_transaction_prepare_commit: failed to get open lock\n"));
+		_tdb_transaction_cancel(tdb);
+		return -1;
+	}
+
+	/* write the recovery data to the end of the file */
+	if (transaction_setup_recovery(tdb, &tdb->transaction->magic_offset) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_prepare_commit: failed to setup recovery data\n"));
+		_tdb_transaction_cancel(tdb);
+		return -1;
+	}
+
+	tdb->transaction->prepared = true;
+
+	/* expand the file to the new size if needed */
+	if (tdb->map_size != tdb->transaction->old_map_size) {
+		if (methods->tdb_expand_file(tdb, tdb->transaction->old_map_size, 
+					     tdb->map_size - 
+					     tdb->transaction->old_map_size) == -1) {
+			tdb->ecode = TDB_ERR_IO;
+			TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_prepare_commit: expansion failed\n"));
+			_tdb_transaction_cancel(tdb);
+			return -1;
+		}
+		tdb->map_size = tdb->transaction->old_map_size;
+		methods->tdb_oob(tdb, tdb->map_size, 1, 1);
+	}
+
+	/* Keep the open lock until the actual commit */
+
+	return 0;
+}
+
+/*
+   prepare to commit the current transaction
+*/
+_PUBLIC_ int tdb_transaction_prepare_commit(struct tdb_context *tdb)
+{
+	tdb_trace(tdb, "tdb_transaction_prepare_commit");
+	return _tdb_transaction_prepare_commit(tdb);
+}
+
+/* A repack is worthwhile if the largest is less than half total free. */
+static bool repack_worthwhile(struct tdb_context *tdb)
+{
+	tdb_off_t ptr;
+	struct tdb_record rec;
+	tdb_len_t total = 0, largest = 0;
+
+	if (tdb_ofs_read(tdb, FREELIST_TOP, &ptr) == -1) {
+		return false;
+	}
+
+	while (ptr != 0 && tdb_rec_free_read(tdb, ptr, &rec) == 0) {
+		total += rec.rec_len;
+		if (rec.rec_len > largest) {
+			largest = rec.rec_len;
+		}
+		ptr = rec.next;
+	}
+
+	return total > largest * 2;
+}
+
+/*
+  commit the current transaction
+*/
+_PUBLIC_ int tdb_transaction_commit(struct tdb_context *tdb)
+{
+	const struct tdb_methods *methods;
+	int i;
+	bool need_repack = false;
+
+	if (tdb->transaction == NULL) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_transaction_commit: no transaction\n"));
+		return -1;
+	}
+
+	tdb_trace(tdb, "tdb_transaction_commit");
+
+	if (tdb->transaction->transaction_error) {
+		tdb->ecode = TDB_ERR_IO;
+		_tdb_transaction_cancel(tdb);
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_transaction_commit: transaction error pending\n"));
+		return -1;
+	}
+
+
+	if (tdb->transaction->nesting != 0) {
+		tdb->transaction->nesting--;
+		return 0;
+	}
+
+	/* check for a null transaction */
+	if (tdb->transaction->blocks == NULL) {
+		_tdb_transaction_cancel(tdb);
+		return 0;
+	}
+
+	if (!tdb->transaction->prepared) {
+		int ret = _tdb_transaction_prepare_commit(tdb);
+		if (ret)
+			return ret;
+	}
+
+	methods = tdb->transaction->io_methods;
+
+	/* perform all the writes */
+	for (i=0;i<tdb->transaction->num_blocks;i++) {
+		tdb_off_t offset;
+		tdb_len_t length;
+
+		if (tdb->transaction->blocks[i] == NULL) {
+			continue;
+		}
+
+		offset = i * tdb->transaction->block_size;
+		length = tdb->transaction->block_size;
+		if (i == tdb->transaction->num_blocks-1) {
+			length = tdb->transaction->last_block_size;
+		}
+
+		if (methods->tdb_write(tdb, offset, tdb->transaction->blocks[i], length) == -1) {
+			TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_commit: write failed during commit\n"));
+
+			/* we've overwritten part of the data and
+			   possibly expanded the file, so we need to
+			   run the crash recovery code */
+			tdb->methods = methods;
+			tdb_transaction_recover(tdb); 
+
+			_tdb_transaction_cancel(tdb);
+
+			TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_commit: write failed\n"));
+			return -1;
+		}
+		SAFE_FREE(tdb->transaction->blocks[i]);
+	} 
+
+	/* Do this before we drop lock or blocks. */
+	if (tdb->transaction->expanded) {
+		need_repack = repack_worthwhile(tdb);
+	}
+
+	SAFE_FREE(tdb->transaction->blocks);
+	tdb->transaction->num_blocks = 0;
+
+	/* ensure the new data is on disk */
+	if (transaction_sync(tdb, 0, tdb->map_size) == -1) {
+		return -1;
+	}
+
+	/*
+	  TODO: maybe write to some dummy hdr field, or write to magic
+	  offset without mmap, before the last sync, instead of the
+	  utime() call
+	*/
+
+	/* on some systems (like Linux 2.6.x) changes via mmap/msync
+	   don't change the mtime of the file, this means the file may
+	   not be backed up (as tdb rounding to block sizes means that
+	   file size changes are quite rare too). The following forces
+	   mtime changes when a transaction completes */
+#ifdef HAVE_UTIME
+	utime(tdb->name, NULL);
+#endif
+
+	/* use a transaction cancel to free memory and remove the
+	   transaction locks */
+	_tdb_transaction_cancel(tdb);
+
+	if (need_repack) {
+		return tdb_repack(tdb);
+	}
+
+	return 0;
+}
+
+
+/*
+  recover from an aborted transaction. Must be called with exclusive
+  database write access already established (including the open
+  lock to prevent new processes attaching)
+*/
+int tdb_transaction_recover(struct tdb_context *tdb)
+{
+	tdb_off_t recovery_head, recovery_eof;
+	unsigned char *data, *p;
+	uint32_t zero = 0;
+	struct tdb_record rec;
+
+	/* find the recovery area */
+	if (tdb_ofs_read(tdb, TDB_RECOVERY_HEAD, &recovery_head) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_recover: failed to read recovery head\n"));
+		tdb->ecode = TDB_ERR_IO;
+		return -1;
+	}
+
+	if (recovery_head == 0) {
+		/* we have never allocated a recovery record */
+		return 0;
+	}
+
+	/* read the recovery record */
+	if (tdb->methods->tdb_read(tdb, recovery_head, &rec, 
+				   sizeof(rec), DOCONV()) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_recover: failed to read recovery record\n"));		
+		tdb->ecode = TDB_ERR_IO;
+		return -1;
+	}
+
+	if (rec.magic != TDB_RECOVERY_MAGIC) {
+		/* there is no valid recovery data */
+		return 0;
+	}
+
+	if (tdb->read_only) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_recover: attempt to recover read only database\n"));
+		tdb->ecode = TDB_ERR_CORRUPT;
+		return -1;
+	}
+
+	recovery_eof = rec.key_len;
+
+	data = (unsigned char *)malloc(rec.data_len);
+	if (data == NULL) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_recover: failed to allocate recovery data\n"));		
+		tdb->ecode = TDB_ERR_OOM;
+		return -1;
+	}
+
+	/* read the full recovery data */
+	if (tdb->methods->tdb_read(tdb, recovery_head + sizeof(rec), data,
+				   rec.data_len, 0) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_recover: failed to read recovery data\n"));		
+		tdb->ecode = TDB_ERR_IO;
+		return -1;
+	}
+
+	/* recover the file data */
+	p = data;
+	while (p+8 < data + rec.data_len) {
+		uint32_t ofs, len;
+		if (DOCONV()) {
+			tdb_convert(p, 8);
+		}
+		memcpy(&ofs, p, 4);
+		memcpy(&len, p+4, 4);
+
+		if (tdb->methods->tdb_write(tdb, ofs, p+8, len) == -1) {
+			free(data);
+			TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_recover: failed to recover %d bytes at offset %d\n", len, ofs));
+			tdb->ecode = TDB_ERR_IO;
+			return -1;
+		}
+		p += 8 + len;
+	}
+
+	free(data);
+
+	if (transaction_sync(tdb, 0, tdb->map_size) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_recover: failed to sync recovery\n"));
+		tdb->ecode = TDB_ERR_IO;
+		return -1;
+	}
+
+	/* if the recovery area is after the recovered eof then remove it */
+	if (recovery_eof <= recovery_head) {
+		if (tdb_ofs_write(tdb, TDB_RECOVERY_HEAD, &zero) == -1) {
+			TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_recover: failed to remove recovery head\n"));
+			tdb->ecode = TDB_ERR_IO;
+			return -1;			
+		}
+	}
+
+	/* remove the recovery magic */
+	if (tdb_ofs_write(tdb, recovery_head + offsetof(struct tdb_record, magic),
+			  &zero) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_recover: failed to remove recovery magic\n"));
+		tdb->ecode = TDB_ERR_IO;
+		return -1;			
+	}
+
+	if (transaction_sync(tdb, 0, recovery_eof) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_recover: failed to sync2 recovery\n"));
+		tdb->ecode = TDB_ERR_IO;
+		return -1;
+	}
+
+	TDB_LOG((tdb, TDB_DEBUG_TRACE, "tdb_transaction_recover: recovered %d byte database\n", 
+		 recovery_eof));
+
+	/* all done */
+	return 0;
+}
+
+/* Any I/O failures we say "needs recovery". */
+bool tdb_needs_recovery(struct tdb_context *tdb)
+{
+	tdb_off_t recovery_head;
+	struct tdb_record rec;
+
+	/* find the recovery area */
+	if (tdb_ofs_read(tdb, TDB_RECOVERY_HEAD, &recovery_head) == -1) {
+		return true;
+	}
+
+	if (recovery_head == 0) {
+		/* we have never allocated a recovery record */
+		return false;
+	}
+
+	/* read the recovery record */
+	if (tdb->methods->tdb_read(tdb, recovery_head, &rec,
+				   sizeof(rec), DOCONV()) == -1) {
+		return true;
+	}
+
+	return (rec.magic == TDB_RECOVERY_MAGIC);
+}
diff --git a/ctdb/lib/tdb/common/traverse.c b/ctdb/lib/tdb/common/traverse.c
new file mode 100644
index 0000000000..517fecb4fc
--- /dev/null
+++ b/ctdb/lib/tdb/common/traverse.c
@@ -0,0 +1,366 @@
+ /* 
+   Unix SMB/CIFS implementation.
+
+   trivial database library
+
+   Copyright (C) Andrew Tridgell              1999-2005
+   Copyright (C) Paul `Rusty' Russell		   2000
+   Copyright (C) Jeremy Allison			   2000-2003
+
+     ** NOTE! The following LGPL license applies to the tdb
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "tdb_private.h"
+
+#define TDB_NEXT_LOCK_ERR ((tdb_off_t)-1)
+
+/* Uses traverse lock: 0 = finish, TDB_NEXT_LOCK_ERR = error,
+   other = record offset */
+static tdb_off_t tdb_next_lock(struct tdb_context *tdb, struct tdb_traverse_lock *tlock,
+			 struct tdb_record *rec)
+{
+	int want_next = (tlock->off != 0);
+
+	/* Lock each chain from the start one. */
+	for (; tlock->hash < tdb->header.hash_size; tlock->hash++) {
+		if (!tlock->off && tlock->hash != 0) {
+			/* this is an optimisation for the common case where
+			   the hash chain is empty, which is particularly
+			   common for the use of tdb with ldb, where large
+			   hashes are used. In that case we spend most of our
+			   time in tdb_brlock(), locking empty hash chains.
+
+			   To avoid this, we do an unlocked pre-check to see
+			   if the hash chain is empty before starting to look
+			   inside it. If it is empty then we can avoid that
+			   hash chain. If it isn't empty then we can't believe
+			   the value we get back, as we read it without a
+			   lock, so instead we get the lock and re-fetch the
+			   value below.
+
+			   Notice that not doing this optimisation on the
+			   first hash chain is critical. We must guarantee
+			   that we have done at least one fcntl lock at the
+			   start of a search to guarantee that memory is
+			   coherent on SMP systems. If records are added by
+			   others during the search then thats OK, and we
+			   could possibly miss those with this trick, but we
+			   could miss them anyway without this trick, so the
+			   semantics don't change.
+
+			   With a non-indexed ldb search this trick gains us a
+			   factor of around 80 in speed on a linux 2.6.x
+			   system (testing using ldbtest).
+			*/
+			tdb->methods->next_hash_chain(tdb, &tlock->hash);
+			if (tlock->hash == tdb->header.hash_size) {
+				continue;
+			}
+		}
+
+		if (tdb_lock(tdb, tlock->hash, tlock->lock_rw) == -1)
+			return TDB_NEXT_LOCK_ERR;
+
+		/* No previous record?  Start at top of chain. */
+		if (!tlock->off) {
+			if (tdb_ofs_read(tdb, TDB_HASH_TOP(tlock->hash),
+				     &tlock->off) == -1)
+				goto fail;
+		} else {
+			/* Otherwise unlock the previous record. */
+			if (tdb_unlock_record(tdb, tlock->off) != 0)
+				goto fail;
+		}
+
+		if (want_next) {
+			/* We have offset of old record: grab next */
+			if (tdb_rec_read(tdb, tlock->off, rec) == -1)
+				goto fail;
+			tlock->off = rec->next;
+		}
+
+		/* Iterate through chain */
+		while( tlock->off) {
+			tdb_off_t current;
+			if (tdb_rec_read(tdb, tlock->off, rec) == -1)
+				goto fail;
+
+			/* Detect infinite loops. From "Shlomi Yaakobovich" <Shlomi@exanet.com>. */
+			if (tlock->off == rec->next) {
+				tdb->ecode = TDB_ERR_CORRUPT;
+				TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_next_lock: loop detected.\n"));
+				goto fail;
+			}
+
+			if (!TDB_DEAD(rec)) {
+				/* Woohoo: we found one! */
+				if (tdb_lock_record(tdb, tlock->off) != 0)
+					goto fail;
+				return tlock->off;
+			}
+
+			/* Try to clean dead ones from old traverses */
+			current = tlock->off;
+			tlock->off = rec->next;
+			if (!(tdb->read_only || tdb->traverse_read) && 
+			    tdb_do_delete(tdb, current, rec) != 0)
+				goto fail;
+		}
+		tdb_unlock(tdb, tlock->hash, tlock->lock_rw);
+		want_next = 0;
+	}
+	/* We finished iteration without finding anything */
+	tdb->ecode = TDB_SUCCESS;
+	return 0;
+
+ fail:
+	tlock->off = 0;
+	if (tdb_unlock(tdb, tlock->hash, tlock->lock_rw) != 0)
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_next_lock: On error unlock failed!\n"));
+	return TDB_NEXT_LOCK_ERR;
+}
+
+/* traverse the entire database - calling fn(tdb, key, data) on each element.
+   return -1 on error or the record count traversed
+   if fn is NULL then it is not called
+   a non-zero return value from fn() indicates that the traversal should stop
+  */
+static int tdb_traverse_internal(struct tdb_context *tdb, 
+				 tdb_traverse_func fn, void *private_data,
+				 struct tdb_traverse_lock *tl)
+{
+	TDB_DATA key, dbuf;
+	struct tdb_record rec;
+	int ret = 0, count = 0;
+	tdb_off_t off;
+
+	/* This was in the initializaton, above, but the IRIX compiler
+	 * did not like it.  crh
+	 */
+	tl->next = tdb->travlocks.next;
+
+	/* fcntl locks don't stack: beware traverse inside traverse */
+	tdb->travlocks.next = tl;
+
+	/* tdb_next_lock places locks on the record returned, and its chain */
+	while ((off = tdb_next_lock(tdb, tl, &rec)) != 0) {
+		if (off == TDB_NEXT_LOCK_ERR) {
+			ret = -1;
+			goto out;
+		}
+		count++;
+		/* now read the full record */
+		key.dptr = tdb_alloc_read(tdb, tl->off + sizeof(rec), 
+					  rec.key_len + rec.data_len);
+		if (!key.dptr) {
+			ret = -1;
+			if (tdb_unlock(tdb, tl->hash, tl->lock_rw) != 0)
+				goto out;
+			if (tdb_unlock_record(tdb, tl->off) != 0)
+				TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_traverse: key.dptr == NULL and unlock_record failed!\n"));
+			goto out;
+		}
+		key.dsize = rec.key_len;
+		dbuf.dptr = key.dptr + rec.key_len;
+		dbuf.dsize = rec.data_len;
+
+		tdb_trace_1rec_retrec(tdb, "traverse", key, dbuf);
+
+		/* Drop chain lock, call out */
+		if (tdb_unlock(tdb, tl->hash, tl->lock_rw) != 0) {
+			ret = -1;
+			SAFE_FREE(key.dptr);
+			goto out;
+		}
+		if (fn && fn(tdb, key, dbuf, private_data)) {
+			/* They want us to terminate traversal */
+			tdb_trace_ret(tdb, "tdb_traverse_end", count);
+			if (tdb_unlock_record(tdb, tl->off) != 0) {
+				TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_traverse: unlock_record failed!\n"));;
+				ret = -1;
+			}
+			SAFE_FREE(key.dptr);
+			goto out;
+		}
+		SAFE_FREE(key.dptr);
+	}
+	tdb_trace(tdb, "tdb_traverse_end");
+out:
+	tdb->travlocks.next = tl->next;
+	if (ret < 0)
+		return -1;
+	else
+		return count;
+}
+
+
+/*
+  a write style traverse - temporarily marks the db read only
+*/
+_PUBLIC_ int tdb_traverse_read(struct tdb_context *tdb, 
+		      tdb_traverse_func fn, void *private_data)
+{
+	struct tdb_traverse_lock tl = { NULL, 0, 0, F_RDLCK };
+	int ret;
+
+	/* we need to get a read lock on the transaction lock here to
+	   cope with the lock ordering semantics of solaris10 */
+	if (tdb_transaction_lock(tdb, F_RDLCK, TDB_LOCK_WAIT)) {
+		return -1;
+	}
+
+	tdb->traverse_read++;
+	tdb_trace(tdb, "tdb_traverse_read_start");
+	ret = tdb_traverse_internal(tdb, fn, private_data, &tl);
+	tdb->traverse_read--;
+
+	tdb_transaction_unlock(tdb, F_RDLCK);
+
+	return ret;
+}
+
+/*
+  a write style traverse - needs to get the transaction lock to
+  prevent deadlocks
+
+  WARNING: The data buffer given to the callback fn does NOT meet the
+  alignment restrictions malloc gives you.
+*/
+_PUBLIC_ int tdb_traverse(struct tdb_context *tdb, 
+		 tdb_traverse_func fn, void *private_data)
+{
+	struct tdb_traverse_lock tl = { NULL, 0, 0, F_WRLCK };
+	int ret;
+
+	if (tdb->read_only || tdb->traverse_read) {
+		return tdb_traverse_read(tdb, fn, private_data);
+	}
+
+	if (tdb_transaction_lock(tdb, F_WRLCK, TDB_LOCK_WAIT)) {
+		return -1;
+	}
+
+	tdb->traverse_write++;
+	tdb_trace(tdb, "tdb_traverse_start");
+	ret = tdb_traverse_internal(tdb, fn, private_data, &tl);
+	tdb->traverse_write--;
+
+	tdb_transaction_unlock(tdb, F_WRLCK);
+
+	return ret;
+}
+
+
+/* find the first entry in the database and return its key */
+_PUBLIC_ TDB_DATA tdb_firstkey(struct tdb_context *tdb)
+{
+	TDB_DATA key;
+	struct tdb_record rec;
+	tdb_off_t off;
+
+	/* release any old lock */
+	if (tdb_unlock_record(tdb, tdb->travlocks.off) != 0)
+		return tdb_null;
+	tdb->travlocks.off = tdb->travlocks.hash = 0;
+	tdb->travlocks.lock_rw = F_RDLCK;
+
+	/* Grab first record: locks chain and returned record. */
+	off = tdb_next_lock(tdb, &tdb->travlocks, &rec);
+	if (off == 0 || off == TDB_NEXT_LOCK_ERR) {
+		tdb_trace_retrec(tdb, "tdb_firstkey", tdb_null);
+		return tdb_null;
+	}
+	/* now read the key */
+	key.dsize = rec.key_len;
+	key.dptr =tdb_alloc_read(tdb,tdb->travlocks.off+sizeof(rec),key.dsize);
+
+	tdb_trace_retrec(tdb, "tdb_firstkey", key);
+
+	/* Unlock the hash chain of the record we just read. */
+	if (tdb_unlock(tdb, tdb->travlocks.hash, tdb->travlocks.lock_rw) != 0)
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_firstkey: error occurred while tdb_unlocking!\n"));
+	return key;
+}
+
+/* find the next entry in the database, returning its key */
+_PUBLIC_ TDB_DATA tdb_nextkey(struct tdb_context *tdb, TDB_DATA oldkey)
+{
+	uint32_t oldhash;
+	TDB_DATA key = tdb_null;
+	struct tdb_record rec;
+	unsigned char *k = NULL;
+	tdb_off_t off;
+
+	/* Is locked key the old key?  If so, traverse will be reliable. */
+	if (tdb->travlocks.off) {
+		if (tdb_lock(tdb,tdb->travlocks.hash,tdb->travlocks.lock_rw))
+			return tdb_null;
+		if (tdb_rec_read(tdb, tdb->travlocks.off, &rec) == -1
+		    || !(k = tdb_alloc_read(tdb,tdb->travlocks.off+sizeof(rec),
+					    rec.key_len))
+		    || memcmp(k, oldkey.dptr, oldkey.dsize) != 0) {
+			/* No, it wasn't: unlock it and start from scratch */
+			if (tdb_unlock_record(tdb, tdb->travlocks.off) != 0) {
+				tdb_trace_1rec_retrec(tdb, "tdb_nextkey",
+						      oldkey, tdb_null);
+				SAFE_FREE(k);
+				return tdb_null;
+			}
+			if (tdb_unlock(tdb, tdb->travlocks.hash, tdb->travlocks.lock_rw) != 0) {
+				SAFE_FREE(k);
+				return tdb_null;
+			}
+			tdb->travlocks.off = 0;
+		}
+
+		SAFE_FREE(k);
+	}
+
+	if (!tdb->travlocks.off) {
+		/* No previous element: do normal find, and lock record */
+		tdb->travlocks.off = tdb_find_lock_hash(tdb, oldkey, tdb->hash_fn(&oldkey), tdb->travlocks.lock_rw, &rec);
+		if (!tdb->travlocks.off) {
+			tdb_trace_1rec_retrec(tdb, "tdb_nextkey", oldkey, tdb_null);
+			return tdb_null;
+		}
+		tdb->travlocks.hash = BUCKET(rec.full_hash);
+		if (tdb_lock_record(tdb, tdb->travlocks.off) != 0) {
+			TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_nextkey: lock_record failed (%s)!\n", strerror(errno)));
+			return tdb_null;
+		}
+	}
+	oldhash = tdb->travlocks.hash;
+
+	/* Grab next record: locks chain and returned record,
+	   unlocks old record */
+	off = tdb_next_lock(tdb, &tdb->travlocks, &rec);
+	if (off != TDB_NEXT_LOCK_ERR && off != 0) {
+		key.dsize = rec.key_len;
+		key.dptr = tdb_alloc_read(tdb, tdb->travlocks.off+sizeof(rec),
+					  key.dsize);
+		/* Unlock the chain of this new record */
+		if (tdb_unlock(tdb, tdb->travlocks.hash, tdb->travlocks.lock_rw) != 0)
+			TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_nextkey: WARNING tdb_unlock failed!\n"));
+	}
+	/* Unlock the chain of old record */
+	if (tdb_unlock(tdb, BUCKET(oldhash), tdb->travlocks.lock_rw) != 0)
+		TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_nextkey: WARNING tdb_unlock failed!\n"));
+	tdb_trace_1rec_retrec(tdb, "tdb_nextkey", oldkey, key);
+	return key;
+}
+
diff --git a/ctdb/lib/tdb/docs/README b/ctdb/lib/tdb/docs/README
new file mode 100644
index 0000000000..fe0e258183
--- /dev/null
+++ b/ctdb/lib/tdb/docs/README
@@ -0,0 +1,273 @@
+tdb - a trivial database system
+tridge@linuxcare.com December 1999
+==================================
+
+This is a simple database API. It was inspired by the realisation that
+in Samba we have several ad-hoc bits of code that essentially
+implement small databases for sharing structures between parts of
+Samba. As I was about to add another I realised that a generic
+database module was called for to replace all the ad-hoc bits.
+
+I based the interface on gdbm. I couldn't use gdbm as we need to be
+able to have multiple writers to the databases at one time.
+
+Compilation
+-----------
+
+add HAVE_MMAP=1 to use mmap instead of read/write
+add NOLOCK=1 to disable locking code
+
+Testing
+-------
+
+Compile tdbtest.c and link with gdbm for testing. tdbtest will perform
+identical operations via tdb and gdbm then make sure the result is the
+same
+
+Also included is tdbtool, which allows simple database manipulation
+on the commandline.
+
+tdbtest and tdbtool are not built as part of Samba, but are included
+for completeness.
+
+Interface
+---------
+
+The interface is very similar to gdbm except for the following:
+
+- different open interface. The tdb_open call is more similar to a
+  traditional open()
+- no tdbm_reorganise() function
+- no tdbm_sync() function. No operations are cached in the library anyway
+- added a tdb_traverse() function for traversing the whole database
+- added transactions support
+
+A general rule for using tdb is that the caller frees any returned
+TDB_DATA structures. Just call free(p.dptr) to free a TDB_DATA
+return value called p. This is the same as gdbm.
+
+here is a full list of tdb functions with brief descriptions.
+
+
+----------------------------------------------------------------------
+TDB_CONTEXT *tdb_open(char *name, int hash_size, int tdb_flags,
+		      int open_flags, mode_t mode)
+
+   open the database, creating it if necessary 
+
+   The open_flags and mode are passed straight to the open call on the database
+   file. A flags value of O_WRONLY is invalid
+
+   The hash size is advisory, use zero for a default value. 
+
+   return is NULL on error
+
+   possible tdb_flags are:
+    TDB_CLEAR_IF_FIRST - clear database if we are the only one with it open
+    TDB_INTERNAL - don't use a file, instaed store the data in
+                   memory. The filename is ignored in this case.
+    TDB_NOLOCK - don't do any locking
+    TDB_NOMMAP - don't use mmap
+    TDB_NOSYNC - don't synchronise transactions to disk
+    TDB_SEQNUM - maintain a sequence number
+    TDB_VOLATILE - activate the per-hashchain freelist, default 5
+    TDB_ALLOW_NESTING - allow transactions to nest
+    TDB_DISALLOW_NESTING - disallow transactions to nest
+
+----------------------------------------------------------------------
+TDB_CONTEXT *tdb_open_ex(char *name, int hash_size, int tdb_flags,
+		         int open_flags, mode_t mode,
+			 const struct tdb_logging_context *log_ctx,
+			 tdb_hash_func hash_fn)
+
+This is like tdb_open(), but allows you to pass an initial logging and
+hash function. Be careful when passing a hash function - all users of
+the database must use the same hash function or you will get data
+corruption.
+
+
+----------------------------------------------------------------------
+char *tdb_error(TDB_CONTEXT *tdb);
+
+     return a error string for the last tdb error
+
+----------------------------------------------------------------------
+int tdb_close(TDB_CONTEXT *tdb);
+
+   close a database
+
+----------------------------------------------------------------------
+TDB_DATA tdb_fetch(TDB_CONTEXT *tdb, TDB_DATA key);
+
+   fetch an entry in the database given a key 
+   if the return value has a null dptr then a error occurred
+
+   caller must free the resulting data
+
+----------------------------------------------------------------------
+int tdb_parse_record(struct tdb_context *tdb, TDB_DATA key,
+		     int (*parser)(TDB_DATA key, TDB_DATA data,
+				   void *private_data),
+		     void *private_data);
+
+   Hand a record to a parser function without allocating it.
+
+   This function is meant as a fast tdb_fetch alternative for large records
+   that are frequently read. The "key" and "data" arguments point directly
+   into the tdb shared memory, they are not aligned at any boundary.
+
+   WARNING: The parser is called while tdb holds a lock on the record. DO NOT
+   call other tdb routines from within the parser. Also, for good performance
+   you should make the parser fast to allow parallel operations.
+
+   tdb_parse_record returns -1 if the record was not found.  If the record was
+   found, the return value of "parser" is passed up to the caller.
+
+----------------------------------------------------------------------
+int tdb_exists(TDB_CONTEXT *tdb, TDB_DATA key);
+
+   check if an entry in the database exists 
+
+   note that 1 is returned if the key is found and 0 is returned if not found
+   this doesn't match the conventions in the rest of this module, but is
+   compatible with gdbm
+
+----------------------------------------------------------------------
+int tdb_traverse(TDB_CONTEXT *tdb, int (*fn)(TDB_CONTEXT *tdb,
+                 TDB_DATA key, TDB_DATA dbuf, void *state), void *state);
+
+   traverse the entire database - calling fn(tdb, key, data, state) on each 
+   element.
+
+   return -1 on error or the record count traversed
+
+   if fn is NULL then it is not called
+
+   a non-zero return value from fn() indicates that the traversal
+   should stop. Traversal callbacks may not start transactions.
+
+   WARNING: The data buffer given to the callback fn does NOT meet the
+   alignment restrictions malloc gives you.
+
+----------------------------------------------------------------------
+int tdb_traverse_read(TDB_CONTEXT *tdb, int (*fn)(TDB_CONTEXT *tdb,
+                     TDB_DATA key, TDB_DATA dbuf, void *state), void *state);
+
+   traverse the entire database - calling fn(tdb, key, data, state) on
+   each element, but marking the database read only during the
+   traversal, so any write operations will fail. This allows tdb to
+   use read locks, which increases the parallelism possible during the
+   traversal.
+
+   return -1 on error or the record count traversed
+
+   if fn is NULL then it is not called
+
+   a non-zero return value from fn() indicates that the traversal
+   should stop. Traversal callbacks may not start transactions.
+
+----------------------------------------------------------------------
+TDB_DATA tdb_firstkey(TDB_CONTEXT *tdb);
+
+   find the first entry in the database and return its key
+
+   the caller must free the returned data
+
+----------------------------------------------------------------------
+TDB_DATA tdb_nextkey(TDB_CONTEXT *tdb, TDB_DATA key);
+
+   find the next entry in the database, returning its key
+
+   the caller must free the returned data
+
+----------------------------------------------------------------------
+int tdb_delete(TDB_CONTEXT *tdb, TDB_DATA key);
+
+   delete an entry in the database given a key
+
+----------------------------------------------------------------------
+int tdb_store(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA dbuf, int flag);
+
+   store an element in the database, replacing any existing element
+   with the same key 
+
+   If flag==TDB_INSERT then don't overwrite an existing entry
+   If flag==TDB_MODIFY then don't create a new entry
+
+   return 0 on success, -1 on failure
+
+----------------------------------------------------------------------
+int tdb_writelock(TDB_CONTEXT *tdb);
+
+   lock the database. If we already have it locked then don't do anything
+
+----------------------------------------------------------------------
+int tdb_writeunlock(TDB_CONTEXT *tdb);
+   unlock the database
+
+----------------------------------------------------------------------
+int tdb_lockchain(TDB_CONTEXT *tdb, TDB_DATA key);
+
+   lock one hash chain. This is meant to be used to reduce locking
+   contention - it cannot guarantee how many records will be locked
+
+----------------------------------------------------------------------
+int tdb_unlockchain(TDB_CONTEXT *tdb, TDB_DATA key);
+
+   unlock one hash chain
+
+----------------------------------------------------------------------
+int tdb_transaction_start(TDB_CONTEXT *tdb)
+
+   start a transaction. All operations after the transaction start can
+   either be committed with tdb_transaction_commit() or cancelled with
+   tdb_transaction_cancel(). 
+
+   If you call tdb_transaction_start() again on the same tdb context
+   while a transaction is in progress, then the same transaction
+   buffer is re-used. The number of tdb_transaction_{commit,cancel}
+   operations must match the number of successful
+   tdb_transaction_start() calls.
+
+   Note that transactions are by default disk synchronous, and use a
+   recover area in the database to automatically recover the database
+   on the next open if the system crashes during a transaction. You
+   can disable the synchronous transaction recovery setup using the
+   TDB_NOSYNC flag, which will greatly speed up operations at the risk
+   of corrupting your database if the system crashes.
+
+   Operations made within a transaction are not visible to other users
+   of the database until a successful commit.
+
+----------------------------------------------------------------------
+int tdb_transaction_cancel(TDB_CONTEXT *tdb)
+
+   cancel a current transaction, discarding all write and lock
+   operations that have been made since the transaction started.
+
+
+----------------------------------------------------------------------
+int tdb_transaction_commit(TDB_CONTEXT *tdb)
+
+   commit a current transaction, updating the database and releasing
+   the transaction locks.
+
+----------------------------------------------------------------------
+int tdb_transaction_prepare_commit(TDB_CONTEXT *tdb)
+
+   prepare to commit a current transaction, for two-phase commits.
+   Once prepared for commit, the only allowed calls are
+   tdb_transaction_commit() or tdb_transaction_cancel(). Preparing
+   allocates disk space for the pending updates, so a subsequent
+   commit should succeed (barring any hardware failures).
+
+----------------------------------------------------------------------
+int tdb_check(TDB_CONTEXT *tdb,
+	      int (*check)(TDB_DATA key, TDB_DATA data, void *private_data),
+	      void *private_data);)
+
+   check the consistency of the database, calling back the check function
+   (if non-NULL) with each record.  If some consistency check fails, or
+   the supplied check function returns -1, tdb_check returns -1, otherwise
+   0.  Note that logging function (if set) will be called with additional
+   information on the corruption found.
diff --git a/ctdb/lib/tdb/docs/mainpage.dox b/ctdb/lib/tdb/docs/mainpage.dox
new file mode 100644
index 0000000000..d130769356
--- /dev/null
+++ b/ctdb/lib/tdb/docs/mainpage.dox
@@ -0,0 +1,61 @@
+/**
+
+@mainpage
+
+This is a simple database API. It was inspired by the realisation that in Samba
+we have several ad-hoc bits of code that essentially implement small databases
+for sharing structures between parts of Samba.
+
+The interface is based on gdbm. gdbm couldn't be use as we needed to be able to
+have multiple writers to the databases at one time.
+
+@section tdb_download Download
+
+You can download the latest releases of tdb from the
+<a href="http://samba.org/ftp/tdb">tdb directory</a> on the samba public source
+archive.
+
+You can download the latest code either via git or rsync.
+
+To fetch via git see the following guide:
+
+<a href="http://wiki.samba.org/index.php/Using_Git_for_Samba_Development">Using Git for Samba Development</a>
+Once you have cloned the tree switch to the master branch and cd into the source/lib/tdb directory.
+
+To fetch via rsync use these commands:
+
+<pre>
+  rsync -Pavz samba.org::ftp/unpacked/standalone_projects/lib/tdb .
+  rsync -Pavz samba.org::ftp/unpacked/standalone_projects/lib/replace .
+</pre>
+
+and build in tdb. It will find the replace library in the directory above
+automatically.
+
+@section tdb_bugs Discussion and bug reports
+
+tdb does not currently have its own mailing list or bug tracking system. For now,
+please use the
+<a href="https://lists.samba.org/mailman/listinfo/samba-technical">samba-technical</a>
+mailing list, and the <a href="http://bugzilla.samba.org/">Samba bugzilla</a> bug
+tracking system.
+
+
+@section tdb_compilation Compilation
+
+add HAVE_MMAP=1 to use mmap instead of read/write
+add NOLOCK=1 to disable locking code
+
+@section tdb_testing Testing
+
+Compile tdbtest.c and link with gdbm for testing. tdbtest will perform
+identical operations via tdb and gdbm then make sure the result is the
+same
+
+Also included is tdbtool, which allows simple database manipulation
+on the commandline.
+
+tdbtest and tdbtool are not built as part of Samba, but are included
+for completeness.
+
+*/
diff --git a/ctdb/lib/tdb/docs/tdb.magic b/ctdb/lib/tdb/docs/tdb.magic
new file mode 100644
index 0000000000..f5619e7327
--- /dev/null
+++ b/ctdb/lib/tdb/docs/tdb.magic
@@ -0,0 +1,10 @@
+# Magic file(1) information about tdb files.
+#
+# Install this into /etc/magic or the corresponding location for your
+# system, or pass as a -m argument to file(1).
+
+# You may use and redistribute this file without restriction.
+
+0	string	TDB\ file		TDB database
+>32	lelong	=0x2601196D		version 6, little-endian
+>>36	lelong	x			hash size %d bytes
diff --git a/ctdb/lib/tdb/docs/tracing.txt b/ctdb/lib/tdb/docs/tracing.txt
new file mode 100644
index 0000000000..98c5db9a51
--- /dev/null
+++ b/ctdb/lib/tdb/docs/tracing.txt
@@ -0,0 +1,46 @@
+How And Why To Use TDB Tracing
+==============================
+
+You can trace all TDB operations, using TDB_TRACE.  It is not complete
+(error conditions which expect to the logged will not always be traced
+correctly, so you should set up a logging function too), but is designed
+to collect benchmark-style traces to allow us to optimize TDB.
+
+Note: tracing is not efficient, and the trace files are huge: a
+traverse of the database is particularly large!  But they compress very
+well with rzip (http://rzip.samba.org)
+
+How to gather trace files:
+--------------------------
+1) Uncomment /* #define TDB_TRACE 1 */ in tdb_private.h.
+2) Rebuild TDB, and everything that uses it.
+3) Run something.
+
+Your trace files will be called <tdbname>.trace.<pid>.  These files
+will not be overwritten: if the same process reopens the same TDB, an
+error will be logged and tracing will be disabled.
+
+How to replay trace files:
+--------------------------
+1) For benchmarking, remember to rebuild tdb with #define TDB_TRACE commented
+   out again!
+2) Grab the latest "replace_trace.c" from CCAN's tdb module (tools/ dir):
+	http://ccan.ozlabs.org/tarballs/tdb.tar.bz2
+3) Compile up replay_trace, munging as necessary.
+4) Run replay_trace <scratch-tdb-name> <tracefiles>...
+
+If given more than one trace file (presumably from the same tdb)
+replay_trace will try to figure out the dependencies between the operations
+and fire off a child to run each trace.  Occasionally it gets stuck, in
+which case it will add another dependency and retry.  Eventually it will
+give a speed value.
+
+replay_trace can intuit the existence of previous data in the tdb (ie.
+activity prior to the trace(s) supplied) and will prepopulate as
+neccessary.
+
+You can run --quiet for straight benchmark results, and -n to run multiple
+times (this saves time, since it need only calculate dependencies once).
+
+Good luck!
+Rusty Russell <rusty@rustcorp.com.au>
diff --git a/ctdb/lib/tdb/doxy.config b/ctdb/lib/tdb/doxy.config
new file mode 100644
index 0000000000..f55e9c3c27
--- /dev/null
+++ b/ctdb/lib/tdb/doxy.config
@@ -0,0 +1,1697 @@
+# Doxyfile 1.7.3
+
+# This file describes the settings to be used by the documentation system
+# doxygen (www.doxygen.org) for a project.
+#
+# All text after a hash (#) is considered a comment and will be ignored.
+# The format is:
+#       TAG = value [value, ...]
+# For lists items can also be appended using:
+#       TAG += value [value, ...]
+# Values that contain spaces should be placed between quotes (" ").
+
+#---------------------------------------------------------------------------
+# Project related configuration options
+#---------------------------------------------------------------------------
+
+# This tag specifies the encoding used for all characters in the config file
+# that follow. The default is UTF-8 which is also the encoding used for all
+# text before the first occurrence of this tag. Doxygen uses libiconv (or the
+# iconv built into libc) for the transcoding. See
+# http://www.gnu.org/software/libiconv for the list of possible encodings.
+
+DOXYFILE_ENCODING      = UTF-8
+
+# The PROJECT_NAME tag is a single word (or a sequence of words surrounded
+# by quotes) that should identify the project.
+
+PROJECT_NAME           = tdb
+
+# The PROJECT_NUMBER tag can be used to enter a project or revision number.
+# This could be handy for archiving the generated documentation or
+# if some version control system is used.
+
+PROJECT_NUMBER         = 1.2.9
+
+# Using the PROJECT_BRIEF tag one can provide an optional one line description for a project that appears at the top of each page and should give viewer a quick idea about the purpose of the project. Keep the description short.
+
+PROJECT_BRIEF          =
+
+# With the PROJECT_LOGO tag one can specify an logo or icon that is
+# included in the documentation. The maximum height of the logo should not
+# exceed 55 pixels and the maximum width should not exceed 200 pixels.
+# Doxygen will copy the logo to the output directory.
+
+PROJECT_LOGO           =
+
+# The OUTPUT_DIRECTORY tag is used to specify the (relative or absolute)
+# base path where the generated documentation will be put.
+# If a relative path is entered, it will be relative to the location
+# where doxygen was started. If left blank the current directory will be used.
+
+OUTPUT_DIRECTORY       = docs
+
+# If the CREATE_SUBDIRS tag is set to YES, then doxygen will create
+# 4096 sub-directories (in 2 levels) under the output directory of each output
+# format and will distribute the generated files over these directories.
+# Enabling this option can be useful when feeding doxygen a huge amount of
+# source files, where putting all generated files in the same directory would
+# otherwise cause performance problems for the file system.
+
+CREATE_SUBDIRS         = NO
+
+# The OUTPUT_LANGUAGE tag is used to specify the language in which all
+# documentation generated by doxygen is written. Doxygen will use this
+# information to generate all constant output in the proper language.
+# The default language is English, other supported languages are:
+# Afrikaans, Arabic, Brazilian, Catalan, Chinese, Chinese-Traditional,
+# Croatian, Czech, Danish, Dutch, Esperanto, Farsi, Finnish, French, German,
+# Greek, Hungarian, Italian, Japanese, Japanese-en (Japanese with English
+# messages), Korean, Korean-en, Lithuanian, Norwegian, Macedonian, Persian,
+# Polish, Portuguese, Romanian, Russian, Serbian, Serbian-Cyrillic, Slovak,
+# Slovene, Spanish, Swedish, Ukrainian, and Vietnamese.
+
+OUTPUT_LANGUAGE        = English
+
+# If the BRIEF_MEMBER_DESC tag is set to YES (the default) Doxygen will
+# include brief member descriptions after the members that are listed in
+# the file and class documentation (similar to JavaDoc).
+# Set to NO to disable this.
+
+BRIEF_MEMBER_DESC      = YES
+
+# If the REPEAT_BRIEF tag is set to YES (the default) Doxygen will prepend
+# the brief description of a member or function before the detailed description.
+# Note: if both HIDE_UNDOC_MEMBERS and BRIEF_MEMBER_DESC are set to NO, the
+# brief descriptions will be completely suppressed.
+
+REPEAT_BRIEF           = YES
+
+# This tag implements a quasi-intelligent brief description abbreviator
+# that is used to form the text in various listings. Each string
+# in this list, if found as the leading text of the brief description, will be
+# stripped from the text and the result after processing the whole list, is
+# used as the annotated text. Otherwise, the brief description is used as-is.
+# If left blank, the following values are used ("$name" is automatically
+# replaced with the name of the entity): "The $name class" "The $name widget"
+# "The $name file" "is" "provides" "specifies" "contains"
+# "represents" "a" "an" "the"
+
+ABBREVIATE_BRIEF       = "The $name class" \
+                         "The $name widget" \
+                         "The $name file" \
+                         is \
+                         provides \
+                         specifies \
+                         contains \
+                         represents \
+                         a \
+                         an \
+                         the
+
+# If the ALWAYS_DETAILED_SEC and REPEAT_BRIEF tags are both set to YES then
+# Doxygen will generate a detailed section even if there is only a brief
+# description.
+
+ALWAYS_DETAILED_SEC    = NO
+
+# If the INLINE_INHERITED_MEMB tag is set to YES, doxygen will show all
+# inherited members of a class in the documentation of that class as if those
+# members were ordinary class members. Constructors, destructors and assignment
+# operators of the base classes will not be shown.
+
+INLINE_INHERITED_MEMB  = NO
+
+# If the FULL_PATH_NAMES tag is set to YES then Doxygen will prepend the full
+# path before files name in the file list and in the header files. If set
+# to NO the shortest path that makes the file name unique will be used.
+
+FULL_PATH_NAMES        = YES
+
+# If the FULL_PATH_NAMES tag is set to YES then the STRIP_FROM_PATH tag
+# can be used to strip a user-defined part of the path. Stripping is
+# only done if one of the specified strings matches the left-hand part of
+# the path. The tag can be used to show relative paths in the file list.
+# If left blank the directory from which doxygen is run is used as the
+# path to strip.
+
+STRIP_FROM_PATH        =
+
+# The STRIP_FROM_INC_PATH tag can be used to strip a user-defined part of
+# the path mentioned in the documentation of a class, which tells
+# the reader which header file to include in order to use a class.
+# If left blank only the name of the header file containing the class
+# definition is used. Otherwise one should specify the include paths that
+# are normally passed to the compiler using the -I flag.
+
+STRIP_FROM_INC_PATH    =
+
+# If the SHORT_NAMES tag is set to YES, doxygen will generate much shorter
+# (but less readable) file names. This can be useful if your file system
+# doesn't support long names like on DOS, Mac, or CD-ROM.
+
+SHORT_NAMES            = NO
+
+# If the JAVADOC_AUTOBRIEF tag is set to YES then Doxygen
+# will interpret the first line (until the first dot) of a JavaDoc-style
+# comment as the brief description. If set to NO, the JavaDoc
+# comments will behave just like regular Qt-style comments
+# (thus requiring an explicit @brief command for a brief description.)
+
+JAVADOC_AUTOBRIEF      = YES
+
+# If the QT_AUTOBRIEF tag is set to YES then Doxygen will
+# interpret the first line (until the first dot) of a Qt-style
+# comment as the brief description. If set to NO, the comments
+# will behave just like regular Qt-style comments (thus requiring
+# an explicit \brief command for a brief description.)
+
+QT_AUTOBRIEF           = NO
+
+# The MULTILINE_CPP_IS_BRIEF tag can be set to YES to make Doxygen
+# treat a multi-line C++ special comment block (i.e. a block of //! or ///
+# comments) as a brief description. This used to be the default behaviour.
+# The new default is to treat a multi-line C++ comment block as a detailed
+# description. Set this tag to YES if you prefer the old behaviour instead.
+
+MULTILINE_CPP_IS_BRIEF = NO
+
+# If the INHERIT_DOCS tag is set to YES (the default) then an undocumented
+# member inherits the documentation from any documented member that it
+# re-implements.
+
+INHERIT_DOCS           = YES
+
+# If the SEPARATE_MEMBER_PAGES tag is set to YES, then doxygen will produce
+# a new page for each member. If set to NO, the documentation of a member will
+# be part of the file/class/namespace that contains it.
+
+SEPARATE_MEMBER_PAGES  = NO
+
+# The TAB_SIZE tag can be used to set the number of spaces in a tab.
+# Doxygen uses this value to replace tabs by spaces in code fragments.
+
+TAB_SIZE               = 8
+
+# This tag can be used to specify a number of aliases that acts
+# as commands in the documentation. An alias has the form "name=value".
+# For example adding "sideeffect=\par Side Effects:\n" will allow you to
+# put the command \sideeffect (or @sideeffect) in the documentation, which
+# will result in a user-defined paragraph with heading "Side Effects:".
+# You can put \n's in the value part of an alias to insert newlines.
+
+ALIASES                =
+
+# Set the OPTIMIZE_OUTPUT_FOR_C tag to YES if your project consists of C
+# sources only. Doxygen will then generate output that is more tailored for C.
+# For instance, some of the names that are used will be different. The list
+# of all members will be omitted, etc.
+
+OPTIMIZE_OUTPUT_FOR_C  = YES
+
+# Set the OPTIMIZE_OUTPUT_JAVA tag to YES if your project consists of Java
+# sources only. Doxygen will then generate output that is more tailored for
+# Java. For instance, namespaces will be presented as packages, qualified
+# scopes will look different, etc.
+
+OPTIMIZE_OUTPUT_JAVA   = NO
+
+# Set the OPTIMIZE_FOR_FORTRAN tag to YES if your project consists of Fortran
+# sources only. Doxygen will then generate output that is more tailored for
+# Fortran.
+
+OPTIMIZE_FOR_FORTRAN   = NO
+
+# Set the OPTIMIZE_OUTPUT_VHDL tag to YES if your project consists of VHDL
+# sources. Doxygen will then generate output that is tailored for
+# VHDL.
+
+OPTIMIZE_OUTPUT_VHDL   = NO
+
+# Doxygen selects the parser to use depending on the extension of the files it
+# parses. With this tag you can assign which parser to use for a given extension.
+# Doxygen has a built-in mapping, but you can override or extend it using this
+# tag. The format is ext=language, where ext is a file extension, and language
+# is one of the parsers supported by doxygen: IDL, Java, Javascript, CSharp, C,
+# C++, D, PHP, Objective-C, Python, Fortran, VHDL, C, C++. For instance to make
+# doxygen treat .inc files as Fortran files (default is PHP), and .f files as C
+# (default is Fortran), use: inc=Fortran f=C. Note that for custom extensions
+# you also need to set FILE_PATTERNS otherwise the files are not read by doxygen.
+
+EXTENSION_MAPPING      =
+
+# If you use STL classes (i.e. std::string, std::vector, etc.) but do not want
+# to include (a tag file for) the STL sources as input, then you should
+# set this tag to YES in order to let doxygen match functions declarations and
+# definitions whose arguments contain STL classes (e.g. func(std::string); v.s.
+# func(std::string) {}). This also makes the inheritance and collaboration
+# diagrams that involve STL classes more complete and accurate.
+
+BUILTIN_STL_SUPPORT    = NO
+
+# If you use Microsoft's C++/CLI language, you should set this option to YES to
+# enable parsing support.
+
+CPP_CLI_SUPPORT        = NO
+
+# Set the SIP_SUPPORT tag to YES if your project consists of sip sources only.
+# Doxygen will parse them like normal C++ but will assume all classes use public
+# instead of private inheritance when no explicit protection keyword is present.
+
+SIP_SUPPORT            = NO
+
+# For Microsoft's IDL there are propget and propput attributes to indicate getter
+# and setter methods for a property. Setting this option to YES (the default)
+# will make doxygen replace the get and set methods by a property in the
+# documentation. This will only work if the methods are indeed getting or
+# setting a simple type. If this is not the case, or you want to show the
+# methods anyway, you should set this option to NO.
+
+IDL_PROPERTY_SUPPORT   = YES
+
+# If member grouping is used in the documentation and the DISTRIBUTE_GROUP_DOC
+# tag is set to YES, then doxygen will reuse the documentation of the first
+# member in the group (if any) for the other members of the group. By default
+# all members of a group must be documented explicitly.
+
+DISTRIBUTE_GROUP_DOC   = NO
+
+# Set the SUBGROUPING tag to YES (the default) to allow class member groups of
+# the same type (for instance a group of public functions) to be put as a
+# subgroup of that type (e.g. under the Public Functions section). Set it to
+# NO to prevent subgrouping. Alternatively, this can be done per class using
+# the \nosubgrouping command.
+
+SUBGROUPING            = YES
+
+# When TYPEDEF_HIDES_STRUCT is enabled, a typedef of a struct, union, or enum
+# is documented as struct, union, or enum with the name of the typedef. So
+# typedef struct TypeS {} TypeT, will appear in the documentation as a struct
+# with name TypeT. When disabled the typedef will appear as a member of a file,
+# namespace, or class. And the struct will be named TypeS. This can typically
+# be useful for C code in case the coding convention dictates that all compound
+# types are typedef'ed and only the typedef is referenced, never the tag name.
+
+TYPEDEF_HIDES_STRUCT   = NO
+
+# The SYMBOL_CACHE_SIZE determines the size of the internal cache use to
+# determine which symbols to keep in memory and which to flush to disk.
+# When the cache is full, less often used symbols will be written to disk.
+# For small to medium size projects (<1000 input files) the default value is
+# probably good enough. For larger projects a too small cache size can cause
+# doxygen to be busy swapping symbols to and from disk most of the time
+# causing a significant performance penalty.
+# If the system has enough physical memory increasing the cache will improve the
+# performance by keeping more symbols in memory. Note that the value works on
+# a logarithmic scale so increasing the size by one will roughly double the
+# memory usage. The cache size is given by this formula:
+# 2^(16+SYMBOL_CACHE_SIZE). The valid range is 0..9, the default is 0,
+# corresponding to a cache size of 2^16 = 65536 symbols
+
+SYMBOL_CACHE_SIZE      = 0
+
+#---------------------------------------------------------------------------
+# Build related configuration options
+#---------------------------------------------------------------------------
+
+# If the EXTRACT_ALL tag is set to YES doxygen will assume all entities in
+# documentation are documented, even if no documentation was available.
+# Private class members and static file members will be hidden unless
+# the EXTRACT_PRIVATE and EXTRACT_STATIC tags are set to YES
+
+EXTRACT_ALL            = NO
+
+# If the EXTRACT_PRIVATE tag is set to YES all private members of a class
+# will be included in the documentation.
+
+EXTRACT_PRIVATE        = NO
+
+# If the EXTRACT_STATIC tag is set to YES all static members of a file
+# will be included in the documentation.
+
+EXTRACT_STATIC         = NO
+
+# If the EXTRACT_LOCAL_CLASSES tag is set to YES classes (and structs)
+# defined locally in source files will be included in the documentation.
+# If set to NO only classes defined in header files are included.
+
+EXTRACT_LOCAL_CLASSES  = NO
+
+# This flag is only useful for Objective-C code. When set to YES local
+# methods, which are defined in the implementation section but not in
+# the interface are included in the documentation.
+# If set to NO (the default) only methods in the interface are included.
+
+EXTRACT_LOCAL_METHODS  = NO
+
+# If this flag is set to YES, the members of anonymous namespaces will be
+# extracted and appear in the documentation as a namespace called
+# 'anonymous_namespace{file}', where file will be replaced with the base
+# name of the file that contains the anonymous namespace. By default
+# anonymous namespaces are hidden.
+
+EXTRACT_ANON_NSPACES   = NO
+
+# If the HIDE_UNDOC_MEMBERS tag is set to YES, Doxygen will hide all
+# undocumented members of documented classes, files or namespaces.
+# If set to NO (the default) these members will be included in the
+# various overviews, but no documentation section is generated.
+# This option has no effect if EXTRACT_ALL is enabled.
+
+HIDE_UNDOC_MEMBERS     = YES
+
+# If the HIDE_UNDOC_CLASSES tag is set to YES, Doxygen will hide all
+# undocumented classes that are normally visible in the class hierarchy.
+# If set to NO (the default) these classes will be included in the various
+# overviews. This option has no effect if EXTRACT_ALL is enabled.
+
+HIDE_UNDOC_CLASSES     = YES
+
+# If the HIDE_FRIEND_COMPOUNDS tag is set to YES, Doxygen will hide all
+# friend (class|struct|union) declarations.
+# If set to NO (the default) these declarations will be included in the
+# documentation.
+
+HIDE_FRIEND_COMPOUNDS  = NO
+
+# If the HIDE_IN_BODY_DOCS tag is set to YES, Doxygen will hide any
+# documentation blocks found inside the body of a function.
+# If set to NO (the default) these blocks will be appended to the
+# function's detailed documentation block.
+
+HIDE_IN_BODY_DOCS      = NO
+
+# The INTERNAL_DOCS tag determines if documentation
+# that is typed after a \internal command is included. If the tag is set
+# to NO (the default) then the documentation will be excluded.
+# Set it to YES to include the internal documentation.
+
+INTERNAL_DOCS          = NO
+
+# If the CASE_SENSE_NAMES tag is set to NO then Doxygen will only generate
+# file names in lower-case letters. If set to YES upper-case letters are also
+# allowed. This is useful if you have classes or files whose names only differ
+# in case and if your file system supports case sensitive file names. Windows
+# and Mac users are advised to set this option to NO.
+
+CASE_SENSE_NAMES       = YES
+
+# If the HIDE_SCOPE_NAMES tag is set to NO (the default) then Doxygen
+# will show members with their full class and namespace scopes in the
+# documentation. If set to YES the scope will be hidden.
+
+HIDE_SCOPE_NAMES       = NO
+
+# If the SHOW_INCLUDE_FILES tag is set to YES (the default) then Doxygen
+# will put a list of the files that are included by a file in the documentation
+# of that file.
+
+SHOW_INCLUDE_FILES     = YES
+
+# If the FORCE_LOCAL_INCLUDES tag is set to YES then Doxygen
+# will list include files with double quotes in the documentation
+# rather than with sharp brackets.
+
+FORCE_LOCAL_INCLUDES   = NO
+
+# If the INLINE_INFO tag is set to YES (the default) then a tag [inline]
+# is inserted in the documentation for inline members.
+
+INLINE_INFO            = YES
+
+# If the SORT_MEMBER_DOCS tag is set to YES (the default) then doxygen
+# will sort the (detailed) documentation of file and class members
+# alphabetically by member name. If set to NO the members will appear in
+# declaration order.
+
+SORT_MEMBER_DOCS       = YES
+
+# If the SORT_BRIEF_DOCS tag is set to YES then doxygen will sort the
+# brief documentation of file, namespace and class members alphabetically
+# by member name. If set to NO (the default) the members will appear in
+# declaration order.
+
+SORT_BRIEF_DOCS        = NO
+
+# If the SORT_MEMBERS_CTORS_1ST tag is set to YES then doxygen
+# will sort the (brief and detailed) documentation of class members so that
+# constructors and destructors are listed first. If set to NO (the default)
+# the constructors will appear in the respective orders defined by
+# SORT_MEMBER_DOCS and SORT_BRIEF_DOCS.
+# This tag will be ignored for brief docs if SORT_BRIEF_DOCS is set to NO
+# and ignored for detailed docs if SORT_MEMBER_DOCS is set to NO.
+
+SORT_MEMBERS_CTORS_1ST = NO
+
+# If the SORT_GROUP_NAMES tag is set to YES then doxygen will sort the
+# hierarchy of group names into alphabetical order. If set to NO (the default)
+# the group names will appear in their defined order.
+
+SORT_GROUP_NAMES       = NO
+
+# If the SORT_BY_SCOPE_NAME tag is set to YES, the class list will be
+# sorted by fully-qualified names, including namespaces. If set to
+# NO (the default), the class list will be sorted only by class name,
+# not including the namespace part.
+# Note: This option is not very useful if HIDE_SCOPE_NAMES is set to YES.
+# Note: This option applies only to the class list, not to the
+# alphabetical list.
+
+SORT_BY_SCOPE_NAME     = NO
+
+# If the STRICT_PROTO_MATCHING option is enabled and doxygen fails to do proper type resolution of all parameters of a function it will reject a
+# match between the prototype and the implementation of a member function even if there is only one candidate or it is obvious which candidate to choose by doing a simple string match. By disabling STRICT_PROTO_MATCHING doxygen
+# will still accept a match between prototype and implementation in such cases.
+
+STRICT_PROTO_MATCHING  = NO
+
+# The GENERATE_TODOLIST tag can be used to enable (YES) or
+# disable (NO) the todo list. This list is created by putting \todo
+# commands in the documentation.
+
+GENERATE_TODOLIST      = YES
+
+# The GENERATE_TESTLIST tag can be used to enable (YES) or
+# disable (NO) the test list. This list is created by putting \test
+# commands in the documentation.
+
+GENERATE_TESTLIST      = YES
+
+# The GENERATE_BUGLIST tag can be used to enable (YES) or
+# disable (NO) the bug list. This list is created by putting \bug
+# commands in the documentation.
+
+GENERATE_BUGLIST       = YES
+
+# The GENERATE_DEPRECATEDLIST tag can be used to enable (YES) or
+# disable (NO) the deprecated list. This list is created by putting
+# \deprecated commands in the documentation.
+
+GENERATE_DEPRECATEDLIST= YES
+
+# The ENABLED_SECTIONS tag can be used to enable conditional
+# documentation sections, marked by \if sectionname ... \endif.
+
+ENABLED_SECTIONS       =
+
+# The MAX_INITIALIZER_LINES tag determines the maximum number of lines
+# the initial value of a variable or macro consists of for it to appear in
+# the documentation. If the initializer consists of more lines than specified
+# here it will be hidden. Use a value of 0 to hide initializers completely.
+# The appearance of the initializer of individual variables and macros in the
+# documentation can be controlled using \showinitializer or \hideinitializer
+# command in the documentation regardless of this setting.
+
+MAX_INITIALIZER_LINES  = 30
+
+# Set the SHOW_USED_FILES tag to NO to disable the list of files generated
+# at the bottom of the documentation of classes and structs. If set to YES the
+# list will mention the files that were used to generate the documentation.
+
+SHOW_USED_FILES        = YES
+
+# If the sources in your project are distributed over multiple directories
+# then setting the SHOW_DIRECTORIES tag to YES will show the directory hierarchy
+# in the documentation. The default is NO.
+
+SHOW_DIRECTORIES       = NO
+
+# Set the SHOW_FILES tag to NO to disable the generation of the Files page.
+# This will remove the Files entry from the Quick Index and from the
+# Folder Tree View (if specified). The default is YES.
+
+SHOW_FILES             = YES
+
+# Set the SHOW_NAMESPACES tag to NO to disable the generation of the
+# Namespaces page.
+# This will remove the Namespaces entry from the Quick Index
+# and from the Folder Tree View (if specified). The default is YES.
+
+SHOW_NAMESPACES        = YES
+
+# The FILE_VERSION_FILTER tag can be used to specify a program or script that
+# doxygen should invoke to get the current version for each file (typically from
+# the version control system). Doxygen will invoke the program by executing (via
+# popen()) the command <command> <input-file>, where <command> is the value of
+# the FILE_VERSION_FILTER tag, and <input-file> is the name of an input file
+# provided by doxygen. Whatever the program writes to standard output
+# is used as the file version. See the manual for examples.
+
+FILE_VERSION_FILTER    =
+
+# The LAYOUT_FILE tag can be used to specify a layout file which will be parsed
+# by doxygen. The layout file controls the global structure of the generated
+# output files in an output format independent way. The create the layout file
+# that represents doxygen's defaults, run doxygen with the -l option.
+# You can optionally specify a file name after the option, if omitted
+# DoxygenLayout.xml will be used as the name of the layout file.
+
+LAYOUT_FILE            =
+
+#---------------------------------------------------------------------------
+# configuration options related to warning and progress messages
+#---------------------------------------------------------------------------
+
+# The QUIET tag can be used to turn on/off the messages that are generated
+# by doxygen. Possible values are YES and NO. If left blank NO is used.
+
+QUIET                  = NO
+
+# The WARNINGS tag can be used to turn on/off the warning messages that are
+# generated by doxygen. Possible values are YES and NO. If left blank
+# NO is used.
+
+WARNINGS               = YES
+
+# If WARN_IF_UNDOCUMENTED is set to YES, then doxygen will generate warnings
+# for undocumented members. If EXTRACT_ALL is set to YES then this flag will
+# automatically be disabled.
+
+WARN_IF_UNDOCUMENTED   = YES
+
+# If WARN_IF_DOC_ERROR is set to YES, doxygen will generate warnings for
+# potential errors in the documentation, such as not documenting some
+# parameters in a documented function, or documenting parameters that
+# don't exist or using markup commands wrongly.
+
+WARN_IF_DOC_ERROR      = YES
+
+# The WARN_NO_PARAMDOC option can be enabled to get warnings for
+# functions that are documented, but have no documentation for their parameters
+# or return value. If set to NO (the default) doxygen will only warn about
+# wrong or incomplete parameter documentation, but not about the absence of
+# documentation.
+
+WARN_NO_PARAMDOC       = NO
+
+# The WARN_FORMAT tag determines the format of the warning messages that
+# doxygen can produce. The string should contain the $file, $line, and $text
+# tags, which will be replaced by the file and line number from which the
+# warning originated and the warning text. Optionally the format may contain
+# $version, which will be replaced by the version of the file (if it could
+# be obtained via FILE_VERSION_FILTER)
+
+WARN_FORMAT            = "$file:$line: $text"
+
+# The WARN_LOGFILE tag can be used to specify a file to which warning
+# and error messages should be written. If left blank the output is written
+# to stderr.
+
+WARN_LOGFILE           =
+
+#---------------------------------------------------------------------------
+# configuration options related to the input files
+#---------------------------------------------------------------------------
+
+# The INPUT tag can be used to specify the files and/or directories that contain
+# documented source files. You may enter file names like "myfile.cpp" or
+# directories like "/usr/src/myproject". Separate the files or directories
+# with spaces.
+
+INPUT                  = include \
+                         docs
+
+# This tag can be used to specify the character encoding of the source files
+# that doxygen parses. Internally doxygen uses the UTF-8 encoding, which is
+# also the default input encoding. Doxygen uses libiconv (or the iconv built
+# into libc) for the transcoding. See http://www.gnu.org/software/libiconv for
+# the list of possible encodings.
+
+INPUT_ENCODING         = UTF-8
+
+# If the value of the INPUT tag contains directories, you can use the
+# FILE_PATTERNS tag to specify one or more wildcard pattern (like *.cpp
+# and *.h) to filter out the source-files in the directories. If left
+# blank the following patterns are tested:
+# *.c *.cc *.cxx *.cpp *.c++ *.d *.java *.ii *.ixx *.ipp *.i++ *.inl *.h *.hh
+# *.hxx *.hpp *.h++ *.idl *.odl *.cs *.php *.php3 *.inc *.m *.mm *.dox *.py
+# *.f90 *.f *.for *.vhd *.vhdl
+
+FILE_PATTERNS          = *.cpp \
+                         *.cc \
+                         *.c \
+                         *.h \
+                         *.hh \
+                         *.hpp \
+                         *.dox
+
+# The RECURSIVE tag can be used to turn specify whether or not subdirectories
+# should be searched for input files as well. Possible values are YES and NO.
+# If left blank NO is used.
+
+RECURSIVE              = NO
+
+# The EXCLUDE tag can be used to specify files and/or directories that should
+# excluded from the INPUT source files. This way you can easily exclude a
+# subdirectory from a directory tree whose root is specified with the INPUT tag.
+
+EXCLUDE                =
+
+# The EXCLUDE_SYMLINKS tag can be used select whether or not files or
+# directories that are symbolic links (a Unix file system feature) are excluded
+# from the input.
+
+EXCLUDE_SYMLINKS       = NO
+
+# If the value of the INPUT tag contains directories, you can use the
+# EXCLUDE_PATTERNS tag to specify one or more wildcard patterns to exclude
+# certain files from those directories. Note that the wildcards are matched
+# against the file with absolute path, so to exclude all test directories
+# for example use the pattern */test/*
+
+EXCLUDE_PATTERNS       = */.git/*
+
+# The EXCLUDE_SYMBOLS tag can be used to specify one or more symbol names
+# (namespaces, classes, functions, etc.) that should be excluded from the
+# output. The symbol name can be a fully qualified name, a word, or if the
+# wildcard * is used, a substring. Examples: ANamespace, AClass,
+# AClass::ANamespace, ANamespace::*Test
+
+EXCLUDE_SYMBOLS        =
+
+# The EXAMPLE_PATH tag can be used to specify one or more files or
+# directories that contain example code fragments that are included (see
+# the \include command).
+
+EXAMPLE_PATH           =
+
+# If the value of the EXAMPLE_PATH tag contains directories, you can use the
+# EXAMPLE_PATTERNS tag to specify one or more wildcard pattern (like *.cpp
+# and *.h) to filter out the source-files in the directories. If left
+# blank all files are included.
+
+EXAMPLE_PATTERNS       =
+
+# If the EXAMPLE_RECURSIVE tag is set to YES then subdirectories will be
+# searched for input files to be used with the \include or \dontinclude
+# commands irrespective of the value of the RECURSIVE tag.
+# Possible values are YES and NO. If left blank NO is used.
+
+EXAMPLE_RECURSIVE      = NO
+
+# The IMAGE_PATH tag can be used to specify one or more files or
+# directories that contain image that are included in the documentation (see
+# the \image command).
+
+IMAGE_PATH             =
+
+# The INPUT_FILTER tag can be used to specify a program that doxygen should
+# invoke to filter for each input file. Doxygen will invoke the filter program
+# by executing (via popen()) the command <filter> <input-file>, where <filter>
+# is the value of the INPUT_FILTER tag, and <input-file> is the name of an
+# input file. Doxygen will then use the output that the filter program writes
+# to standard output.
+# If FILTER_PATTERNS is specified, this tag will be
+# ignored.
+
+INPUT_FILTER           =
+
+# The FILTER_PATTERNS tag can be used to specify filters on a per file pattern
+# basis.
+# Doxygen will compare the file name with each pattern and apply the
+# filter if there is a match.
+# The filters are a list of the form:
+# pattern=filter (like *.cpp=my_cpp_filter). See INPUT_FILTER for further
+# info on how filters are used. If FILTER_PATTERNS is empty or if
+# non of the patterns match the file name, INPUT_FILTER is applied.
+
+FILTER_PATTERNS        =
+
+# If the FILTER_SOURCE_FILES tag is set to YES, the input filter (if set using
+# INPUT_FILTER) will be used to filter the input files when producing source
+# files to browse (i.e. when SOURCE_BROWSER is set to YES).
+
+FILTER_SOURCE_FILES    = NO
+
+# The FILTER_SOURCE_PATTERNS tag can be used to specify source filters per file
+# pattern. A pattern will override the setting for FILTER_PATTERN (if any)
+# and it is also possible to disable source filtering for a specific pattern
+# using *.ext= (so without naming a filter). This option only has effect when
+# FILTER_SOURCE_FILES is enabled.
+
+FILTER_SOURCE_PATTERNS =
+
+#---------------------------------------------------------------------------
+# configuration options related to source browsing
+#---------------------------------------------------------------------------
+
+# If the SOURCE_BROWSER tag is set to YES then a list of source files will
+# be generated. Documented entities will be cross-referenced with these sources.
+# Note: To get rid of all source code in the generated output, make sure also
+# VERBATIM_HEADERS is set to NO.
+
+SOURCE_BROWSER         = NO
+
+# Setting the INLINE_SOURCES tag to YES will include the body
+# of functions and classes directly in the documentation.
+
+INLINE_SOURCES         = NO
+
+# Setting the STRIP_CODE_COMMENTS tag to YES (the default) will instruct
+# doxygen to hide any special comment blocks from generated source code
+# fragments. Normal C and C++ comments will always remain visible.
+
+STRIP_CODE_COMMENTS    = YES
+
+# If the REFERENCED_BY_RELATION tag is set to YES
+# then for each documented function all documented
+# functions referencing it will be listed.
+
+REFERENCED_BY_RELATION = NO
+
+# If the REFERENCES_RELATION tag is set to YES
+# then for each documented function all documented entities
+# called/used by that function will be listed.
+
+REFERENCES_RELATION    = NO
+
+# If the REFERENCES_LINK_SOURCE tag is set to YES (the default)
+# and SOURCE_BROWSER tag is set to YES, then the hyperlinks from
+# functions in REFERENCES_RELATION and REFERENCED_BY_RELATION lists will
+# link to the source code.
+# Otherwise they will link to the documentation.
+
+REFERENCES_LINK_SOURCE = YES
+
+# If the USE_HTAGS tag is set to YES then the references to source code
+# will point to the HTML generated by the htags(1) tool instead of doxygen
+# built-in source browser. The htags tool is part of GNU's global source
+# tagging system (see http://www.gnu.org/software/global/global.html). You
+# will need version 4.8.6 or higher.
+
+USE_HTAGS              = NO
+
+# If the VERBATIM_HEADERS tag is set to YES (the default) then Doxygen
+# will generate a verbatim copy of the header file for each class for
+# which an include is specified. Set to NO to disable this.
+
+VERBATIM_HEADERS       = YES
+
+#---------------------------------------------------------------------------
+# configuration options related to the alphabetical class index
+#---------------------------------------------------------------------------
+
+# If the ALPHABETICAL_INDEX tag is set to YES, an alphabetical index
+# of all compounds will be generated. Enable this if the project
+# contains a lot of classes, structs, unions or interfaces.
+
+ALPHABETICAL_INDEX     = NO
+
+# If the alphabetical index is enabled (see ALPHABETICAL_INDEX) then
+# the COLS_IN_ALPHA_INDEX tag can be used to specify the number of columns
+# in which this list will be split (can be a number in the range [1..20])
+
+COLS_IN_ALPHA_INDEX    = 5
+
+# In case all classes in a project start with a common prefix, all
+# classes will be put under the same header in the alphabetical index.
+# The IGNORE_PREFIX tag can be used to specify one or more prefixes that
+# should be ignored while generating the index headers.
+
+IGNORE_PREFIX          =
+
+#---------------------------------------------------------------------------
+# configuration options related to the HTML output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_HTML tag is set to YES (the default) Doxygen will
+# generate HTML output.
+
+GENERATE_HTML          = YES
+
+# The HTML_OUTPUT tag is used to specify where the HTML docs will be put.
+# If a relative path is entered the value of OUTPUT_DIRECTORY will be
+# put in front of it. If left blank `html' will be used as the default path.
+
+HTML_OUTPUT            = html
+
+# The HTML_FILE_EXTENSION tag can be used to specify the file extension for
+# each generated HTML page (for example: .htm,.php,.asp). If it is left blank
+# doxygen will generate files with .html extension.
+
+HTML_FILE_EXTENSION    = .html
+
+# The HTML_HEADER tag can be used to specify a personal HTML header for
+# each generated HTML page. If it is left blank doxygen will generate a
+# standard header.
+
+HTML_HEADER            =
+
+# The HTML_FOOTER tag can be used to specify a personal HTML footer for
+# each generated HTML page. If it is left blank doxygen will generate a
+# standard footer.
+
+HTML_FOOTER            =
+
+# The HTML_STYLESHEET tag can be used to specify a user-defined cascading
+# style sheet that is used by each HTML page. It can be used to
+# fine-tune the look of the HTML output. If the tag is left blank doxygen
+# will generate a default style sheet. Note that doxygen will try to copy
+# the style sheet file to the HTML output directory, so don't put your own
+# stylesheet in the HTML output directory as well, or it will be erased!
+
+HTML_STYLESHEET        =
+
+# The HTML_COLORSTYLE_HUE tag controls the color of the HTML output.
+# Doxygen will adjust the colors in the stylesheet and background images
+# according to this color. Hue is specified as an angle on a colorwheel,
+# see http://en.wikipedia.org/wiki/Hue for more information.
+# For instance the value 0 represents red, 60 is yellow, 120 is green,
+# 180 is cyan, 240 is blue, 300 purple, and 360 is red again.
+# The allowed range is 0 to 359.
+
+HTML_COLORSTYLE_HUE    = 220
+
+# The HTML_COLORSTYLE_SAT tag controls the purity (or saturation) of
+# the colors in the HTML output. For a value of 0 the output will use
+# grayscales only. A value of 255 will produce the most vivid colors.
+
+HTML_COLORSTYLE_SAT    = 100
+
+# The HTML_COLORSTYLE_GAMMA tag controls the gamma correction applied to
+# the luminance component of the colors in the HTML output. Values below
+# 100 gradually make the output lighter, whereas values above 100 make
+# the output darker. The value divided by 100 is the actual gamma applied,
+# so 80 represents a gamma of 0.8, The value 220 represents a gamma of 2.2,
+# and 100 does not change the gamma.
+
+HTML_COLORSTYLE_GAMMA  = 80
+
+# If the HTML_TIMESTAMP tag is set to YES then the footer of each generated HTML
+# page will contain the date and time when the page was generated. Setting
+# this to NO can help when comparing the output of multiple runs.
+
+HTML_TIMESTAMP         = NO
+
+# If the HTML_ALIGN_MEMBERS tag is set to YES, the members of classes,
+# files or namespaces will be aligned in HTML using tables. If set to
+# NO a bullet list will be used.
+
+HTML_ALIGN_MEMBERS     = YES
+
+# If the HTML_DYNAMIC_SECTIONS tag is set to YES then the generated HTML
+# documentation will contain sections that can be hidden and shown after the
+# page has loaded. For this to work a browser that supports
+# JavaScript and DHTML is required (for instance Mozilla 1.0+, Firefox
+# Netscape 6.0+, Internet explorer 5.0+, Konqueror, or Safari).
+
+HTML_DYNAMIC_SECTIONS  = NO
+
+# If the GENERATE_DOCSET tag is set to YES, additional index files
+# will be generated that can be used as input for Apple's Xcode 3
+# integrated development environment, introduced with OSX 10.5 (Leopard).
+# To create a documentation set, doxygen will generate a Makefile in the
+# HTML output directory. Running make will produce the docset in that
+# directory and running "make install" will install the docset in
+# ~/Library/Developer/Shared/Documentation/DocSets so that Xcode will find
+# it at startup.
+# See http://developer.apple.com/tools/creatingdocsetswithdoxygen.html
+# for more information.
+
+GENERATE_DOCSET        = NO
+
+# When GENERATE_DOCSET tag is set to YES, this tag determines the name of the
+# feed. A documentation feed provides an umbrella under which multiple
+# documentation sets from a single provider (such as a company or product suite)
+# can be grouped.
+
+DOCSET_FEEDNAME        = "Doxygen generated docs"
+
+# When GENERATE_DOCSET tag is set to YES, this tag specifies a string that
+# should uniquely identify the documentation set bundle. This should be a
+# reverse domain-name style string, e.g. com.mycompany.MyDocSet. Doxygen
+# will append .docset to the name.
+
+DOCSET_BUNDLE_ID       = org.doxygen.Project
+
+# When GENERATE_PUBLISHER_ID tag specifies a string that should uniquely identify
+# the documentation publisher. This should be a reverse domain-name style
+# string, e.g. com.mycompany.MyDocSet.documentation.
+
+DOCSET_PUBLISHER_ID    = org.doxygen.Publisher
+
+# The GENERATE_PUBLISHER_NAME tag identifies the documentation publisher.
+
+DOCSET_PUBLISHER_NAME  = Publisher
+
+# If the GENERATE_HTMLHELP tag is set to YES, additional index files
+# will be generated that can be used as input for tools like the
+# Microsoft HTML help workshop to generate a compiled HTML help file (.chm)
+# of the generated HTML documentation.
+
+GENERATE_HTMLHELP      = NO
+
+# If the GENERATE_HTMLHELP tag is set to YES, the CHM_FILE tag can
+# be used to specify the file name of the resulting .chm file. You
+# can add a path in front of the file if the result should not be
+# written to the html output directory.
+
+CHM_FILE               =
+
+# If the GENERATE_HTMLHELP tag is set to YES, the HHC_LOCATION tag can
+# be used to specify the location (absolute path including file name) of
+# the HTML help compiler (hhc.exe). If non-empty doxygen will try to run
+# the HTML help compiler on the generated index.hhp.
+
+HHC_LOCATION           =
+
+# If the GENERATE_HTMLHELP tag is set to YES, the GENERATE_CHI flag
+# controls if a separate .chi index file is generated (YES) or that
+# it should be included in the master .chm file (NO).
+
+GENERATE_CHI           = NO
+
+# If the GENERATE_HTMLHELP tag is set to YES, the CHM_INDEX_ENCODING
+# is used to encode HtmlHelp index (hhk), content (hhc) and project file
+# content.
+
+CHM_INDEX_ENCODING     =
+
+# If the GENERATE_HTMLHELP tag is set to YES, the BINARY_TOC flag
+# controls whether a binary table of contents is generated (YES) or a
+# normal table of contents (NO) in the .chm file.
+
+BINARY_TOC             = NO
+
+# The TOC_EXPAND flag can be set to YES to add extra items for group members
+# to the contents of the HTML help documentation and to the tree view.
+
+TOC_EXPAND             = NO
+
+# If the GENERATE_QHP tag is set to YES and both QHP_NAMESPACE and
+# QHP_VIRTUAL_FOLDER are set, an additional index file will be generated
+# that can be used as input for Qt's qhelpgenerator to generate a
+# Qt Compressed Help (.qch) of the generated HTML documentation.
+
+GENERATE_QHP           = NO
+
+# If the QHG_LOCATION tag is specified, the QCH_FILE tag can
+# be used to specify the file name of the resulting .qch file.
+# The path specified is relative to the HTML output folder.
+
+QCH_FILE               =
+
+# The QHP_NAMESPACE tag specifies the namespace to use when generating
+# Qt Help Project output. For more information please see
+# http://doc.trolltech.com/qthelpproject.html#namespace
+
+QHP_NAMESPACE          =
+
+# The QHP_VIRTUAL_FOLDER tag specifies the namespace to use when generating
+# Qt Help Project output. For more information please see
+# http://doc.trolltech.com/qthelpproject.html#virtual-folders
+
+QHP_VIRTUAL_FOLDER     = doc
+
+# If QHP_CUST_FILTER_NAME is set, it specifies the name of a custom filter to
+# add. For more information please see
+# http://doc.trolltech.com/qthelpproject.html#custom-filters
+
+QHP_CUST_FILTER_NAME   =
+
+# The QHP_CUST_FILT_ATTRS tag specifies the list of the attributes of the
+# custom filter to add. For more information please see
+# <a href="http://doc.trolltech.com/qthelpproject.html#custom-filters">
+# Qt Help Project / Custom Filters</a>.
+
+QHP_CUST_FILTER_ATTRS  =
+
+# The QHP_SECT_FILTER_ATTRS tag specifies the list of the attributes this
+# project's
+# filter section matches.
+# <a href="http://doc.trolltech.com/qthelpproject.html#filter-attributes">
+# Qt Help Project / Filter Attributes</a>.
+
+QHP_SECT_FILTER_ATTRS  =
+
+# If the GENERATE_QHP tag is set to YES, the QHG_LOCATION tag can
+# be used to specify the location of Qt's qhelpgenerator.
+# If non-empty doxygen will try to run qhelpgenerator on the generated
+# .qhp file.
+
+QHG_LOCATION           =
+
+# If the GENERATE_ECLIPSEHELP tag is set to YES, additional index files
+#  will be generated, which together with the HTML files, form an Eclipse help
+# plugin. To install this plugin and make it available under the help contents
+# menu in Eclipse, the contents of the directory containing the HTML and XML
+# files needs to be copied into the plugins directory of eclipse. The name of
+# the directory within the plugins directory should be the same as
+# the ECLIPSE_DOC_ID value. After copying Eclipse needs to be restarted before
+# the help appears.
+
+GENERATE_ECLIPSEHELP   = NO
+
+# A unique identifier for the eclipse help plugin. When installing the plugin
+# the directory name containing the HTML and XML files should also have
+# this name.
+
+ECLIPSE_DOC_ID         = org.doxygen.Project
+
+# The DISABLE_INDEX tag can be used to turn on/off the condensed index at
+# top of each HTML page. The value NO (the default) enables the index and
+# the value YES disables it.
+
+DISABLE_INDEX          = NO
+
+# This tag can be used to set the number of enum values (range [0,1..20])
+# that doxygen will group on one line in the generated HTML documentation.
+# Note that a value of 0 will completely suppress the enum values from appearing in the overview section.
+
+ENUM_VALUES_PER_LINE   = 4
+
+# The GENERATE_TREEVIEW tag is used to specify whether a tree-like index
+# structure should be generated to display hierarchical information.
+# If the tag value is set to YES, a side panel will be generated
+# containing a tree-like index structure (just like the one that
+# is generated for HTML Help). For this to work a browser that supports
+# JavaScript, DHTML, CSS and frames is required (i.e. any modern browser).
+# Windows users are probably better off using the HTML help feature.
+
+GENERATE_TREEVIEW      = NONE
+
+# By enabling USE_INLINE_TREES, doxygen will generate the Groups, Directories,
+# and Class Hierarchy pages using a tree view instead of an ordered list.
+
+USE_INLINE_TREES       = NO
+
+# If the treeview is enabled (see GENERATE_TREEVIEW) then this tag can be
+# used to set the initial width (in pixels) of the frame in which the tree
+# is shown.
+
+TREEVIEW_WIDTH         = 250
+
+# When the EXT_LINKS_IN_WINDOW option is set to YES doxygen will open
+# links to external symbols imported via tag files in a separate window.
+
+EXT_LINKS_IN_WINDOW    = NO
+
+# Use this tag to change the font size of Latex formulas included
+# as images in the HTML documentation. The default is 10. Note that
+# when you change the font size after a successful doxygen run you need
+# to manually remove any form_*.png images from the HTML output directory
+# to force them to be regenerated.
+
+FORMULA_FONTSIZE       = 10
+
+# Use the FORMULA_TRANPARENT tag to determine whether or not the images
+# generated for formulas are transparent PNGs. Transparent PNGs are
+# not supported properly for IE 6.0, but are supported on all modern browsers.
+# Note that when changing this option you need to delete any form_*.png files
+# in the HTML output before the changes have effect.
+
+FORMULA_TRANSPARENT    = YES
+
+# Enable the USE_MATHJAX option to render LaTeX formulas using MathJax
+# (see http://www.mathjax.org) which uses client side Javascript for the
+# rendering instead of using prerendered bitmaps. Use this if you do not
+# have LaTeX installed or if you want to formulas look prettier in the HTML
+# output. When enabled you also need to install MathJax separately and
+# configure the path to it using the MATHJAX_RELPATH option.
+
+USE_MATHJAX            = NO
+
+# When MathJax is enabled you need to specify the location relative to the
+# HTML output directory using the MATHJAX_RELPATH option. The destination
+# directory should contain the MathJax.js script. For instance, if the mathjax
+# directory is located at the same level as the HTML output directory, then
+# MATHJAX_RELPATH should be ../mathjax. The default value points to the mathjax.org site, so you can quickly see the result without installing
+# MathJax, but it is strongly recommended to install a local copy of MathJax
+# before deployment.
+
+MATHJAX_RELPATH        = http://www.mathjax.org/mathjax
+
+# When the SEARCHENGINE tag is enabled doxygen will generate a search box
+# for the HTML output. The underlying search engine uses javascript
+# and DHTML and should work on any modern browser. Note that when using
+# HTML help (GENERATE_HTMLHELP), Qt help (GENERATE_QHP), or docsets
+# (GENERATE_DOCSET) there is already a search function so this one should
+# typically be disabled. For large projects the javascript based search engine
+# can be slow, then enabling SERVER_BASED_SEARCH may provide a better solution.
+
+SEARCHENGINE           = NO
+
+# When the SERVER_BASED_SEARCH tag is enabled the search engine will be
+# implemented using a PHP enabled web server instead of at the web client
+# using Javascript. Doxygen will generate the search PHP script and index
+# file to put on the web server. The advantage of the server
+# based approach is that it scales better to large projects and allows
+# full text search. The disadvantages are that it is more difficult to setup
+# and does not have live searching capabilities.
+
+SERVER_BASED_SEARCH    = NO
+
+#---------------------------------------------------------------------------
+# configuration options related to the LaTeX output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_LATEX tag is set to YES (the default) Doxygen will
+# generate Latex output.
+
+GENERATE_LATEX         = YES
+
+# The LATEX_OUTPUT tag is used to specify where the LaTeX docs will be put.
+# If a relative path is entered the value of OUTPUT_DIRECTORY will be
+# put in front of it. If left blank `latex' will be used as the default path.
+
+LATEX_OUTPUT           = latex
+
+# The LATEX_CMD_NAME tag can be used to specify the LaTeX command name to be
+# invoked. If left blank `latex' will be used as the default command name.
+# Note that when enabling USE_PDFLATEX this option is only used for
+# generating bitmaps for formulas in the HTML output, but not in the
+# Makefile that is written to the output directory.
+
+LATEX_CMD_NAME         = latex
+
+# The MAKEINDEX_CMD_NAME tag can be used to specify the command name to
+# generate index for LaTeX. If left blank `makeindex' will be used as the
+# default command name.
+
+MAKEINDEX_CMD_NAME     = makeindex
+
+# If the COMPACT_LATEX tag is set to YES Doxygen generates more compact
+# LaTeX documents. This may be useful for small projects and may help to
+# save some trees in general.
+
+COMPACT_LATEX          = NO
+
+# The PAPER_TYPE tag can be used to set the paper type that is used
+# by the printer. Possible values are: a4, letter, legal and
+# executive. If left blank a4wide will be used.
+
+PAPER_TYPE             = a4wide
+
+# The EXTRA_PACKAGES tag can be to specify one or more names of LaTeX
+# packages that should be included in the LaTeX output.
+
+EXTRA_PACKAGES         =
+
+# The LATEX_HEADER tag can be used to specify a personal LaTeX header for
+# the generated latex document. The header should contain everything until
+# the first chapter. If it is left blank doxygen will generate a
+# standard header. Notice: only use this tag if you know what you are doing!
+
+LATEX_HEADER           =
+
+# If the PDF_HYPERLINKS tag is set to YES, the LaTeX that is generated
+# is prepared for conversion to pdf (using ps2pdf). The pdf file will
+# contain links (just like the HTML output) instead of page references
+# This makes the output suitable for online browsing using a pdf viewer.
+
+PDF_HYPERLINKS         = YES
+
+# If the USE_PDFLATEX tag is set to YES, pdflatex will be used instead of
+# plain latex in the generated Makefile. Set this option to YES to get a
+# higher quality PDF documentation.
+
+USE_PDFLATEX           = YES
+
+# If the LATEX_BATCHMODE tag is set to YES, doxygen will add the \\batchmode.
+# command to the generated LaTeX files. This will instruct LaTeX to keep
+# running if errors occur, instead of asking the user for help.
+# This option is also used when generating formulas in HTML.
+
+LATEX_BATCHMODE        = NO
+
+# If LATEX_HIDE_INDICES is set to YES then doxygen will not
+# include the index chapters (such as File Index, Compound Index, etc.)
+# in the output.
+
+LATEX_HIDE_INDICES     = NO
+
+# If LATEX_SOURCE_CODE is set to YES then doxygen will include
+# source code with syntax highlighting in the LaTeX output.
+# Note that which sources are shown also depends on other settings
+# such as SOURCE_BROWSER.
+
+LATEX_SOURCE_CODE      = NO
+
+#---------------------------------------------------------------------------
+# configuration options related to the RTF output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_RTF tag is set to YES Doxygen will generate RTF output
+# The RTF output is optimized for Word 97 and may not look very pretty with
+# other RTF readers or editors.
+
+GENERATE_RTF           = NO
+
+# The RTF_OUTPUT tag is used to specify where the RTF docs will be put.
+# If a relative path is entered the value of OUTPUT_DIRECTORY will be
+# put in front of it. If left blank `rtf' will be used as the default path.
+
+RTF_OUTPUT             = rtf
+
+# If the COMPACT_RTF tag is set to YES Doxygen generates more compact
+# RTF documents. This may be useful for small projects and may help to
+# save some trees in general.
+
+COMPACT_RTF            = NO
+
+# If the RTF_HYPERLINKS tag is set to YES, the RTF that is generated
+# will contain hyperlink fields. The RTF file will
+# contain links (just like the HTML output) instead of page references.
+# This makes the output suitable for online browsing using WORD or other
+# programs which support those fields.
+# Note: wordpad (write) and others do not support links.
+
+RTF_HYPERLINKS         = NO
+
+# Load stylesheet definitions from file. Syntax is similar to doxygen's
+# config file, i.e. a series of assignments. You only have to provide
+# replacements, missing definitions are set to their default value.
+
+RTF_STYLESHEET_FILE    =
+
+# Set optional variables used in the generation of an rtf document.
+# Syntax is similar to doxygen's config file.
+
+RTF_EXTENSIONS_FILE    =
+
+#---------------------------------------------------------------------------
+# configuration options related to the man page output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_MAN tag is set to YES (the default) Doxygen will
+# generate man pages
+
+GENERATE_MAN           = YES
+
+# The MAN_OUTPUT tag is used to specify where the man pages will be put.
+# If a relative path is entered the value of OUTPUT_DIRECTORY will be
+# put in front of it. If left blank `man' will be used as the default path.
+
+MAN_OUTPUT             = man
+
+# The MAN_EXTENSION tag determines the extension that is added to
+# the generated man pages (default is the subroutine's section .3)
+
+MAN_EXTENSION          = .3
+
+# If the MAN_LINKS tag is set to YES and Doxygen generates man output,
+# then it will generate one additional man file for each entity
+# documented in the real man page(s). These additional files
+# only source the real man page, but without them the man command
+# would be unable to find the correct page. The default is NO.
+
+MAN_LINKS              = NO
+
+#---------------------------------------------------------------------------
+# configuration options related to the XML output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_XML tag is set to YES Doxygen will
+# generate an XML file that captures the structure of
+# the code including all documentation.
+
+GENERATE_XML           = NO
+
+# The XML_OUTPUT tag is used to specify where the XML pages will be put.
+# If a relative path is entered the value of OUTPUT_DIRECTORY will be
+# put in front of it. If left blank `xml' will be used as the default path.
+
+XML_OUTPUT             = xml
+
+# The XML_SCHEMA tag can be used to specify an XML schema,
+# which can be used by a validating XML parser to check the
+# syntax of the XML files.
+
+XML_SCHEMA             =
+
+# The XML_DTD tag can be used to specify an XML DTD,
+# which can be used by a validating XML parser to check the
+# syntax of the XML files.
+
+XML_DTD                =
+
+# If the XML_PROGRAMLISTING tag is set to YES Doxygen will
+# dump the program listings (including syntax highlighting
+# and cross-referencing information) to the XML output. Note that
+# enabling this will significantly increase the size of the XML output.
+
+XML_PROGRAMLISTING     = YES
+
+#---------------------------------------------------------------------------
+# configuration options for the AutoGen Definitions output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_AUTOGEN_DEF tag is set to YES Doxygen will
+# generate an AutoGen Definitions (see autogen.sf.net) file
+# that captures the structure of the code including all
+# documentation. Note that this feature is still experimental
+# and incomplete at the moment.
+
+GENERATE_AUTOGEN_DEF   = NO
+
+#---------------------------------------------------------------------------
+# configuration options related to the Perl module output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_PERLMOD tag is set to YES Doxygen will
+# generate a Perl module file that captures the structure of
+# the code including all documentation. Note that this
+# feature is still experimental and incomplete at the
+# moment.
+
+GENERATE_PERLMOD       = NO
+
+# If the PERLMOD_LATEX tag is set to YES Doxygen will generate
+# the necessary Makefile rules, Perl scripts and LaTeX code to be able
+# to generate PDF and DVI output from the Perl module output.
+
+PERLMOD_LATEX          = NO
+
+# If the PERLMOD_PRETTY tag is set to YES the Perl module output will be
+# nicely formatted so it can be parsed by a human reader.
+# This is useful
+# if you want to understand what is going on.
+# On the other hand, if this
+# tag is set to NO the size of the Perl module output will be much smaller
+# and Perl will parse it just the same.
+
+PERLMOD_PRETTY         = YES
+
+# The names of the make variables in the generated doxyrules.make file
+# are prefixed with the string contained in PERLMOD_MAKEVAR_PREFIX.
+# This is useful so different doxyrules.make files included by the same
+# Makefile don't overwrite each other's variables.
+
+PERLMOD_MAKEVAR_PREFIX =
+
+#---------------------------------------------------------------------------
+# Configuration options related to the preprocessor
+#---------------------------------------------------------------------------
+
+# If the ENABLE_PREPROCESSING tag is set to YES (the default) Doxygen will
+# evaluate all C-preprocessor directives found in the sources and include
+# files.
+
+ENABLE_PREPROCESSING   = YES
+
+# If the MACRO_EXPANSION tag is set to YES Doxygen will expand all macro
+# names in the source code. If set to NO (the default) only conditional
+# compilation will be performed. Macro expansion can be done in a controlled
+# way by setting EXPAND_ONLY_PREDEF to YES.
+
+MACRO_EXPANSION        = YES
+
+# If the EXPAND_ONLY_PREDEF and MACRO_EXPANSION tags are both set to YES
+# then the macro expansion is limited to the macros specified with the
+# PREDEFINED and EXPAND_AS_DEFINED tags.
+
+EXPAND_ONLY_PREDEF     = YES
+
+# If the SEARCH_INCLUDES tag is set to YES (the default) the includes files
+# in the INCLUDE_PATH (see below) will be search if a #include is found.
+
+SEARCH_INCLUDES        = YES
+
+# The INCLUDE_PATH tag can be used to specify one or more directories that
+# contain include files that are not input files but should be processed by
+# the preprocessor.
+
+INCLUDE_PATH           =
+
+# You can use the INCLUDE_FILE_PATTERNS tag to specify one or more wildcard
+# patterns (like *.h and *.hpp) to filter out the header-files in the
+# directories. If left blank, the patterns specified with FILE_PATTERNS will
+# be used.
+
+INCLUDE_FILE_PATTERNS  =
+
+# The PREDEFINED tag can be used to specify one or more macro names that
+# are defined before the preprocessor is started (similar to the -D option of
+# gcc). The argument of the tag is a list of macros of the form: name
+# or name=definition (no spaces). If the definition and the = are
+# omitted =1 is assumed. To prevent a macro definition from being
+# undefined via #undef or recursively expanded use the := operator
+# instead of the = operator.
+
+PREDEFINED             = DOXYGEN \
+                         PRINTF_ATTRIBUTE(x,y)=
+
+# If the MACRO_EXPANSION and EXPAND_ONLY_PREDEF tags are set to YES then
+# this tag can be used to specify a list of macro names that should be expanded.
+# The macro definition that is found in the sources will be used.
+# Use the PREDEFINED tag if you want to use a different macro definition that overrules the definition found in the source code.
+
+EXPAND_AS_DEFINED      =
+
+# If the SKIP_FUNCTION_MACROS tag is set to YES (the default) then
+# doxygen's preprocessor will remove all references to function-like macros
+# that are alone on a line, have an all uppercase name, and do not end with a
+# semicolon, because these will confuse the parser if not removed.
+
+SKIP_FUNCTION_MACROS   = YES
+
+#---------------------------------------------------------------------------
+# Configuration::additions related to external references
+#---------------------------------------------------------------------------
+
+# The TAGFILES option can be used to specify one or more tagfiles.
+# Optionally an initial location of the external documentation
+# can be added for each tagfile. The format of a tag file without
+# this location is as follows:
+#
+# TAGFILES = file1 file2 ...
+# Adding location for the tag files is done as follows:
+#
+# TAGFILES = file1=loc1 "file2 = loc2" ...
+# where "loc1" and "loc2" can be relative or absolute paths or
+# URLs. If a location is present for each tag, the installdox tool
+# does not have to be run to correct the links.
+# Note that each tag file must have a unique name
+# (where the name does NOT include the path)
+# If a tag file is not located in the directory in which doxygen
+# is run, you must also specify the path to the tagfile here.
+
+TAGFILES               =
+
+# When a file name is specified after GENERATE_TAGFILE, doxygen will create
+# a tag file that is based on the input files it reads.
+
+GENERATE_TAGFILE       =
+
+# If the ALLEXTERNALS tag is set to YES all external classes will be listed
+# in the class index. If set to NO only the inherited external classes
+# will be listed.
+
+ALLEXTERNALS           = NO
+
+# If the EXTERNAL_GROUPS tag is set to YES all external groups will be listed
+# in the modules index. If set to NO, only the current project's groups will
+# be listed.
+
+EXTERNAL_GROUPS        = YES
+
+# The PERL_PATH should be the absolute path and name of the perl script
+# interpreter (i.e. the result of `which perl').
+
+PERL_PATH              = /usr/bin/perl
+
+#---------------------------------------------------------------------------
+# Configuration options related to the dot tool
+#---------------------------------------------------------------------------
+
+# If the CLASS_DIAGRAMS tag is set to YES (the default) Doxygen will
+# generate a inheritance diagram (in HTML, RTF and LaTeX) for classes with base
+# or super classes. Setting the tag to NO turns the diagrams off. Note that
+# this option also works with HAVE_DOT disabled, but it is recommended to
+# install and use dot, since it yields more powerful graphs.
+
+CLASS_DIAGRAMS         = YES
+
+# You can define message sequence charts within doxygen comments using the \msc
+# command. Doxygen will then run the mscgen tool (see
+# http://www.mcternan.me.uk/mscgen/) to produce the chart and insert it in the
+# documentation. The MSCGEN_PATH tag allows you to specify the directory where
+# the mscgen tool resides. If left empty the tool is assumed to be found in the
+# default search path.
+
+MSCGEN_PATH            =
+
+# If set to YES, the inheritance and collaboration graphs will hide
+# inheritance and usage relations if the target is undocumented
+# or is not a class.
+
+HIDE_UNDOC_RELATIONS   = YES
+
+# If you set the HAVE_DOT tag to YES then doxygen will assume the dot tool is
+# available from the path. This tool is part of Graphviz, a graph visualization
+# toolkit from AT&T and Lucent Bell Labs. The other options in this section
+# have no effect if this option is set to NO (the default)
+
+HAVE_DOT               = NO
+
+# The DOT_NUM_THREADS specifies the number of dot invocations doxygen is
+# allowed to run in parallel. When set to 0 (the default) doxygen will
+# base this on the number of processors available in the system. You can set it
+# explicitly to a value larger than 0 to get control over the balance
+# between CPU load and processing speed.
+
+DOT_NUM_THREADS        = 0
+
+# By default doxygen will write a font called Helvetica to the output
+# directory and reference it in all dot files that doxygen generates.
+# When you want a differently looking font you can specify the font name
+# using DOT_FONTNAME. You need to make sure dot is able to find the font,
+# which can be done by putting it in a standard location or by setting the
+# DOTFONTPATH environment variable or by setting DOT_FONTPATH to the directory
+# containing the font.
+
+DOT_FONTNAME           = FreeSans
+
+# The DOT_FONTSIZE tag can be used to set the size of the font of dot graphs.
+# The default size is 10pt.
+
+DOT_FONTSIZE           = 10
+
+# By default doxygen will tell dot to use the output directory to look for the
+# FreeSans.ttf font (which doxygen will put there itself). If you specify a
+# different font using DOT_FONTNAME you can set the path where dot
+# can find it using this tag.
+
+DOT_FONTPATH           =
+
+# If the CLASS_GRAPH and HAVE_DOT tags are set to YES then doxygen
+# will generate a graph for each documented class showing the direct and
+# indirect inheritance relations. Setting this tag to YES will force the
+# the CLASS_DIAGRAMS tag to NO.
+
+CLASS_GRAPH            = YES
+
+# If the COLLABORATION_GRAPH and HAVE_DOT tags are set to YES then doxygen
+# will generate a graph for each documented class showing the direct and
+# indirect implementation dependencies (inheritance, containment, and
+# class references variables) of the class with other documented classes.
+
+COLLABORATION_GRAPH    = YES
+
+# If the GROUP_GRAPHS and HAVE_DOT tags are set to YES then doxygen
+# will generate a graph for groups, showing the direct groups dependencies
+
+GROUP_GRAPHS           = YES
+
+# If the UML_LOOK tag is set to YES doxygen will generate inheritance and
+# collaboration diagrams in a style similar to the OMG's Unified Modeling
+# Language.
+
+UML_LOOK               = NO
+
+# If set to YES, the inheritance and collaboration graphs will show the
+# relations between templates and their instances.
+
+TEMPLATE_RELATIONS     = NO
+
+# If the ENABLE_PREPROCESSING, SEARCH_INCLUDES, INCLUDE_GRAPH, and HAVE_DOT
+# tags are set to YES then doxygen will generate a graph for each documented
+# file showing the direct and indirect include dependencies of the file with
+# other documented files.
+
+INCLUDE_GRAPH          = YES
+
+# If the ENABLE_PREPROCESSING, SEARCH_INCLUDES, INCLUDED_BY_GRAPH, and
+# HAVE_DOT tags are set to YES then doxygen will generate a graph for each
+# documented header file showing the documented files that directly or
+# indirectly include this file.
+
+INCLUDED_BY_GRAPH      = YES
+
+# If the CALL_GRAPH and HAVE_DOT options are set to YES then
+# doxygen will generate a call dependency graph for every global function
+# or class method. Note that enabling this option will significantly increase
+# the time of a run. So in most cases it will be better to enable call graphs
+# for selected functions only using the \callgraph command.
+
+CALL_GRAPH             = NO
+
+# If the CALLER_GRAPH and HAVE_DOT tags are set to YES then
+# doxygen will generate a caller dependency graph for every global function
+# or class method. Note that enabling this option will significantly increase
+# the time of a run. So in most cases it will be better to enable caller
+# graphs for selected functions only using the \callergraph command.
+
+CALLER_GRAPH           = NO
+
+# If the GRAPHICAL_HIERARCHY and HAVE_DOT tags are set to YES then doxygen
+# will generate a graphical hierarchy of all classes instead of a textual one.
+
+GRAPHICAL_HIERARCHY    = YES
+
+# If the DIRECTORY_GRAPH, SHOW_DIRECTORIES and HAVE_DOT tags are set to YES
+# then doxygen will show the dependencies a directory has on other directories
+# in a graphical way. The dependency relations are determined by the #include
+# relations between the files in the directories.
+
+DIRECTORY_GRAPH        = YES
+
+# The DOT_IMAGE_FORMAT tag can be used to set the image format of the images
+# generated by dot. Possible values are png, svg, gif or svg.
+# If left blank png will be used.
+
+DOT_IMAGE_FORMAT       = png
+
+# The tag DOT_PATH can be used to specify the path where the dot tool can be
+# found. If left blank, it is assumed the dot tool can be found in the path.
+
+DOT_PATH               =
+
+# The DOTFILE_DIRS tag can be used to specify one or more directories that
+# contain dot files that are included in the documentation (see the
+# \dotfile command).
+
+DOTFILE_DIRS           =
+
+# The MSCFILE_DIRS tag can be used to specify one or more directories that
+# contain msc files that are included in the documentation (see the
+# \mscfile command).
+
+MSCFILE_DIRS           =
+
+# The DOT_GRAPH_MAX_NODES tag can be used to set the maximum number of
+# nodes that will be shown in the graph. If the number of nodes in a graph
+# becomes larger than this value, doxygen will truncate the graph, which is
+# visualized by representing a node as a red box. Note that doxygen if the
+# number of direct children of the root node in a graph is already larger than
+# DOT_GRAPH_MAX_NODES then the graph will not be shown at all. Also note
+# that the size of a graph can be further restricted by MAX_DOT_GRAPH_DEPTH.
+
+DOT_GRAPH_MAX_NODES    = 50
+
+# The MAX_DOT_GRAPH_DEPTH tag can be used to set the maximum depth of the
+# graphs generated by dot. A depth value of 3 means that only nodes reachable
+# from the root by following a path via at most 3 edges will be shown. Nodes
+# that lay further from the root node will be omitted. Note that setting this
+# option to 1 or 2 may greatly reduce the computation time needed for large
+# code bases. Also note that the size of a graph can be further restricted by
+# DOT_GRAPH_MAX_NODES. Using a depth of 0 means no depth restriction.
+
+MAX_DOT_GRAPH_DEPTH    = 0
+
+# Set the DOT_TRANSPARENT tag to YES to generate images with a transparent
+# background. This is disabled by default, because dot on Windows does not
+# seem to support this out of the box. Warning: Depending on the platform used,
+# enabling this option may lead to badly anti-aliased labels on the edges of
+# a graph (i.e. they become hard to read).
+
+DOT_TRANSPARENT        = YES
+
+# Set the DOT_MULTI_TARGETS tag to YES allow dot to generate multiple output
+# files in one run (i.e. multiple -o and -T options on the command line). This
+# makes dot run faster, but since only newer versions of dot (>1.8.10)
+# support this, this feature is disabled by default.
+
+DOT_MULTI_TARGETS      = NO
+
+# If the GENERATE_LEGEND tag is set to YES (the default) Doxygen will
+# generate a legend page explaining the meaning of the various boxes and
+# arrows in the dot generated graphs.
+
+GENERATE_LEGEND        = YES
+
+# If the DOT_CLEANUP tag is set to YES (the default) Doxygen will
+# remove the intermediate dot files that are used to generate
+# the various graphs.
+
+DOT_CLEANUP            = YES
diff --git a/ctdb/lib/tdb/include/tdb.h b/ctdb/lib/tdb/include/tdb.h
new file mode 100644
index 0000000000..d19439e760
--- /dev/null
+++ b/ctdb/lib/tdb/include/tdb.h
@@ -0,0 +1,869 @@
+#ifndef __TDB_H__
+#define __TDB_H__
+
+/* 
+   Unix SMB/CIFS implementation.
+
+   trivial database library
+
+   Copyright (C) Andrew Tridgell 1999-2004
+   
+     ** NOTE! The following LGPL license applies to the tdb
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#ifdef  __cplusplus
+extern "C" {
+#endif
+
+#include <signal.h>
+
+/**
+ * @defgroup tdb The tdb API
+ *
+ * tdb is a Trivial database. In concept, it is very much like GDBM, and BSD's
+ * DB except that it allows multiple simultaneous writers and uses locking
+ * internally to keep writers from trampling on each other. tdb is also
+ * extremely small.
+ *
+ * @section tdb_interface Interface
+ *
+ * The interface is very similar to gdbm except for the following:
+ *
+ * <ul>
+ * <li>different open interface. The tdb_open call is more similar to a
+ * traditional open()</li>
+ * <li>no tdbm_reorganise() function</li>
+ * <li>no tdbm_sync() function. No operations are cached in the library
+ *     anyway</li>
+ * <li>added a tdb_traverse() function for traversing the whole database</li>
+ * <li>added transactions support</li>
+ * </ul>
+ *
+ * A general rule for using tdb is that the caller frees any returned TDB_DATA
+ * structures. Just call free(p.dptr) to free a TDB_DATA return value called p.
+ * This is the same as gdbm.
+ *
+ * @{
+ */
+
+/** Flags to tdb_store() */
+#define TDB_REPLACE 1		/** Unused */
+#define TDB_INSERT 2 		/** Don't overwrite an existing entry */
+#define TDB_MODIFY 3		/** Don't create an existing entry    */
+
+/** Flags for tdb_open() */
+#define TDB_DEFAULT 0 /** just a readability place holder */
+#define TDB_CLEAR_IF_FIRST 1 /** If this is the first open, wipe the db */
+#define TDB_INTERNAL 2 /** Don't store on disk */
+#define TDB_NOLOCK   4 /** Don't do any locking */
+#define TDB_NOMMAP   8 /** Don't use mmap */
+#define TDB_CONVERT 16 /** Convert endian (internal use) */
+#define TDB_BIGENDIAN 32 /** Header is big-endian (internal use) */
+#define TDB_NOSYNC   64 /** Don't use synchronous transactions */
+#define TDB_SEQNUM   128 /** Maintain a sequence number */
+#define TDB_VOLATILE   256 /** Activate the per-hashchain freelist, default 5 */
+#define TDB_ALLOW_NESTING 512 /** Allow transactions to nest */
+#define TDB_DISALLOW_NESTING 1024 /** Disallow transactions to nest */
+#define TDB_INCOMPATIBLE_HASH 2048 /** Better hashing: can't be opened by tdb < 1.2.6. */
+
+/** The tdb error codes */
+enum TDB_ERROR {TDB_SUCCESS=0, TDB_ERR_CORRUPT, TDB_ERR_IO, TDB_ERR_LOCK, 
+		TDB_ERR_OOM, TDB_ERR_EXISTS, TDB_ERR_NOLOCK, TDB_ERR_LOCK_TIMEOUT,
+		TDB_ERR_NOEXIST, TDB_ERR_EINVAL, TDB_ERR_RDONLY,
+		TDB_ERR_NESTING};
+
+/** Debugging uses one of the following levels */
+enum tdb_debug_level {TDB_DEBUG_FATAL = 0, TDB_DEBUG_ERROR, 
+		      TDB_DEBUG_WARNING, TDB_DEBUG_TRACE};
+
+/** The tdb data structure */
+typedef struct TDB_DATA {
+	unsigned char *dptr;
+	size_t dsize;
+} TDB_DATA;
+
+#ifndef PRINTF_ATTRIBUTE
+#if (__GNUC__ >= 3)
+/** Use gcc attribute to check printf fns.  a1 is the 1-based index of
+ * the parameter containing the format, and a2 the index of the first
+ * argument. Note that some gcc 2.x versions don't handle this
+ * properly **/
+#define PRINTF_ATTRIBUTE(a1, a2) __attribute__ ((format (__printf__, a1, a2)))
+#else
+#define PRINTF_ATTRIBUTE(a1, a2)
+#endif
+#endif
+
+/** This is the context structure that is returned from a db open. */
+typedef struct tdb_context TDB_CONTEXT;
+
+typedef int (*tdb_traverse_func)(struct tdb_context *, TDB_DATA, TDB_DATA, void *);
+typedef void (*tdb_log_func)(struct tdb_context *, enum tdb_debug_level, const char *, ...) PRINTF_ATTRIBUTE(3, 4);
+typedef unsigned int (*tdb_hash_func)(TDB_DATA *key);
+
+struct tdb_logging_context {
+        tdb_log_func log_fn;
+        void *log_private;
+};
+
+/**
+ * @brief Open the database and creating it if necessary.
+ *
+ * @param[in]  name     The name of the db to open.
+ *
+ * @param[in]  hash_size The hash size is advisory, use zero for a default
+ *                       value.
+ *
+ * @param[in]  tdb_flags The flags to use to open the db:\n\n
+ *                         TDB_CLEAR_IF_FIRST - Clear database if we are the
+ *                                              only one with it open\n
+ *                         TDB_INTERNAL - Don't use a file, instaed store the
+ *                                        data in memory. The filename is
+ *                                        ignored in this case.\n
+ *                         TDB_NOLOCK - Don't do any locking\n
+ *                         TDB_NOMMAP - Don't use mmap\n
+ *                         TDB_NOSYNC - Don't synchronise transactions to disk\n
+ *                         TDB_SEQNUM - Maintain a sequence number\n
+ *                         TDB_VOLATILE - activate the per-hashchain freelist,
+ *                                        default 5.\n
+ *                         TDB_ALLOW_NESTING - Allow transactions to nest.\n
+ *                         TDB_DISALLOW_NESTING - Disallow transactions to nest.\n
+ *
+ * @param[in]  open_flags Flags for the open(2) function.
+ *
+ * @param[in]  mode     The mode for the open(2) function.
+ *
+ * @return              A tdb context structure, NULL on error.
+ */
+struct tdb_context *tdb_open(const char *name, int hash_size, int tdb_flags,
+		      int open_flags, mode_t mode);
+
+/**
+ * @brief Open the database and creating it if necessary.
+ *
+ * This is like tdb_open(), but allows you to pass an initial logging and
+ * hash function. Be careful when passing a hash function - all users of the
+ * database must use the same hash function or you will get data corruption.
+ *
+ * @param[in]  name     The name of the db to open.
+ *
+ * @param[in]  hash_size The hash size is advisory, use zero for a default
+ *                       value.
+ *
+ * @param[in]  tdb_flags The flags to use to open the db:\n\n
+ *                         TDB_CLEAR_IF_FIRST - Clear database if we are the
+ *                                              only one with it open\n
+ *                         TDB_INTERNAL - Don't use a file, instaed store the
+ *                                        data in memory. The filename is
+ *                                        ignored in this case.\n
+ *                         TDB_NOLOCK - Don't do any locking\n
+ *                         TDB_NOMMAP - Don't use mmap\n
+ *                         TDB_NOSYNC - Don't synchronise transactions to disk\n
+ *                         TDB_SEQNUM - Maintain a sequence number\n
+ *                         TDB_VOLATILE - activate the per-hashchain freelist,
+ *                                        default 5.\n
+ *                         TDB_ALLOW_NESTING - Allow transactions to nest.\n
+ *                         TDB_DISALLOW_NESTING - Disallow transactions to nest.\n
+ *
+ * @param[in]  open_flags Flags for the open(2) function.
+ *
+ * @param[in]  mode     The mode for the open(2) function.
+ *
+ * @param[in]  log_ctx  The logging function to use.
+ *
+ * @param[in]  hash_fn  The hash function you want to use.
+ *
+ * @return              A tdb context structure, NULL on error.
+ *
+ * @see tdb_open()
+ */
+struct tdb_context *tdb_open_ex(const char *name, int hash_size, int tdb_flags,
+			 int open_flags, mode_t mode,
+			 const struct tdb_logging_context *log_ctx,
+			 tdb_hash_func hash_fn);
+
+/**
+ * @brief Set the maximum number of dead records per hash chain.
+ *
+ * @param[in]  tdb      The database handle to set the maximum.
+ *
+ * @param[in]  max_dead The maximum number of dead records per hash chain.
+ */
+void tdb_set_max_dead(struct tdb_context *tdb, int max_dead);
+
+/**
+ * @brief Reopen a tdb.
+ *
+ * This can be used after a fork to ensure that we have an independent seek
+ * pointer from our parent and to re-establish locks.
+ *
+ * @param[in]  tdb      The database to reopen.
+ *
+ * @return              0 on success, -1 on error.
+ */
+int tdb_reopen(struct tdb_context *tdb);
+
+/**
+ * @brief Reopen all tdb's
+ *
+ * If the parent is longlived (ie. a parent daemon architecture), we know it
+ * will keep it's active lock on a tdb opened with CLEAR_IF_FIRST. Thus for
+ * child processes we don't have to add an active lock. This is essential to
+ * improve performance on systems that keep POSIX locks as a non-scalable data
+ * structure in the kernel.
+ *
+ * @param[in]  parent_longlived Wether the parent is longlived or not.
+ *
+ * @return              0 on success, -1 on error.
+ */
+int tdb_reopen_all(int parent_longlived);
+
+/**
+ * @brief Set a different tdb logging function.
+ *
+ * @param[in]  tdb      The tdb to set the logging function.
+ *
+ * @param[in]  log_ctx  The logging function to set.
+ */
+void tdb_set_logging_function(struct tdb_context *tdb, const struct tdb_logging_context *log_ctx);
+
+/**
+ * @brief Get the tdb last error code.
+ *
+ * @param[in]  tdb      The tdb to get the error code from.
+ *
+ * @return              A TDB_ERROR code.
+ *
+ * @see TDB_ERROR
+ */
+enum TDB_ERROR tdb_error(struct tdb_context *tdb);
+
+/**
+ * @brief Get a error string for the last tdb error
+ *
+ * @param[in]  tdb      The tdb to get the error code from.
+ *
+ * @return              An error string.
+ */
+const char *tdb_errorstr(struct tdb_context *tdb);
+
+/**
+ * @brief Fetch an entry in the database given a key.
+ *
+ * The caller must free the resulting data.
+ *
+ * @param[in]  tdb      The tdb to fetch the key.
+ *
+ * @param[in]  key      The key to fetch.
+ *
+ * @return              The key entry found in the database, NULL on error with
+ *                      TDB_ERROR set.
+ *
+ * @see tdb_error()
+ * @see tdb_errorstr()
+ */
+TDB_DATA tdb_fetch(struct tdb_context *tdb, TDB_DATA key);
+
+/**
+ * @brief Hand a record to a parser function without allocating it.
+ *
+ * This function is meant as a fast tdb_fetch alternative for large records
+ * that are frequently read. The "key" and "data" arguments point directly
+ * into the tdb shared memory, they are not aligned at any boundary.
+ *
+ * @warning The parser is called while tdb holds a lock on the record. DO NOT
+ * call other tdb routines from within the parser. Also, for good performance
+ * you should make the parser fast to allow parallel operations.
+ *
+ * @param[in]  tdb      The tdb to parse the record.
+ *
+ * @param[in]  key      The key to parse.
+ *
+ * @param[in]  parser   The parser to use to parse the data.
+ *
+ * @param[in]  private_data A private data pointer which is passed to the parser
+ *                          function.
+ *
+ * @return              -1 if the record was not found. If the record was found,
+ *                      the return value of "parser" is passed up to the caller.
+ */
+int tdb_parse_record(struct tdb_context *tdb, TDB_DATA key,
+			      int (*parser)(TDB_DATA key, TDB_DATA data,
+					    void *private_data),
+			      void *private_data);
+
+/**
+ * @brief Delete an entry in the database given a key.
+ *
+ * @param[in]  tdb      The tdb to delete the key.
+ *
+ * @param[in]  key      The key to delete.
+ *
+ * @return              0 on success, -1 if the key doesn't exist.
+ */
+int tdb_delete(struct tdb_context *tdb, TDB_DATA key);
+
+/**
+ * @brief Store an element in the database.
+ *
+ * This replaces any existing element with the same key.
+ *
+ * @param[in]  tdb      The tdb to store the entry.
+ *
+ * @param[in]  key      The key to use to store the entry.
+ *
+ * @param[in]  dbuf     The data to store under the key.
+ *
+ * @param[in]  flag     The flags to store the key:\n\n
+ *                      TDB_INSERT: Don't overwrite an existing entry.\n
+ *                      TDB_MODIFY: Don't create a new entry\n
+ *
+ * @return              0 on success, -1 on error with error code set.
+ *
+ * @see tdb_error()
+ * @see tdb_errorstr()
+ */
+int tdb_store(struct tdb_context *tdb, TDB_DATA key, TDB_DATA dbuf, int flag);
+
+/**
+ * @brief Append data to an entry.
+ *
+ * If the entry doesn't exist, it will create a new one.
+ *
+ * @param[in]  tdb      The database to use.
+ *
+ * @param[in]  key      The key to append the data.
+ *
+ * @param[in]  new_dbuf The data to append to the key.
+ *
+ * @return              0 on success, -1 on error with error code set.
+ *
+ * @see tdb_error()
+ * @see tdb_errorstr()
+ */
+int tdb_append(struct tdb_context *tdb, TDB_DATA key, TDB_DATA new_dbuf);
+
+/**
+ * @brief Close a database.
+ *
+ * @param[in]  tdb      The database to close.
+ *
+ * @return              0 for success, -1 on error.
+ */
+int tdb_close(struct tdb_context *tdb);
+
+/**
+ * @brief Find the first entry in the database and return its key.
+ *
+ * The caller must free the returned data.
+ *
+ * @param[in]  tdb      The database to use.
+ *
+ * @return              The first entry of the database, an empty TDB_DATA entry
+ *                      if the database is empty.
+ */
+TDB_DATA tdb_firstkey(struct tdb_context *tdb);
+
+/**
+ * @brief Find the next entry in the database, returning its key.
+ *
+ * The caller must free the returned data.
+ *
+ * @param[in]  tdb      The database to use.
+ *
+ * @param[in]  key      The key from which you want the next key.
+ *
+ * @return              The next entry of the current key, an empty TDB_DATA
+ *                      entry if there is no entry.
+ */
+TDB_DATA tdb_nextkey(struct tdb_context *tdb, TDB_DATA key);
+
+/**
+ * @brief Traverse the entire database.
+ *
+ * While travering the function fn(tdb, key, data, state) is called on each
+ * element. If fn is NULL then it is not called. A non-zero return value from
+ * fn() indicates that the traversal should stop. Traversal callbacks may not
+ * start transactions.
+ *
+ * @warning The data buffer given to the callback fn does NOT meet the alignment
+ * restrictions malloc gives you.
+ *
+ * @param[in]  tdb      The database to traverse.
+ *
+ * @param[in]  fn       The function to call on each entry.
+ *
+ * @param[in]  private_data The private data which should be passed to the
+ *                          traversing function.
+ *
+ * @return              The record count traversed, -1 on error.
+ */
+int tdb_traverse(struct tdb_context *tdb, tdb_traverse_func fn, void *private_data);
+
+/**
+ * @brief Traverse the entire database.
+ *
+ * While traversing the database the function fn(tdb, key, data, state) is
+ * called on each element, but marking the database read only during the
+ * traversal, so any write operations will fail. This allows tdb to use read
+ * locks, which increases the parallelism possible during the traversal.
+ *
+ * @param[in]  tdb      The database to traverse.
+ *
+ * @param[in]  fn       The function to call on each entry.
+ *
+ * @param[in]  private_data The private data which should be passed to the
+ *                          traversing function.
+ *
+ * @return              The record count traversed, -1 on error.
+ */
+int tdb_traverse_read(struct tdb_context *tdb, tdb_traverse_func fn, void *private_data);
+
+/**
+ * @brief Check if an entry in the database exists.
+ *
+ * @note 1 is returned if the key is found and 0 is returned if not found this
+ * doesn't match the conventions in the rest of this module, but is compatible
+ * with gdbm.
+ *
+ * @param[in]  tdb      The database to check if the entry exists.
+ *
+ * @param[in]  key      The key to check if the entry exists.
+ *
+ * @return              1 if the key is found, 0 if not.
+ */
+int tdb_exists(struct tdb_context *tdb, TDB_DATA key);
+
+/**
+ * @brief Lock entire database with a write lock.
+ *
+ * @param[in]  tdb      The database to lock.
+ *
+ * @return              0 on success, -1 on error with error code set.
+ *
+ * @see tdb_error()
+ * @see tdb_errorstr()
+ */
+int tdb_lockall(struct tdb_context *tdb);
+
+/**
+ * @brief Lock entire database with a write lock.
+ *
+ * This is the non-blocking call.
+ *
+ * @param[in]  tdb      The database to lock.
+ *
+ * @return              0 on success, -1 on error with error code set.
+ *
+ * @see tdb_lockall()
+ * @see tdb_error()
+ * @see tdb_errorstr()
+ */
+int tdb_lockall_nonblock(struct tdb_context *tdb);
+
+/**
+ * @brief Unlock entire database with write lock.
+ *
+ * @param[in]  tdb      The database to unlock.
+ *
+ * @return              0 on success, -1 on error with error code set.
+ *
+ * @see tdb_lockall()
+ * @see tdb_error()
+ * @see tdb_errorstr()
+ */
+int tdb_unlockall(struct tdb_context *tdb);
+
+/**
+ * @brief Lock entire database with a read lock.
+ *
+ * @param[in]  tdb      The database to lock.
+ *
+ * @return              0 on success, -1 on error with error code set.
+ *
+ * @see tdb_error()
+ * @see tdb_errorstr()
+ */
+int tdb_lockall_read(struct tdb_context *tdb);
+
+/**
+ * @brief Lock entire database with a read lock.
+ *
+ * This is the non-blocking call.
+ *
+ * @param[in]  tdb      The database to lock.
+ *
+ * @return              0 on success, -1 on error with error code set.
+ *
+ * @see tdb_lockall_read()
+ * @see tdb_error()
+ * @see tdb_errorstr()
+ */
+int tdb_lockall_read_nonblock(struct tdb_context *tdb);
+
+/**
+ * @brief Unlock entire database with read lock.
+ *
+ * @param[in]  tdb      The database to unlock.
+ *
+ * @return              0 on success, -1 on error with error code set.
+ *
+ * @see tdb_lockall_read()
+ * @see tdb_error()
+ * @see tdb_errorstr()
+ */
+int tdb_unlockall_read(struct tdb_context *tdb);
+
+/**
+ * @brief Lock entire database with write lock - mark only.
+ *
+ * @todo Add more details.
+ *
+ * @param[in]  tdb      The database to mark.
+ *
+ * @return              0 on success, -1 on error with error code set.
+ *
+ * @see tdb_error()
+ * @see tdb_errorstr()
+ */
+int tdb_lockall_mark(struct tdb_context *tdb);
+
+/**
+ * @brief Lock entire database with write lock - unmark only.
+ *
+ * @todo Add more details.
+ *
+ * @param[in]  tdb      The database to mark.
+ *
+ * @return              0 on success, -1 on error with error code set.
+ *
+ * @see tdb_error()
+ * @see tdb_errorstr()
+ */
+int tdb_lockall_unmark(struct tdb_context *tdb);
+
+/**
+ * @brief Get the name of the current tdb file.
+ *
+ * This is useful for external logging functions.
+ *
+ * @param[in]  tdb      The database to get the name from.
+ *
+ * @return              The name of the database.
+ */
+const char *tdb_name(struct tdb_context *tdb);
+
+/**
+ * @brief Get the underlying file descriptor being used by tdb.
+ *
+ * This is useful for external routines that want to check the device/inode
+ * of the fd.
+ *
+ * @param[in]  tdb      The database to get the fd from.
+ *
+ * @return              The file descriptor or -1.
+ */
+int tdb_fd(struct tdb_context *tdb);
+
+/**
+ * @brief Get the current logging function.
+ *
+ * This is useful for external tdb routines that wish to log tdb errors.
+ *
+ * @param[in]  tdb      The database to get the logging function from.
+ *
+ * @return              The logging function of the database.
+ *
+ * @see tdb_get_logging_private()
+ */
+tdb_log_func tdb_log_fn(struct tdb_context *tdb);
+
+/**
+ * @brief Get the private data of the logging function.
+ *
+ * @param[in]  tdb      The database to get the data from.
+ *
+ * @return              The private data pointer of the logging function.
+ *
+ * @see tdb_log_fn()
+ */
+void *tdb_get_logging_private(struct tdb_context *tdb);
+
+/**
+ * @brief Start a transaction.
+ *
+ * All operations after the transaction start can either be committed with
+ * tdb_transaction_commit() or cancelled with tdb_transaction_cancel().
+ *
+ * If you call tdb_transaction_start() again on the same tdb context while a
+ * transaction is in progress, then the same transaction buffer is re-used. The
+ * number of tdb_transaction_{commit,cancel} operations must match the number
+ * of successful tdb_transaction_start() calls.
+ *
+ * Note that transactions are by default disk synchronous, and use a recover
+ * area in the database to automatically recover the database on the next open
+ * if the system crashes during a transaction. You can disable the synchronous
+ * transaction recovery setup using the TDB_NOSYNC flag, which will greatly
+ * speed up operations at the risk of corrupting your database if the system
+ * crashes.
+ *
+ * Operations made within a transaction are not visible to other users of the
+ * database until a successful commit.
+ *
+ * @param[in]  tdb      The database to start the transaction.
+ *
+ * @return              0 on success, -1 on error with error code set.
+ *
+ * @see tdb_error()
+ * @see tdb_errorstr()
+ */
+int tdb_transaction_start(struct tdb_context *tdb);
+
+/**
+ * @brief Start a transaction, non-blocking.
+ *
+ * @param[in]  tdb      The database to start the transaction.
+ *
+ * @return              0 on success, -1 on error with error code set.
+ *
+ * @see tdb_error()
+ * @see tdb_errorstr()
+ * @see tdb_transaction_start()
+ */
+int tdb_transaction_start_nonblock(struct tdb_context *tdb);
+
+/**
+ * @brief Prepare to commit a current transaction, for two-phase commits.
+ *
+ * Once prepared for commit, the only allowed calls are tdb_transaction_commit()
+ * or tdb_transaction_cancel(). Preparing allocates disk space for the pending
+ * updates, so a subsequent commit should succeed (barring any hardware
+ * failures).
+ *
+ * @param[in]  tdb      The database to prepare the commit.
+ *
+ * @return              0 on success, -1 on error with error code set.
+ *
+ * @see tdb_error()
+ * @see tdb_errorstr()
+ */
+int tdb_transaction_prepare_commit(struct tdb_context *tdb);
+
+/**
+ * @brief Commit a current transaction.
+ *
+ * This updates the database and releases the current transaction locks.
+ *
+ * @param[in]  tdb      The database to commit the transaction.
+ *
+ * @return              0 on success, -1 on error with error code set.
+ *
+ * @see tdb_error()
+ * @see tdb_errorstr()
+ */
+int tdb_transaction_commit(struct tdb_context *tdb);
+
+/**
+ * @brief Cancel a current transaction.
+ *
+ * This discards all write and lock operations that have been made since the
+ * transaction started.
+ *
+ * @param[in]  tdb      The tdb to cancel the transaction on.
+ *
+ * @return              0 on success, -1 on error with error code set.
+ *
+ * @see tdb_error()
+ * @see tdb_errorstr()
+ */
+int tdb_transaction_cancel(struct tdb_context *tdb);
+
+/**
+ * @brief Get the tdb sequence number.
+ *
+ * Only makes sense if the writers opened with TDB_SEQNUM set. Note that this
+ * sequence number will wrap quite quickly, so it should only be used for a
+ * 'has something changed' test, not for code that relies on the count of the
+ * number of changes made. If you want a counter then use a tdb record.
+ *
+ * The aim of this sequence number is to allow for a very lightweight test of a
+ * possible tdb change.
+ *
+ * @param[in]  tdb      The database to get the sequence number from.
+ *
+ * @return              The sequence number or 0.
+ *
+ * @see tdb_open()
+ * @see tdb_enable_seqnum()
+ */
+int tdb_get_seqnum(struct tdb_context *tdb);
+
+/**
+ * @brief Get the hash size.
+ *
+ * @param[in]  tdb      The database to get the hash size from.
+ *
+ * @return              The hash size.
+ */
+int tdb_hash_size(struct tdb_context *tdb);
+
+/**
+ * @brief Get the map size.
+ *
+ * @param[in]  tdb     The database to get the map size from.
+ *
+ * @return             The map size.
+ */
+size_t tdb_map_size(struct tdb_context *tdb);
+
+/**
+ * @brief Get the tdb flags set during open.
+ *
+ * @param[in]  tdb      The database to get the flags form.
+ *
+ * @return              The flags set to on the database.
+ */
+int tdb_get_flags(struct tdb_context *tdb);
+
+/**
+ * @brief Add flags to the database.
+ *
+ * @param[in]  tdb      The database to add the flags.
+ *
+ * @param[in]  flag     The tdb flags to add.
+ */
+void tdb_add_flags(struct tdb_context *tdb, unsigned flag);
+
+/**
+ * @brief Remove flags from the database.
+ *
+ * @param[in]  tdb      The database to remove the flags.
+ *
+ * @param[in]  flag     The tdb flags to remove.
+ */
+void tdb_remove_flags(struct tdb_context *tdb, unsigned flag);
+
+/**
+ * @brief Enable sequence number handling on an open tdb.
+ *
+ * @param[in]  tdb      The database to enable sequence number handling.
+ *
+ * @see tdb_get_seqnum()
+ */
+void tdb_enable_seqnum(struct tdb_context *tdb);
+
+/**
+ * @brief Increment the tdb sequence number.
+ *
+ * This only works if the tdb has been opened using the TDB_SEQNUM flag or
+ * enabled useing tdb_enable_seqnum().
+ *
+ * @param[in]  tdb      The database to increment the sequence number.
+ *
+ * @see tdb_enable_seqnum()
+ * @see tdb_get_seqnum()
+ */
+void tdb_increment_seqnum_nonblock(struct tdb_context *tdb);
+
+/**
+ * @brief Create a hash of the key.
+ *
+ * @param[in]  key      The key to hash
+ *
+ * @return              The hash.
+ */
+unsigned int tdb_jenkins_hash(TDB_DATA *key);
+
+/**
+ * @brief Check the consistency of the database.
+ *
+ * This check the consistency of the database calling back the check function
+ * (if non-NULL) on each record.  If some consistency check fails, or the
+ * supplied check function returns -1, tdb_check returns -1, otherwise 0.
+ *
+ * @note The logging function (if set) will be called with additional
+ * information on the corruption found.
+ *
+ * @param[in]  tdb      The database to check.
+ *
+ * @param[in]  check    The check function to use.
+ *
+ * @param[in]  private_data the private data to pass to the check function.
+ *
+ * @return              0 on success, -1 on error with error code set.
+ *
+ * @see tdb_error()
+ * @see tdb_errorstr()
+ */
+int tdb_check(struct tdb_context *tdb,
+	      int (*check) (TDB_DATA key, TDB_DATA data, void *private_data),
+	      void *private_data);
+
+/**
+ * @brief Dump all possible records in a corrupt database.
+ *
+ * This is the only way to get data out of a database where tdb_check() fails.
+ * It will call walk() with anything which looks like a database record; this
+ * may well include invalid, incomplete or duplicate records.
+ *
+ * @param[in]  tdb      The database to check.
+ *
+ * @param[in]  walk     The walk function to use.
+ *
+ * @param[in]  private_data the private data to pass to the walk function.
+ *
+ * @return              0 on success, -1 on error with error code set.
+ *
+ * @see tdb_error()
+ * @see tdb_errorstr()
+ */
+int tdb_rescue(struct tdb_context *tdb,
+	       void (*walk) (TDB_DATA key, TDB_DATA data, void *private_data),
+	       void *private_data);
+
+/* @} ******************************************************************/
+
+/* Low level locking functions: use with care */
+int tdb_chainlock(struct tdb_context *tdb, TDB_DATA key);
+int tdb_chainlock_nonblock(struct tdb_context *tdb, TDB_DATA key);
+int tdb_chainunlock(struct tdb_context *tdb, TDB_DATA key);
+int tdb_chainlock_read(struct tdb_context *tdb, TDB_DATA key);
+int tdb_chainunlock_read(struct tdb_context *tdb, TDB_DATA key);
+int tdb_chainlock_mark(struct tdb_context *tdb, TDB_DATA key);
+int tdb_chainlock_unmark(struct tdb_context *tdb, TDB_DATA key);
+
+void tdb_setalarm_sigptr(struct tdb_context *tdb, volatile sig_atomic_t *sigptr);
+
+/* wipe and repack */
+int tdb_wipe_all(struct tdb_context *tdb);
+int tdb_repack(struct tdb_context *tdb);
+
+/* Debug functions. Not used in production. */
+void tdb_dump_all(struct tdb_context *tdb);
+int tdb_printfreelist(struct tdb_context *tdb);
+int tdb_validate_freelist(struct tdb_context *tdb, int *pnum_entries);
+int tdb_freelist_size(struct tdb_context *tdb);
+char *tdb_summary(struct tdb_context *tdb);
+
+extern TDB_DATA tdb_null;
+
+#ifdef  __cplusplus
+}
+#endif
+
+#endif /* tdb.h */
diff --git a/ctdb/lib/tdb/libtdb.m4 b/ctdb/lib/tdb/libtdb.m4
new file mode 100644
index 0000000000..6503bf15f4
--- /dev/null
+++ b/ctdb/lib/tdb/libtdb.m4
@@ -0,0 +1,59 @@
+dnl Check to see if we should use the included tdb
+
+INCLUDED_TDB=auto
+AC_ARG_WITH(included-tdb,
+    [AC_HELP_STRING([--with-included-tdb], [use bundled tdb library, not from system])],
+    [ INCLUDED_TDB=$withval ])
+
+AC_SUBST(TDB_LIBS)
+AC_SUBST(TDB_CFLAGS)
+
+if test x"$INCLUDED_TDB" != x"yes" ; then
+    AC_CHECK_HEADERS(tdb.h)
+    AC_CHECK_LIB(tdb, tdb_transaction_write_lock_mark, [ TDB_LIBS="-ltdb" ])
+    if test x"$ac_cv_header_tdb_h" = x"no" -o x"$ac_cv_lib_tdb_tdb_transaction_write_lock_mark" = x"no" ; then
+        INCLUDED_TDB=yes
+        TDB_CFLAGS=""
+    else
+        INCLUDED_TDB=no
+    fi
+fi
+
+AC_MSG_CHECKING(whether to use included tdb)
+AC_MSG_RESULT($INCLUDED_TDB)
+if test x"$INCLUDED_TDB" != x"no" ; then
+    dnl find the tdb sources. This is meant to work both for 
+    dnl tdb standalone builds, and builds of packages using tdb
+    tdbdir=""
+    tdbpaths=". lib/tdb tdb ../tdb ../lib/tdb"
+    for d in $tdbpaths; do
+    	if test -f "$srcdir/$d/common/tdb.c"; then
+    		tdbdir="$d"		
+    		AC_SUBST(tdbdir)
+    		break;
+    	fi
+    done
+    if test x"$tdbdir" = "x"; then
+       AC_MSG_ERROR([cannot find tdb source in $tdbpaths])
+    fi
+    TDB_OBJ="common/tdb.o common/dump.o common/transaction.o common/error.o common/traverse.o"
+    TDB_OBJ="$TDB_OBJ common/freelist.o common/freelistcheck.o common/io.o common/lock.o common/open.o common/check.o common/hash.o common/summary.o common/rescue.o"
+    AC_SUBST(TDB_OBJ)
+
+    TDB_LIBS=""
+    AC_SUBST(TDB_LIBS)
+
+    TDB_CFLAGS="-I$tdbdir/include"
+    AC_SUBST(TDB_CFLAGS)
+fi
+
+AC_CHECK_FUNCS(mmap pread pwrite getpagesize utime)
+AC_CHECK_HEADERS(getopt.h sys/select.h sys/time.h)
+
+AC_HAVE_DECL(pread, [#include <unistd.h>])
+AC_HAVE_DECL(pwrite, [#include <unistd.h>])
+
+if test x"$VERSIONSCRIPT" != "x"; then
+    EXPORTSFILE=tdb.exports
+    AC_SUBST(EXPORTSFILE)
+fi
diff --git a/ctdb/lib/tdb/manpages/tdbbackup.8.xml b/ctdb/lib/tdb/manpages/tdbbackup.8.xml
new file mode 100644
index 0000000000..78fe32eb8e
--- /dev/null
+++ b/ctdb/lib/tdb/manpages/tdbbackup.8.xml
@@ -0,0 +1,136 @@
+<?xml version="1.0" encoding="iso-8859-1"?>
+<!DOCTYPE refentry PUBLIC "-//OASIS//DTD DocBook XML V4.2//EN" "http://www.oasis-open.org/docbook/xml/4.2/docbookx.dtd">
+<refentry id="tdbbackup.8">
+
+<refmeta>
+	<refentrytitle>tdbbackup</refentrytitle>
+	<manvolnum>8</manvolnum>
+	<refmiscinfo class="source">Samba</refmiscinfo>
+	<refmiscinfo class="manual">System Administration tools</refmiscinfo>
+	<refmiscinfo class="version">3.6</refmiscinfo>
+</refmeta>
+
+
+<refnamediv>
+	<refname>tdbbackup</refname>
+	<refpurpose>tool for backing up and for validating the integrity of samba .tdb files</refpurpose>
+</refnamediv>
+
+<refsynopsisdiv>
+	<cmdsynopsis>
+		<command>tdbbackup</command>
+		<arg choice="opt">-s suffix</arg>
+		<arg choice="opt">-v</arg>
+		<arg choice="opt">-h</arg>
+	</cmdsynopsis>
+</refsynopsisdiv>
+
+<refsect1>
+	<title>DESCRIPTION</title>
+
+	<para>This tool is part of the <citerefentry><refentrytitle>samba</refentrytitle>
+	<manvolnum>1</manvolnum></citerefentry> suite.</para>
+
+	<para><command>tdbbackup</command> is a tool that may be used to backup samba .tdb
+	files. This tool may also be used to verify the integrity of the .tdb files prior
+	to samba startup or during normal operation. If it finds file damage and it finds 
+	a prior backup the backup file will be restored. 
+	</para>
+</refsect1>
+
+
+<refsect1>
+	<title>OPTIONS</title>
+
+	<variablelist>
+
+		<varlistentry>
+		<term>-h</term>
+		<listitem><para>
+		Get help information.
+		</para></listitem>
+		</varlistentry>
+
+		<varlistentry>
+		<term>-s suffix</term>
+		<listitem><para>
+		The <command>-s</command> option allows the adminisistrator to specify a file
+		backup extension. This way it is possible to keep a history of tdb backup
+		files by using a new suffix for each backup.
+		</para> </listitem>
+		</varlistentry>
+
+		<varlistentry>
+		<term>-v</term>
+		<listitem><para>
+		The <command>-v</command> will check the database for damages (currupt data)
+		which if detected causes the backup to be restored.
+		</para></listitem>
+		</varlistentry>
+
+	</variablelist>
+</refsect1>
+
+
+<refsect1>
+	<title>COMMANDS</title>
+
+	<para><emphasis>GENERAL INFORMATION</emphasis></para>
+
+	<para>
+	The <command>tdbbackup</command> utility can safely be run at any time. It was designed so
+	that it can be used at any time to validate the integrity of tdb files, even during Samba
+	operation. Typical usage for the command will be:
+	</para>
+
+	<para>tdbbackup [-s suffix] *.tdb</para>
+
+	<para>
+	Before restarting samba the following command may be run to validate .tdb files:
+	</para>
+
+	<para>tdbbackup -v [-s suffix] *.tdb</para>
+
+	<para>
+	Samba .tdb files are stored in various locations, be sure to run backup all
+	.tdb file on the system. Important files includes:
+	</para>
+
+	<itemizedlist>
+		<listitem><para>
+		<command>secrets.tdb</command> - usual location is in the /usr/local/samba/private
+		directory, or on some systems in /etc/samba.
+		</para></listitem>
+
+		<listitem><para>
+		<command>passdb.tdb</command> - usual location is in the /usr/local/samba/private
+		directory, or on some systems in /etc/samba.
+		</para></listitem>
+
+		<listitem><para>
+		<command>*.tdb</command> located in the /usr/local/samba/var directory or on some
+		systems in the /var/cache or /var/lib/samba directories.
+		</para></listitem>
+	</itemizedlist>
+
+</refsect1>
+
+<refsect1>
+	<title>VERSION</title>
+
+	<para>This man page is correct for version 3 of the Samba suite.</para>
+</refsect1>
+
+<refsect1>
+	<title>AUTHOR</title>
+
+	<para>
+	The original Samba software and related utilities were created by Andrew Tridgell.
+	Samba is now developed by the Samba Team as an Open Source project similar to the way
+	the Linux kernel is developed.
+	</para> 
+
+	<para>The tdbbackup man page was written by John H Terpstra.</para>
+</refsect1>
+
+</refentry>
diff --git a/ctdb/lib/tdb/manpages/tdbdump.8.xml b/ctdb/lib/tdb/manpages/tdbdump.8.xml
new file mode 100644
index 0000000000..3420193243
--- /dev/null
+++ b/ctdb/lib/tdb/manpages/tdbdump.8.xml
@@ -0,0 +1,92 @@
+<?xml version="1.0" encoding="iso-8859-1"?>
+<!DOCTYPE refentry PUBLIC "-//OASIS//DTD DocBook XML V4.2//EN" "http://www.oasis-open.org/docbook/xml/4.2/docbookx.dtd">
+<refentry id="tdbdump.8">
+
+<refmeta>
+	<refentrytitle>tdbdump</refentrytitle>
+	<manvolnum>8</manvolnum>
+	<refmiscinfo class="source">Samba</refmiscinfo>
+	<refmiscinfo class="manual">System Administration tools</refmiscinfo>
+	<refmiscinfo class="version">3.6</refmiscinfo>
+</refmeta>
+
+
+<refnamediv>
+	<refname>tdbdump</refname>
+	<refpurpose>tool for printing the contents of a TDB file</refpurpose>
+</refnamediv>
+
+<refsynopsisdiv>
+	<cmdsynopsis>
+		<command>tdbdump</command>
+		<arg choice="opt">-k <replaceable>keyname</replaceable></arg>
+		<arg choice="opt">-e</arg>
+		<arg choice="opt">-h</arg>
+		<arg choice="req">filename</arg>
+	</cmdsynopsis>
+</refsynopsisdiv>
+
+<refsect1>
+	<title>DESCRIPTION</title>
+
+	<para>This tool is part of the <citerefentry><refentrytitle>samba</refentrytitle>
+	<manvolnum>1</manvolnum></citerefentry> suite.</para>
+
+	<para><command>tdbdump</command> is a very simple utility that 'dumps' the 
+		contents of a TDB (Trivial DataBase) file to standard output in a 
+		human-readable format.
+	</para>
+
+	<para>This tool can be used when debugging problems with TDB files. It is 
+		intended for those who are somewhat familiar with Samba internals.
+	</para>
+</refsect1>
+
+<refsect1>
+	<title>OPTIONS</title>
+
+	<variablelist>
+
+		<varlistentry>
+		<term>-h</term>
+		<listitem><para>
+		Get help information.
+		</para></listitem>
+		</varlistentry>
+
+		<varlistentry>
+		<term>-k <replaceable>keyname</replaceable></term>
+		<listitem><para>
+		The <command>-k</command> option restricts dumping to a single key, if found.
+		</para> </listitem>
+		</varlistentry>
+
+		<varlistentry>
+		<term>-e</term>
+		<listitem><para>
+		The <command>-e</command> tries to dump out from a corrupt database.  Naturally, such a dump is unreliable, at best.
+		</para></listitem>
+		</varlistentry>
+
+	</variablelist>
+</refsect1>
+
+<refsect1>
+	<title>VERSION</title>
+
+	<para>This man page is correct for version 3 of the Samba suite.</para>
+</refsect1>
+
+<refsect1>
+	<title>AUTHOR</title>
+
+	<para>
+	The original Samba software and related utilities were created by Andrew Tridgell.
+	Samba is now developed by the Samba Team as an Open Source project similar to the way
+	the Linux kernel is developed.
+	</para> 
+
+	<para>The tdbdump man page was written by Jelmer Vernooij.</para>
+</refsect1>
+
+</refentry>
diff --git a/ctdb/lib/tdb/manpages/tdbrestore.8.xml b/ctdb/lib/tdb/manpages/tdbrestore.8.xml
new file mode 100644
index 0000000000..64c0ba2dac
--- /dev/null
+++ b/ctdb/lib/tdb/manpages/tdbrestore.8.xml
@@ -0,0 +1,66 @@
+<?xml version="1.0" encoding="iso-8859-1"?>
+<!DOCTYPE refentry PUBLIC "-//OASIS//DTD DocBook XML V4.2//EN" "http://www.oasis-open.org/docbook/xml/4.2/docbookx.dtd">
+<refentry id="tdbrestore.8">
+
+<refmeta>
+	<refentrytitle>tdbrestore</refentrytitle>
+	<manvolnum>8</manvolnum>
+	<refmiscinfo class="source">Samba</refmiscinfo>
+	<refmiscinfo class="manual">System Administration tools</refmiscinfo>
+	<refmiscinfo class="version">3.6</refmiscinfo>
+</refmeta>
+
+
+<refnamediv>
+	<refname>tdbrestore</refname>
+	<refpurpose>tool for creating a TDB file out of a tdbdump output</refpurpose>
+</refnamediv>
+
+<refsynopsisdiv>
+	<cmdsynopsis>
+		<command>tdbrestore</command>
+		<arg choice="req">tdbfilename</arg>
+	</cmdsynopsis>
+</refsynopsisdiv>
+
+<refsect1>
+	<title>DESCRIPTION</title>
+
+	<para>This tool is part of the <citerefentry><refentrytitle>samba</refentrytitle>
+	<manvolnum>1</manvolnum></citerefentry> suite.</para>
+
+	<para><command>tdbrestore</command> is a very simple utility that 'restores' the
+		contents of dump file into TDB (Trivial DataBase) file. The dump file is obtained from the tdbdump
+		command.
+	</para>
+
+	<para>This tool wait on the standard input for the content of the dump and will write the tdb in the tdbfilename
+  parameter.
+	</para>
+	<para>This tool can be used for unpacking the content of tdb as backup mean.
+	</para>
+</refsect1>
+
+
+<refsect1>
+	<title>VERSION</title>
+
+	<para>This man page is correct for version 3 of the Samba suite.</para>
+</refsect1>
+
+<refsect1>
+	<title>AUTHOR</title>
+
+	<para>
+	The original Samba software and related utilities were created by Andrew Tridgell.
+	Samba is now developed by the Samba Team as an Open Source project similar to the way
+	the Linux kernel is developed.
+
+        This tool was initially written by Volker Lendecke based on an
+        idea by Simon McVittie.
+	</para>
+
+	<para>The tdbrestore man page was written by Matthieu Patou.</para>
+</refsect1>
+
+</refentry>
diff --git a/ctdb/lib/tdb/manpages/tdbtool.8.xml b/ctdb/lib/tdb/manpages/tdbtool.8.xml
new file mode 100644
index 0000000000..9f96db277d
--- /dev/null
+++ b/ctdb/lib/tdb/manpages/tdbtool.8.xml
@@ -0,0 +1,235 @@
+<?xml version="1.0" encoding="iso-8859-1"?>
+<!DOCTYPE refentry PUBLIC "-//OASIS//DTD DocBook XML V4.2//EN" "http://www.oasis-open.org/docbook/xml/4.2/docbookx.dtd">
+<refentry id="tdbtool.8">
+
+<refmeta>
+	<refentrytitle>tdbtool</refentrytitle>
+	<manvolnum>8</manvolnum>
+	<refmiscinfo class="source">Samba</refmiscinfo>
+	<refmiscinfo class="manual">System Administration tools</refmiscinfo>
+	<refmiscinfo class="version">3.6</refmiscinfo>
+</refmeta>
+
+
+<refnamediv>
+	<refname>tdbtool</refname>
+	<refpurpose>manipulate the contents TDB files</refpurpose>
+</refnamediv>
+
+<refsynopsisdiv>
+
+	<cmdsynopsis>
+		<command>tdbtool</command>
+	</cmdsynopsis>
+
+	<cmdsynopsis>
+		<command>tdbtool</command>
+		<arg choice="plain">
+		<replaceable>TDBFILE</replaceable>
+		</arg>
+		<arg rep="repeat" choice="opt">
+		<replaceable>COMMANDS</replaceable>
+		</arg>
+	</cmdsynopsis>
+
+</refsynopsisdiv>
+
+<refsect1>
+	<title>DESCRIPTION</title>
+
+	<para>This tool is part of the
+	<citerefentry><refentrytitle>samba</refentrytitle>
+	<manvolnum>1</manvolnum></citerefentry> suite.</para>
+
+	<para><command>tdbtool</command> a tool for displaying and
+	altering the contents of Samba TDB (Trivial DataBase) files. Each
+	of the commands listed below can be entered interactively or
+	provided on the command line.</para>
+
+</refsect1>
+
+
+<refsect1>
+	<title>COMMANDS</title>
+
+	<variablelist>
+
+		<varlistentry>
+		<term><option>create</option>
+		<replaceable>TDBFILE</replaceable></term>
+		<listitem><para>Create a new database named
+		<replaceable>TDBFILE</replaceable>.
+		</para></listitem>
+		</varlistentry>
+
+		<varlistentry>
+		<term><option>open</option>
+		<replaceable>TDBFILE</replaceable></term>
+		<listitem><para>Open an existing database named
+		<replaceable>TDBFILE</replaceable>.
+		</para></listitem>
+		</varlistentry>
+
+		<varlistentry>
+		<term><option>erase</option></term>
+		<listitem><para>Erase the current database.
+		</para></listitem>
+		</varlistentry>
+
+		<varlistentry>
+		<term><option>dump</option></term>
+		<listitem><para>Dump the current database as strings.
+		</para></listitem>
+		</varlistentry>
+
+		<varlistentry>
+		<term><option>cdump</option></term>
+		<listitem><para>Dump the current database as connection records.
+		</para></listitem>
+		</varlistentry>
+
+		<varlistentry>
+		<term><option>keys</option></term>
+		<listitem><para>Dump the current database keys as strings.
+		</para></listitem>
+		</varlistentry>
+
+		<varlistentry>
+		<term><option>hexkeys</option></term>
+		<listitem><para>Dump the current database keys as hex values.
+		</para></listitem>
+		</varlistentry>
+
+		<varlistentry>
+		<term><option>info</option></term>
+		<listitem><para>Print summary information about the
+		current database.
+		</para></listitem>
+		</varlistentry>
+
+		<varlistentry>
+		<term><option>insert</option>
+		<replaceable>KEY</replaceable>
+		<replaceable>DATA</replaceable>
+		</term>
+		<listitem><para>Insert a record into the
+		current database.
+		</para></listitem>
+		</varlistentry>
+
+		<varlistentry>
+		<term><option>move</option>
+		<replaceable>KEY</replaceable>
+		<replaceable>TDBFILE</replaceable>
+		</term>
+		<listitem><para>Move a record from the 
+		current database into <replaceable>TDBFILE</replaceable>.
+		</para></listitem>
+		</varlistentry>
+
+		<varlistentry>
+		<term><option>store</option>
+		<replaceable>KEY</replaceable>
+		<replaceable>DATA</replaceable>
+		</term>
+		<listitem><para>Store (replace) a record in the
+		current database.
+		</para></listitem>
+		</varlistentry>
+
+		<varlistentry>
+		<term><option>show</option>
+		<replaceable>KEY</replaceable>
+		</term>
+		<listitem><para>Show a record by key.
+		</para></listitem>
+		</varlistentry>
+
+		<varlistentry>
+		<term><option>delete</option>
+		<replaceable>KEY</replaceable>
+		</term>
+		<listitem><para>Delete a record by key.
+		</para></listitem>
+		</varlistentry>
+
+		<varlistentry>
+		<term><option>list</option>
+		</term>
+		<listitem><para>Print the current database hash table and free list.
+		</para></listitem>
+		</varlistentry>
+
+		<varlistentry>
+		<term><option>free</option>
+		</term>
+		<listitem><para>Print the current database and free list.
+		</para></listitem>
+		</varlistentry>
+
+		<varlistentry>
+		<term><option>!</option>
+		<replaceable>COMMAND</replaceable>
+		</term>
+		<listitem><para>Execute the given system command.
+		</para></listitem>
+		</varlistentry>
+
+		<varlistentry>
+		<term>
+		<option>first</option>
+		</term>
+		<listitem><para>Print the first record in the current database.
+		</para></listitem>
+		</varlistentry>
+
+		<varlistentry>
+		<term>
+		<option>next</option>
+		</term>
+		<listitem><para>Print the next record in the current database.
+		</para></listitem>
+		</varlistentry>
+
+		<varlistentry>
+		<term>
+		<option>check</option>
+		</term>
+		<listitem><para>Check the integrity of the current database.
+		</para></listitem>
+		</varlistentry>
+
+		<varlistentry>
+		<term>
+		<option>quit</option>
+		</term>
+		<listitem><para>Exit <command>tdbtool</command>.
+		</para></listitem>
+		</varlistentry>
+
+	</variablelist>
+</refsect1>
+
+<refsect1>
+	<title>CAVEATS</title>
+	<para>The contents of the Samba TDB files are private
+	to the implementation and should not be altered with
+	<command>tdbtool</command>.
+	</para>
+</refsect1>
+
+<refsect1>
+	<title>VERSION</title>
+	<para>This man page is correct for version 3.0.25 of the Samba suite.</para>
+</refsect1>
+
+<refsect1>
+	<title>AUTHOR</title>
+
+	<para> The original Samba software and related utilities were
+	created by Andrew Tridgell.  Samba is now developed by the
+	Samba Team as an Open Source project similar to the way the
+	Linux kernel is developed.</para>
+</refsect1>
+
+</refentry>
diff --git a/ctdb/lib/tdb/pytdb.c b/ctdb/lib/tdb/pytdb.c
new file mode 100644
index 0000000000..ae0e6f808d
--- /dev/null
+++ b/ctdb/lib/tdb/pytdb.c
@@ -0,0 +1,689 @@
+/* 
+   Unix SMB/CIFS implementation.
+
+   Python interface to tdb.
+
+   Copyright (C) 2004-2006 Tim Potter <tpot@samba.org>
+   Copyright (C) 2007-2008 Jelmer Vernooij <jelmer@samba.org>
+
+     ** NOTE! The following LGPL license applies to the tdb
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include <Python.h>
+#include "replace.h"
+#include "system/filesys.h"
+
+#ifndef Py_RETURN_NONE
+#define Py_RETURN_NONE return Py_INCREF(Py_None), Py_None
+#endif
+
+/* Include tdb headers */
+#include <tdb.h>
+
+typedef struct {
+	PyObject_HEAD
+	TDB_CONTEXT *ctx;
+	bool closed;
+} PyTdbObject;
+
+staticforward PyTypeObject PyTdb;
+
+static void PyErr_SetTDBError(TDB_CONTEXT *tdb)
+{
+	PyErr_SetObject(PyExc_RuntimeError, 
+		Py_BuildValue("(i,s)", tdb_error(tdb), tdb_errorstr(tdb)));
+}
+
+static TDB_DATA PyString_AsTDB_DATA(PyObject *data)
+{
+	TDB_DATA ret;
+	ret.dptr = (unsigned char *)PyString_AsString(data);
+	ret.dsize = PyString_Size(data);
+	return ret;
+}
+
+static PyObject *PyString_FromTDB_DATA(TDB_DATA data)
+{
+	if (data.dptr == NULL && data.dsize == 0) {
+		Py_RETURN_NONE;
+	} else {
+		PyObject *ret = PyString_FromStringAndSize((const char *)data.dptr, 
+												   data.dsize);
+		free(data.dptr);
+		return ret;
+    }
+}
+
+#define PyErr_TDB_ERROR_IS_ERR_RAISE(ret, tdb) \
+	if (ret != 0) { \
+		PyErr_SetTDBError(tdb); \
+		return NULL; \
+	}
+
+#define PyErr_TDB_RAISE_IF_CLOSED(self) \
+	if (self->closed) {						\
+	        PyErr_SetObject(PyExc_RuntimeError,				\
+				Py_BuildValue("(i,s)", TDB_ERR_IO, "Database is already closed")); \
+		return NULL;						\
+	}
+
+#define PyErr_TDB_RAISE_RETURN_MINUS_1_IF_CLOSED(self) \
+	if (self->closed) {						\
+	        PyErr_SetObject(PyExc_RuntimeError,				\
+				Py_BuildValue("(i,s)", TDB_ERR_IO, "Database is already closed")); \
+		return -1;						\
+	}
+
+static PyObject *py_tdb_open(PyTypeObject *type, PyObject *args, PyObject *kwargs)
+{
+	char *name = NULL;
+	int hash_size = 0, tdb_flags = TDB_DEFAULT, flags = O_RDWR, mode = 0600;
+	TDB_CONTEXT *ctx;
+	PyTdbObject *ret;
+	const char *kwnames[] = { "name", "hash_size", "tdb_flags", "flags", "mode", NULL };
+
+	if (!PyArg_ParseTupleAndKeywords(args, kwargs, "|siiii", (char **)kwnames, &name, &hash_size, &tdb_flags, &flags, &mode))
+		return NULL;
+
+	if (name == NULL) {
+		tdb_flags |= TDB_INTERNAL;
+	}
+
+	ctx = tdb_open(name, hash_size, tdb_flags, flags, mode);
+	if (ctx == NULL) {
+		PyErr_SetFromErrno(PyExc_IOError);
+		return NULL;
+	}
+
+	ret = PyObject_New(PyTdbObject, &PyTdb);
+	if (!ret) {
+		tdb_close(ctx);
+		return NULL;
+	}
+
+	ret->ctx = ctx;
+	ret->closed = false;
+	return (PyObject *)ret;
+}
+
+static PyObject *obj_transaction_cancel(PyTdbObject *self)
+{
+	int ret;
+
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+
+	ret = tdb_transaction_cancel(self->ctx);
+	PyErr_TDB_ERROR_IS_ERR_RAISE(ret, self->ctx);
+	Py_RETURN_NONE;
+}
+
+static PyObject *obj_transaction_commit(PyTdbObject *self)
+{
+	int ret;
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+	ret = tdb_transaction_commit(self->ctx);
+	PyErr_TDB_ERROR_IS_ERR_RAISE(ret, self->ctx);
+	Py_RETURN_NONE;
+}
+
+static PyObject *obj_transaction_prepare_commit(PyTdbObject *self)
+{
+	int ret;
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+	ret = tdb_transaction_prepare_commit(self->ctx);
+	PyErr_TDB_ERROR_IS_ERR_RAISE(ret, self->ctx);
+	Py_RETURN_NONE;
+}
+
+static PyObject *obj_transaction_start(PyTdbObject *self)
+{
+	int ret;
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+	ret = tdb_transaction_start(self->ctx);
+	PyErr_TDB_ERROR_IS_ERR_RAISE(ret, self->ctx);
+	Py_RETURN_NONE;
+}
+
+static PyObject *obj_reopen(PyTdbObject *self)
+{
+	int ret;
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+	ret = tdb_reopen(self->ctx);
+	PyErr_TDB_ERROR_IS_ERR_RAISE(ret, self->ctx);
+	Py_RETURN_NONE;
+}
+
+static PyObject *obj_lockall(PyTdbObject *self)
+{
+	int ret;
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+	ret = tdb_lockall(self->ctx);
+	PyErr_TDB_ERROR_IS_ERR_RAISE(ret, self->ctx);
+	Py_RETURN_NONE;
+}
+
+static PyObject *obj_unlockall(PyTdbObject *self)
+{
+	int ret;
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+	ret = tdb_unlockall(self->ctx);
+	PyErr_TDB_ERROR_IS_ERR_RAISE(ret, self->ctx);
+	Py_RETURN_NONE;
+}
+
+static PyObject *obj_lockall_read(PyTdbObject *self)
+{
+	int ret;
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+	ret = tdb_lockall_read(self->ctx);
+	PyErr_TDB_ERROR_IS_ERR_RAISE(ret, self->ctx);
+	Py_RETURN_NONE;
+}
+
+static PyObject *obj_unlockall_read(PyTdbObject *self)
+{
+	int ret = tdb_unlockall_read(self->ctx);
+	PyErr_TDB_ERROR_IS_ERR_RAISE(ret, self->ctx);
+	Py_RETURN_NONE;
+}
+
+static PyObject *obj_close(PyTdbObject *self)
+{
+	int ret;
+	if (self->closed)
+		Py_RETURN_NONE;
+	ret = tdb_close(self->ctx);
+	self->closed = true;
+	PyErr_TDB_ERROR_IS_ERR_RAISE(ret, self->ctx);
+	Py_RETURN_NONE;
+}
+
+static PyObject *obj_get(PyTdbObject *self, PyObject *args)
+{
+	TDB_DATA key;
+	PyObject *py_key;
+
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+
+	if (!PyArg_ParseTuple(args, "O", &py_key))
+		return NULL;
+
+	key = PyString_AsTDB_DATA(py_key);
+	if (!key.dptr)
+		return NULL;
+
+	return PyString_FromTDB_DATA(tdb_fetch(self->ctx, key));
+}
+
+static PyObject *obj_append(PyTdbObject *self, PyObject *args)
+{
+	TDB_DATA key, data;
+	PyObject *py_key, *py_data;
+	int ret;
+
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+
+	if (!PyArg_ParseTuple(args, "OO", &py_key, &py_data))
+		return NULL;
+
+	key = PyString_AsTDB_DATA(py_key);
+	if (!key.dptr)
+		return NULL;
+	data = PyString_AsTDB_DATA(py_data);
+	if (!data.dptr)
+		return NULL;
+
+	ret = tdb_append(self->ctx, key, data);
+	PyErr_TDB_ERROR_IS_ERR_RAISE(ret, self->ctx);
+	Py_RETURN_NONE;
+}
+
+static PyObject *obj_firstkey(PyTdbObject *self)
+{
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+
+	return PyString_FromTDB_DATA(tdb_firstkey(self->ctx));
+}
+
+static PyObject *obj_nextkey(PyTdbObject *self, PyObject *args)
+{
+	TDB_DATA key;
+	PyObject *py_key;
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+
+	if (!PyArg_ParseTuple(args, "O", &py_key))
+		return NULL;
+
+	key = PyString_AsTDB_DATA(py_key);
+	if (!key.dptr)
+		return NULL;
+	
+	return PyString_FromTDB_DATA(tdb_nextkey(self->ctx, key));
+}
+
+static PyObject *obj_delete(PyTdbObject *self, PyObject *args)
+{
+	TDB_DATA key;
+	PyObject *py_key;
+	int ret;
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+
+	if (!PyArg_ParseTuple(args, "O", &py_key))
+		return NULL;
+
+	key = PyString_AsTDB_DATA(py_key);
+	if (!key.dptr)
+		return NULL;
+	ret = tdb_delete(self->ctx, key);
+	PyErr_TDB_ERROR_IS_ERR_RAISE(ret, self->ctx);
+	Py_RETURN_NONE;
+}
+
+static PyObject *obj_has_key(PyTdbObject *self, PyObject *args)
+{
+	TDB_DATA key;
+	int ret;
+	PyObject *py_key;
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+
+	if (!PyArg_ParseTuple(args, "O", &py_key))
+		return NULL;
+
+	key = PyString_AsTDB_DATA(py_key);
+	if (!key.dptr)
+		return NULL;
+	ret = tdb_exists(self->ctx, key);
+	if (ret != TDB_ERR_NOEXIST) {
+		PyErr_TDB_ERROR_IS_ERR_RAISE(ret, self->ctx);
+	}
+
+	return (ret == TDB_ERR_NOEXIST)?Py_False:Py_True;
+}
+
+static PyObject *obj_store(PyTdbObject *self, PyObject *args)
+{
+	TDB_DATA key, value;
+	int ret;
+	int flag = TDB_REPLACE;
+	PyObject *py_key, *py_value;
+
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+
+	if (!PyArg_ParseTuple(args, "OO|i", &py_key, &py_value, &flag))
+		return NULL;
+
+	key = PyString_AsTDB_DATA(py_key);
+	if (!key.dptr)
+		return NULL;
+	value = PyString_AsTDB_DATA(py_value);
+	if (!value.dptr)
+		return NULL;
+
+	ret = tdb_store(self->ctx, key, value, flag);
+	PyErr_TDB_ERROR_IS_ERR_RAISE(ret, self->ctx);
+	Py_RETURN_NONE;
+}
+
+static PyObject *obj_add_flags(PyTdbObject *self, PyObject *args)
+{
+	unsigned flags;
+
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+
+	if (!PyArg_ParseTuple(args, "I", &flags))
+		return NULL;
+
+	tdb_add_flags(self->ctx, flags);
+	Py_RETURN_NONE;
+}
+
+static PyObject *obj_remove_flags(PyTdbObject *self, PyObject *args)
+{
+	unsigned flags;
+
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+
+	if (!PyArg_ParseTuple(args, "I", &flags))
+		return NULL;
+
+	tdb_remove_flags(self->ctx, flags);
+	Py_RETURN_NONE;
+}
+
+typedef struct {
+	PyObject_HEAD
+	TDB_DATA current;
+	PyTdbObject *iteratee;
+} PyTdbIteratorObject;
+
+static PyObject *tdb_iter_next(PyTdbIteratorObject *self)
+{
+	TDB_DATA current;
+	PyObject *ret;
+	if (self->current.dptr == NULL && self->current.dsize == 0)
+		return NULL;
+	current = self->current;
+	self->current = tdb_nextkey(self->iteratee->ctx, self->current);
+	ret = PyString_FromTDB_DATA(current);
+	return ret;
+}
+
+static void tdb_iter_dealloc(PyTdbIteratorObject *self)
+{
+	Py_DECREF(self->iteratee);
+	PyObject_Del(self);
+}
+
+PyTypeObject PyTdbIterator = {
+	.tp_name = "Iterator",
+	.tp_basicsize = sizeof(PyTdbIteratorObject),
+	.tp_iternext = (iternextfunc)tdb_iter_next,
+	.tp_dealloc = (destructor)tdb_iter_dealloc,
+	.tp_flags = Py_TPFLAGS_DEFAULT,
+	.tp_iter = PyObject_SelfIter,
+};
+
+static PyObject *tdb_object_iter(PyTdbObject *self)
+{
+	PyTdbIteratorObject *ret;	
+
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+
+	ret = PyObject_New(PyTdbIteratorObject, &PyTdbIterator);
+	if (!ret)
+		return NULL;
+	ret->current = tdb_firstkey(self->ctx);
+	ret->iteratee = self;
+	Py_INCREF(self);
+	return (PyObject *)ret;
+}
+
+static PyObject *obj_clear(PyTdbObject *self)
+{
+	int ret;
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+	ret = tdb_wipe_all(self->ctx);
+	PyErr_TDB_ERROR_IS_ERR_RAISE(ret, self->ctx);
+	Py_RETURN_NONE;
+}
+
+static PyObject *obj_repack(PyTdbObject *self)
+{
+	int ret;
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+	ret = tdb_repack(self->ctx);
+	PyErr_TDB_ERROR_IS_ERR_RAISE(ret, self->ctx);
+	Py_RETURN_NONE;
+}
+
+static PyObject *obj_enable_seqnum(PyTdbObject *self)
+{
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+	tdb_enable_seqnum(self->ctx);
+	Py_RETURN_NONE;
+}
+
+static PyObject *obj_increment_seqnum_nonblock(PyTdbObject *self)
+{
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+	tdb_increment_seqnum_nonblock(self->ctx);
+	Py_RETURN_NONE;
+}
+
+static PyMethodDef tdb_object_methods[] = {
+	{ "transaction_cancel", (PyCFunction)obj_transaction_cancel, METH_NOARGS, 
+		"S.transaction_cancel() -> None\n"
+		"Cancel the currently active transaction." },
+	{ "transaction_commit", (PyCFunction)obj_transaction_commit, METH_NOARGS,
+		"S.transaction_commit() -> None\n"
+		"Commit the currently active transaction." },
+	{ "transaction_prepare_commit", (PyCFunction)obj_transaction_prepare_commit, METH_NOARGS,
+		"S.transaction_prepare_commit() -> None\n"
+		"Prepare to commit the currently active transaction" },
+	{ "transaction_start", (PyCFunction)obj_transaction_start, METH_NOARGS,
+		"S.transaction_start() -> None\n"
+		"Start a new transaction." },
+	{ "reopen", (PyCFunction)obj_reopen, METH_NOARGS, "Reopen this file." },
+	{ "lock_all", (PyCFunction)obj_lockall, METH_NOARGS, NULL },
+	{ "unlock_all", (PyCFunction)obj_unlockall, METH_NOARGS, NULL },
+	{ "read_lock_all", (PyCFunction)obj_lockall_read, METH_NOARGS, NULL },
+	{ "read_unlock_all", (PyCFunction)obj_unlockall_read, METH_NOARGS, NULL },
+	{ "close", (PyCFunction)obj_close, METH_NOARGS, NULL },
+	{ "get", (PyCFunction)obj_get, METH_VARARGS, "S.get(key) -> value\n"
+		"Fetch a value." },
+	{ "append", (PyCFunction)obj_append, METH_VARARGS, "S.append(key, value) -> None\n"
+		"Append data to an existing key." },
+	{ "firstkey", (PyCFunction)obj_firstkey, METH_NOARGS, "S.firstkey() -> data\n"
+		"Return the first key in this database." },
+	{ "nextkey", (PyCFunction)obj_nextkey, METH_NOARGS, "S.nextkey(key) -> data\n"
+		"Return the next key in this database." },
+	{ "delete", (PyCFunction)obj_delete, METH_VARARGS, "S.delete(key) -> None\n"
+		"Delete an entry." },
+	{ "has_key", (PyCFunction)obj_has_key, METH_VARARGS, "S.has_key(key) -> None\n"
+		"Check whether key exists in this database." },
+	{ "store", (PyCFunction)obj_store, METH_VARARGS, "S.store(key, data, flag=REPLACE) -> None"
+		"Store data." },
+	{ "add_flags", (PyCFunction)obj_add_flags, METH_VARARGS, "S.add_flags(flags) -> None" },
+	{ "remove_flags", (PyCFunction)obj_remove_flags, METH_VARARGS, "S.remove_flags(flags) -> None" },
+	{ "iterkeys", (PyCFunction)tdb_object_iter, METH_NOARGS, "S.iterkeys() -> iterator" },
+	{ "clear", (PyCFunction)obj_clear, METH_NOARGS, "S.clear() -> None\n"
+		"Wipe the entire database." },
+	{ "repack", (PyCFunction)obj_repack, METH_NOARGS, "S.repack() -> None\n"
+		"Repack the entire database." },
+	{ "enable_seqnum", (PyCFunction)obj_enable_seqnum, METH_NOARGS,
+		"S.enable_seqnum() -> None" },
+	{ "increment_seqnum_nonblock", (PyCFunction)obj_increment_seqnum_nonblock, METH_NOARGS,
+		"S.increment_seqnum_nonblock() -> None" },
+	{ NULL }
+};
+
+static PyObject *obj_get_hash_size(PyTdbObject *self, void *closure)
+{
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+	return PyInt_FromLong(tdb_hash_size(self->ctx));
+}
+
+static int obj_set_max_dead(PyTdbObject *self, PyObject *max_dead, void *closure)
+{
+	PyErr_TDB_RAISE_RETURN_MINUS_1_IF_CLOSED(self);
+	if (!PyInt_Check(max_dead))
+		return -1;
+	tdb_set_max_dead(self->ctx, PyInt_AsLong(max_dead));
+	return 0;
+}
+
+static PyObject *obj_get_map_size(PyTdbObject *self, void *closure)
+{
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+	return PyInt_FromLong(tdb_map_size(self->ctx));
+}
+
+static PyObject *obj_get_freelist_size(PyTdbObject *self, void *closure)
+{
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+	return PyInt_FromLong(tdb_freelist_size(self->ctx));
+}
+
+static PyObject *obj_get_flags(PyTdbObject *self, void *closure)
+{
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+	return PyInt_FromLong(tdb_get_flags(self->ctx));
+}
+
+static PyObject *obj_get_filename(PyTdbObject *self, void *closure)
+{
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+	return PyString_FromString(tdb_name(self->ctx));
+}
+
+static PyObject *obj_get_seqnum(PyTdbObject *self, void *closure)
+{
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+	return PyInt_FromLong(tdb_get_seqnum(self->ctx));
+}
+
+
+static PyGetSetDef tdb_object_getsetters[] = {
+	{ (char *)"hash_size", (getter)obj_get_hash_size, NULL, NULL },
+	{ (char *)"map_size", (getter)obj_get_map_size, NULL, NULL },
+	{ (char *)"freelist_size", (getter)obj_get_freelist_size, NULL, NULL },
+	{ (char *)"flags", (getter)obj_get_flags, NULL, NULL },
+	{ (char *)"max_dead", NULL, (setter)obj_set_max_dead, NULL },
+	{ (char *)"filename", (getter)obj_get_filename, NULL, (char *)"The filename of this TDB file."},
+	{ (char *)"seqnum", (getter)obj_get_seqnum, NULL, NULL },
+	{ NULL }
+};
+
+static PyObject *tdb_object_repr(PyTdbObject *self)
+{
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+	if (tdb_get_flags(self->ctx) & TDB_INTERNAL) {
+		return PyString_FromString("Tdb(<internal>)");
+	} else {
+		return PyString_FromFormat("Tdb('%s')", tdb_name(self->ctx));
+	}
+}
+
+static void tdb_object_dealloc(PyTdbObject *self)
+{
+	if (!self->closed)
+		tdb_close(self->ctx);
+	self->ob_type->tp_free(self);
+}
+
+static PyObject *obj_getitem(PyTdbObject *self, PyObject *key)
+{
+	TDB_DATA tkey, val;
+	PyErr_TDB_RAISE_IF_CLOSED(self);
+	if (!PyString_Check(key)) {
+		PyErr_SetString(PyExc_TypeError, "Expected string as key");
+		return NULL;
+	}
+
+	tkey.dptr = (unsigned char *)PyString_AsString(key);
+	tkey.dsize = PyString_Size(key);
+
+	val = tdb_fetch(self->ctx, tkey);
+	if (val.dptr == NULL) {
+		PyErr_SetString(PyExc_KeyError, "No such TDB entry");
+		return NULL;
+	} else {
+		return PyString_FromTDB_DATA(val);
+	}
+}
+
+static int obj_setitem(PyTdbObject *self, PyObject *key, PyObject *value)
+{
+	TDB_DATA tkey, tval;
+	int ret;
+	PyErr_TDB_RAISE_RETURN_MINUS_1_IF_CLOSED(self);
+	if (!PyString_Check(key)) {
+		PyErr_SetString(PyExc_TypeError, "Expected string as key");
+		return -1;
+	}
+
+	tkey = PyString_AsTDB_DATA(key);
+
+	if (value == NULL) { 
+		ret = tdb_delete(self->ctx, tkey);
+	} else { 
+		if (!PyString_Check(value)) {
+			PyErr_SetString(PyExc_TypeError, "Expected string as value");
+			return -1;
+		}
+
+		tval = PyString_AsTDB_DATA(value);
+
+		ret = tdb_store(self->ctx, tkey, tval, TDB_REPLACE);
+	}
+
+	if (ret != 0) {
+		PyErr_SetTDBError(self->ctx);
+		return -1;
+	} 
+
+	return ret;
+}
+
+static PyMappingMethods tdb_object_mapping = {
+	.mp_subscript = (binaryfunc)obj_getitem,
+	.mp_ass_subscript = (objobjargproc)obj_setitem,
+};
+static PyTypeObject PyTdb = {
+	.tp_name = "tdb.Tdb",
+	.tp_basicsize = sizeof(PyTdbObject),
+	.tp_methods = tdb_object_methods,
+	.tp_getset = tdb_object_getsetters,
+	.tp_new = py_tdb_open,
+	.tp_doc = "A TDB file",
+	.tp_repr = (reprfunc)tdb_object_repr,
+	.tp_dealloc = (destructor)tdb_object_dealloc,
+	.tp_as_mapping = &tdb_object_mapping,
+	.tp_flags = Py_TPFLAGS_DEFAULT|Py_TPFLAGS_BASETYPE|Py_TPFLAGS_HAVE_ITER,
+	.tp_iter = (getiterfunc)tdb_object_iter,
+};
+
+static PyMethodDef tdb_methods[] = {
+	{ "open", (PyCFunction)py_tdb_open, METH_VARARGS|METH_KEYWORDS, "open(name, hash_size=0, tdb_flags=TDB_DEFAULT, flags=O_RDWR, mode=0600)\n"
+		"Open a TDB file." },
+	{ NULL }
+};
+
+void inittdb(void);
+void inittdb(void)
+{
+	PyObject *m;
+
+	if (PyType_Ready(&PyTdb) < 0)
+		return;
+
+	if (PyType_Ready(&PyTdbIterator) < 0)
+		return;
+
+	m = Py_InitModule3("tdb", tdb_methods,
+		"simple key-value database that supports multiple writers.");
+	if (m == NULL)
+		return;
+
+	PyModule_AddObject(m, "REPLACE", PyInt_FromLong(TDB_REPLACE));
+	PyModule_AddObject(m, "INSERT", PyInt_FromLong(TDB_INSERT));
+	PyModule_AddObject(m, "MODIFY", PyInt_FromLong(TDB_MODIFY));
+
+	PyModule_AddObject(m, "DEFAULT", PyInt_FromLong(TDB_DEFAULT));
+	PyModule_AddObject(m, "CLEAR_IF_FIRST", PyInt_FromLong(TDB_CLEAR_IF_FIRST));
+	PyModule_AddObject(m, "INTERNAL", PyInt_FromLong(TDB_INTERNAL));
+	PyModule_AddObject(m, "NOLOCK", PyInt_FromLong(TDB_NOLOCK));
+	PyModule_AddObject(m, "NOMMAP", PyInt_FromLong(TDB_NOMMAP));
+	PyModule_AddObject(m, "CONVERT", PyInt_FromLong(TDB_CONVERT));
+	PyModule_AddObject(m, "BIGENDIAN", PyInt_FromLong(TDB_BIGENDIAN));
+	PyModule_AddObject(m, "NOSYNC", PyInt_FromLong(TDB_NOSYNC));
+	PyModule_AddObject(m, "SEQNUM", PyInt_FromLong(TDB_SEQNUM));
+	PyModule_AddObject(m, "VOLATILE", PyInt_FromLong(TDB_VOLATILE));
+	PyModule_AddObject(m, "ALLOW_NESTING", PyInt_FromLong(TDB_ALLOW_NESTING));
+	PyModule_AddObject(m, "DISALLOW_NESTING", PyInt_FromLong(TDB_DISALLOW_NESTING));
+	PyModule_AddObject(m, "INCOMPATIBLE_HASH", PyInt_FromLong(TDB_INCOMPATIBLE_HASH));
+
+	PyModule_AddObject(m, "__docformat__", PyString_FromString("restructuredText"));
+
+	PyModule_AddObject(m, "__version__", PyString_FromString(PACKAGE_VERSION));
+
+	Py_INCREF(&PyTdb);
+	PyModule_AddObject(m, "Tdb", (PyObject *)&PyTdb);
+
+	Py_INCREF(&PyTdbIterator);
+}
diff --git a/ctdb/lib/tdb/python/tdbdump.py b/ctdb/lib/tdb/python/tdbdump.py
new file mode 100644
index 0000000000..01859ebce2
--- /dev/null
+++ b/ctdb/lib/tdb/python/tdbdump.py
@@ -0,0 +1,12 @@
+#!/usr/bin/env python
+# Trivial reimplementation of tdbdump in Python
+
+import tdb, sys
+
+if len(sys.argv) < 2:
+    print "Usage: tdbdump.py <tdb-file>"
+    sys.exit(1)
+
+db = tdb.Tdb(sys.argv[1])
+for (k, v) in db.iteritems():
+    print "{\nkey(%d) = %r\ndata(%d) = %r\n}" % (len(k), k, len(v), v)
diff --git a/ctdb/lib/tdb/python/tests/simple.py b/ctdb/lib/tdb/python/tests/simple.py
new file mode 100644
index 0000000000..7e295a8c95
--- /dev/null
+++ b/ctdb/lib/tdb/python/tests/simple.py
@@ -0,0 +1,204 @@
+#!/usr/bin/env python
+# Some simple tests for the Python bindings for TDB
+# Note that this tests the interface of the Python bindings
+# It does not test tdb itself.
+#
+# Copyright (C) 2007-2008 Jelmer Vernooij <jelmer@samba.org>
+# Published under the GNU LGPLv3 or later
+
+import tdb
+from unittest import TestCase
+import os, tempfile
+
+
+class OpenTdbTests(TestCase):
+
+    def test_nonexistent_read(self):
+        self.assertRaises(IOError, tdb.Tdb, "/some/nonexistent/file", 0,
+                tdb.DEFAULT, os.O_RDWR)
+
+class CloseTdbTests(TestCase):
+
+    def test_double_close(self):
+        # No hash size in tdb2.
+        if tdb.__version__.startswith("2"):
+            self.tdb = tdb.Tdb(tempfile.mkstemp()[1], tdb.DEFAULT,
+                               os.O_CREAT|os.O_RDWR)
+        else:
+            self.tdb = tdb.Tdb(tempfile.mkstemp()[1], 0, tdb.DEFAULT,
+                               os.O_CREAT|os.O_RDWR)
+        self.assertNotEqual(None, self.tdb)
+
+        # ensure that double close does not crash python
+        self.tdb.close()
+        self.tdb.close()
+
+        # Check that further operations do not crash python
+        self.assertRaises(RuntimeError, lambda: self.tdb.transaction_start())
+
+        self.assertRaises(RuntimeError, lambda: self.tdb["bar"])
+
+
+class InternalTdbTests(TestCase):
+
+    def test_repr(self):
+        self.tdb = tdb.Tdb()
+
+        # repr used to crash on internal db
+        self.assertEquals(repr(self.tdb), "Tdb(<internal>)")
+
+
+class SimpleTdbTests(TestCase):
+
+    def setUp(self):
+        super(SimpleTdbTests, self).setUp()
+        if tdb.__version__.startswith("2"):
+            self.tdb = tdb.Tdb(tempfile.mkstemp()[1], tdb.DEFAULT,
+                               os.O_CREAT|os.O_RDWR)
+        else:
+            self.tdb = tdb.Tdb(tempfile.mkstemp()[1], 0, tdb.DEFAULT,
+                               os.O_CREAT|os.O_RDWR)
+        self.assertNotEqual(None, self.tdb)
+
+    def tearDown(self):
+        del self.tdb
+
+    def test_repr(self):
+        self.assertTrue(repr(self.tdb).startswith("Tdb('"))
+
+    def test_lockall(self):
+        self.tdb.lock_all()
+
+    def test_max_dead(self):
+        if not tdb.__version__.startswith("2"):
+            self.tdb.max_dead = 20
+
+    def test_unlockall(self):
+        self.tdb.lock_all()
+        self.tdb.unlock_all()
+
+    def test_lockall_read(self):
+        self.tdb.read_lock_all()
+        self.tdb.read_unlock_all()
+
+    def test_reopen(self):
+        if not tdb.__version__.startswith("2"):
+            self.tdb.reopen()
+
+    def test_store(self):
+        self.tdb.store("bar", "bla")
+        self.assertEquals("bla", self.tdb.get("bar"))
+
+    def test_getitem(self):
+        self.tdb["bar"] = "foo"
+        if not tdb.__version__.startswith("2"):
+            self.tdb.reopen()
+        self.assertEquals("foo", self.tdb["bar"])
+
+    def test_delete(self):
+        self.tdb["bar"] = "foo"
+        del self.tdb["bar"]
+        self.assertRaises(KeyError, lambda: self.tdb["bar"])
+    
+    def test_contains(self):
+        self.tdb["bla"] = "bloe"
+        self.assertTrue("bla" in self.tdb)
+
+    def test_keyerror(self):
+        self.assertRaises(KeyError, lambda: self.tdb["bla"])
+
+    def test_hash_size(self):
+        if not tdb.__version__.startswith("2"):
+            self.tdb.hash_size
+
+    def test_map_size(self):
+        if not tdb.__version__.startswith("2"):
+            self.tdb.map_size
+
+    def test_freelist_size(self):
+        if not tdb.__version__.startswith("2"):
+            self.tdb.freelist_size
+
+    def test_name(self):
+        self.tdb.filename
+
+    def test_iterator(self):
+        self.tdb["bla"] = "1"
+        self.tdb["brainslug"] = "2"
+        l = list(self.tdb)
+        l.sort()
+        self.assertEquals(["bla", "brainslug"], l)
+
+    def test_transaction_cancel(self):
+        self.tdb["bloe"] = "2"
+        self.tdb.transaction_start()
+        self.tdb["bloe"] = "1"
+        self.tdb.transaction_cancel()
+        self.assertEquals("2", self.tdb["bloe"])
+
+    def test_transaction_commit(self):
+        self.tdb["bloe"] = "2"
+        self.tdb.transaction_start()
+        self.tdb["bloe"] = "1"
+        self.tdb.transaction_commit()
+        self.assertEquals("1", self.tdb["bloe"])
+
+    def test_transaction_prepare_commit(self):
+        self.tdb["bloe"] = "2"
+        self.tdb.transaction_start()
+        self.tdb["bloe"] = "1"
+        self.tdb.transaction_prepare_commit()
+        self.tdb.transaction_commit()
+        self.assertEquals("1", self.tdb["bloe"])
+
+    def test_iterkeys(self):
+        self.tdb["bloe"] = "2"
+        self.tdb["bla"] = "25"
+        i = self.tdb.iterkeys()
+        self.assertEquals(set(["bloe", "bla"]), set([i.next(), i.next()]))
+
+    def test_clear(self):
+        self.tdb["bloe"] = "2"
+        self.tdb["bla"] = "25"
+        self.assertEquals(2, len(list(self.tdb)))
+        self.tdb.clear()
+        self.assertEquals(0, len(list(self.tdb)))
+
+    def test_repack(self):
+        if not tdb.__version__.startswith("2"):
+            self.tdb["foo"] = "abc"
+            self.tdb["bar"] = "def"
+            del self.tdb["foo"]
+            self.tdb.repack()
+
+    def test_seqnum(self):
+        if not tdb.__version__.startswith("2"):
+            self.tdb.enable_seqnum()
+            seq1 = self.tdb.seqnum
+            self.tdb.increment_seqnum_nonblock()
+            seq2 = self.tdb.seqnum
+            self.assertEquals(seq2-seq1, 1)
+
+    def test_len(self):
+        self.assertEquals(0, len(list(self.tdb)))
+        self.tdb["entry"] = "value"
+        self.assertEquals(1, len(list(self.tdb)))
+
+    def test_add_flags(self):
+        if tdb.__version__.startswith("2"):
+            self.tdb.add_flag(tdb.NOMMAP)
+            self.tdb.remove_flag(tdb.NOMMAP)
+        else:
+            self.tdb.add_flags(tdb.NOMMAP)
+            self.tdb.remove_flags(tdb.NOMMAP)
+
+
+class VersionTests(TestCase):
+
+    def test_present(self):
+        self.assertTrue(isinstance(tdb.__version__, str))
+
+
+if __name__ == '__main__':
+    import unittest
+    unittest.TestProgram()
diff --git a/ctdb/lib/tdb/tdb.pc.in b/ctdb/lib/tdb/tdb.pc.in
new file mode 100644
index 0000000000..b78419ea78
--- /dev/null
+++ b/ctdb/lib/tdb/tdb.pc.in
@@ -0,0 +1,11 @@
+prefix=@prefix@
+exec_prefix=@exec_prefix@
+libdir=@libdir@
+includedir=@includedir@
+
+Name: tdb
+Description: A trivial database
+Version: @PACKAGE_VERSION@
+Libs: @LIB_RPATH@ -L${libdir} -ltdb
+Cflags: -I${includedir} 
+URL: http://tdb.samba.org/
diff --git a/ctdb/lib/tdb/test/external-agent.c b/ctdb/lib/tdb/test/external-agent.c
new file mode 100644
index 0000000000..8140e70ead
--- /dev/null
+++ b/ctdb/lib/tdb/test/external-agent.c
@@ -0,0 +1,198 @@
+#include "external-agent.h"
+#include "lock-tracking.h"
+#include "logging.h"
+#include <sys/types.h>
+#include <sys/wait.h>
+#include <unistd.h>
+#include <fcntl.h>
+#include <stdlib.h>
+#include <limits.h>
+#include <string.h>
+#include <errno.h>
+#include "../common/tdb_private.h"
+#include "tap-interface.h"
+#include <stdio.h>
+#include <stdarg.h>
+
+static struct tdb_context *tdb;
+
+static enum agent_return do_operation(enum operation op, const char *name)
+{
+	TDB_DATA k;
+	enum agent_return ret;
+	TDB_DATA data;
+
+	if (op != OPEN && op != OPEN_WITH_CLEAR_IF_FIRST && !tdb) {
+		diag("external: No tdb open!");
+		return OTHER_FAILURE;
+	}
+
+	k.dptr = (void *)name;
+	k.dsize = strlen(name);
+
+	locking_would_block = 0;
+	switch (op) {
+	case OPEN:
+		if (tdb) {
+			diag("Already have tdb %s open", tdb_name(tdb));
+			return OTHER_FAILURE;
+		}
+		tdb = tdb_open_ex(name, 0, TDB_DEFAULT, O_RDWR, 0,
+				  &taplogctx, NULL);
+		if (!tdb) {
+			if (!locking_would_block)
+				diag("Opening tdb gave %s", strerror(errno));
+			ret = OTHER_FAILURE;
+		} else
+			ret = SUCCESS;
+		break;
+	case OPEN_WITH_CLEAR_IF_FIRST:
+		if (tdb)
+			return OTHER_FAILURE;
+		tdb = tdb_open_ex(name, 0, TDB_CLEAR_IF_FIRST, O_RDWR, 0,
+				  &taplogctx, NULL);
+		ret = tdb ? SUCCESS : OTHER_FAILURE;
+		break;
+	case TRANSACTION_START:
+		ret = tdb_transaction_start(tdb) == 0 ? SUCCESS : OTHER_FAILURE;
+		break;
+	case FETCH:
+		data = tdb_fetch(tdb, k);
+		if (data.dptr == NULL) {
+			if (tdb_error(tdb) == TDB_ERR_NOEXIST)
+				ret = FAILED;
+			else
+				ret = OTHER_FAILURE;
+		} else if (data.dsize != k.dsize
+			   || memcmp(data.dptr, k.dptr, k.dsize) != 0) {
+			ret = OTHER_FAILURE;
+		} else {
+			ret = SUCCESS;
+		}
+		free(data.dptr);
+		break;
+	case STORE:
+		ret = tdb_store(tdb, k, k, 0) == 0 ? SUCCESS : OTHER_FAILURE;
+		break;
+	case TRANSACTION_COMMIT:
+		ret = tdb_transaction_commit(tdb)==0 ? SUCCESS : OTHER_FAILURE;
+		break;
+	case CHECK:
+		ret = tdb_check(tdb, NULL, NULL) == 0 ? SUCCESS : OTHER_FAILURE;
+		break;
+	case NEEDS_RECOVERY:
+		ret = tdb_needs_recovery(tdb) ? SUCCESS : FAILED;
+		break;
+	case CLOSE:
+		ret = tdb_close(tdb) == 0 ? SUCCESS : OTHER_FAILURE;
+		tdb = NULL;
+		break;
+	default:
+		ret = OTHER_FAILURE;
+	}
+
+	if (locking_would_block)
+		ret = WOULD_HAVE_BLOCKED;
+
+	return ret;
+}
+
+struct agent {
+	int cmdfd, responsefd;
+};
+
+/* Do this before doing any tdb stuff.  Return handle, or NULL. */
+struct agent *prepare_external_agent(void)
+{
+	int pid, ret;
+	int command[2], response[2];
+	char name[1+PATH_MAX];
+
+	if (pipe(command) != 0 || pipe(response) != 0) {
+		fprintf(stderr, "pipe failed: %s\n", strerror(errno));
+		exit(1);
+	}
+
+	pid = fork();
+	if (pid < 0) {
+		fprintf(stderr, "fork failed: %s\n", strerror(errno));
+		exit(1);
+	}
+
+	if (pid != 0) {
+		struct agent *agent = malloc(sizeof(*agent));
+
+		close(command[0]);
+		close(response[1]);
+		agent->cmdfd = command[1];
+		agent->responsefd = response[0];
+		return agent;
+	}
+
+	close(command[1]);
+	close(response[0]);
+
+	/* We want to fail, not block. */
+	nonblocking_locks = true;
+	log_prefix = "external: ";
+	while ((ret = read(command[0], name, sizeof(name))) > 0) {
+		enum agent_return result;
+
+		result = do_operation(name[0], name+1);
+		if (write(response[1], &result, sizeof(result))
+		    != sizeof(result))
+			abort();
+	}
+	exit(0);
+}
+
+/* Ask the external agent to try to do an operation. */
+enum agent_return external_agent_operation(struct agent *agent,
+					   enum operation op,
+					   const char *name)
+{
+	enum agent_return res;
+	unsigned int len;
+	char *string;
+
+	if (!name)
+		name = "";
+	len = 1 + strlen(name) + 1;
+	string = malloc(len);
+
+	string[0] = op;
+	strcpy(string+1, name);
+
+	if (write(agent->cmdfd, string, len) != len
+	    || read(agent->responsefd, &res, sizeof(res)) != sizeof(res))
+		res = AGENT_DIED;
+
+	free(string);
+	return res;
+}
+
+const char *agent_return_name(enum agent_return ret)
+{
+	return ret == SUCCESS ? "SUCCESS"
+		: ret == WOULD_HAVE_BLOCKED ? "WOULD_HAVE_BLOCKED"
+		: ret == AGENT_DIED ? "AGENT_DIED"
+		: ret == FAILED ? "FAILED"
+		: ret == OTHER_FAILURE ? "OTHER_FAILURE"
+		: "**INVALID**";
+}
+
+const char *operation_name(enum operation op)
+{
+	switch (op) {
+	case OPEN: return "OPEN";
+	case OPEN_WITH_CLEAR_IF_FIRST: return "OPEN_WITH_CLEAR_IF_FIRST";
+	case TRANSACTION_START: return "TRANSACTION_START";
+	case FETCH: return "FETCH";
+	case STORE: return "STORE";
+	case TRANSACTION_COMMIT: return "TRANSACTION_COMMIT";
+	case CHECK: return "CHECK";
+	case NEEDS_RECOVERY: return "NEEDS_RECOVERY";
+	case CLOSE: return "CLOSE";
+	}
+	return "**INVALID**";
+}
diff --git a/ctdb/lib/tdb/test/external-agent.h b/ctdb/lib/tdb/test/external-agent.h
new file mode 100644
index 0000000000..dffdca962f
--- /dev/null
+++ b/ctdb/lib/tdb/test/external-agent.h
@@ -0,0 +1,41 @@
+#ifndef TDB_TEST_EXTERNAL_AGENT_H
+#define TDB_TEST_EXTERNAL_AGENT_H
+
+/* For locking tests, we need a different process to try things at
+ * various times. */
+enum operation {
+	OPEN,
+	OPEN_WITH_CLEAR_IF_FIRST,
+	TRANSACTION_START,
+	FETCH,
+	STORE,
+	TRANSACTION_COMMIT,
+	CHECK,
+	NEEDS_RECOVERY,
+	CLOSE,
+};
+
+/* Do this before doing any tdb stuff.  Return handle, or -1. */
+struct agent *prepare_external_agent(void);
+
+enum agent_return {
+	SUCCESS,
+	WOULD_HAVE_BLOCKED,
+	AGENT_DIED,
+	FAILED, /* For fetch, or NEEDS_RECOVERY */
+	OTHER_FAILURE,
+};
+
+/* Ask the external agent to try to do an operation.
+ * name == tdb name for OPEN/OPEN_WITH_CLEAR_IF_FIRST,
+ * record name for FETCH/STORE (store stores name as data too)
+ */
+enum agent_return external_agent_operation(struct agent *handle,
+					   enum operation op,
+					   const char *name);
+
+/* Mapping enum -> string. */
+const char *agent_return_name(enum agent_return ret);
+const char *operation_name(enum operation op);
+
+#endif /* TDB_TEST_EXTERNAL_AGENT_H */
diff --git a/ctdb/lib/tdb/test/jenkins-be-hash.tdb b/ctdb/lib/tdb/test/jenkins-be-hash.tdb
new file mode 100644
index 0000000000..b652840414
--- /dev/null
+++ b/ctdb/lib/tdb/test/jenkins-be-hash.tdb
diff --git a/ctdb/lib/tdb/test/jenkins-le-hash.tdb b/ctdb/lib/tdb/test/jenkins-le-hash.tdb
new file mode 100644
index 0000000000..007e0a3368
--- /dev/null
+++ b/ctdb/lib/tdb/test/jenkins-le-hash.tdb
diff --git a/ctdb/lib/tdb/test/lock-tracking.c b/ctdb/lib/tdb/test/lock-tracking.c
new file mode 100644
index 0000000000..90a07f8ef1
--- /dev/null
+++ b/ctdb/lib/tdb/test/lock-tracking.c
@@ -0,0 +1,146 @@
+/* We save the locks so we can reaquire them. */
+#include "../common/tdb_private.h"
+#include <unistd.h>
+#include <fcntl.h>
+#include <stdarg.h>
+#include <stdlib.h>
+#include "tap-interface.h"
+#include "lock-tracking.h"
+
+struct testlock {
+	struct testlock *next;
+	unsigned int off;
+	unsigned int len;
+	int type;
+};
+static struct testlock *testlocks;
+int locking_errors = 0;
+bool suppress_lockcheck = false;
+bool nonblocking_locks;
+int locking_would_block = 0;
+void (*unlock_callback)(int fd);
+
+int fcntl_with_lockcheck(int fd, int cmd, ... /* arg */ )
+{
+	va_list ap;
+	int ret, arg3;
+	struct flock *fl;
+	bool may_block = false;
+
+	if (cmd != F_SETLK && cmd != F_SETLKW) {
+		/* This may be totally bogus, but we don't know in general. */
+		va_start(ap, cmd);
+		arg3 = va_arg(ap, int);
+		va_end(ap);
+
+		return fcntl(fd, cmd, arg3);
+	}
+
+	va_start(ap, cmd);
+	fl = va_arg(ap, struct flock *);
+	va_end(ap);
+
+	if (cmd == F_SETLKW && nonblocking_locks) {
+		cmd = F_SETLK;
+		may_block = true;
+	}
+	ret = fcntl(fd, cmd, fl);
+
+	/* Detect when we failed, but might have been OK if we waited. */
+	if (may_block && ret == -1 && (errno == EAGAIN || errno == EACCES)) {
+		locking_would_block++;
+	}
+
+	if (fl->l_type == F_UNLCK) {
+		struct testlock **l;
+		struct testlock *old = NULL;
+
+		for (l = &testlocks; *l; l = &(*l)->next) {
+			if ((*l)->off == fl->l_start
+			    && (*l)->len == fl->l_len) {
+				if (ret == 0) {
+					old = *l;
+					*l = (*l)->next;
+					free(old);
+				}
+				break;
+			}
+		}
+		if (!old && !suppress_lockcheck) {
+			diag("Unknown unlock %u@%u - %i",
+			     (int)fl->l_len, (int)fl->l_start, ret);
+			locking_errors++;
+		}
+	} else {
+		struct testlock *new, *i;
+		unsigned int fl_end = fl->l_start + fl->l_len;
+		if (fl->l_len == 0)
+			fl_end = (unsigned int)-1;
+
+		/* Check for overlaps: we shouldn't do this. */
+		for (i = testlocks; i; i = i->next) {
+			unsigned int i_end = i->off + i->len;
+			if (i->len == 0)
+				i_end = (unsigned int)-1;
+
+			if (fl->l_start >= i->off && fl->l_start < i_end)
+				break;
+			if (fl_end >= i->off && fl_end < i_end)
+				break;
+
+			/* tdb_allrecord_lock does this, handle adjacent: */
+			if (fl->l_start == i_end && fl->l_type == i->type) {
+				if (ret == 0) {
+					i->len = fl->l_len
+						? i->len + fl->l_len
+						: 0;
+				}
+				goto done;
+			}
+		}
+		if (i) {
+			/* Special case: upgrade of allrecord lock. */
+			if (i->type == F_RDLCK && fl->l_type == F_WRLCK
+			    && i->off == FREELIST_TOP
+			    && fl->l_start == FREELIST_TOP
+			    && i->len == 0
+			    && fl->l_len == 0) {
+				if (ret == 0)
+					i->type = F_WRLCK;
+				goto done;
+			}
+			if (!suppress_lockcheck) {
+				diag("%s testlock %u@%u overlaps %u@%u",
+				     fl->l_type == F_WRLCK ? "write" : "read",
+				     (int)fl->l_len, (int)fl->l_start,
+				     i->len, (int)i->off);
+				locking_errors++;
+			}
+		}
+
+		if (ret == 0) {
+			new = malloc(sizeof *new);
+			new->off = fl->l_start;
+			new->len = fl->l_len;
+			new->type = fl->l_type;
+			new->next = testlocks;
+			testlocks = new;
+		}
+	}
+done:
+	if (ret == 0 && fl->l_type == F_UNLCK && unlock_callback)
+		unlock_callback(fd);
+	return ret;
+}
+
+unsigned int forget_locking(void)
+{
+	unsigned int num = 0;
+	while (testlocks) {
+		struct testlock *next = testlocks->next;
+		free(testlocks);
+		testlocks = next;
+		num++;
+	}
+	return num;
+}
diff --git a/ctdb/lib/tdb/test/lock-tracking.h b/ctdb/lib/tdb/test/lock-tracking.h
new file mode 100644
index 0000000000..f2c9c44653
--- /dev/null
+++ b/ctdb/lib/tdb/test/lock-tracking.h
@@ -0,0 +1,25 @@
+#ifndef LOCK_TRACKING_H
+#define LOCK_TRACKING_H
+#include <stdbool.h>
+
+/* Set this if you want a callback after fnctl unlock. */
+extern void (*unlock_callback)(int fd);
+
+/* Replacement fcntl. */
+int fcntl_with_lockcheck(int fd, int cmd, ... /* arg */ );
+
+/* Discard locking info: returns number of locks outstanding. */
+unsigned int forget_locking(void);
+
+/* Number of errors in locking. */
+extern int locking_errors;
+
+/* Suppress lock checking. */
+extern bool suppress_lockcheck;
+
+/* Make all locks non-blocking. */
+extern bool nonblocking_locks;
+
+/* Number of times we failed a lock because we made it non-blocking. */
+extern int locking_would_block;
+#endif /* LOCK_TRACKING_H */
diff --git a/ctdb/lib/tdb/test/logging.c b/ctdb/lib/tdb/test/logging.c
new file mode 100644
index 0000000000..dfab4868d2
--- /dev/null
+++ b/ctdb/lib/tdb/test/logging.c
@@ -0,0 +1,33 @@
+#include "logging.h"
+#include "tap-interface.h"
+#include <stdarg.h>
+#include <stdlib.h>
+#include <string.h>
+#include <stdio.h>
+
+bool suppress_logging = false;
+const char *log_prefix = "";
+
+/* Turn log messages into tap diag messages. */
+static void taplog(struct tdb_context *tdb,
+		   enum tdb_debug_level level,
+		   const char *fmt, ...)
+{
+	va_list ap;
+	char line[200];
+
+	if (suppress_logging)
+		return;
+
+	va_start(ap, fmt);
+	vsprintf(line, fmt, ap);
+	va_end(ap);
+
+	/* Strip trailing \n: diag adds it. */
+	if (line[0] && line[strlen(line)-1] == '\n')
+		diag("%s%.*s", log_prefix, (unsigned)strlen(line)-1, line);
+	else
+		diag("%s%s", log_prefix, line);
+}
+
+struct tdb_logging_context taplogctx = { taplog, NULL };
diff --git a/ctdb/lib/tdb/test/logging.h b/ctdb/lib/tdb/test/logging.h
new file mode 100644
index 0000000000..89e77b2108
--- /dev/null
+++ b/ctdb/lib/tdb/test/logging.h
@@ -0,0 +1,11 @@
+#ifndef TDB_TEST_LOGGING_H
+#define TDB_TEST_LOGGING_H
+#include "replace.h"
+#include "../include/tdb.h"
+#include <stdbool.h>
+
+extern bool suppress_logging;
+extern const char *log_prefix;
+extern struct tdb_logging_context taplogctx;
+
+#endif /* TDB_TEST_LOGGING_H */
diff --git a/ctdb/lib/tdb/test/old-nohash-be.tdb b/ctdb/lib/tdb/test/old-nohash-be.tdb
new file mode 100644
index 0000000000..1c49116c1d
--- /dev/null
+++ b/ctdb/lib/tdb/test/old-nohash-be.tdb
diff --git a/ctdb/lib/tdb/test/old-nohash-le.tdb b/ctdb/lib/tdb/test/old-nohash-le.tdb
new file mode 100644
index 0000000000..0655072d88
--- /dev/null
+++ b/ctdb/lib/tdb/test/old-nohash-le.tdb
diff --git a/ctdb/lib/tdb/test/run-3G-file.c b/ctdb/lib/tdb/test/run-3G-file.c
new file mode 100644
index 0000000000..3ee9de15ed
--- /dev/null
+++ b/ctdb/lib/tdb/test/run-3G-file.c
@@ -0,0 +1,144 @@
+#include "../common/tdb_private.h"
+#include "../common/io.c"
+#include "../common/tdb.c"
+#include "../common/lock.c"
+#include "../common/freelist.c"
+#include "../common/traverse.c"
+#include "../common/transaction.c"
+#include "../common/error.c"
+#include "../common/open.c"
+#include "../common/check.c"
+#include "../common/hash.c"
+#include "tap-interface.h"
+#include <stdlib.h>
+#include "logging.h"
+
+static int tdb_expand_file_sparse(struct tdb_context *tdb,
+				  tdb_off_t size,
+				  tdb_off_t addition)
+{
+	if (tdb->read_only || tdb->traverse_read) {
+		tdb->ecode = TDB_ERR_RDONLY;
+		return -1;
+	}
+
+	if (ftruncate(tdb->fd, size+addition) == -1) {
+		char b = 0;
+		ssize_t written = pwrite(tdb->fd,  &b, 1, (size+addition) - 1);
+		if (written == 0) {
+			/* try once more, potentially revealing errno */
+			written = pwrite(tdb->fd,  &b, 1, (size+addition) - 1);
+		}
+		if (written == 0) {
+			/* again - give up, guessing errno */
+			errno = ENOSPC;
+		}
+		if (written != 1) {
+			TDB_LOG((tdb, TDB_DEBUG_FATAL, "expand_file to %d failed (%s)\n",
+				 size+addition, strerror(errno)));
+			return -1;
+		}
+	}
+
+	return 0;
+}
+
+static const struct tdb_methods large_io_methods = {
+	tdb_read,
+	tdb_write,
+	tdb_next_hash_chain,
+	tdb_oob,
+	tdb_expand_file_sparse
+};
+
+static int test_traverse(struct tdb_context *tdb, TDB_DATA key, TDB_DATA data,
+			 void *_data)
+{
+	TDB_DATA *expect = _data;
+	ok1(key.dsize == strlen("hi"));
+	ok1(memcmp(key.dptr, "hi", strlen("hi")) == 0);
+	ok1(data.dsize == expect->dsize);
+	ok1(memcmp(data.dptr, expect->dptr, data.dsize) == 0);
+	return 0;
+}
+
+int main(int argc, char *argv[])
+{
+	struct tdb_context *tdb;
+	TDB_DATA key, orig_data, data;
+	uint32_t hash;
+	tdb_off_t rec_ptr;
+	struct tdb_record rec;
+	int ret;
+
+	plan_tests(24);
+	tdb = tdb_open_ex("run-36-file.tdb", 1024, TDB_CLEAR_IF_FIRST,
+			  O_CREAT|O_TRUNC|O_RDWR, 0600, &taplogctx, NULL);
+
+	ok1(tdb);
+	tdb->methods = &large_io_methods;
+
+	key.dsize = strlen("hi");
+	key.dptr = (void *)"hi";
+	orig_data.dsize = strlen("world");
+	orig_data.dptr = (void *)"world";
+
+	/* Enlarge the file (internally multiplies by 2). */
+	ret = tdb_expand(tdb, 1500000000);
+#ifdef HAVE_INCOHERENT_MMAP
+	/* This can fail due to mmap failure on 32 bit systems. */
+	if (ret == -1) {
+		/* These should now fail. */
+		ok1(tdb_store(tdb, key, orig_data, TDB_INSERT) == -1);
+		data = tdb_fetch(tdb, key);
+		ok1(data.dptr == NULL);
+		ok1(tdb_traverse(tdb, test_traverse, &orig_data) == -1);
+		ok1(tdb_delete(tdb, key) == -1);
+		ok1(tdb_traverse(tdb, test_traverse, NULL) == -1);
+		/* Skip the rest... */
+		for (ret = 0; ret < 24 - 6; ret++)
+			ok1(1);
+		tdb_close(tdb);
+		return exit_status();
+	}
+#endif
+	ok1(ret == 0);
+
+	/* Put an entry in, and check it. */
+	ok1(tdb_store(tdb, key, orig_data, TDB_INSERT) == 0);
+
+	data = tdb_fetch(tdb, key);
+	ok1(data.dsize == strlen("world"));
+	ok1(memcmp(data.dptr, "world", strlen("world")) == 0);
+	free(data.dptr);
+
+	/* That currently fills at the end, make sure that's true. */
+	hash = tdb->hash_fn(&key);
+	rec_ptr = tdb_find_lock_hash(tdb, key, hash, F_RDLCK, &rec);
+	ok1(rec_ptr);
+	ok1(rec_ptr > 2U*1024*1024*1024);
+	tdb_unlock(tdb, BUCKET(rec.full_hash), F_RDLCK);
+
+	/* Traverse must work. */
+	ok1(tdb_traverse(tdb, test_traverse, &orig_data) == 1);
+
+	/* Delete should work. */
+	ok1(tdb_delete(tdb, key) == 0);
+
+	ok1(tdb_traverse(tdb, test_traverse, NULL) == 0);
+
+	/* Transactions should work. */
+	ok1(tdb_transaction_start(tdb) == 0);
+	ok1(tdb_store(tdb, key, orig_data, TDB_INSERT) == 0);
+
+	data = tdb_fetch(tdb, key);
+	ok1(data.dsize == strlen("world"));
+	ok1(memcmp(data.dptr, "world", strlen("world")) == 0);
+	free(data.dptr);
+	ok1(tdb_transaction_commit(tdb) == 0);
+
+	ok1(tdb_traverse(tdb, test_traverse, &orig_data) == 1);
+	tdb_close(tdb);
+
+	return exit_status();
+}
diff --git a/ctdb/lib/tdb/test/run-bad-tdb-header.c b/ctdb/lib/tdb/test/run-bad-tdb-header.c
new file mode 100644
index 0000000000..b00fb8934a
--- /dev/null
+++ b/ctdb/lib/tdb/test/run-bad-tdb-header.c
@@ -0,0 +1,58 @@
+#include "../common/tdb_private.h"
+#include "../common/io.c"
+#include "../common/tdb.c"
+#include "../common/lock.c"
+#include "../common/freelist.c"
+#include "../common/traverse.c"
+#include "../common/transaction.c"
+#include "../common/error.c"
+#include "../common/open.c"
+#include "../common/check.c"
+#include "../common/hash.c"
+#include "tap-interface.h"
+#include <stdlib.h>
+#include "logging.h"
+
+int main(int argc, char *argv[])
+{
+	struct tdb_context *tdb;
+	struct tdb_header hdr;
+	int fd;
+
+	plan_tests(11);
+	/* Can open fine if complete crap, as long as O_CREAT. */
+	fd = open("run-bad-tdb-header.tdb", O_RDWR|O_CREAT|O_TRUNC, 0600);
+	ok1(fd >= 0);
+	ok1(write(fd, "hello world", 11) == 11);
+	close(fd);
+	tdb = tdb_open_ex("run-bad-tdb-header.tdb", 1024, 0, O_RDWR, 0,
+			  &taplogctx, NULL);
+	ok1(!tdb);
+	tdb = tdb_open_ex("run-bad-tdb-header.tdb", 1024, 0, O_CREAT|O_RDWR,
+			  0600, &taplogctx, NULL);
+	ok1(tdb);
+	tdb_close(tdb);
+
+	/* Now, with wrong version it should *not* overwrite. */
+	fd = open("run-bad-tdb-header.tdb", O_RDWR);
+	ok1(fd >= 0);
+	ok1(read(fd, &hdr, sizeof(hdr)) == sizeof(hdr));
+	ok1(hdr.version == TDB_VERSION);
+	hdr.version++;
+	lseek(fd, 0, SEEK_SET);
+	ok1(write(fd, &hdr, sizeof(hdr)) == sizeof(hdr));
+	close(fd);
+
+	tdb = tdb_open_ex("run-bad-tdb-header.tdb", 1024, 0, O_RDWR|O_CREAT,
+			  0600, &taplogctx, NULL);
+	ok1(errno == EIO);
+	ok1(!tdb);
+
+	/* With truncate, will be fine. */
+	tdb = tdb_open_ex("run-bad-tdb-header.tdb", 1024, 0,
+			  O_RDWR|O_CREAT|O_TRUNC, 0600, &taplogctx, NULL);
+	ok1(tdb);
+	tdb_close(tdb);
+
+	return exit_status();
+}
diff --git a/ctdb/lib/tdb/test/run-check.c b/ctdb/lib/tdb/test/run-check.c
new file mode 100644
index 0000000000..05f7aecd08
--- /dev/null
+++ b/ctdb/lib/tdb/test/run-check.c
@@ -0,0 +1,64 @@
+#include "../common/tdb_private.h"
+#include "../common/io.c"
+#include "../common/tdb.c"
+#include "../common/lock.c"
+#include "../common/freelist.c"
+#include "../common/traverse.c"
+#include "../common/transaction.c"
+#include "../common/error.c"
+#include "../common/open.c"
+#include "../common/check.c"
+#include "../common/hash.c"
+#include "tap-interface.h"
+#include <stdlib.h>
+#include "logging.h"
+
+int main(int argc, char *argv[])
+{
+	struct tdb_context *tdb;
+	TDB_DATA key, data;
+
+	plan_tests(13);
+	tdb = tdb_open_ex("run-check.tdb", 1, TDB_CLEAR_IF_FIRST,
+			  O_CREAT|O_TRUNC|O_RDWR, 0600, &taplogctx, NULL);
+
+	ok1(tdb);
+	ok1(tdb_check(tdb, NULL, NULL) == 0);
+
+	key.dsize = strlen("hi");
+	key.dptr = (void *)"hi";
+	data.dsize = strlen("world");
+	data.dptr = (void *)"world";
+
+	ok1(tdb_store(tdb, key, data, TDB_INSERT) == 0);
+	ok1(tdb_check(tdb, NULL, NULL) == 0);
+	tdb_close(tdb);
+
+	tdb = tdb_open_ex("run-check.tdb", 1024, 0, O_RDWR, 0,
+			  &taplogctx, NULL);
+	ok1(tdb);
+	ok1(tdb_check(tdb, NULL, NULL) == 0);
+	tdb_close(tdb);
+
+	tdb = tdb_open_ex("test/tdb.corrupt", 1024, 0, O_RDWR, 0,
+			  &taplogctx, NULL);
+	ok1(tdb);
+	ok1(tdb_check(tdb, NULL, NULL) == -1);
+	ok1(tdb_error(tdb) == TDB_ERR_CORRUPT);
+	tdb_close(tdb);
+
+	/* Big and little endian should work! */
+	tdb = tdb_open_ex("test/old-nohash-le.tdb", 1024, 0, O_RDWR, 0,
+			  &taplogctx, NULL);
+	ok1(tdb);
+	ok1(tdb_check(tdb, NULL, NULL) == 0);
+	tdb_close(tdb);
+
+	tdb = tdb_open_ex("test/old-nohash-be.tdb", 1024, 0, O_RDWR, 0,
+			  &taplogctx, NULL);
+	ok1(tdb);
+	ok1(tdb_check(tdb, NULL, NULL) == 0);
+	tdb_close(tdb);
+
+	return exit_status();
+}
diff --git a/ctdb/lib/tdb/test/run-corrupt.c b/ctdb/lib/tdb/test/run-corrupt.c
new file mode 100644
index 0000000000..1a3c769183
--- /dev/null
+++ b/ctdb/lib/tdb/test/run-corrupt.c
@@ -0,0 +1,131 @@
+#include "../common/tdb_private.h"
+#include "../common/io.c"
+#include "../common/tdb.c"
+#include "../common/lock.c"
+#include "../common/freelist.c"
+#include "../common/traverse.c"
+#include "../common/transaction.c"
+#include "../common/error.c"
+#include "../common/open.c"
+#include "../common/check.c"
+#include "../common/hash.c"
+#include "tap-interface.h"
+#include <stdlib.h>
+#include "logging.h"
+
+static int check(TDB_DATA key, TDB_DATA data, void *private)
+{
+	unsigned int *sizes = private;
+
+	if (key.dsize > strlen("hello"))
+		return -1;
+	if (memcmp(key.dptr, "hello", key.dsize) != 0)
+		return -1;
+
+	if (data.dsize != strlen("world"))
+		return -1;
+	if (memcmp(data.dptr, "world", data.dsize) != 0)
+		return -1;
+
+	sizes[0] += key.dsize;
+	sizes[1] += data.dsize;
+	return 0;
+}
+
+static void tdb_flip_bit(struct tdb_context *tdb, unsigned int bit)
+{
+	unsigned int off = bit / CHAR_BIT;
+	unsigned char mask = (1 << (bit % CHAR_BIT));
+
+	if (tdb->map_ptr)
+		((unsigned char *)tdb->map_ptr)[off] ^= mask;
+	else {
+		unsigned char c;
+		if (pread(tdb->fd, &c, 1, off) != 1) {
+			fprintf(stderr, "pread: %s\n", strerror(errno));
+			exit(1);
+		}
+		c ^= mask;
+		if (pwrite(tdb->fd, &c, 1, off) != 1) {
+			fprintf(stderr, "pwrite: %s\n", strerror(errno));
+			exit(1);
+		}
+	}
+}
+
+static void check_test(struct tdb_context *tdb)
+{
+	TDB_DATA key, data;
+	unsigned int i, verifiable, corrupt, sizes[2], dsize, ksize;
+
+	ok1(tdb_check(tdb, NULL, NULL) == 0);
+
+	key.dptr = (void *)"hello";
+	data.dsize = strlen("world");
+	data.dptr = (void *)"world";
+
+	/* Key and data size respectively. */
+	dsize = ksize = 0;
+
+	/* 5 keys in hash size 2 means we'll have multichains. */
+	for (key.dsize = 1; key.dsize <= 5; key.dsize++) {
+		ksize += key.dsize;
+		dsize += data.dsize;
+		if (tdb_store(tdb, key, data, TDB_INSERT) != 0)
+			abort();
+	}
+
+	/* This is how many bytes we expect to be verifiable. */
+	/* From the file header. */
+	verifiable = strlen(TDB_MAGIC_FOOD) + 1
+		+ 2 * sizeof(uint32_t) + 2 * sizeof(tdb_off_t)
+		+ 2 * sizeof(uint32_t);
+	/* From the free list chain and hash chains. */
+	verifiable += 3 * sizeof(tdb_off_t);
+	/* From the record headers & tailer */
+	verifiable += 5 * (sizeof(struct tdb_record) + sizeof(uint32_t));
+	/* The free block: we ignore datalen, keylen, full_hash. */
+	verifiable += sizeof(struct tdb_record) - 3*sizeof(uint32_t) +
+		sizeof(uint32_t);
+	/* Our check function verifies the key and data. */
+	verifiable += ksize + dsize;
+
+	/* Flip one bit at a time, make sure it detects verifiable bytes. */
+	for (i = 0, corrupt = 0; i < tdb->map_size * CHAR_BIT; i++) {
+		tdb_flip_bit(tdb, i);
+		memset(sizes, 0, sizeof(sizes));
+		if (tdb_check(tdb, check, sizes) != 0)
+			corrupt++;
+		else if (sizes[0] != ksize || sizes[1] != dsize)
+			corrupt++;
+		tdb_flip_bit(tdb, i);
+	}
+	ok(corrupt == verifiable * CHAR_BIT, "corrupt %u should be %u",
+	   corrupt, verifiable * CHAR_BIT);
+}
+
+int main(int argc, char *argv[])
+{
+	struct tdb_context *tdb;
+
+	plan_tests(4);
+	/* This should use mmap. */
+	tdb = tdb_open_ex("run-corrupt.tdb", 2, TDB_CLEAR_IF_FIRST,
+			  O_CREAT|O_TRUNC|O_RDWR, 0600, &taplogctx, NULL);
+
+	if (!tdb)
+		abort();
+	check_test(tdb);
+	tdb_close(tdb);
+
+	/* This should not. */
+	tdb = tdb_open_ex("run-corrupt.tdb", 2, TDB_CLEAR_IF_FIRST|TDB_NOMMAP,
+			  O_CREAT|O_TRUNC|O_RDWR, 0600, &taplogctx, NULL);
+
+	if (!tdb)
+		abort();
+	check_test(tdb);
+	tdb_close(tdb);
+
+	return exit_status();
+}
diff --git a/ctdb/lib/tdb/test/run-die-during-transaction.c b/ctdb/lib/tdb/test/run-die-during-transaction.c
new file mode 100644
index 0000000000..6e3a70d4ae
--- /dev/null
+++ b/ctdb/lib/tdb/test/run-die-during-transaction.c
@@ -0,0 +1,231 @@
+#include "../common/tdb_private.h"
+#include "lock-tracking.h"
+static ssize_t pwrite_check(int fd, const void *buf, size_t count, off_t offset);
+static ssize_t write_check(int fd, const void *buf, size_t count);
+static int ftruncate_check(int fd, off_t length);
+
+#define pwrite pwrite_check
+#define write write_check
+#define fcntl fcntl_with_lockcheck
+#define ftruncate ftruncate_check
+
+#include "../common/io.c"
+#include "../common/tdb.c"
+#include "../common/lock.c"
+#include "../common/freelist.c"
+#include "../common/traverse.c"
+#include "../common/transaction.c"
+#include "../common/error.c"
+#include "../common/open.c"
+#include "../common/check.c"
+#include "../common/hash.c"
+#include "tap-interface.h"
+#include <stdlib.h>
+#include <stdbool.h>
+#include <stdarg.h>
+#include <setjmp.h>
+#include "external-agent.h"
+#include "logging.h"
+
+#undef write
+#undef pwrite
+#undef fcntl
+#undef ftruncate
+
+static bool in_transaction;
+static int target, current;
+static jmp_buf jmpbuf;
+#define TEST_DBNAME "run-die-during-transaction.tdb"
+#define KEY_STRING "helloworld"
+
+static void maybe_die(int fd)
+{
+	if (in_transaction && current++ == target) {
+		longjmp(jmpbuf, 1);
+	}
+}
+
+static ssize_t pwrite_check(int fd,
+			    const void *buf, size_t count, off_t offset)
+{
+	ssize_t ret;
+
+	maybe_die(fd);
+
+	ret = pwrite(fd, buf, count, offset);
+	if (ret != count)
+		return ret;
+
+	maybe_die(fd);
+	return ret;
+}
+
+static ssize_t write_check(int fd, const void *buf, size_t count)
+{
+	ssize_t ret;
+
+	maybe_die(fd);
+
+	ret = write(fd, buf, count);
+	if (ret != count)
+		return ret;
+
+	maybe_die(fd);
+	return ret;
+}
+
+static int ftruncate_check(int fd, off_t length)
+{
+	int ret;
+
+	maybe_die(fd);
+
+	ret = ftruncate(fd, length);
+
+	maybe_die(fd);
+	return ret;
+}
+
+static bool test_death(enum operation op, struct agent *agent)
+{
+	struct tdb_context *tdb = NULL;
+	TDB_DATA key;
+	enum agent_return ret;
+	int needed_recovery = 0;
+
+	current = target = 0;
+reset:
+	unlink(TEST_DBNAME);
+	tdb = tdb_open_ex(TEST_DBNAME, 1024, TDB_NOMMAP,
+			  O_CREAT|O_TRUNC|O_RDWR, 0600, &taplogctx, NULL);
+
+	if (setjmp(jmpbuf) != 0) {
+		/* We're partway through.  Simulate our death. */
+		close(tdb->fd);
+		forget_locking();
+		in_transaction = false;
+
+		ret = external_agent_operation(agent, NEEDS_RECOVERY, "");
+		if (ret == SUCCESS)
+			needed_recovery++;
+		else if (ret != FAILED) {
+			diag("Step %u agent NEEDS_RECOVERY = %s", current,
+			     agent_return_name(ret));
+			return false;
+		}
+
+		ret = external_agent_operation(agent, op, KEY_STRING);
+		if (ret != SUCCESS) {
+			diag("Step %u op %s failed = %s", current,
+			     operation_name(op),
+			     agent_return_name(ret));
+			return false;
+		}
+
+		ret = external_agent_operation(agent, NEEDS_RECOVERY, "");
+		if (ret != FAILED) {
+			diag("Still needs recovery after step %u = %s",
+			     current, agent_return_name(ret));
+			return false;
+		}
+
+		ret = external_agent_operation(agent, CHECK, "");
+		if (ret != SUCCESS) {
+			diag("Step %u check failed = %s", current,
+			     agent_return_name(ret));
+			return false;
+		}
+
+		ret = external_agent_operation(agent, CLOSE, "");
+		if (ret != SUCCESS) {
+			diag("Step %u close failed = %s", current,
+			     agent_return_name(ret));
+			return false;
+		}
+
+		/* Suppress logging as this tries to use closed fd. */
+		suppress_logging = true;
+		suppress_lockcheck = true;
+		tdb_close(tdb);
+		suppress_logging = false;
+		suppress_lockcheck = false;
+		target++;
+		current = 0;
+		goto reset;
+	}
+
+	/* Put key for agent to fetch. */
+	key.dsize = strlen(KEY_STRING);
+	key.dptr = (void *)KEY_STRING;
+	if (tdb_store(tdb, key, key, TDB_INSERT) != 0)
+		return false;
+
+	/* This is the key we insert in transaction. */
+	key.dsize--;
+
+	ret = external_agent_operation(agent, OPEN, TEST_DBNAME);
+	if (ret != SUCCESS) {
+		fprintf(stderr, "Agent failed to open: %s\n",
+			agent_return_name(ret));
+		exit(1);
+	}
+
+	ret = external_agent_operation(agent, FETCH, KEY_STRING);
+	if (ret != SUCCESS) {
+		fprintf(stderr, "Agent failed find key: %s\n",
+			agent_return_name(ret));
+		exit(1);
+	}
+
+	in_transaction = true;
+	if (tdb_transaction_start(tdb) != 0)
+		return false;
+
+	if (tdb_store(tdb, key, key, TDB_INSERT) != 0)
+		return false;
+
+	if (tdb_transaction_commit(tdb) != 0)
+		return false;
+
+	in_transaction = false;
+
+	/* We made it! */
+	diag("Completed %u runs", current);
+	tdb_close(tdb);
+	ret = external_agent_operation(agent, CLOSE, "");
+	if (ret != SUCCESS) {
+		diag("Step %u close failed = %s", current,
+		     agent_return_name(ret));
+		return false;
+	}
+
+#ifdef HAVE_INCOHERENT_MMAP
+	/* This means we always mmap, which makes this test a noop. */
+	ok1(1);
+#else
+	ok1(needed_recovery);
+#endif
+	ok1(locking_errors == 0);
+	ok1(forget_locking() == 0);
+	locking_errors = 0;
+	return true;
+}
+
+int main(int argc, char *argv[])
+{
+	enum operation ops[] = { FETCH, STORE, TRANSACTION_START };
+	struct agent *agent;
+	int i;
+
+	plan_tests(12);
+	unlock_callback = maybe_die;
+
+	agent = prepare_external_agent();
+
+	for (i = 0; i < sizeof(ops)/sizeof(ops[0]); i++) {
+		diag("Testing %s after death", operation_name(ops[i]));
+		ok1(test_death(ops[i], agent));
+	}
+
+	return exit_status();
+}
diff --git a/ctdb/lib/tdb/test/run-endian.c b/ctdb/lib/tdb/test/run-endian.c
new file mode 100644
index 0000000000..b19ffd373f
--- /dev/null
+++ b/ctdb/lib/tdb/test/run-endian.c
@@ -0,0 +1,63 @@
+#include "../common/tdb_private.h"
+#include "../common/io.c"
+#include "../common/tdb.c"
+#include "../common/lock.c"
+#include "../common/freelist.c"
+#include "../common/traverse.c"
+#include "../common/transaction.c"
+#include "../common/error.c"
+#include "../common/open.c"
+#include "../common/check.c"
+#include "../common/hash.c"
+#include "tap-interface.h"
+#include <stdlib.h>
+#include "logging.h"
+
+int main(int argc, char *argv[])
+{
+	struct tdb_context *tdb;
+	TDB_DATA key, data;
+
+	plan_tests(13);
+	tdb = tdb_open_ex("run-endian.tdb", 1024,
+			  TDB_CLEAR_IF_FIRST|TDB_CONVERT,
+			  O_CREAT|O_TRUNC|O_RDWR, 0600, &taplogctx, NULL);
+
+	ok1(tdb);
+	key.dsize = strlen("hi");
+	key.dptr = (void *)"hi";
+	data.dsize = strlen("world");
+	data.dptr = (void *)"world";
+
+	ok1(tdb_store(tdb, key, data, TDB_MODIFY) < 0);
+	ok1(tdb_error(tdb) == TDB_ERR_NOEXIST);
+	ok1(tdb_store(tdb, key, data, TDB_INSERT) == 0);
+	ok1(tdb_store(tdb, key, data, TDB_INSERT) < 0);
+	ok1(tdb_error(tdb) == TDB_ERR_EXISTS);
+	ok1(tdb_store(tdb, key, data, TDB_MODIFY) == 0);
+
+	data = tdb_fetch(tdb, key);
+	ok1(data.dsize == strlen("world"));
+	ok1(memcmp(data.dptr, "world", strlen("world")) == 0);
+	free(data.dptr);
+
+	key.dsize++;
+	data = tdb_fetch(tdb, key);
+	ok1(data.dptr == NULL);
+	tdb_close(tdb);
+
+	/* Reopen: should read it */
+	tdb = tdb_open_ex("run-endian.tdb", 1024, 0, O_RDWR, 0,
+			  &taplogctx, NULL);
+	ok1(tdb);
+
+	key.dsize = strlen("hi");
+	key.dptr = (void *)"hi";
+	data = tdb_fetch(tdb, key);
+	ok1(data.dsize == strlen("world"));
+	ok1(memcmp(data.dptr, "world", strlen("world")) == 0);
+	free(data.dptr);
+	tdb_close(tdb);
+
+	return exit_status();
+}
diff --git a/ctdb/lib/tdb/test/run-incompatible.c b/ctdb/lib/tdb/test/run-incompatible.c
new file mode 100644
index 0000000000..628927c886
--- /dev/null
+++ b/ctdb/lib/tdb/test/run-incompatible.c
@@ -0,0 +1,185 @@
+#include "../common/tdb_private.h"
+#include "../common/io.c"
+#include "../common/tdb.c"
+#include "../common/lock.c"
+#include "../common/freelist.c"
+#include "../common/traverse.c"
+#include "../common/transaction.c"
+#include "../common/error.c"
+#include "../common/open.c"
+#include "../common/check.c"
+#include "../common/hash.c"
+#include "tap-interface.h"
+#include <stdlib.h>
+
+static unsigned int tdb_dumb_hash(TDB_DATA *key)
+{
+	return key->dsize;
+}
+
+static void log_fn(struct tdb_context *tdb, enum tdb_debug_level level, const char *fmt, ...)
+{
+	unsigned int *count = tdb_get_logging_private(tdb);
+	if (strstr(fmt, "hash"))
+		(*count)++;
+}
+
+static unsigned int hdr_rwlocks(const char *fname)
+{
+	struct tdb_header hdr;
+
+	int fd = open(fname, O_RDONLY);
+	if (fd == -1)
+		return -1;
+
+	if (read(fd, &hdr, sizeof(hdr)) != sizeof(hdr))
+		return -1;
+
+	close(fd);
+	return hdr.rwlocks;
+}
+
+int main(int argc, char *argv[])
+{
+	struct tdb_context *tdb;
+	unsigned int log_count, flags;
+	TDB_DATA d, r;
+	struct tdb_logging_context log_ctx = { log_fn, &log_count };
+
+	plan_tests(38 * 2);
+
+	for (flags = 0; flags <= TDB_CONVERT; flags += TDB_CONVERT) {
+		unsigned int rwmagic = TDB_HASH_RWLOCK_MAGIC;
+
+		if (flags & TDB_CONVERT)
+			tdb_convert(&rwmagic, sizeof(rwmagic));
+
+		/* Create an old-style hash. */
+		log_count = 0;
+		tdb = tdb_open_ex("run-incompatible.tdb", 0, flags,
+				  O_CREAT|O_RDWR|O_TRUNC, 0600, &log_ctx,
+				  NULL);
+		ok1(tdb);
+		ok1(log_count == 0);
+		d.dptr = (void *)"Hello";
+		d.dsize = 5;
+		ok1(tdb_store(tdb, d, d, TDB_INSERT) == 0);
+		tdb_close(tdb);
+
+		/* Should not have marked rwlocks field. */
+		ok1(hdr_rwlocks("run-incompatible.tdb") == 0);
+
+		/* We can still open any old-style with incompat flag. */
+		log_count = 0;
+		tdb = tdb_open_ex("run-incompatible.tdb", 0,
+				  TDB_INCOMPATIBLE_HASH,
+				  O_RDWR, 0600, &log_ctx, NULL);
+		ok1(tdb);
+		ok1(log_count == 0);
+		r = tdb_fetch(tdb, d);
+		ok1(r.dsize == 5);
+		free(r.dptr);
+		ok1(tdb_check(tdb, NULL, NULL) == 0);
+		tdb_close(tdb);
+
+		log_count = 0;
+		tdb = tdb_open_ex("test/jenkins-le-hash.tdb", 0, 0, O_RDONLY,
+				  0, &log_ctx, tdb_jenkins_hash);
+		ok1(tdb);
+		ok1(log_count == 0);
+		ok1(tdb_check(tdb, NULL, NULL) == 0);
+		tdb_close(tdb);
+
+		log_count = 0;
+		tdb = tdb_open_ex("test/jenkins-be-hash.tdb", 0, 0, O_RDONLY,
+				  0, &log_ctx, tdb_jenkins_hash);
+		ok1(tdb);
+		ok1(log_count == 0);
+		ok1(tdb_check(tdb, NULL, NULL) == 0);
+		tdb_close(tdb);
+
+		/* OK, now create with incompatible flag, default hash. */
+		log_count = 0;
+		tdb = tdb_open_ex("run-incompatible.tdb", 0,
+				  flags|TDB_INCOMPATIBLE_HASH,
+				  O_CREAT|O_RDWR|O_TRUNC, 0600, &log_ctx,
+				  NULL);
+		ok1(tdb);
+		ok1(log_count == 0);
+		d.dptr = (void *)"Hello";
+		d.dsize = 5;
+		ok1(tdb_store(tdb, d, d, TDB_INSERT) == 0);
+		tdb_close(tdb);
+
+		/* Should have marked rwlocks field. */
+		ok1(hdr_rwlocks("run-incompatible.tdb") == rwmagic);
+
+		/* Cannot open with old hash. */
+		log_count = 0;
+		tdb = tdb_open_ex("run-incompatible.tdb", 0, 0,
+				  O_RDWR, 0600, &log_ctx, tdb_old_hash);
+		ok1(!tdb);
+		ok1(log_count == 1);
+
+		/* Can open with jenkins hash. */
+		log_count = 0;
+		tdb = tdb_open_ex("run-incompatible.tdb", 0, 0,
+				  O_RDWR, 0600, &log_ctx, tdb_jenkins_hash);
+		ok1(tdb);
+		ok1(log_count == 0);
+		r = tdb_fetch(tdb, d);
+		ok1(r.dsize == 5);
+		free(r.dptr);
+		ok1(tdb_check(tdb, NULL, NULL) == 0);
+		tdb_close(tdb);
+
+		/* Can open by letting it figure it out itself. */
+		log_count = 0;
+		tdb = tdb_open_ex("run-incompatible.tdb", 0, 0,
+				  O_RDWR, 0600, &log_ctx, NULL);
+		ok1(tdb);
+		ok1(log_count == 0);
+		r = tdb_fetch(tdb, d);
+		ok1(r.dsize == 5);
+		free(r.dptr);
+		ok1(tdb_check(tdb, NULL, NULL) == 0);
+		tdb_close(tdb);
+
+		/* We can also use incompatible hash with other hashes. */
+		log_count = 0;
+		tdb = tdb_open_ex("run-incompatible.tdb", 0,
+				  flags|TDB_INCOMPATIBLE_HASH,
+				  O_CREAT|O_RDWR|O_TRUNC, 0600, &log_ctx,
+				  tdb_dumb_hash);
+		ok1(tdb);
+		ok1(log_count == 0);
+		d.dptr = (void *)"Hello";
+		d.dsize = 5;
+		ok1(tdb_store(tdb, d, d, TDB_INSERT) == 0);
+		tdb_close(tdb);
+
+		/* Should have marked rwlocks field. */
+		ok1(hdr_rwlocks("run-incompatible.tdb") == rwmagic);
+
+		/* It should not open if we don't specify. */
+		log_count = 0;
+		tdb = tdb_open_ex("run-incompatible.tdb", 0, 0, O_RDWR, 0,
+				  &log_ctx, NULL);
+		ok1(!tdb);
+		ok1(log_count == 1);
+
+		/* Should reopen with correct hash. */
+		log_count = 0;
+		tdb = tdb_open_ex("run-incompatible.tdb", 0, 0, O_RDWR, 0,
+				  &log_ctx, tdb_dumb_hash);
+		ok1(tdb);
+		ok1(log_count == 0);
+		r = tdb_fetch(tdb, d);
+		ok1(r.dsize == 5);
+		free(r.dptr);
+		ok1(tdb_check(tdb, NULL, NULL) == 0);
+		tdb_close(tdb);
+	}
+
+	return exit_status();
+}
diff --git a/ctdb/lib/tdb/test/run-nested-transactions.c b/ctdb/lib/tdb/test/run-nested-transactions.c
new file mode 100644
index 0000000000..8c84bcac83
--- /dev/null
+++ b/ctdb/lib/tdb/test/run-nested-transactions.c
@@ -0,0 +1,78 @@
+#include "../common/tdb_private.h"
+#include "../common/io.c"
+#include "../common/tdb.c"
+#include "../common/lock.c"
+#include "../common/freelist.c"
+#include "../common/traverse.c"
+#include "../common/transaction.c"
+#include "../common/error.c"
+#include "../common/open.c"
+#include "../common/check.c"
+#include "../common/hash.c"
+#include "tap-interface.h"
+#include <stdlib.h>
+#include <stdbool.h>
+#include "logging.h"
+
+int main(int argc, char *argv[])
+{
+	struct tdb_context *tdb;
+	TDB_DATA key, data;
+
+	plan_tests(27);
+	key.dsize = strlen("hi");
+	key.dptr = (void *)"hi";
+
+	tdb = tdb_open_ex("run-nested-transactions.tdb",
+			  1024, TDB_CLEAR_IF_FIRST|TDB_DISALLOW_NESTING,
+			  O_CREAT|O_TRUNC|O_RDWR, 0600, &taplogctx, NULL);
+	ok1(tdb);
+
+	/* Nesting disallowed. */
+	ok1(tdb_transaction_start(tdb) == 0);
+	data.dptr = (void *)"world";
+	data.dsize = strlen("world");
+	ok1(tdb_store(tdb, key, data, TDB_INSERT) == 0);
+	data = tdb_fetch(tdb, key);
+	ok1(data.dsize == strlen("world"));
+	ok1(memcmp(data.dptr, "world", strlen("world")) == 0);
+	free(data.dptr);
+	ok1(tdb_transaction_start(tdb) != 0);
+	ok1(tdb_error(tdb) == TDB_ERR_NESTING);
+
+	data = tdb_fetch(tdb, key);
+	ok1(data.dsize == strlen("world"));
+	ok1(memcmp(data.dptr, "world", strlen("world")) == 0);
+	free(data.dptr);
+	ok1(tdb_transaction_commit(tdb) == 0);
+	data = tdb_fetch(tdb, key);
+	ok1(data.dsize == strlen("world"));
+	ok1(memcmp(data.dptr, "world", strlen("world")) == 0);
+	free(data.dptr);
+	tdb_close(tdb);
+
+	/* Nesting allowed by default */
+	tdb = tdb_open_ex("run-nested-transactions.tdb",
+			  1024, TDB_DEFAULT, O_RDWR, 0, &taplogctx, NULL);
+	ok1(tdb);
+
+	ok1(tdb_transaction_start(tdb) == 0);
+	ok1(tdb_transaction_start(tdb) == 0);
+	ok1(tdb_delete(tdb, key) == 0);
+	ok1(tdb_transaction_commit(tdb) == 0);
+	ok1(!tdb_exists(tdb, key));
+	ok1(tdb_transaction_cancel(tdb) == 0);
+	/* Surprise! Kills inner "committed" transaction. */
+	ok1(tdb_exists(tdb, key));
+
+	ok1(tdb_transaction_start(tdb) == 0);
+	ok1(tdb_transaction_start(tdb) == 0);
+	ok1(tdb_delete(tdb, key) == 0);
+	ok1(tdb_transaction_commit(tdb) == 0);
+	ok1(!tdb_exists(tdb, key));
+	ok1(tdb_transaction_commit(tdb) == 0);
+	ok1(!tdb_exists(tdb, key));
+	tdb_close(tdb);
+
+	return exit_status();
+}
diff --git a/ctdb/lib/tdb/test/run-nested-traverse.c b/ctdb/lib/tdb/test/run-nested-traverse.c
new file mode 100644
index 0000000000..37d57c0125
--- /dev/null
+++ b/ctdb/lib/tdb/test/run-nested-traverse.c
@@ -0,0 +1,87 @@
+#include "../common/tdb_private.h"
+#include "lock-tracking.h"
+#define fcntl fcntl_with_lockcheck
+#include "../common/io.c"
+#include "../common/tdb.c"
+#include "../common/lock.c"
+#include "../common/freelist.c"
+#include "../common/traverse.c"
+#include "../common/transaction.c"
+#include "../common/error.c"
+#include "../common/open.c"
+#include "../common/check.c"
+#include "../common/hash.c"
+#include "tap-interface.h"
+#undef fcntl
+#include <stdlib.h>
+#include <stdbool.h>
+#include "external-agent.h"
+#include "logging.h"
+
+static struct agent *agent;
+
+static bool correct_key(TDB_DATA key)
+{
+	return key.dsize == strlen("hi")
+		&& memcmp(key.dptr, "hi", key.dsize) == 0;
+}
+
+static bool correct_data(TDB_DATA data)
+{
+	return data.dsize == strlen("world")
+		&& memcmp(data.dptr, "world", data.dsize) == 0;
+}
+
+static int traverse2(struct tdb_context *tdb, TDB_DATA key, TDB_DATA data,
+		     void *p)
+{
+	ok1(correct_key(key));
+	ok1(correct_data(data));
+	return 0;
+}
+
+static int traverse1(struct tdb_context *tdb, TDB_DATA key, TDB_DATA data,
+		     void *p)
+{
+	ok1(correct_key(key));
+	ok1(correct_data(data));
+	ok1(external_agent_operation(agent, TRANSACTION_START, tdb_name(tdb))
+	    == WOULD_HAVE_BLOCKED);
+	tdb_traverse(tdb, traverse2, NULL);
+
+	/* That should *not* release the transaction lock! */
+	ok1(external_agent_operation(agent, TRANSACTION_START, tdb_name(tdb))
+	    == WOULD_HAVE_BLOCKED);
+	return 0;
+}
+
+int main(int argc, char *argv[])
+{
+	struct tdb_context *tdb;
+	TDB_DATA key, data;
+
+	plan_tests(17);
+	agent = prepare_external_agent();
+
+	tdb = tdb_open_ex("run-nested-traverse.tdb", 1024, TDB_CLEAR_IF_FIRST,
+			  O_CREAT|O_TRUNC|O_RDWR, 0600, &taplogctx, NULL);
+	ok1(tdb);
+
+	ok1(external_agent_operation(agent, OPEN, tdb_name(tdb)) == SUCCESS);
+	ok1(external_agent_operation(agent, TRANSACTION_START, tdb_name(tdb))
+	    == SUCCESS);
+	ok1(external_agent_operation(agent, TRANSACTION_COMMIT, tdb_name(tdb))
+	    == SUCCESS);
+
+	key.dsize = strlen("hi");
+	key.dptr = (void *)"hi";
+	data.dptr = (void *)"world";
+	data.dsize = strlen("world");
+
+	ok1(tdb_store(tdb, key, data, TDB_INSERT) == 0);
+	tdb_traverse(tdb, traverse1, NULL);
+	tdb_traverse_read(tdb, traverse1, NULL);
+	tdb_close(tdb);
+
+	return exit_status();
+}
diff --git a/ctdb/lib/tdb/test/run-no-lock-during-traverse.c b/ctdb/lib/tdb/test/run-no-lock-during-traverse.c
new file mode 100644
index 0000000000..0a72282eb6
--- /dev/null
+++ b/ctdb/lib/tdb/test/run-no-lock-during-traverse.c
@@ -0,0 +1,113 @@
+#include "../common/tdb_private.h"
+#include "lock-tracking.h"
+
+#define fcntl fcntl_with_lockcheck
+
+#include "../common/io.c"
+#include "../common/tdb.c"
+#include "../common/lock.c"
+#include "../common/freelist.c"
+#include "../common/traverse.c"
+#include "../common/transaction.c"
+#include "../common/error.c"
+#include "../common/open.c"
+#include "../common/check.c"
+#include "../common/hash.c"
+#include "tap-interface.h"
+#include <stdlib.h>
+#include "logging.h"
+
+#undef fcntl
+
+#define NUM_ENTRIES 10
+
+static bool prepare_entries(struct tdb_context *tdb)
+{
+	unsigned int i;
+	TDB_DATA key, data;
+
+	for (i = 0; i < NUM_ENTRIES; i++) {
+		key.dsize = sizeof(i);
+		key.dptr = (void *)&i;
+		data.dsize = strlen("world");
+		data.dptr = (void *)"world";
+
+		if (tdb_store(tdb, key, data, 0) != 0)
+			return false;
+	}
+	return true;
+}
+
+static void delete_entries(struct tdb_context *tdb)
+{
+	unsigned int i;
+	TDB_DATA key;
+
+	for (i = 0; i < NUM_ENTRIES; i++) {
+		key.dsize = sizeof(i);
+		key.dptr = (void *)&i;
+
+		ok1(tdb_delete(tdb, key) == 0);
+	}
+}
+
+/* We don't know how many times this will run. */
+static int delete_other(struct tdb_context *tdb, TDB_DATA key, TDB_DATA data,
+			void *private_data)
+{
+	unsigned int i;
+	memcpy(&i, key.dptr, 4);
+	i = (i + 1) % NUM_ENTRIES;
+	key.dptr = (void *)&i;
+	if (tdb_delete(tdb, key) != 0)
+		(*(int *)private_data)++;
+	return 0;
+}
+
+static int delete_self(struct tdb_context *tdb, TDB_DATA key, TDB_DATA data,
+			void *private_data)
+{
+	ok1(tdb_delete(tdb, key) == 0);
+	return 0;
+}
+
+int main(int argc, char *argv[])
+{
+	struct tdb_context *tdb;
+	int errors = 0;
+
+	plan_tests(41);
+	tdb = tdb_open_ex("run-no-lock-during-traverse.tdb",
+			  1024, TDB_CLEAR_IF_FIRST, O_CREAT|O_TRUNC|O_RDWR,
+			  0600, &taplogctx, NULL);
+
+	ok1(tdb);
+	ok1(prepare_entries(tdb));
+	ok1(locking_errors == 0);
+	ok1(tdb_lockall(tdb) == 0);
+	ok1(locking_errors == 0);
+	tdb_traverse(tdb, delete_other, &errors);
+	ok1(errors == 0);
+	ok1(locking_errors == 0);
+	ok1(tdb_unlockall(tdb) == 0);
+
+	ok1(prepare_entries(tdb));
+	ok1(locking_errors == 0);
+	ok1(tdb_lockall(tdb) == 0);
+	ok1(locking_errors == 0);
+	tdb_traverse(tdb, delete_self, NULL);
+	ok1(locking_errors == 0);
+	ok1(tdb_unlockall(tdb) == 0);
+
+	ok1(prepare_entries(tdb));
+	ok1(locking_errors == 0);
+	ok1(tdb_lockall(tdb) == 0);
+	ok1(locking_errors == 0);
+	delete_entries(tdb);
+	ok1(locking_errors == 0);
+	ok1(tdb_unlockall(tdb) == 0);
+
+	ok1(tdb_close(tdb) == 0);
+
+	return exit_status();
+}
diff --git a/ctdb/lib/tdb/test/run-oldhash.c b/ctdb/lib/tdb/test/run-oldhash.c
new file mode 100644
index 0000000000..535336cb47
--- /dev/null
+++ b/ctdb/lib/tdb/test/run-oldhash.c
@@ -0,0 +1,49 @@
+#include "../common/tdb_private.h"
+#include "../common/io.c"
+#include "../common/tdb.c"
+#include "../common/lock.c"
+#include "../common/freelist.c"
+#include "../common/traverse.c"
+#include "../common/transaction.c"
+#include "../common/error.c"
+#include "../common/open.c"
+#include "../common/check.c"
+#include "../common/hash.c"
+#include "tap-interface.h"
+#include <stdlib.h>
+#include "logging.h"
+
+int main(int argc, char *argv[])
+{
+	struct tdb_context *tdb;
+
+	plan_tests(8);
+
+	/* Old format (with zeroes in the hash magic fields) should
+	 * open with any hash (since we don't know what hash they used). */
+	tdb = tdb_open_ex("test/old-nohash-le.tdb", 0, 0, O_RDWR, 0,
+			  &taplogctx, NULL);
+	ok1(tdb);
+	ok1(tdb_check(tdb, NULL, NULL) == 0);
+	tdb_close(tdb);
+
+	tdb = tdb_open_ex("test/old-nohash-be.tdb", 0, 0, O_RDWR, 0,
+			  &taplogctx, NULL);
+	ok1(tdb);
+	ok1(tdb_check(tdb, NULL, NULL) == 0);
+	tdb_close(tdb);
+
+	tdb = tdb_open_ex("test/old-nohash-le.tdb", 0, 0, O_RDWR, 0,
+			  &taplogctx, tdb_jenkins_hash);
+	ok1(tdb);
+	ok1(tdb_check(tdb, NULL, NULL) == 0);
+	tdb_close(tdb);
+
+	tdb = tdb_open_ex("test/old-nohash-be.tdb", 0, 0, O_RDWR, 0,
+			  &taplogctx, tdb_jenkins_hash);
+	ok1(tdb);
+	ok1(tdb_check(tdb, NULL, NULL) == 0);
+	tdb_close(tdb);
+
+	return exit_status();
+}
diff --git a/ctdb/lib/tdb/test/run-open-during-transaction.c b/ctdb/lib/tdb/test/run-open-during-transaction.c
new file mode 100644
index 0000000000..a825e6269d
--- /dev/null
+++ b/ctdb/lib/tdb/test/run-open-during-transaction.c
@@ -0,0 +1,181 @@
+#include "../common/tdb_private.h"
+#include "lock-tracking.h"
+
+static ssize_t pwrite_check(int fd, const void *buf, size_t count, off_t offset);
+static ssize_t write_check(int fd, const void *buf, size_t count);
+static int ftruncate_check(int fd, off_t length);
+
+#define pwrite pwrite_check
+#define write write_check
+#define fcntl fcntl_with_lockcheck
+#define ftruncate ftruncate_check
+
+#include "../common/io.c"
+#include "../common/tdb.c"
+#include "../common/lock.c"
+#include "../common/freelist.c"
+#include "../common/traverse.c"
+#include "../common/transaction.c"
+#include "../common/error.c"
+#include "../common/open.c"
+#include "../common/check.c"
+#include "../common/hash.c"
+#include "tap-interface.h"
+#include <stdlib.h>
+#include <stdbool.h>
+#include <stdarg.h>
+#include "external-agent.h"
+#include "logging.h"
+
+static struct agent *agent;
+static bool opened;
+static int errors = 0;
+static bool clear_if_first;
+#define TEST_DBNAME "run-open-during-transaction.tdb"
+
+#undef write
+#undef pwrite
+#undef fcntl
+#undef ftruncate
+
+static bool is_same(const char *snapshot, const char *latest, off_t len)
+{
+	unsigned i;
+
+	for (i = 0; i < len; i++) {
+		if (snapshot[i] != latest[i])
+			return false;
+	}
+	return true;
+}
+
+static bool compare_file(int fd, const char *snapshot, off_t snapshot_len)
+{
+	char *contents;
+	bool same;
+
+	/* over-length read serves as length check. */
+	contents = malloc(snapshot_len+1);
+	same = pread(fd, contents, snapshot_len+1, 0) == snapshot_len
+		&& is_same(snapshot, contents, snapshot_len);
+	free(contents);
+	return same;
+}
+
+static void check_file_intact(int fd)
+{
+	enum agent_return ret;
+	struct stat st;
+	char *contents;
+
+	fstat(fd, &st);
+	contents = malloc(st.st_size);
+	if (pread(fd, contents, st.st_size, 0) != st.st_size) {
+		diag("Read fail");
+		errors++;
+		return;
+	}
+
+	/* Ask agent to open file. */
+	ret = external_agent_operation(agent, clear_if_first ?
+				       OPEN_WITH_CLEAR_IF_FIRST :
+				       OPEN,
+				       TEST_DBNAME);
+
+	/* It's OK to open it, but it must not have changed! */
+	if (!compare_file(fd, contents, st.st_size)) {
+		diag("Agent changed file after opening %s",
+		     agent_return_name(ret));
+		errors++;
+	}
+
+	if (ret == SUCCESS) {
+		ret = external_agent_operation(agent, CLOSE, NULL);
+		if (ret != SUCCESS) {
+			diag("Agent failed to close tdb: %s",
+			     agent_return_name(ret));
+			errors++;
+		}
+	} else if (ret != WOULD_HAVE_BLOCKED) {
+		diag("Agent opening file gave %s",
+		     agent_return_name(ret));
+		errors++;
+	}
+
+	free(contents);
+}
+
+static void after_unlock(int fd)
+{
+	if (opened)
+		check_file_intact(fd);
+}
+
+static ssize_t pwrite_check(int fd,
+			    const void *buf, size_t count, off_t offset)
+{
+	if (opened)
+		check_file_intact(fd);
+
+	return pwrite(fd, buf, count, offset);
+}
+
+static ssize_t write_check(int fd, const void *buf, size_t count)
+{
+	if (opened)
+		check_file_intact(fd);
+
+	return write(fd, buf, count);
+}
+
+static int ftruncate_check(int fd, off_t length)
+{
+	if (opened)
+		check_file_intact(fd);
+
+	return ftruncate(fd, length);
+
+}
+
+int main(int argc, char *argv[])
+{
+	const int flags[] = { TDB_DEFAULT,
+			      TDB_CLEAR_IF_FIRST,
+			      TDB_NOMMAP,
+			      TDB_CLEAR_IF_FIRST | TDB_NOMMAP };
+	int i;
+	struct tdb_context *tdb;
+	TDB_DATA key, data;
+
+	plan_tests(20);
+	agent = prepare_external_agent();
+
+	unlock_callback = after_unlock;
+	for (i = 0; i < sizeof(flags)/sizeof(flags[0]); i++) {
+		clear_if_first = (flags[i] & TDB_CLEAR_IF_FIRST);
+		diag("Test with %s and %s\n",
+		     clear_if_first ? "CLEAR" : "DEFAULT",
+		     (flags[i] & TDB_NOMMAP) ? "no mmap" : "mmap");
+		unlink(TEST_DBNAME);
+		tdb = tdb_open_ex(TEST_DBNAME, 1024, flags[i],
+				  O_CREAT|O_TRUNC|O_RDWR, 0600,
+				  &taplogctx, NULL);
+		ok1(tdb);
+
+		opened = true;
+		ok1(tdb_transaction_start(tdb) == 0);
+		key.dsize = strlen("hi");
+		key.dptr = (void *)"hi";
+		data.dptr = (void *)"world";
+		data.dsize = strlen("world");
+
+		ok1(tdb_store(tdb, key, data, TDB_INSERT) == 0);
+		ok1(tdb_transaction_commit(tdb) == 0);
+		ok(!errors, "We had %u open errors", errors);
+
+		opened = false;
+		tdb_close(tdb);
+	}
+
+	return exit_status();
+}
diff --git a/ctdb/lib/tdb/test/run-readonly-check.c b/ctdb/lib/tdb/test/run-readonly-check.c
new file mode 100644
index 0000000000..fdd9507cda
--- /dev/null
+++ b/ctdb/lib/tdb/test/run-readonly-check.c
@@ -0,0 +1,52 @@
+/* We should be able to tdb_check a O_RDONLY tdb, and we were previously allowed
+ * to tdb_check() inside a transaction (though that's paranoia!). */
+#include "../common/tdb_private.h"
+#include "../common/io.c"
+#include "../common/tdb.c"
+#include "../common/lock.c"
+#include "../common/freelist.c"
+#include "../common/traverse.c"
+#include "../common/transaction.c"
+#include "../common/error.c"
+#include "../common/open.c"
+#include "../common/check.c"
+#include "../common/hash.c"
+#include "tap-interface.h"
+#include <stdlib.h>
+#include "logging.h"
+
+int main(int argc, char *argv[])
+{
+	struct tdb_context *tdb;
+	TDB_DATA key, data;
+
+	plan_tests(11);
+	tdb = tdb_open_ex("run-readonly-check.tdb", 1024,
+			  TDB_DEFAULT,
+			  O_CREAT|O_TRUNC|O_RDWR, 0600, &taplogctx, NULL);
+
+	ok1(tdb);
+	key.dsize = strlen("hi");
+	key.dptr = (void *)"hi";
+	data.dsize = strlen("world");
+	data.dptr = (void *)"world";
+
+	ok1(tdb_store(tdb, key, data, TDB_INSERT) == 0);
+	ok1(tdb_check(tdb, NULL, NULL) == 0);
+
+	/* We are also allowed to do a check inside a transaction. */
+	ok1(tdb_transaction_start(tdb) == 0);
+	ok1(tdb_check(tdb, NULL, NULL) == 0);
+	ok1(tdb_close(tdb) == 0);
+
+	tdb = tdb_open_ex("run-readonly-check.tdb", 1024,
+			  TDB_DEFAULT, O_RDONLY, 0, &taplogctx, NULL);
+
+	ok1(tdb);
+	ok1(tdb_store(tdb, key, data, TDB_MODIFY) == -1);
+	ok1(tdb_error(tdb) == TDB_ERR_RDONLY);
+	ok1(tdb_check(tdb, NULL, NULL) == 0);
+	ok1(tdb_close(tdb) == 0);
+
+	return exit_status();
+}
diff --git a/ctdb/lib/tdb/test/run-rescue-find_entry.c b/ctdb/lib/tdb/test/run-rescue-find_entry.c
new file mode 100644
index 0000000000..25f4f1c05f
--- /dev/null
+++ b/ctdb/lib/tdb/test/run-rescue-find_entry.c
@@ -0,0 +1,50 @@
+#include "../common/tdb_private.h"
+#include "../common/io.c"
+#include "../common/tdb.c"
+#include "../common/lock.c"
+#include "../common/freelist.c"
+#include "../common/traverse.c"
+#include "../common/transaction.c"
+#include "../common/error.c"
+#include "../common/open.c"
+#include "../common/check.c"
+#include "../common/hash.c"
+#include "../common/rescue.c"
+#include "tap-interface.h"
+#include <stdlib.h>
+#include "logging.h"
+
+#define NUM 20
+
+/* Binary searches are deceptively simple: easy to screw up! */
+int main(int argc, char *argv[])
+{
+	unsigned int i, j, n;
+	struct found f[NUM+1];
+	struct found_table table;
+
+	/* Set up array for searching. */
+	for (i = 0; i < NUM+1; i++) {
+		f[i].head = i * 3;
+	}
+	table.arr = f;
+
+	for (i = 0; i < NUM; i++) {
+		table.num = i;
+		for (j = 0; j < (i + 2) * 3; j++) {
+			n = find_entry(&table, j);
+			ok1(n <= i);
+
+			/* If we were searching for something too large... */
+			if (j > i*3)
+				ok1(n == i);
+			else {
+				/* It must give us something after j */
+				ok1(f[n].head >= j);
+				ok1(n == 0 || f[n-1].head < j);
+			}
+		}
+	}
+
+	return exit_status();
+}
diff --git a/ctdb/lib/tdb/test/run-rescue.c b/ctdb/lib/tdb/test/run-rescue.c
new file mode 100644
index 0000000000..a26c493972
--- /dev/null
+++ b/ctdb/lib/tdb/test/run-rescue.c
@@ -0,0 +1,126 @@
+#include "../common/tdb_private.h"
+#include "../common/io.c"
+#include "../common/tdb.c"
+#include "../common/lock.c"
+#include "../common/freelist.c"
+#include "../common/traverse.c"
+#include "../common/transaction.c"
+#include "../common/error.c"
+#include "../common/open.c"
+#include "../common/check.c"
+#include "../common/hash.c"
+#include "../common/rescue.c"
+#include "tap-interface.h"
+#include <stdlib.h>
+#include "logging.h"
+
+struct walk_data {
+	TDB_DATA key;
+	TDB_DATA data;
+	bool fail;
+	unsigned count;
+};
+
+static inline bool tdb_deq(TDB_DATA a, TDB_DATA b)
+{
+	return a.dsize == b.dsize && memcmp(a.dptr, b.dptr, a.dsize) == 0;
+}
+
+static inline TDB_DATA tdb_mkdata(const void *p, size_t len)
+{
+	TDB_DATA d;
+	d.dptr = (void *)p;
+	d.dsize = len;
+	return d;
+}
+
+static void walk(TDB_DATA key, TDB_DATA data, void *_wd)
+{
+	struct walk_data *wd = _wd;
+
+	if (!tdb_deq(key, wd->key)) {
+		wd->fail = true;
+	}
+
+	if (!tdb_deq(data, wd->data)) {
+		wd->fail = true;
+	}
+	wd->count++;
+}
+
+static void count_records(TDB_DATA key, TDB_DATA data, void *_wd)
+{
+	struct walk_data *wd = _wd;
+
+	if (!tdb_deq(key, wd->key) || !tdb_deq(data, wd->data))
+		diag("%.*s::%.*s\n",
+		     (int)key.dsize, key.dptr, (int)data.dsize, data.dptr);
+	wd->count++;
+}
+
+static void log_fn(struct tdb_context *tdb, enum tdb_debug_level level, const char *fmt, ...)
+{
+	unsigned int *count = tdb_get_logging_private(tdb);
+	(*count)++;
+}
+
+int main(int argc, char *argv[])
+{
+	struct tdb_context *tdb;
+	struct walk_data wd;
+	unsigned int i, size, log_count = 0;
+	struct tdb_logging_context log_ctx = { log_fn, &log_count };
+
+	plan_tests(8);
+	tdb = tdb_open_ex("run-rescue.tdb", 1, TDB_CLEAR_IF_FIRST,
+			  O_CREAT|O_TRUNC|O_RDWR, 0600, &log_ctx, NULL);
+
+	wd.key.dsize = strlen("hi");
+	wd.key.dptr = (void *)"hi";
+	wd.data.dsize = strlen("world");
+	wd.data.dptr = (void *)"world";
+	wd.count = 0;
+	wd.fail = false;
+
+	ok1(tdb_store(tdb, wd.key, wd.data, TDB_INSERT) == 0);
+
+	ok1(tdb_rescue(tdb, walk, &wd) == 0);
+	ok1(!wd.fail);
+	ok1(wd.count == 1);
+
+	/* Corrupt the database, walk should either get it or not. */
+	size = tdb->map_size;
+	for (i = sizeof(struct tdb_header); i < size; i++) {
+		char c;
+		if (tdb->methods->tdb_read(tdb, i, &c, 1, false) != 0)
+			fail("Reading offset %i", i);
+		if (tdb->methods->tdb_write(tdb, i, "X", 1) != 0)
+			fail("Writing X at offset %i", i);
+
+		wd.count = 0;
+		if (tdb_rescue(tdb, count_records, &wd) != 0) {
+			wd.fail = true;
+			break;
+		}
+		/* Could be 0 or 1. */
+		if (wd.count > 1) {
+			wd.fail = true;
+			break;
+		}
+		if (tdb->methods->tdb_write(tdb, i, &c, 1) != 0)
+			fail("Restoring offset %i", i);
+	}
+	ok1(log_count == 0);
+	ok1(!wd.fail);
+	tdb_close(tdb);
+
+	/* Now try our known-corrupt db. */
+	tdb = tdb_open_ex("test/tdb.corrupt", 1024, 0, O_RDWR, 0,
+			  &taplogctx, NULL);
+	wd.count = 0;
+	ok1(tdb_rescue(tdb, count_records, &wd) == 0);
+	ok1(wd.count == 1627);
+	tdb_close(tdb);
+
+	return exit_status();
+}
diff --git a/ctdb/lib/tdb/test/run-rwlock-check.c b/ctdb/lib/tdb/test/run-rwlock-check.c
new file mode 100644
index 0000000000..8b8072db1e
--- /dev/null
+++ b/ctdb/lib/tdb/test/run-rwlock-check.c
@@ -0,0 +1,45 @@
+#include "../common/tdb_private.h"
+#include "../common/io.c"
+#include "../common/tdb.c"
+#include "../common/lock.c"
+#include "../common/freelist.c"
+#include "../common/traverse.c"
+#include "../common/transaction.c"
+#include "../common/error.c"
+#include "../common/open.c"
+#include "../common/check.c"
+#include "../common/hash.c"
+#include "tap-interface.h"
+#include <stdlib.h>
+
+static void log_fn(struct tdb_context *tdb, enum tdb_debug_level level, const char *fmt, ...)
+{
+	unsigned int *count = tdb_get_logging_private(tdb);
+	if (strstr(fmt, "spinlocks"))
+		(*count)++;
+}
+
+/* The code should barf on TDBs created with rwlocks. */
+int main(int argc, char *argv[])
+{
+	struct tdb_context *tdb;
+	unsigned int log_count;
+	struct tdb_logging_context log_ctx = { log_fn, &log_count };
+
+	plan_tests(4);
+
+	/* We should fail to open rwlock-using tdbs of either endian. */
+	log_count = 0;
+	tdb = tdb_open_ex("test/rwlock-le.tdb", 0, 0, O_RDWR, 0,
+			  &log_ctx, NULL);
+	ok1(!tdb);
+	ok1(log_count == 1);
+
+	log_count = 0;
+	tdb = tdb_open_ex("test/rwlock-be.tdb", 0, 0, O_RDWR, 0,
+			  &log_ctx, NULL);
+	ok1(!tdb);
+	ok1(log_count == 1);
+
+	return exit_status();
+}
diff --git a/ctdb/lib/tdb/test/run-summary.c b/ctdb/lib/tdb/test/run-summary.c
new file mode 100644
index 0000000000..22312843e7
--- /dev/null
+++ b/ctdb/lib/tdb/test/run-summary.c
@@ -0,0 +1,64 @@
+#include "../common/tdb_private.h"
+#include "../common/io.c"
+#include "../common/tdb.c"
+#include "../common/lock.c"
+#include "../common/freelist.c"
+#include "../common/traverse.c"
+#include "../common/transaction.c"
+#include "../common/error.c"
+#include "../common/open.c"
+#include "../common/check.c"
+#include "../common/hash.c"
+#include "../common/summary.c"
+#include "tap-interface.h"
+#include <stdlib.h>
+
+int main(int argc, char *argv[])
+{
+	unsigned int i, j;
+	struct tdb_context *tdb;
+	int flags[] = { TDB_INTERNAL, TDB_DEFAULT, TDB_NOMMAP,
+			TDB_INTERNAL|TDB_CONVERT, TDB_CONVERT,
+			TDB_NOMMAP|TDB_CONVERT };
+	TDB_DATA key = { (unsigned char *)&j, sizeof(j) };
+	TDB_DATA data = { (unsigned char *)&j, sizeof(j) };
+	char *summary;
+
+	plan_tests(sizeof(flags) / sizeof(flags[0]) * 14);
+	for (i = 0; i < sizeof(flags) / sizeof(flags[0]); i++) {
+		tdb = tdb_open("run-summary.tdb", 131, flags[i],
+			       O_RDWR|O_CREAT|O_TRUNC, 0600);
+		ok1(tdb);
+		if (!tdb)
+			continue;
+
+		/* Put some stuff in there. */
+		for (j = 0; j < 500; j++) {
+			/* Make sure padding varies to we get some graphs! */
+			data.dsize = j % (sizeof(j) + 1);
+			if (tdb_store(tdb, key, data, TDB_REPLACE) != 0)
+				fail("Storing in tdb");
+		}
+
+		summary = tdb_summary(tdb);
+		diag("%s", summary);
+		ok1(strstr(summary, "Size of file/data: "));
+		ok1(strstr(summary, "Number of records: 500\n"));
+		ok1(strstr(summary, "Smallest/average/largest keys: 4/4/4\n"));
+		ok1(strstr(summary, "Smallest/average/largest data: 0/2/4\n"));
+		ok1(strstr(summary, "Smallest/average/largest padding: "));
+		ok1(strstr(summary, "Number of dead records: 0\n"));
+		ok1(strstr(summary, "Number of free records: 1\n"));
+		ok1(strstr(summary, "Smallest/average/largest free records: "));
+		ok1(strstr(summary, "Number of hash chains: 131\n"));
+		ok1(strstr(summary, "Smallest/average/largest hash chains: "));
+		ok1(strstr(summary, "Number of uncoalesced records: 0\n"));
+		ok1(strstr(summary, "Smallest/average/largest uncoalesced runs: 0/0/0\n"));
+		ok1(strstr(summary, "Percentage keys/data/padding/free/dead/rechdrs&tailers/hashes: "));
+
+		free(summary);
+		tdb_close(tdb);
+	}
+
+	return exit_status();
+}
diff --git a/ctdb/lib/tdb/test/run-transaction-expand.c b/ctdb/lib/tdb/test/run-transaction-expand.c
new file mode 100644
index 0000000000..1271d92b33
--- /dev/null
+++ b/ctdb/lib/tdb/test/run-transaction-expand.c
@@ -0,0 +1,119 @@
+#include "../common/tdb_private.h"
+
+/* Speed up the tests, but do the actual sync tests. */
+static unsigned int sync_counts = 0;
+static inline int fake_fsync(int fd)
+{
+	sync_counts++;
+	return 0;
+}
+#define fsync fake_fsync
+
+#ifdef MS_SYNC
+static inline int fake_msync(void *addr, size_t length, int flags)
+{
+	sync_counts++;
+	return 0;
+}
+#define msync fake_msync
+#endif
+
+#ifdef HAVE_FDATASYNC
+static inline int fake_fdatasync(int fd)
+{
+	sync_counts++;
+	return 0;
+}
+#define fdatasync fake_fdatasync
+#endif
+
+#include "../common/io.c"
+#include "../common/tdb.c"
+#include "../common/lock.c"
+#include "../common/freelist.c"
+#include "../common/traverse.c"
+#include "../common/transaction.c"
+#include "../common/error.c"
+#include "../common/open.c"
+#include "../common/check.c"
+#include "../common/hash.c"
+#include "tap-interface.h"
+#include <stdlib.h>
+#include "logging.h"
+
+static void write_record(struct tdb_context *tdb, size_t extra_len,
+			 TDB_DATA *data)
+{
+	TDB_DATA key;
+	key.dsize = strlen("hi");
+	key.dptr = (void *)"hi";
+
+	data->dsize += extra_len;
+	tdb_transaction_start(tdb);
+	tdb_store(tdb, key, *data, TDB_REPLACE);
+	tdb_transaction_commit(tdb);
+}
+
+int main(int argc, char *argv[])
+{
+	struct tdb_context *tdb;
+	size_t i;
+	TDB_DATA data;
+	struct tdb_record rec;
+	tdb_off_t off;
+
+	/* Do *not* suppress sync for this test; we do it ourselves. */
+	unsetenv("TDB_NO_FSYNC");
+
+	plan_tests(5);
+	tdb = tdb_open_ex("run-transaction-expand.tdb",
+			  1024, TDB_CLEAR_IF_FIRST,
+			  O_CREAT|O_TRUNC|O_RDWR, 0600, &taplogctx, NULL);
+	ok1(tdb);
+
+	data.dsize = 0;
+	data.dptr = calloc(1000, getpagesize());
+
+	/* Simulate a slowly growing record. */
+	for (i = 0; i < 1000; i++)
+		write_record(tdb, getpagesize(), &data);
+
+	tdb_ofs_read(tdb, TDB_RECOVERY_HEAD, &off);
+	tdb_read(tdb, off, &rec, sizeof(rec), DOCONV());
+	diag("TDB size = %zu, recovery = %llu-%llu",
+	     (size_t)tdb->map_size, (unsigned long long)off, (unsigned long long)(off + sizeof(rec) + rec.rec_len));
+
+	/* We should only be about 5 times larger than largest record. */
+	ok1(tdb->map_size < 6 * i * getpagesize());
+	tdb_close(tdb);
+
+	tdb = tdb_open_ex("run-transaction-expand.tdb",
+			  1024, TDB_CLEAR_IF_FIRST,
+			  O_CREAT|O_TRUNC|O_RDWR, 0600, &taplogctx, NULL);
+	ok1(tdb);
+
+	data.dsize = 0;
+
+	/* Simulate a slowly growing record, repacking to keep
+	 * recovery area at end. */
+	for (i = 0; i < 1000; i++) {
+		write_record(tdb, getpagesize(), &data);
+		if (i % 10 == 0)
+			tdb_repack(tdb);
+	}
+
+	tdb_ofs_read(tdb, TDB_RECOVERY_HEAD, &off);
+	tdb_read(tdb, off, &rec, sizeof(rec), DOCONV());
+	diag("TDB size = %zu, recovery = %llu-%llu",
+	     (size_t)tdb->map_size, (unsigned long long)off, (unsigned long long)(off + sizeof(rec) + rec.rec_len));
+
+	/* We should only be about 4 times larger than largest record. */
+	ok1(tdb->map_size < 5 * i * getpagesize());
+
+	/* We should have synchronized multiple times. */
+	ok1(sync_counts);
+	tdb_close(tdb);
+	free(data.dptr);
+
+	return exit_status();
+}
diff --git a/ctdb/lib/tdb/test/run-traverse-in-transaction.c b/ctdb/lib/tdb/test/run-traverse-in-transaction.c
new file mode 100644
index 0000000000..bcdc3545cd
--- /dev/null
+++ b/ctdb/lib/tdb/test/run-traverse-in-transaction.c
@@ -0,0 +1,86 @@
+#include "lock-tracking.h"
+#include "../common/tdb_private.h"
+#define fcntl fcntl_with_lockcheck
+#include "../common/io.c"
+#include "../common/tdb.c"
+#include "../common/lock.c"
+#include "../common/freelist.c"
+#include "../common/traverse.c"
+#include "../common/transaction.c"
+#include "../common/error.c"
+#include "../common/open.c"
+#include "../common/check.c"
+#include "../common/hash.c"
+#include "tap-interface.h"
+#undef fcntl_with_lockcheck
+#include <stdlib.h>
+#include <stdbool.h>
+#include "external-agent.h"
+#include "logging.h"
+
+static struct agent *agent;
+
+static bool correct_key(TDB_DATA key)
+{
+	return key.dsize == strlen("hi")
+		&& memcmp(key.dptr, "hi", key.dsize) == 0;
+}
+
+static bool correct_data(TDB_DATA data)
+{
+	return data.dsize == strlen("world")
+		&& memcmp(data.dptr, "world", data.dsize) == 0;
+}
+
+static int traverse(struct tdb_context *tdb, TDB_DATA key, TDB_DATA data,
+		     void *p)
+{
+	ok1(correct_key(key));
+	ok1(correct_data(data));
+	return 0;
+}
+
+int main(int argc, char *argv[])
+{
+	struct tdb_context *tdb;
+	TDB_DATA key, data;
+
+	plan_tests(13);
+	agent = prepare_external_agent();
+
+	tdb = tdb_open_ex("run-traverse-in-transaction.tdb",
+			  1024, TDB_CLEAR_IF_FIRST, O_CREAT|O_TRUNC|O_RDWR,
+			  0600, &taplogctx, NULL);
+	ok1(tdb);
+
+	key.dsize = strlen("hi");
+	key.dptr = (void *)"hi";
+	data.dptr = (void *)"world";
+	data.dsize = strlen("world");
+
+	ok1(tdb_store(tdb, key, data, TDB_INSERT) == 0);
+
+	ok1(external_agent_operation(agent, OPEN, tdb_name(tdb)) == SUCCESS);
+
+	ok1(tdb_transaction_start(tdb) == 0);
+	ok1(external_agent_operation(agent, TRANSACTION_START, tdb_name(tdb))
+	    == WOULD_HAVE_BLOCKED);
+	tdb_traverse(tdb, traverse, NULL);
+
+	/* That should *not* release the transaction lock! */
+	ok1(external_agent_operation(agent, TRANSACTION_START, tdb_name(tdb))
+	    == WOULD_HAVE_BLOCKED);
+	tdb_traverse_read(tdb, traverse, NULL);
+
+	/* That should *not* release the transaction lock! */
+	ok1(external_agent_operation(agent, TRANSACTION_START, tdb_name(tdb))
+	    == WOULD_HAVE_BLOCKED);
+	ok1(tdb_transaction_commit(tdb) == 0);
+	/* Now we should be fine. */
+	ok1(external_agent_operation(agent, TRANSACTION_START, tdb_name(tdb))
+	    == SUCCESS);
+
+	tdb_close(tdb);
+
+	return exit_status();
+}
diff --git a/ctdb/lib/tdb/test/run-wronghash-fail.c b/ctdb/lib/tdb/test/run-wronghash-fail.c
new file mode 100644
index 0000000000..74bbc30ba6
--- /dev/null
+++ b/ctdb/lib/tdb/test/run-wronghash-fail.c
@@ -0,0 +1,120 @@
+#include "../common/tdb_private.h"
+#include "../common/io.c"
+#include "../common/tdb.c"
+#include "../common/lock.c"
+#include "../common/freelist.c"
+#include "../common/traverse.c"
+#include "../common/transaction.c"
+#include "../common/error.c"
+#include "../common/open.c"
+#include "../common/check.c"
+#include "../common/hash.c"
+#include "tap-interface.h"
+#include <stdlib.h>
+
+static void log_fn(struct tdb_context *tdb, enum tdb_debug_level level, const char *fmt, ...)
+{
+	unsigned int *count = tdb_get_logging_private(tdb);
+	if (strstr(fmt, "hash"))
+		(*count)++;
+}
+
+int main(int argc, char *argv[])
+{
+	struct tdb_context *tdb;
+	unsigned int log_count;
+	TDB_DATA d;
+	struct tdb_logging_context log_ctx = { log_fn, &log_count };
+
+	plan_tests(28);
+
+	/* Create with default hash. */
+	log_count = 0;
+	tdb = tdb_open_ex("run-wronghash-fail.tdb", 0, 0,
+			  O_CREAT|O_RDWR|O_TRUNC, 0600, &log_ctx, NULL);
+	ok1(tdb);
+	ok1(log_count == 0);
+	d.dptr = (void *)"Hello";
+	d.dsize = 5;
+	ok1(tdb_store(tdb, d, d, TDB_INSERT) == 0);
+	tdb_close(tdb);
+
+	/* Fail to open with different hash. */
+	tdb = tdb_open_ex("run-wronghash-fail.tdb", 0, 0, O_RDWR, 0,
+			  &log_ctx, tdb_jenkins_hash);
+	ok1(!tdb);
+	ok1(log_count == 1);
+
+	/* Create with different hash. */
+	log_count = 0;
+	tdb = tdb_open_ex("run-wronghash-fail.tdb", 0, 0,
+			  O_CREAT|O_RDWR|O_TRUNC,
+			  0600, &log_ctx, tdb_jenkins_hash);
+	ok1(tdb);
+	ok1(log_count == 0);
+	tdb_close(tdb);
+
+	/* Endian should be no problem. */
+	log_count = 0;
+	tdb = tdb_open_ex("test/jenkins-le-hash.tdb", 0, 0, O_RDWR, 0,
+			  &log_ctx, tdb_old_hash);
+	ok1(!tdb);
+	ok1(log_count == 1);
+
+	log_count = 0;
+	tdb = tdb_open_ex("test/jenkins-be-hash.tdb", 0, 0, O_RDWR, 0,
+			  &log_ctx, tdb_old_hash);
+	ok1(!tdb);
+	ok1(log_count == 1);
+
+	log_count = 0;
+	/* Fail to open with old default hash. */
+	tdb = tdb_open_ex("run-wronghash-fail.tdb", 0, 0, O_RDWR, 0,
+			  &log_ctx, tdb_old_hash);
+	ok1(!tdb);
+	ok1(log_count == 1);
+
+	log_count = 0;
+	tdb = tdb_open_ex("test/jenkins-le-hash.tdb", 0, 0, O_RDONLY,
+			  0, &log_ctx, tdb_jenkins_hash);
+	ok1(tdb);
+	ok1(log_count == 0);
+	ok1(tdb_check(tdb, NULL, NULL) == 0);
+	tdb_close(tdb);
+
+	log_count = 0;
+	tdb = tdb_open_ex("test/jenkins-be-hash.tdb", 0, 0, O_RDONLY,
+			  0, &log_ctx, tdb_jenkins_hash);
+	ok1(tdb);
+	ok1(log_count == 0);
+	ok1(tdb_check(tdb, NULL, NULL) == 0);
+	tdb_close(tdb);
+
+	/* It should open with jenkins hash if we don't specify. */
+	log_count = 0;
+	tdb = tdb_open_ex("test/jenkins-le-hash.tdb", 0, 0, O_RDWR, 0,
+			  &log_ctx, NULL);
+	ok1(tdb);
+	ok1(log_count == 0);
+	ok1(tdb_check(tdb, NULL, NULL) == 0);
+	tdb_close(tdb);
+
+	log_count = 0;
+	tdb = tdb_open_ex("test/jenkins-be-hash.tdb", 0, 0, O_RDWR, 0,
+			  &log_ctx, NULL);
+	ok1(tdb);
+	ok1(log_count == 0);
+	ok1(tdb_check(tdb, NULL, NULL) == 0);
+	tdb_close(tdb);
+
+	log_count = 0;
+	tdb = tdb_open_ex("run-wronghash-fail.tdb", 0, 0, O_RDONLY,
+			  0, &log_ctx, NULL);
+	ok1(tdb);
+	ok1(log_count == 0);
+	ok1(tdb_check(tdb, NULL, NULL) == 0);
+	tdb_close(tdb);
+
+
+	return exit_status();
+}
diff --git a/ctdb/lib/tdb/test/run-zero-append.c b/ctdb/lib/tdb/test/run-zero-append.c
new file mode 100644
index 0000000000..36bf6990f5
--- /dev/null
+++ b/ctdb/lib/tdb/test/run-zero-append.c
@@ -0,0 +1,40 @@
+#include "../common/tdb_private.h"
+#include "../common/io.c"
+#include "../common/tdb.c"
+#include "../common/lock.c"
+#include "../common/freelist.c"
+#include "../common/traverse.c"
+#include "../common/transaction.c"
+#include "../common/error.c"
+#include "../common/open.c"
+#include "../common/check.c"
+#include "../common/hash.c"
+#include "tap-interface.h"
+#include <stdlib.h>
+#include "logging.h"
+
+int main(int argc, char *argv[])
+{
+	struct tdb_context *tdb;
+	TDB_DATA key, data;
+
+	plan_tests(4);
+	tdb = tdb_open_ex(NULL, 1024, TDB_INTERNAL, O_CREAT|O_TRUNC|O_RDWR,
+			  0600, &taplogctx, NULL);
+	ok1(tdb);
+
+	/* Tickle bug on appending zero length buffer to zero length buffer. */
+	key.dsize = strlen("hi");
+	key.dptr = (void *)"hi";
+	data.dptr = (void *)"world";
+	data.dsize = 0;
+
+	ok1(tdb_append(tdb, key, data) == 0);
+	ok1(tdb_append(tdb, key, data) == 0);
+	data = tdb_fetch(tdb, key);
+	ok1(data.dsize == 0);
+	tdb_close(tdb);
+	free(data.dptr);
+
+	return exit_status();
+}
diff --git a/ctdb/lib/tdb/test/run.c b/ctdb/lib/tdb/test/run.c
new file mode 100644
index 0000000000..f49e85055e
--- /dev/null
+++ b/ctdb/lib/tdb/test/run.c
@@ -0,0 +1,49 @@
+#include "../common/tdb_private.h"
+#include "../common/io.c"
+#include "../common/tdb.c"
+#include "../common/lock.c"
+#include "../common/freelist.c"
+#include "../common/traverse.c"
+#include "../common/transaction.c"
+#include "../common/error.c"
+#include "../common/open.c"
+#include "../common/check.c"
+#include "../common/hash.c"
+#include "tap-interface.h"
+#include <stdlib.h>
+#include "logging.h"
+
+int main(int argc, char *argv[])
+{
+	struct tdb_context *tdb;
+	TDB_DATA key, data;
+
+	plan_tests(10);
+	tdb = tdb_open_ex("run.tdb", 1024, TDB_CLEAR_IF_FIRST,
+			  O_CREAT|O_TRUNC|O_RDWR, 0600, &taplogctx, NULL);
+
+	ok1(tdb);
+	key.dsize = strlen("hi");
+	key.dptr = (void *)"hi";
+	data.dsize = strlen("world");
+	data.dptr = (void *)"world";
+
+	ok1(tdb_store(tdb, key, data, TDB_MODIFY) < 0);
+	ok1(tdb_error(tdb) == TDB_ERR_NOEXIST);
+	ok1(tdb_store(tdb, key, data, TDB_INSERT) == 0);
+	ok1(tdb_store(tdb, key, data, TDB_INSERT) < 0);
+	ok1(tdb_error(tdb) == TDB_ERR_EXISTS);
+	ok1(tdb_store(tdb, key, data, TDB_MODIFY) == 0);
+
+	data = tdb_fetch(tdb, key);
+	ok1(data.dsize == strlen("world"));
+	ok1(memcmp(data.dptr, "world", strlen("world")) == 0);
+	free(data.dptr);
+
+	key.dsize++;
+	data = tdb_fetch(tdb, key);
+	ok1(data.dptr == NULL);
+	tdb_close(tdb);
+
+	return exit_status();
+}
diff --git a/ctdb/lib/tdb/test/rwlock-be.tdb b/ctdb/lib/tdb/test/rwlock-be.tdb
new file mode 100644
index 0000000000..45b5f09a1b
--- /dev/null
+++ b/ctdb/lib/tdb/test/rwlock-be.tdb
diff --git a/ctdb/lib/tdb/test/rwlock-le.tdb b/ctdb/lib/tdb/test/rwlock-le.tdb
new file mode 100644
index 0000000000..45b5f09a1b
--- /dev/null
+++ b/ctdb/lib/tdb/test/rwlock-le.tdb
diff --git a/ctdb/lib/tdb/test/tap-interface.h b/ctdb/lib/tdb/test/tap-interface.h
new file mode 100644
index 0000000000..d9ed6e84ea
--- /dev/null
+++ b/ctdb/lib/tdb/test/tap-interface.h
@@ -0,0 +1,39 @@
+/* 
+   Unix SMB/CIFS implementation.
+   Simplistic implementation of tap interface.
+
+   Copyright (C) Rusty Russell 2012
+   
+     ** NOTE! The following LGPL license applies to the talloc
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+#include <stdio.h>
+
+#ifndef __location__
+#define __TAP_STRING_LINE1__(s)    #s
+#define __TAP_STRING_LINE2__(s)   __TAP_STRING_LINE1__(s)
+#define __TAP_STRING_LINE3__  __TAP_STRING_LINE2__(__LINE__)
+#define __location__ __FILE__ ":" __TAP_STRING_LINE3__
+#endif
+
+#define plan_tests(num)
+#define ok(e, ...) do { if (e) { (void)printf("."); } else { fprintf(stderr, __VA_ARGS__); exit(1); } } while(0)
+#define ok1(e) ok((e), "%s:%s", __location__, #e)
+#define pass(...) printf(".")
+#define fail(...) do { fprintf(stderr, __VA_ARGS__); exit(1); } while(0)
+#define diag printf
+#define exit_status() 0
diff --git a/ctdb/lib/tdb/test/tap-to-subunit.h b/ctdb/lib/tdb/test/tap-to-subunit.h
new file mode 100644
index 0000000000..a5cf74fb04
--- /dev/null
+++ b/ctdb/lib/tdb/test/tap-to-subunit.h
@@ -0,0 +1,155 @@
+#ifndef TAP_TO_SUBUNIT_H
+#define TAP_TO_SUBUNIT_H
+/*
+ * tap-style wrapper for subunit.
+ *
+ * Copyright (c) 2011 Rusty Russell
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+#include "replace.h"
+
+/**
+ * plan_tests - announce the number of tests you plan to run
+ * @tests: the number of tests
+ *
+ * This should be the first call in your test program: it allows tracing
+ * of failures which mean that not all tests are run.
+ *
+ * If you don't know how many tests will actually be run, assume all of them
+ * and use skip() if you don't actually run some tests.
+ *
+ * Example:
+ *	plan_tests(13);
+ */
+void plan_tests(unsigned int tests);
+
+/**
+ * ok1 - Simple conditional test
+ * @e: the expression which we expect to be true.
+ *
+ * This is the simplest kind of test: if the expression is true, the
+ * test passes.  The name of the test which is printed will simply be
+ * file name, line number, and the expression itself.
+ *
+ * Example:
+ *	ok1(somefunc() == 1);
+ */
+# define ok1(e) ((e) ?							\
+		 _gen_result(1, __func__, __FILE__, __LINE__, "%s", #e) : \
+		 _gen_result(0, __func__, __FILE__, __LINE__, "%s", #e))
+
+/**
+ * ok - Conditional test with a name
+ * @e: the expression which we expect to be true.
+ * @...: the printf-style name of the test.
+ *
+ * If the expression is true, the test passes.  The name of the test will be
+ * the filename, line number, and the printf-style string.  This can be clearer
+ * than simply the expression itself.
+ *
+ * Example:
+ *	ok1(somefunc() == 1);
+ *	ok(somefunc() == 0, "Second somefunc() should fail");
+ */
+# define ok(e, ...) ((e) ?						\
+		     _gen_result(1, __func__, __FILE__, __LINE__,	\
+				 __VA_ARGS__) :				\
+		     _gen_result(0, __func__, __FILE__, __LINE__,	\
+				 __VA_ARGS__))
+
+/**
+ * pass - Note that a test passed
+ * @...: the printf-style name of the test.
+ *
+ * For complicated code paths, it can be easiest to simply call pass() in one
+ * branch and fail() in another.
+ *
+ * Example:
+ *	int x = somefunc();
+ *	if (x > 0)
+ *		pass("somefunc() returned a valid value");
+ *	else
+ *		fail("somefunc() returned an invalid value");
+ */
+# define pass(...) ok(1, __VA_ARGS__)
+
+/**
+ * fail - Note that a test failed
+ * @...: the printf-style name of the test.
+ *
+ * For complicated code paths, it can be easiest to simply call pass() in one
+ * branch and fail() in another.
+ */
+# define fail(...) ok(0, __VA_ARGS__)
+
+unsigned int _gen_result(int, const char *, const char *, unsigned int,
+   const char *, ...) PRINTF_ATTRIBUTE(5, 6);
+
+/**
+ * diag - print a diagnostic message (use instead of printf/fprintf)
+ * @fmt: the format of the printf-style message
+ *
+ * diag ensures that the output will not be considered to be a test
+ * result by the TAP test harness.  It will append '\n' for you.
+ *
+ * Example:
+ *	diag("Now running complex tests");
+ */
+void diag(const char *fmt, ...) PRINTF_ATTRIBUTE(1, 2);
+
+/**
+ * skip - print a diagnostic message (use instead of printf/fprintf)
+ * @n: number of tests you're skipping.
+ * @fmt: the format of the reason you're skipping the tests.
+ *
+ * Sometimes tests cannot be run because the test system lacks some feature:
+ * you should explicitly document that you're skipping tests using skip().
+ *
+ * From the Test::More documentation:
+ *   If it's something the user might not be able to do, use SKIP.  This
+ *   includes optional modules that aren't installed, running under an OS that
+ *   doesn't have some feature (like fork() or symlinks), or maybe you need an
+ *   Internet connection and one isn't available.
+ *
+ * Example:
+ *	#ifdef HAVE_SOME_FEATURE
+ *	ok1(somefunc());
+ *	#else
+ *	skip(1, "Don't have SOME_FEATURE");
+ *	#endif
+ */
+void skip(unsigned int n, const char *fmt, ...) PRINTF_ATTRIBUTE(2, 3);
+
+/**
+ * exit_status - the value that main should return.
+ *
+ * For maximum compatibility your test program should return a particular exit
+ * code (ie. 0 if all tests were run, and every test which was expected to
+ * succeed succeeded).
+ *
+ * Example:
+ *	exit(exit_status());
+ */
+int exit_status(void);
+#endif /* CCAN_TAP_H */
diff --git a/ctdb/lib/tdb/test/tdb.corrupt b/ctdb/lib/tdb/test/tdb.corrupt
new file mode 100644
index 0000000000..83d6677454
--- /dev/null
+++ b/ctdb/lib/tdb/test/tdb.corrupt
diff --git a/ctdb/lib/tdb/tools/tdbbackup.c b/ctdb/lib/tdb/tools/tdbbackup.c
new file mode 100644
index 0000000000..11ecaa0290
--- /dev/null
+++ b/ctdb/lib/tdb/tools/tdbbackup.c
@@ -0,0 +1,342 @@
+/* 
+   Unix SMB/CIFS implementation.
+   low level tdb backup and restore utility
+   Copyright (C) Andrew Tridgell              2002
+
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.
+*/
+
+/*
+
+  This program is meant for backup/restore of tdb databases. Typical usage would be:
+     tdbbackup *.tdb
+  when Samba shuts down cleanly, which will make a backup of all the local databases
+  to *.bak files. Then on Samba startup you would use:
+     tdbbackup -v *.tdb
+  and this will check the databases for corruption and if corruption is detected then
+  the backup will be restored.
+
+  You may also like to do a backup on a regular basis while Samba is
+  running, perhaps using cron.
+
+  The reason this program is needed is to cope with power failures
+  while Samba is running. A power failure could lead to database
+  corruption and Samba will then not start correctly.
+
+  Note that many of the databases in Samba are transient and thus
+  don't need to be backed up, so you can optimise the above a little
+  by only running the backup on the critical databases.
+
+ */
+
+#include "replace.h"
+#include "system/locale.h"
+#include "system/time.h"
+#include "system/filesys.h"
+#include "system/wait.h"
+#include "tdb.h"
+
+#ifdef HAVE_GETOPT_H
+#include <getopt.h>
+#endif
+
+static int failed;
+
+static struct tdb_logging_context log_ctx;
+
+#ifdef PRINTF_ATTRIBUTE
+static void tdb_log(struct tdb_context *tdb, enum tdb_debug_level level, const char *format, ...) PRINTF_ATTRIBUTE(3,4);
+#endif
+static void tdb_log(struct tdb_context *tdb, enum tdb_debug_level level, const char *format, ...)
+{
+	va_list ap;
+    
+	va_start(ap, format);
+	vfprintf(stdout, format, ap);
+	va_end(ap);
+	fflush(stdout);
+}
+
+static char *add_suffix(const char *name, const char *suffix)
+{
+	char *ret;
+	int len = strlen(name) + strlen(suffix) + 1;
+	ret = (char *)malloc(len);
+	if (!ret) {
+		fprintf(stderr,"Out of memory!\n");
+		exit(1);
+	}
+	snprintf(ret, len, "%s%s", name, suffix);
+	return ret;
+}
+
+static int copy_fn(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA dbuf, void *state)
+{
+	TDB_CONTEXT *tdb_new = (TDB_CONTEXT *)state;
+
+	if (tdb_store(tdb_new, key, dbuf, TDB_INSERT) != 0) {
+		fprintf(stderr,"Failed to insert into %s\n", tdb_name(tdb_new));
+		failed = 1;
+		return 1;
+	}
+	return 0;
+}
+
+
+static int test_fn(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA dbuf, void *state)
+{
+	return 0;
+}
+
+/*
+  carefully backup a tdb, validating the contents and
+  only doing the backup if its OK
+  this function is also used for restore
+*/
+static int backup_tdb(const char *old_name, const char *new_name, int hash_size)
+{
+	TDB_CONTEXT *tdb;
+	TDB_CONTEXT *tdb_new;
+	char *tmp_name;
+	struct stat st;
+	int count1, count2;
+
+	tmp_name = add_suffix(new_name, ".tmp");
+
+	/* stat the old tdb to find its permissions */
+	if (stat(old_name, &st) != 0) {
+		perror(old_name);
+		free(tmp_name);
+		return 1;
+	}
+
+	/* open the old tdb */
+	tdb = tdb_open_ex(old_name, 0, 0, 
+			  O_RDWR, 0, &log_ctx, NULL);
+	if (!tdb) {
+		printf("Failed to open %s\n", old_name);
+		free(tmp_name);
+		return 1;
+	}
+
+	/* create the new tdb */
+	unlink(tmp_name);
+	tdb_new = tdb_open_ex(tmp_name, 
+			      hash_size ? hash_size : tdb_hash_size(tdb), 
+			      TDB_DEFAULT, 
+			      O_RDWR|O_CREAT|O_EXCL, st.st_mode & 0777, 
+			      &log_ctx, NULL);
+	if (!tdb_new) {
+		perror(tmp_name);
+		free(tmp_name);
+		return 1;
+	}
+
+	if (tdb_transaction_start(tdb) != 0) {
+		printf("Failed to start transaction on old tdb\n");
+		tdb_close(tdb);
+		tdb_close(tdb_new);
+		unlink(tmp_name);
+		free(tmp_name);
+		return 1;
+	}
+
+	/* lock the backup tdb so that nobody else can change it */
+	if (tdb_lockall(tdb_new) != 0) {
+		printf("Failed to lock backup tdb\n");
+		tdb_close(tdb);
+		tdb_close(tdb_new);
+		unlink(tmp_name);
+		free(tmp_name);
+		return 1;
+	}
+
+	failed = 0;
+
+	/* traverse and copy */
+	count1 = tdb_traverse(tdb, copy_fn, (void *)tdb_new);
+	if (count1 < 0 || failed) {
+		fprintf(stderr,"failed to copy %s\n", old_name);
+		tdb_close(tdb);
+		tdb_close(tdb_new);
+		unlink(tmp_name);
+		free(tmp_name);
+		return 1;
+	}
+
+	/* close the old tdb */
+	tdb_close(tdb);
+
+	/* copy done, unlock the backup tdb */
+	tdb_unlockall(tdb_new);
+
+#ifdef HAVE_FDATASYNC
+	if (fdatasync(tdb_fd(tdb_new)) != 0) {
+#else
+	if (fsync(tdb_fd(tdb_new)) != 0) {
+#endif
+		/* not fatal */
+		fprintf(stderr, "failed to fsync backup file\n");
+	}
+
+	/* close the new tdb and re-open read-only */
+	tdb_close(tdb_new);
+	tdb_new = tdb_open_ex(tmp_name, 
+			      0,
+			      TDB_DEFAULT, 
+			      O_RDONLY, 0,
+			      &log_ctx, NULL);
+	if (!tdb_new) {
+		fprintf(stderr,"failed to reopen %s\n", tmp_name);
+		unlink(tmp_name);
+		perror(tmp_name);
+		free(tmp_name);
+		return 1;
+	}
+	
+	/* traverse the new tdb to confirm */
+	count2 = tdb_traverse(tdb_new, test_fn, NULL);
+	if (count2 != count1) {
+		fprintf(stderr,"failed to copy %s\n", old_name);
+		tdb_close(tdb_new);
+		unlink(tmp_name);
+		free(tmp_name);
+		return 1;
+	}
+
+	/* close the new tdb and rename it to .bak */
+	tdb_close(tdb_new);
+	if (rename(tmp_name, new_name) != 0) {
+		perror(new_name);
+		free(tmp_name);
+		return 1;
+	}
+
+	free(tmp_name);
+
+	return 0;
+}
+
+/*
+  verify a tdb and if it is corrupt then restore from *.bak
+*/
+static int verify_tdb(const char *fname, const char *bak_name)
+{
+	TDB_CONTEXT *tdb;
+	int count = -1;
+
+	/* open the tdb */
+	tdb = tdb_open_ex(fname, 0, 0, 
+			  O_RDONLY, 0, &log_ctx, NULL);
+
+	/* traverse the tdb, then close it */
+	if (tdb) {
+		count = tdb_traverse(tdb, test_fn, NULL);
+		tdb_close(tdb);
+	}
+
+	/* count is < 0 means an error */
+	if (count < 0) {
+		printf("restoring %s\n", fname);
+		return backup_tdb(bak_name, fname, 0);
+	}
+
+	printf("%s : %d records\n", fname, count);
+
+	return 0;
+}
+
+/*
+  see if one file is newer than another
+*/
+static int file_newer(const char *fname1, const char *fname2)
+{
+	struct stat st1, st2;
+	if (stat(fname1, &st1) != 0) {
+		return 0;
+	}
+	if (stat(fname2, &st2) != 0) {
+		return 1;
+	}
+	return (st1.st_mtime > st2.st_mtime);
+}
+
+static void usage(void)
+{
+	printf("Usage: tdbbackup [options] <fname...>\n\n");
+	printf("   -h            this help message\n");
+	printf("   -s suffix     set the backup suffix\n");
+	printf("   -v            verify mode (restore if corrupt)\n");
+	printf("   -n hashsize   set the new hash size for the backup\n");
+}
+		
+
+ int main(int argc, char *argv[])
+{
+	int i;
+	int ret = 0;
+	int c;
+	int verify = 0;
+	int hashsize = 0;
+	const char *suffix = ".bak";
+
+	log_ctx.log_fn = tdb_log;
+
+	while ((c = getopt(argc, argv, "vhs:n:")) != -1) {
+		switch (c) {
+		case 'h':
+			usage();
+			exit(0);
+		case 'v':
+			verify = 1;
+			break;
+		case 's':
+			suffix = optarg;
+			break;
+		case 'n':
+			hashsize = atoi(optarg);
+			break;
+		}
+	}
+
+	argc -= optind;
+	argv += optind;
+
+	if (argc < 1) {
+		usage();
+		exit(1);
+	}
+
+	for (i=0; i<argc; i++) {
+		const char *fname = argv[i];
+		char *bak_name;
+
+		bak_name = add_suffix(fname, suffix);
+
+		if (verify) {
+			if (verify_tdb(fname, bak_name) != 0) {
+				ret = 1;
+			}
+		} else {
+			if (file_newer(fname, bak_name) &&
+			    backup_tdb(fname, bak_name, hashsize) != 0) {
+				ret = 1;
+			}
+		}
+
+		free(bak_name);
+	}
+
+	return ret;
+}
diff --git a/ctdb/lib/tdb/tools/tdbdump.c b/ctdb/lib/tdb/tools/tdbdump.c
new file mode 100644
index 0000000000..bcd395fd5a
--- /dev/null
+++ b/ctdb/lib/tdb/tools/tdbdump.c
@@ -0,0 +1,171 @@
+/* 
+   Unix SMB/CIFS implementation.
+   simple tdb dump util
+   Copyright (C) Andrew Tridgell              2001
+
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "replace.h"
+#include "system/locale.h"
+#include "system/time.h"
+#include "system/filesys.h"
+#include "system/wait.h"
+#include "tdb.h"
+
+static void print_data(TDB_DATA d)
+{
+	unsigned char *p = (unsigned char *)d.dptr;
+	int len = d.dsize;
+	while (len--) {
+		if (isprint(*p) && !strchr("\"\\", *p)) {
+			fputc(*p, stdout);
+		} else {
+			printf("\\%02X", *p);
+		}
+		p++;
+	}
+}
+
+static int traverse_fn(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA dbuf, void *state)
+{
+	printf("{\n");
+	printf("key(%d) = \"", (int)key.dsize);
+	print_data(key);
+	printf("\"\n");
+	printf("data(%d) = \"", (int)dbuf.dsize);
+	print_data(dbuf);
+	printf("\"\n");
+	printf("}\n");
+	return 0;
+}
+
+static void log_stderr(struct tdb_context *tdb, enum tdb_debug_level level,
+		       const char *fmt, ...)
+{
+	va_list ap;
+	char *ptr = NULL;
+	int debuglevel = 0;
+	int ret;
+	const char *name = tdb_name(tdb);
+	const char *prefix = "";
+
+	if (!name)
+		name = "unnamed";
+
+	switch (level) {
+	case TDB_DEBUG_ERROR:
+		prefix = "ERROR: ";
+		break;
+	case TDB_DEBUG_WARNING:
+		prefix = "WARNING: ";
+		break;
+	case TDB_DEBUG_TRACE:
+		return;
+
+	default:
+	case TDB_DEBUG_FATAL:
+		prefix = "FATAL: ";
+		break;
+	}
+
+	va_start(ap, fmt);
+	fprintf(stderr, "tdb(%s): %s", name, prefix);
+	vfprintf(stderr, fmt, ap);
+	va_end(ap);
+}
+
+static void emergency_walk(TDB_DATA key, TDB_DATA dbuf, void *keyname)
+{
+	if (keyname) {
+		if (key.dsize != strlen(keyname))
+			return;
+		if (memcmp(key.dptr, keyname, key.dsize) != 0)
+			return;
+	}
+	traverse_fn(NULL, key, dbuf, NULL);
+}
+
+static int dump_tdb(const char *fname, const char *keyname, bool emergency)
+{
+	TDB_CONTEXT *tdb;
+	TDB_DATA key, value;
+	struct tdb_logging_context logfn = { log_stderr };
+	
+	tdb = tdb_open_ex(fname, 0, 0, O_RDONLY, 0, &logfn, NULL);
+	if (!tdb) {
+		printf("Failed to open %s\n", fname);
+		return 1;
+	}
+
+	if (emergency) {
+		return tdb_rescue(tdb, emergency_walk, keyname) == 0;
+	}
+	if (!keyname) {
+		return tdb_traverse(tdb, traverse_fn, NULL) == -1 ? 1 : 0;
+	} else {
+		key.dptr = discard_const_p(uint8_t, keyname);
+		key.dsize = strlen(keyname);
+		value = tdb_fetch(tdb, key);
+		if (!value.dptr) {
+			return 1;
+		} else {
+			print_data(value);
+			free(value.dptr);
+		}
+	}
+
+	return 0;
+}
+
+static void usage( void)
+{
+	printf( "Usage: tdbdump [options] <filename>\n\n");
+	printf( "   -h          this help message\n");
+	printf( "   -k keyname  dumps value of keyname\n");
+	printf( "   -e          emergency dump, for corrupt databases\n");
+}
+
+ int main(int argc, char *argv[])
+{
+	char *fname, *keyname=NULL;
+	bool emergency = false;
+	int c;
+
+	if (argc < 2) {
+		printf("Usage: tdbdump <fname>\n");
+		exit(1);
+	}
+
+	while ((c = getopt( argc, argv, "hk:e")) != -1) {
+		switch (c) {
+		case 'h':
+			usage();
+			exit( 0);
+		case 'k':
+			keyname = optarg;
+			break;
+		case 'e':
+			emergency = true;
+			break;
+		default:
+			usage();
+			exit( 1);
+		}
+	}
+
+	fname = argv[optind];
+
+	return dump_tdb(fname, keyname, emergency);
+}
diff --git a/ctdb/lib/tdb/tools/tdbrestore.c b/ctdb/lib/tdb/tools/tdbrestore.c
new file mode 100644
index 0000000000..f65b36fa12
--- /dev/null
+++ b/ctdb/lib/tdb/tools/tdbrestore.c
@@ -0,0 +1,222 @@
+/*
+   tdbrestore -- construct a tdb from tdbdump output.
+   Copyright (C) Volker Lendecke		2010
+   Copyright (C) Simon McVittie			2005
+
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include <assert.h>
+#include "replace.h"
+#include "system/locale.h"
+#include "system/time.h"
+#include "system/filesys.h"
+#include "system/wait.h"
+#include "tdb.h"
+
+static int read_linehead(FILE *f)
+{
+	int i, c;
+	int num_bytes;
+	char prefix[128];
+
+	while (1) {
+		c = getc(f);
+		if (c == EOF) {
+			return -1;
+		}
+		if (c == '(') {
+			break;
+		}
+	}
+	for (i=0; i<sizeof(prefix); i++) {
+		c = getc(f);
+		if (c == EOF) {
+			return -1;
+		}
+		prefix[i] = c;
+		if (c == '"') {
+			break;
+		}
+	}
+	if (i == sizeof(prefix)) {
+		return -1;
+	}
+	prefix[i] = '\0';
+
+	if (sscanf(prefix, "%d) = ", &num_bytes) != 1) {
+		return -1;
+	}
+	return num_bytes;
+}
+
+static int read_hex(void) {
+	int c;
+	c = getchar();
+	if (c == EOF) {
+		fprintf(stderr, "Unexpected EOF in data\n");
+		return -1;
+	} else if (c == '"') {
+		fprintf(stderr, "Unexpected \\\" sequence\n");
+		return -1;
+	} else if ('0' <= c && c <= '9')  {
+		return c - '0';
+	} else if ('A' <= c && c <= 'F')  {
+		return c - 'A' + 10;
+	} else if ('a' <= c && c <= 'f')  {
+		return c - 'a' + 10;
+	} else {
+		fprintf(stderr, "Invalid hex: %c\n", c);
+		return -1;
+	}
+}
+
+static int read_data(FILE *f, TDB_DATA *d, size_t size) {
+	int c, low, high;
+	int i;
+
+	d->dptr = (unsigned char *)malloc(size);
+	if (d->dptr == NULL) {
+		return -1;
+	}
+	d->dsize = size;
+
+	for (i=0; i<size; i++) {
+		c = getc(f);
+		if (c == EOF) {
+			fprintf(stderr, "Unexpected EOF in data\n");
+			return 1;
+		} else if (c == '"') {
+			return 0;
+		} else if (c == '\\') {
+			high = read_hex();
+			if (high < 0) {
+				return -1;
+			}
+			high = high << 4;
+			assert(high == (high & 0xf0));
+			low = read_hex();
+			if (low < 0) {
+				return -1;
+			}
+			assert(low == (low & 0x0f));
+			d->dptr[i] = (low|high);
+		} else {
+			d->dptr[i] = c;
+		}
+	}
+	return 0;
+}
+
+static int swallow(FILE *f, const char *s, int *eof)
+{
+	char line[128];
+
+	if (fgets(line, sizeof(line), f) == NULL) {
+		if (eof != NULL) {
+			*eof = 1;
+		}
+		return -1;
+	}
+	if (strcmp(line, s) != 0) {
+		return -1;
+	}
+	return 0;
+}
+
+static int read_rec(FILE *f, TDB_CONTEXT *tdb, int *eof)
+{
+	int length;
+	TDB_DATA key, data;
+	int ret = -1;
+
+	key.dptr = NULL;
+	data.dptr = NULL;
+
+	if (swallow(f, "{\n", eof) == -1) {
+		goto fail;
+	}
+	length = read_linehead(f);
+	if (length == -1) {
+		goto fail;
+	}
+	if (read_data(f, &key, length) == -1) {
+		goto fail;
+	}
+	if (swallow(f, "\"\n", NULL) == -1) {
+		goto fail;
+	}
+	length = read_linehead(f);
+	if (length == -1) {
+		goto fail;
+	}
+	if (read_data(f, &data, length) == -1) {
+		goto fail;
+	}
+	if ((swallow(f, "\"\n", NULL) == -1)
+	    || (swallow(f, "}\n", NULL) == -1)) {
+		goto fail;
+	}
+	if (tdb_store(tdb, key, data, TDB_INSERT) != 0) {
+		fprintf(stderr, "TDB error: %s\n", tdb_errorstr(tdb));
+		goto fail;
+	}
+
+	ret = 0;
+fail:
+	free(key.dptr);
+	free(data.dptr);
+	return ret;
+}
+
+static int restore_tdb(const char *fname)
+{
+	TDB_CONTEXT *tdb;
+
+	tdb = tdb_open(fname, 0, 0, O_RDWR|O_CREAT|O_EXCL, 0666);
+	if (!tdb) {
+		perror("tdb_open");
+		fprintf(stderr, "Failed to open %s\n", fname);
+		return 1;
+	}
+
+	while (1) {
+		int eof = 0;
+		if (read_rec(stdin, tdb, &eof) == -1) {
+			if (eof) {
+				break;
+			}
+			return 1;
+		}
+	}
+	if (tdb_close(tdb)) {
+		fprintf(stderr, "Error closing tdb\n");
+		return 1;
+	}
+	return 0;
+}
+
+int main(int argc, char *argv[])
+{
+	char *fname;
+
+	if (argc < 2) {
+		printf("Usage: %s dbname < tdbdump_output\n", argv[0]);
+		exit(1);
+	}
+
+	fname = argv[1];
+
+	return restore_tdb(fname);
+}
diff --git a/ctdb/lib/tdb/tools/tdbtest.c b/ctdb/lib/tdb/tools/tdbtest.c
new file mode 100644
index 0000000000..44c78efda5
--- /dev/null
+++ b/ctdb/lib/tdb/tools/tdbtest.c
@@ -0,0 +1,290 @@
+/* a test program for tdb - the trivial database */
+
+#include "replace.h"
+#include "tdb.h"
+#include "system/filesys.h"
+#include "system/time.h"
+
+#include <gdbm.h>
+
+
+#define DELETE_PROB 7
+#define STORE_PROB 5
+
+static struct tdb_context *db;
+static GDBM_FILE gdbm;
+
+struct timeval tp1,tp2;
+
+static void _start_timer(void)
+{
+	gettimeofday(&tp1,NULL);
+}
+
+static double _end_timer(void)
+{
+	gettimeofday(&tp2,NULL);
+	return((tp2.tv_sec - tp1.tv_sec) + 
+	       (tp2.tv_usec - tp1.tv_usec)*1.0e-6);
+}
+
+static void fatal(const char *why)
+{
+	perror(why);
+	exit(1);
+}
+
+#ifdef PRINTF_ATTRIBUTE
+static void tdb_log(struct tdb_context *tdb, int level, const char *format, ...) PRINTF_ATTRIBUTE(3,4);
+#endif
+static void tdb_log(struct tdb_context *tdb, int level, const char *format, ...)
+{
+	va_list ap;
+    
+	va_start(ap, format);
+	vfprintf(stdout, format, ap);
+	va_end(ap);
+	fflush(stdout);
+}
+
+static void compare_db(void)
+{
+	TDB_DATA d, key, nextkey;
+	datum gd, gkey, gnextkey;
+
+	key = tdb_firstkey(db);
+	while (key.dptr) {
+		d = tdb_fetch(db, key);
+		gkey.dptr = key.dptr;
+		gkey.dsize = key.dsize;
+
+		gd = gdbm_fetch(gdbm, gkey);
+
+		if (!gd.dptr) fatal("key not in gdbm");
+		if (gd.dsize != d.dsize) fatal("data sizes differ");
+		if (memcmp(gd.dptr, d.dptr, d.dsize)) {
+			fatal("data differs");
+		}
+
+		nextkey = tdb_nextkey(db, key);
+		free(key.dptr);
+		free(d.dptr);
+		free(gd.dptr);
+		key = nextkey;
+	}
+
+	gkey = gdbm_firstkey(gdbm);
+	while (gkey.dptr) {
+		gd = gdbm_fetch(gdbm, gkey);
+		key.dptr = gkey.dptr;
+		key.dsize = gkey.dsize;
+
+		d = tdb_fetch(db, key);
+
+		if (!d.dptr) fatal("key not in db");
+		if (d.dsize != gd.dsize) fatal("data sizes differ");
+		if (memcmp(d.dptr, gd.dptr, gd.dsize)) {
+			fatal("data differs");
+		}
+
+		gnextkey = gdbm_nextkey(gdbm, gkey);
+		free(gkey.dptr);
+		free(gd.dptr);
+		free(d.dptr);
+		gkey = gnextkey;
+	}
+}
+
+static char *randbuf(int len)
+{
+	char *buf;
+	int i;
+	buf = (char *)malloc(len+1);
+
+	for (i=0;i<len;i++) {
+		buf[i] = 'a' + (rand() % 26);
+	}
+	buf[i] = 0;
+	return buf;
+}
+
+static void addrec_db(void)
+{
+	int klen, dlen;
+	char *k, *d;
+	TDB_DATA key, data;
+
+	klen = 1 + (rand() % 4);
+	dlen = 1 + (rand() % 100);
+
+	k = randbuf(klen);
+	d = randbuf(dlen);
+
+	key.dptr = k;
+	key.dsize = klen+1;
+
+	data.dptr = d;
+	data.dsize = dlen+1;
+
+	if (rand() % DELETE_PROB == 0) {
+		tdb_delete(db, key);
+	} else if (rand() % STORE_PROB == 0) {
+		if (tdb_store(db, key, data, TDB_REPLACE) != 0) {
+			fatal("tdb_store failed");
+		}
+	} else {
+		data = tdb_fetch(db, key);
+		if (data.dptr) free(data.dptr);
+	}
+
+	free(k);
+	free(d);
+}
+
+static void addrec_gdbm(void)
+{
+	int klen, dlen;
+	char *k, *d;
+	datum key, data;
+
+	klen = 1 + (rand() % 4);
+	dlen = 1 + (rand() % 100);
+
+	k = randbuf(klen);
+	d = randbuf(dlen);
+
+	key.dptr = k;
+	key.dsize = klen+1;
+
+	data.dptr = d;
+	data.dsize = dlen+1;
+
+	if (rand() % DELETE_PROB == 0) {
+		gdbm_delete(gdbm, key);
+	} else if (rand() % STORE_PROB == 0) {
+		if (gdbm_store(gdbm, key, data, GDBM_REPLACE) != 0) {
+			fatal("gdbm_store failed");
+		}
+	} else {
+		data = gdbm_fetch(gdbm, key);
+		if (data.dptr) free(data.dptr);
+	}
+
+	free(k);
+	free(d);
+}
+
+static int traverse_fn(struct tdb_context *tdb, TDB_DATA key, TDB_DATA dbuf, void *state)
+{
+#if 0
+	printf("[%s] [%s]\n", key.dptr, dbuf.dptr);
+#endif
+	tdb_delete(tdb, key);
+	return 0;
+}
+
+static void merge_test(void)
+{
+	int i;
+	char keys[5][2];
+	char tdata[] = "test";
+	TDB_DATA key, data;
+	
+	for (i = 0; i < 5; i++) {
+		snprintf(keys[i],2, "%d", i);
+		key.dptr = keys[i];
+		key.dsize = 2;
+		
+		data.dptr = tdata;
+		data.dsize = 4;
+		
+		if (tdb_store(db, key, data, TDB_REPLACE) != 0) {
+			fatal("tdb_store failed");
+		}
+	}
+
+	key.dptr = keys[0];
+	tdb_delete(db, key);
+	key.dptr = keys[4];
+	tdb_delete(db, key);
+	key.dptr = keys[2];
+	tdb_delete(db, key);
+	key.dptr = keys[1];
+	tdb_delete(db, key);
+	key.dptr = keys[3];
+	tdb_delete(db, key);
+}
+
+static char *test_path(const char *filename)
+{
+	const char *prefix = getenv("TEST_DATA_PREFIX");
+
+	if (prefix) {
+		char *path = NULL;
+		int ret;
+
+		ret = asprintf(&path, "%s/%s", prefix, filename);
+		if (ret == -1) {
+			return NULL;
+		}
+		return path;
+	}
+
+	return strdup(filename);
+}
+
+ int main(int argc, const char *argv[])
+{
+	int i, seed=0;
+	int loops = 10000;
+	int num_entries;
+	char test_gdbm[1] = "test.gdbm";
+	char *test_tdb;
+
+	test_gdbm[0] = test_path("test.gdbm");
+	test_tdb = test_path("test.tdb");
+
+	unlink(test_gdbm[0]);
+
+	db = tdb_open(test_tdb, 0, TDB_CLEAR_IF_FIRST,
+		      O_RDWR | O_CREAT | O_TRUNC, 0600);
+	gdbm = gdbm_open(test_gdbm, 512, GDBM_WRITER|GDBM_NEWDB|GDBM_FAST, 
+			 0600, NULL);
+
+	if (!db || !gdbm) {
+		fatal("db open failed");
+	}
+
+#if 1
+	srand(seed);
+	_start_timer();
+	for (i=0;i<loops;i++) addrec_gdbm();
+	printf("gdbm got %.2f ops/sec\n", i/_end_timer());
+#endif
+
+	merge_test();
+
+	srand(seed);
+	_start_timer();
+	for (i=0;i<loops;i++) addrec_db();
+	printf("tdb got %.2f ops/sec\n", i/_end_timer());
+
+	if (tdb_validate_freelist(db, &num_entries) == -1) {
+		printf("tdb freelist is corrupt\n");
+	} else {
+		printf("tdb freelist is good (%d entries)\n", num_entries);
+	}
+
+	compare_db();
+
+	printf("traversed %d records\n", tdb_traverse(db, traverse_fn, NULL));
+	printf("traversed %d records\n", tdb_traverse(db, traverse_fn, NULL));
+
+	tdb_close(db);
+	gdbm_close(gdbm);
+
+	free(test_gdbm[0]);
+	free(test_tdb);
+
+	return 0;
+}
diff --git a/ctdb/lib/tdb/tools/tdbtool.c b/ctdb/lib/tdb/tools/tdbtool.c
new file mode 100644
index 0000000000..dc5747f87a
--- /dev/null
+++ b/ctdb/lib/tdb/tools/tdbtool.c
@@ -0,0 +1,785 @@
+/* 
+   Unix SMB/CIFS implementation.
+   Samba database functions
+   Copyright (C) Andrew Tridgell              1999-2000
+   Copyright (C) Paul `Rusty' Russell		   2000
+   Copyright (C) Jeremy Allison			   2000
+   Copyright (C) Andrew Esh                        2001
+
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "replace.h"
+#include "system/locale.h"
+#include "system/time.h"
+#include "system/filesys.h"
+#include "system/wait.h"
+#include "tdb.h"
+
+static int do_command(void);
+const char *cmdname;
+char *arg1, *arg2;
+size_t arg1len, arg2len;
+int bIterate = 0;
+char *line;
+TDB_DATA iterate_kbuf;
+char cmdline[1024];
+static int disable_mmap;
+
+enum commands {
+	CMD_CREATE_TDB,
+	CMD_OPEN_TDB,
+	CMD_TRANSACTION_START,
+	CMD_TRANSACTION_COMMIT,
+	CMD_TRANSACTION_CANCEL,
+	CMD_ERASE,
+	CMD_DUMP,
+	CMD_INSERT,
+	CMD_MOVE,
+	CMD_STORE,
+	CMD_SHOW,
+	CMD_KEYS,
+	CMD_HEXKEYS,
+	CMD_DELETE,
+	CMD_LIST_HASH_FREE,
+	CMD_LIST_FREE,
+	CMD_INFO,
+	CMD_MMAP,
+	CMD_SPEED,
+	CMD_FIRST,
+	CMD_NEXT,
+	CMD_SYSTEM,
+	CMD_CHECK,
+	CMD_REPACK,
+	CMD_QUIT,
+	CMD_HELP
+};
+
+typedef struct {
+	const char *name;
+	enum commands cmd;
+} COMMAND_TABLE;
+
+COMMAND_TABLE cmd_table[] = {
+	{"create",	CMD_CREATE_TDB},
+	{"open",	CMD_OPEN_TDB},
+	{"transaction_start",	CMD_TRANSACTION_START},
+	{"transaction_commit",	CMD_TRANSACTION_COMMIT},
+	{"transaction_cancel",	CMD_TRANSACTION_CANCEL},
+	{"erase",	CMD_ERASE},
+	{"dump",	CMD_DUMP},
+	{"insert",	CMD_INSERT},
+	{"move",	CMD_MOVE},
+	{"store",	CMD_STORE},
+	{"show",	CMD_SHOW},
+	{"keys",	CMD_KEYS},
+	{"hexkeys",	CMD_HEXKEYS},
+	{"delete",	CMD_DELETE},
+	{"list",	CMD_LIST_HASH_FREE},
+	{"free",	CMD_LIST_FREE},
+	{"info",	CMD_INFO},
+	{"speed",	CMD_SPEED},
+	{"mmap",	CMD_MMAP},
+	{"first",	CMD_FIRST},
+	{"1",		CMD_FIRST},
+	{"next",	CMD_NEXT},
+	{"n",		CMD_NEXT},
+	{"check",	CMD_CHECK},
+	{"quit",	CMD_QUIT},
+	{"q",		CMD_QUIT},
+	{"!",		CMD_SYSTEM},
+	{"repack",	CMD_REPACK},
+	{NULL,		CMD_HELP}
+};
+
+struct timeval tp1,tp2;
+
+static void _start_timer(void)
+{
+	gettimeofday(&tp1,NULL);
+}
+
+static double _end_timer(void)
+{
+	gettimeofday(&tp2,NULL);
+	return((tp2.tv_sec - tp1.tv_sec) + 
+	       (tp2.tv_usec - tp1.tv_usec)*1.0e-6);
+}
+
+#ifdef PRINTF_ATTRIBUTE
+static void tdb_log(struct tdb_context *tdb, enum tdb_debug_level level, const char *format, ...) PRINTF_ATTRIBUTE(3,4);
+#endif
+static void tdb_log(struct tdb_context *tdb, enum tdb_debug_level level, const char *format, ...)
+{
+	va_list ap;
+
+	va_start(ap, format);
+	vfprintf(stderr, format, ap);
+	va_end(ap);
+}
+
+/* a tdb tool for manipulating a tdb database */
+
+static TDB_CONTEXT *tdb;
+
+static int print_rec(TDB_CONTEXT *the_tdb, TDB_DATA key, TDB_DATA dbuf, void *state);
+static int print_key(TDB_CONTEXT *the_tdb, TDB_DATA key, TDB_DATA dbuf, void *state);
+static int print_hexkey(TDB_CONTEXT *the_tdb, TDB_DATA key, TDB_DATA dbuf, void *state);
+
+static void print_asc(const char *buf,int len)
+{
+	int i;
+
+	/* We're probably printing ASCII strings so don't try to display
+	   the trailing NULL character. */
+
+	if (buf[len - 1] == 0)
+	        len--;
+
+	for (i=0;i<len;i++)
+		printf("%c",isprint(buf[i])?buf[i]:'.');
+}
+
+static void print_data(const char *buf,int len)
+{
+	int i=0;
+	if (len<=0) return;
+	printf("[%03X] ",i);
+	for (i=0;i<len;) {
+		printf("%02X ",(int)((unsigned char)buf[i]));
+		i++;
+		if (i%8 == 0) printf(" ");
+		if (i%16 == 0) {      
+			print_asc(&buf[i-16],8); printf(" ");
+			print_asc(&buf[i-8],8); printf("\n");
+			if (i<len) printf("[%03X] ",i);
+		}
+	}
+	if (i%16) {
+		int n;
+		
+		n = 16 - (i%16);
+		printf(" ");
+		if (n>8) printf(" ");
+		while (n--) printf("   ");
+		
+		n = i%16;
+		if (n > 8) n = 8;
+		print_asc(&buf[i-(i%16)],n); printf(" ");
+		n = (i%16) - n;
+		if (n>0) print_asc(&buf[i-n],n); 
+		printf("\n");    
+	}
+}
+
+static void help(void)
+{
+	printf("\n"
+"tdbtool: \n"
+"  create    dbname     : create a database\n"
+"  open      dbname     : open an existing database\n"
+"  transaction_start    : start a transaction\n"
+"  transaction_commit   : commit a transaction\n"
+"  transaction_cancel   : cancel a transaction\n"
+"  erase                : erase the database\n"
+"  dump                 : dump the database as strings\n"
+"  keys                 : dump the database keys as strings\n"
+"  hexkeys              : dump the database keys as hex values\n"
+"  info                 : print summary info about the database\n"
+"  insert    key  data  : insert a record\n"
+"  move      key  file  : move a record to a destination tdb\n"
+"  store     key  data  : store a record (replace)\n"
+"  show      key        : show a record by key\n"
+"  delete    key        : delete a record by key\n"
+"  list                 : print the database hash table and freelist\n"
+"  free                 : print the database freelist\n"
+"  check                : check the integrity of an opened database\n"
+"  repack               : repack the database\n"
+"  speed                : perform speed tests on the database\n"
+"  ! command            : execute system command\n"
+"  1 | first            : print the first record\n"
+"  n | next             : print the next record\n"
+"  q | quit             : terminate\n"
+"  \\n                   : repeat 'next' command\n"
+"\n");
+}
+
+static void terror(const char *why)
+{
+	printf("%s\n", why);
+}
+
+static void create_tdb(const char *tdbname)
+{
+	struct tdb_logging_context log_ctx;
+	log_ctx.log_fn = tdb_log;
+
+	if (tdb) tdb_close(tdb);
+	tdb = tdb_open_ex(tdbname, 0, TDB_CLEAR_IF_FIRST | (disable_mmap?TDB_NOMMAP:0),
+			  O_RDWR | O_CREAT | O_TRUNC, 0600, &log_ctx, NULL);
+	if (!tdb) {
+		printf("Could not create %s: %s\n", tdbname, strerror(errno));
+	}
+}
+
+static void open_tdb(const char *tdbname)
+{
+	struct tdb_logging_context log_ctx;
+	log_ctx.log_fn = tdb_log;
+
+	if (tdb) tdb_close(tdb);
+	tdb = tdb_open_ex(tdbname, 0, disable_mmap?TDB_NOMMAP:0, O_RDWR, 0600,
+			  &log_ctx, NULL);
+	if (!tdb) {
+		printf("Could not open %s: %s\n", tdbname, strerror(errno));
+	}
+}
+
+static void insert_tdb(char *keyname, size_t keylen, char* data, size_t datalen)
+{
+	TDB_DATA key, dbuf;
+
+	if ((keyname == NULL) || (keylen == 0)) {
+		terror("need key");
+		return;
+	}
+
+	key.dptr = (unsigned char *)keyname;
+	key.dsize = keylen;
+	dbuf.dptr = (unsigned char *)data;
+	dbuf.dsize = datalen;
+
+	if (tdb_store(tdb, key, dbuf, TDB_INSERT) != 0) {
+		terror("insert failed");
+	}
+}
+
+static void store_tdb(char *keyname, size_t keylen, char* data, size_t datalen)
+{
+	TDB_DATA key, dbuf;
+
+	if ((keyname == NULL) || (keylen == 0)) {
+		terror("need key");
+		return;
+	}
+
+	if ((data == NULL) || (datalen == 0)) {
+		terror("need data");
+		return;
+	}
+
+	key.dptr = (unsigned char *)keyname;
+	key.dsize = keylen;
+	dbuf.dptr = (unsigned char *)data;
+	dbuf.dsize = datalen;
+
+	printf("Storing key:\n");
+	print_rec(tdb, key, dbuf, NULL);
+
+	if (tdb_store(tdb, key, dbuf, TDB_REPLACE) != 0) {
+		terror("store failed");
+	}
+}
+
+static void show_tdb(char *keyname, size_t keylen)
+{
+	TDB_DATA key, dbuf;
+
+	if ((keyname == NULL) || (keylen == 0)) {
+		terror("need key");
+		return;
+	}
+
+	key.dptr = (unsigned char *)keyname;
+	key.dsize = keylen;
+
+	dbuf = tdb_fetch(tdb, key);
+	if (!dbuf.dptr) {
+	    terror("fetch failed");
+	    return;
+	}
+	
+	print_rec(tdb, key, dbuf, NULL);
+	
+	free( dbuf.dptr );
+	
+	return;
+}
+
+static void delete_tdb(char *keyname, size_t keylen)
+{
+	TDB_DATA key;
+
+	if ((keyname == NULL) || (keylen == 0)) {
+		terror("need key");
+		return;
+	}
+
+	key.dptr = (unsigned char *)keyname;
+	key.dsize = keylen;
+
+	if (tdb_delete(tdb, key) != 0) {
+		terror("delete failed");
+	}
+}
+
+static void move_rec(char *keyname, size_t keylen, char* tdbname)
+{
+	TDB_DATA key, dbuf;
+	TDB_CONTEXT *dst_tdb;
+
+	if ((keyname == NULL) || (keylen == 0)) {
+		terror("need key");
+		return;
+	}
+
+	if ( !tdbname ) {
+		terror("need destination tdb name");
+		return;
+	}
+
+	key.dptr = (unsigned char *)keyname;
+	key.dsize = keylen;
+
+	dbuf = tdb_fetch(tdb, key);
+	if (!dbuf.dptr) {
+		terror("fetch failed");
+		return;
+	}
+	
+	print_rec(tdb, key, dbuf, NULL);
+	
+	dst_tdb = tdb_open(tdbname, 0, 0, O_RDWR, 0600);
+	if ( !dst_tdb ) {
+		terror("unable to open destination tdb");
+		return;
+	}
+	
+	if (tdb_store( dst_tdb, key, dbuf, TDB_REPLACE ) != 0) {
+		terror("failed to move record");
+	}
+	else
+		printf("record moved\n");
+	
+	tdb_close( dst_tdb );
+	
+	return;
+}
+
+static int print_rec(TDB_CONTEXT *the_tdb, TDB_DATA key, TDB_DATA dbuf, void *state)
+{
+	printf("\nkey %d bytes\n", (int)key.dsize);
+	print_asc((const char *)key.dptr, key.dsize);
+	printf("\ndata %d bytes\n", (int)dbuf.dsize);
+	print_data((const char *)dbuf.dptr, dbuf.dsize);
+	return 0;
+}
+
+static int print_key(TDB_CONTEXT *the_tdb, TDB_DATA key, TDB_DATA dbuf, void *state)
+{
+	printf("key %d bytes: ", (int)key.dsize);
+	print_asc((const char *)key.dptr, key.dsize);
+	printf("\n");
+	return 0;
+}
+
+static int print_hexkey(TDB_CONTEXT *the_tdb, TDB_DATA key, TDB_DATA dbuf, void *state)
+{
+	printf("key %d bytes\n", (int)key.dsize);
+	print_data((const char *)key.dptr, key.dsize);
+	printf("\n");
+	return 0;
+}
+
+static int total_bytes;
+
+static int traverse_fn(TDB_CONTEXT *the_tdb, TDB_DATA key, TDB_DATA dbuf, void *state)
+{
+	total_bytes += dbuf.dsize;
+	return 0;
+}
+
+static void info_tdb(void)
+{
+	char *summary = tdb_summary(tdb);
+
+	if (!summary) {
+		printf("Error = %s\n", tdb_errorstr(tdb));
+	} else {
+		printf("%s", summary);
+		free(summary);
+	}
+}
+
+static void speed_tdb(const char *tlimit)
+{
+	const char *str = "store test", *str2 = "transaction test";
+	unsigned timelimit = tlimit?atoi(tlimit):0;
+	double t;
+	int ops;
+	if (timelimit == 0) timelimit = 5;
+
+	ops = 0;
+	printf("Testing store speed for %u seconds\n", timelimit);
+	_start_timer();
+	do {
+		long int r = random();
+		TDB_DATA key, dbuf;
+		key.dptr = discard_const_p(uint8_t, str);
+		key.dsize = strlen((char *)key.dptr);
+		dbuf.dptr = (uint8_t *) &r;
+		dbuf.dsize = sizeof(r);
+		tdb_store(tdb, key, dbuf, TDB_REPLACE);
+		t = _end_timer();
+		ops++;
+	} while (t < timelimit);
+	printf("%10.3f ops/sec\n", ops/t);
+
+	ops = 0;
+	printf("Testing fetch speed for %u seconds\n", timelimit);
+	_start_timer();
+	do {
+		TDB_DATA key;
+		key.dptr = discard_const_p(uint8_t, str);
+		key.dsize = strlen((char *)key.dptr);
+		tdb_fetch(tdb, key);
+		t = _end_timer();
+		ops++;
+	} while (t < timelimit);
+	printf("%10.3f ops/sec\n", ops/t);
+
+	ops = 0;
+	printf("Testing transaction speed for %u seconds\n", timelimit);
+	_start_timer();
+	do {
+		long int r = random();
+		TDB_DATA key, dbuf;
+		key.dptr = discard_const_p(uint8_t, str2);
+		key.dsize = strlen((char *)key.dptr);
+		dbuf.dptr = (uint8_t *) &r;
+		dbuf.dsize = sizeof(r);
+		tdb_transaction_start(tdb);
+		tdb_store(tdb, key, dbuf, TDB_REPLACE);
+		tdb_transaction_commit(tdb);
+		t = _end_timer();
+		ops++;
+	} while (t < timelimit);
+	printf("%10.3f ops/sec\n", ops/t);
+
+	ops = 0;
+	printf("Testing traverse speed for %u seconds\n", timelimit);
+	_start_timer();
+	do {
+		tdb_traverse(tdb, traverse_fn, NULL);
+		t = _end_timer();
+		ops++;
+	} while (t < timelimit);
+	printf("%10.3f ops/sec\n", ops/t);
+}
+
+static void toggle_mmap(void)
+{
+	disable_mmap = !disable_mmap;
+	if (disable_mmap) {
+		printf("mmap is disabled\n");
+	} else {
+		printf("mmap is enabled\n");
+	}
+}
+
+static char *tdb_getline(const char *prompt)
+{
+	static char thisline[1024];
+	char *p;
+	fputs(prompt, stdout);
+	thisline[0] = 0;
+	p = fgets(thisline, sizeof(thisline)-1, stdin);
+	if (p) p = strchr(p, '\n');
+	if (p) *p = 0;
+	return p?thisline:NULL;
+}
+
+static int do_delete_fn(TDB_CONTEXT *the_tdb, TDB_DATA key, TDB_DATA dbuf,
+                     void *state)
+{
+    return tdb_delete(the_tdb, key);
+}
+
+static void first_record(TDB_CONTEXT *the_tdb, TDB_DATA *pkey)
+{
+	TDB_DATA dbuf;
+	*pkey = tdb_firstkey(the_tdb);
+	
+	dbuf = tdb_fetch(the_tdb, *pkey);
+	if (!dbuf.dptr) terror("fetch failed");
+	else {
+		print_rec(the_tdb, *pkey, dbuf, NULL);
+	}
+}
+
+static void next_record(TDB_CONTEXT *the_tdb, TDB_DATA *pkey)
+{
+	TDB_DATA dbuf;
+	*pkey = tdb_nextkey(the_tdb, *pkey);
+	
+	dbuf = tdb_fetch(the_tdb, *pkey);
+	if (!dbuf.dptr) 
+		terror("fetch failed");
+	else
+		print_rec(the_tdb, *pkey, dbuf, NULL);
+}
+
+static int count(TDB_DATA key, TDB_DATA data, void *private_data)
+{
+	(*(unsigned int *)private_data)++;
+	return 0;
+}
+
+static void check_db(TDB_CONTEXT *the_tdb)
+{
+	int tdbcount = 0;
+	if (!the_tdb)
+		printf("Error: No database opened!\n");
+	else if (tdb_check(the_tdb, count, &tdbcount) == -1)
+		printf("Integrity check for the opened database failed.\n");
+	else
+		printf("Database integrity is OK and has %d records.\n",
+		       tdbcount);
+}
+
+static int do_command(void)
+{
+	COMMAND_TABLE *ctp = cmd_table;
+	enum commands mycmd = CMD_HELP;
+	int cmd_len;
+
+	if (cmdname && strlen(cmdname) == 0) {
+		mycmd = CMD_NEXT;
+	} else {
+		while (ctp->name) {
+			cmd_len = strlen(ctp->name);
+			if (strncmp(ctp->name,cmdname,cmd_len) == 0) {
+				mycmd = ctp->cmd;
+				break;
+			}
+			ctp++;
+		}
+	}
+
+	switch (mycmd) {
+	case CMD_CREATE_TDB:
+		bIterate = 0;
+		create_tdb(arg1);
+		return 0;
+	case CMD_OPEN_TDB:
+		bIterate = 0;
+		open_tdb(arg1);
+		return 0;
+	case CMD_SYSTEM:
+		/* Shell command */
+		if (system(arg1) == -1) {
+			terror("system() call failed\n");
+		}
+		return 0;
+	case CMD_QUIT:
+		return 1;
+	default:
+		/* all the rest require a open database */
+		if (!tdb) {
+			bIterate = 0;
+			terror("database not open");
+			help();
+			return 0;
+		}
+		switch (mycmd) {
+		case CMD_TRANSACTION_START:
+			bIterate = 0;
+			tdb_transaction_start(tdb);
+			return 0;
+		case CMD_TRANSACTION_COMMIT:
+			bIterate = 0;
+			tdb_transaction_commit(tdb);
+			return 0;
+		case CMD_REPACK:
+			bIterate = 0;
+			tdb_repack(tdb);
+			return 0;
+		case CMD_TRANSACTION_CANCEL:
+			bIterate = 0;
+			tdb_transaction_cancel(tdb);
+			return 0;
+		case CMD_ERASE:
+			bIterate = 0;
+			tdb_traverse(tdb, do_delete_fn, NULL);
+			return 0;
+		case CMD_DUMP:
+			bIterate = 0;
+			tdb_traverse(tdb, print_rec, NULL);
+			return 0;
+		case CMD_INSERT:
+			bIterate = 0;
+			insert_tdb(arg1, arg1len,arg2,arg2len);
+			return 0;
+		case CMD_MOVE:
+			bIterate = 0;
+			move_rec(arg1,arg1len,arg2);
+			return 0;
+		case CMD_STORE:
+			bIterate = 0;
+			store_tdb(arg1,arg1len,arg2,arg2len);
+			return 0;
+		case CMD_SHOW:
+			bIterate = 0;
+			show_tdb(arg1, arg1len);
+			return 0;
+		case CMD_KEYS:
+			tdb_traverse(tdb, print_key, NULL);
+			return 0;
+		case CMD_HEXKEYS:
+			tdb_traverse(tdb, print_hexkey, NULL);
+			return 0;
+		case CMD_DELETE:
+			bIterate = 0;
+			delete_tdb(arg1,arg1len);
+			return 0;
+		case CMD_LIST_HASH_FREE:
+			tdb_dump_all(tdb);
+			return 0;
+		case CMD_LIST_FREE:
+			tdb_printfreelist(tdb);
+			return 0;
+		case CMD_INFO:
+			info_tdb();
+			return 0;
+		case CMD_SPEED:
+			speed_tdb(arg1);
+			return 0;
+		case CMD_MMAP:
+			toggle_mmap();
+			return 0;
+		case CMD_FIRST:
+			bIterate = 1;
+			first_record(tdb, &iterate_kbuf);
+			return 0;
+		case CMD_NEXT:
+			if (bIterate)
+				next_record(tdb, &iterate_kbuf);
+			return 0;
+		case CMD_CHECK:
+			check_db(tdb);
+			return 0;
+		case CMD_HELP:
+			help();
+			return 0;
+		case CMD_CREATE_TDB:
+		case CMD_OPEN_TDB:
+		case CMD_SYSTEM:
+		case CMD_QUIT:
+			/*
+			 * unhandled commands.  cases included here to avoid compiler
+			 * warnings.
+			 */
+			return 0;
+		}
+	}
+
+	return 0;
+}
+
+static char *tdb_convert_string(char *instring, size_t *sizep)
+{
+	size_t length = 0;
+	char *outp, *inp;
+	char temp[3];
+
+	outp = inp = instring;
+
+	while (*inp) {
+		if (*inp == '\\') {
+			inp++;
+			if (*inp && strchr("0123456789abcdefABCDEF",(int)*inp)) {
+				temp[0] = *inp++;
+				temp[1] = '\0';
+				if (*inp && strchr("0123456789abcdefABCDEF",(int)*inp)) {
+					temp[1] = *inp++;
+					temp[2] = '\0';
+				}
+				*outp++ = (char)strtol((const char *)temp,NULL,16);
+			} else {
+				*outp++ = *inp++;
+			}
+		} else {
+			*outp++ = *inp++;
+		}
+		length++;
+	}
+	*sizep = length;
+	return instring;
+}
+
+int main(int argc, char *argv[])
+{
+	cmdname = "";
+	arg1 = NULL;
+	arg1len = 0;
+	arg2 = NULL;
+	arg2len = 0;
+
+	if (argv[1]) {
+		cmdname = "open";
+		arg1 = argv[1];
+		do_command();
+		cmdname =  "";
+		arg1 = NULL;
+	}
+
+	switch (argc) {
+	case 1:
+	case 2:
+		/* Interactive mode */
+		while ((cmdname = tdb_getline("tdb> "))) {
+			arg2 = arg1 = NULL;
+			if ((arg1 = strchr((const char *)cmdname,' ')) != NULL) {
+				arg1++;
+				arg2 = arg1;
+				while (*arg2) {
+					if (*arg2 == ' ') {
+						*arg2++ = '\0';
+						break;
+					}
+					if ((*arg2++ == '\\') && (*arg2 == ' ')) {
+						arg2++;
+					}
+				}
+			}
+			if (arg1) arg1 = tdb_convert_string(arg1,&arg1len);
+			if (arg2) arg2 = tdb_convert_string(arg2,&arg2len);
+			if (do_command()) break;
+		}
+		break;
+	case 5:
+		arg2 = tdb_convert_string(argv[4],&arg2len);
+	case 4:
+		arg1 = tdb_convert_string(argv[3],&arg1len);
+	case 3:
+		cmdname = argv[2];
+	default:
+		do_command();
+		break;
+	}
+
+	if (tdb) tdb_close(tdb);
+
+	return 0;
+}
diff --git a/ctdb/lib/tdb/tools/tdbtorture.c b/ctdb/lib/tdb/tools/tdbtorture.c
new file mode 100644
index 0000000000..64c5043441
--- /dev/null
+++ b/ctdb/lib/tdb/tools/tdbtorture.c
@@ -0,0 +1,453 @@
+/* this tests tdb by doing lots of ops from several simultaneous
+   writers - that stresses the locking code. 
+*/
+
+#include "replace.h"
+#include "system/time.h"
+#include "system/wait.h"
+#include "system/filesys.h"
+#include "tdb.h"
+
+#ifdef HAVE_GETOPT_H
+#include <getopt.h>
+#endif
+
+
+#define REOPEN_PROB 30
+#define DELETE_PROB 8
+#define STORE_PROB 4
+#define APPEND_PROB 6
+#define TRANSACTION_PROB 10
+#define TRANSACTION_PREPARE_PROB 2
+#define LOCKSTORE_PROB 5
+#define TRAVERSE_PROB 20
+#define TRAVERSE_READ_PROB 20
+#define CULL_PROB 100
+#define KEYLEN 3
+#define DATALEN 100
+
+static struct tdb_context *db;
+static int in_transaction;
+static int error_count;
+static int always_transaction = 0;
+static int hash_size = 2;
+static int loopnum;
+static int count_pipe;
+static struct tdb_logging_context log_ctx;
+
+#ifdef PRINTF_ATTRIBUTE
+static void tdb_log(struct tdb_context *tdb, enum tdb_debug_level level, const char *format, ...) PRINTF_ATTRIBUTE(3,4);
+#endif
+static void tdb_log(struct tdb_context *tdb, enum tdb_debug_level level, const char *format, ...)
+{
+	va_list ap;
+
+	/* trace level messages do not indicate an error */
+	if (level != TDB_DEBUG_TRACE) {
+		error_count++;
+	}
+
+	va_start(ap, format);
+	vfprintf(stdout, format, ap);
+	va_end(ap);
+	fflush(stdout);
+#if 0
+	if (level != TDB_DEBUG_TRACE) {
+		char *ptr;
+		signal(SIGUSR1, SIG_IGN);
+		asprintf(&ptr,"xterm -e gdb /proc/%d/exe %d", getpid(), getpid());
+		system(ptr);
+		free(ptr);
+	}
+#endif	
+}
+
+static void fatal(const char *why)
+{
+	perror(why);
+	error_count++;
+}
+
+static char *randbuf(int len)
+{
+	char *buf;
+	int i;
+	buf = (char *)malloc(len+1);
+
+	for (i=0;i<len;i++) {
+		buf[i] = 'a' + (rand() % 26);
+	}
+	buf[i] = 0;
+	return buf;
+}
+
+static int cull_traverse(struct tdb_context *tdb, TDB_DATA key, TDB_DATA dbuf,
+			 void *state)
+{
+#if CULL_PROB
+	if (random() % CULL_PROB == 0) {
+		tdb_delete(tdb, key);
+	}
+#endif
+	return 0;
+}
+
+static void addrec_db(void)
+{
+	int klen, dlen;
+	char *k, *d;
+	TDB_DATA key, data;
+
+	klen = 1 + (rand() % KEYLEN);
+	dlen = 1 + (rand() % DATALEN);
+
+	k = randbuf(klen);
+	d = randbuf(dlen);
+
+	key.dptr = (unsigned char *)k;
+	key.dsize = klen+1;
+
+	data.dptr = (unsigned char *)d;
+	data.dsize = dlen+1;
+
+#if REOPEN_PROB
+	if (in_transaction == 0 && random() % REOPEN_PROB == 0) {
+		tdb_reopen_all(0);
+		goto next;
+	}
+#endif
+
+#if TRANSACTION_PROB
+	if (in_transaction == 0 &&
+	    (always_transaction || random() % TRANSACTION_PROB == 0)) {
+		if (tdb_transaction_start(db) != 0) {
+			fatal("tdb_transaction_start failed");
+		}
+		in_transaction++;
+		goto next;
+	}
+	if (in_transaction && random() % TRANSACTION_PROB == 0) {
+		if (random() % TRANSACTION_PREPARE_PROB == 0) {
+			if (tdb_transaction_prepare_commit(db) != 0) {
+				fatal("tdb_transaction_prepare_commit failed");
+			}
+		}
+		if (tdb_transaction_commit(db) != 0) {
+			fatal("tdb_transaction_commit failed");
+		}
+		in_transaction--;
+		goto next;
+	}
+	if (in_transaction && random() % TRANSACTION_PROB == 0) {
+		if (tdb_transaction_cancel(db) != 0) {
+			fatal("tdb_transaction_cancel failed");
+		}
+		in_transaction--;
+		goto next;
+	}
+#endif
+
+#if DELETE_PROB
+	if (random() % DELETE_PROB == 0) {
+		tdb_delete(db, key);
+		goto next;
+	}
+#endif
+
+#if STORE_PROB
+	if (random() % STORE_PROB == 0) {
+		if (tdb_store(db, key, data, TDB_REPLACE) != 0) {
+			fatal("tdb_store failed");
+		}
+		goto next;
+	}
+#endif
+
+#if APPEND_PROB
+	if (random() % APPEND_PROB == 0) {
+		if (tdb_append(db, key, data) != 0) {
+			fatal("tdb_append failed");
+		}
+		goto next;
+	}
+#endif
+
+#if LOCKSTORE_PROB
+	if (random() % LOCKSTORE_PROB == 0) {
+		tdb_chainlock(db, key);
+		data = tdb_fetch(db, key);
+		if (tdb_store(db, key, data, TDB_REPLACE) != 0) {
+			fatal("tdb_store failed");
+		}
+		if (data.dptr) free(data.dptr);
+		tdb_chainunlock(db, key);
+		goto next;
+	} 
+#endif
+
+#if TRAVERSE_PROB
+	if (random() % TRAVERSE_PROB == 0) {
+		tdb_traverse(db, cull_traverse, NULL);
+		goto next;
+	}
+#endif
+
+#if TRAVERSE_READ_PROB
+	if (random() % TRAVERSE_READ_PROB == 0) {
+		tdb_traverse_read(db, NULL, NULL);
+		goto next;
+	}
+#endif
+
+	data = tdb_fetch(db, key);
+	if (data.dptr) free(data.dptr);
+
+next:
+	free(k);
+	free(d);
+}
+
+static int traverse_fn(struct tdb_context *tdb, TDB_DATA key, TDB_DATA dbuf,
+                       void *state)
+{
+	tdb_delete(tdb, key);
+	return 0;
+}
+
+static void usage(void)
+{
+	printf("Usage: tdbtorture [-t] [-k] [-n NUM_PROCS] [-l NUM_LOOPS] [-s SEED] [-H HASH_SIZE]\n");
+	exit(0);
+}
+
+static void send_count_and_suicide(int sig)
+{
+	/* This ensures our successor can continue where we left off. */
+	write(count_pipe, &loopnum, sizeof(loopnum));
+	/* This gives a unique signature. */
+	kill(getpid(), SIGUSR2);
+}
+
+static int run_child(const char *filename, int i, int seed, unsigned num_loops, unsigned start)
+{
+	db = tdb_open_ex(filename, hash_size, TDB_DEFAULT,
+			 O_RDWR | O_CREAT, 0600, &log_ctx, NULL);
+	if (!db) {
+		fatal("db open failed");
+	}
+
+	srand(seed + i);
+	srandom(seed + i);
+
+	/* Set global, then we're ready to handle being killed. */
+	loopnum = start;
+	signal(SIGUSR1, send_count_and_suicide);
+
+	for (;loopnum<num_loops && error_count == 0;loopnum++) {
+		addrec_db();
+	}
+
+	if (error_count == 0) {
+		tdb_traverse_read(db, NULL, NULL);
+		if (always_transaction) {
+			while (in_transaction) {
+				tdb_transaction_cancel(db);
+				in_transaction--;
+			}
+			if (tdb_transaction_start(db) != 0)
+				fatal("tdb_transaction_start failed");
+		}
+		tdb_traverse(db, traverse_fn, NULL);
+		tdb_traverse(db, traverse_fn, NULL);
+		if (always_transaction) {
+			if (tdb_transaction_commit(db) != 0)
+				fatal("tdb_transaction_commit failed");
+		}
+	}
+
+	tdb_close(db);
+
+	return (error_count < 100 ? error_count : 100);
+}
+
+static char *test_path(const char *filename)
+{
+	const char *prefix = getenv("TEST_DATA_PREFIX");
+
+	if (prefix) {
+		char *path = NULL;
+		int ret;
+
+		ret = asprintf(&path, "%s/%s", prefix, filename);
+		if (ret == -1) {
+			return NULL;
+		}
+		return path;
+	}
+
+	return strdup(filename);
+}
+
+int main(int argc, char * const *argv)
+{
+	int i, seed = -1;
+	int num_loops = 5000;
+	int num_procs = 3;
+	int c, pfds[2];
+	extern char *optarg;
+	pid_t *pids;
+	int kill_random = 0;
+	int *done;
+	char *test_tdb;
+
+	log_ctx.log_fn = tdb_log;
+
+	while ((c = getopt(argc, argv, "n:l:s:H:thk")) != -1) {
+		switch (c) {
+		case 'n':
+			num_procs = strtol(optarg, NULL, 0);
+			break;
+		case 'l':
+			num_loops = strtol(optarg, NULL, 0);
+			break;
+		case 'H':
+			hash_size = strtol(optarg, NULL, 0);
+			break;
+		case 's':
+			seed = strtol(optarg, NULL, 0);
+			break;
+		case 't':
+			always_transaction = 1;
+			break;
+		case 'k':
+			kill_random = 1;
+			break;
+		default:
+			usage();
+		}
+	}
+
+	test_tdb = test_path("torture.tdb");
+
+	unlink(test_tdb);
+
+	if (seed == -1) {
+		seed = (getpid() + time(NULL)) & 0x7FFFFFFF;
+	}
+
+	if (num_procs == 1 && !kill_random) {
+		/* Don't fork for this case, makes debugging easier. */
+		error_count = run_child(test_tdb, 0, seed, num_loops, 0);
+		goto done;
+	}
+
+	pids = (pid_t *)calloc(sizeof(pid_t), num_procs);
+	done = (int *)calloc(sizeof(int), num_procs);
+
+	if (pipe(pfds) != 0) {
+		perror("Creating pipe");
+		exit(1);
+	}
+	count_pipe = pfds[1];
+
+	for (i=0;i<num_procs;i++) {
+		if ((pids[i]=fork()) == 0) {
+			close(pfds[0]);
+			if (i == 0) {
+				printf("Testing with %d processes, %d loops, %d hash_size, seed=%d%s\n",
+				       num_procs, num_loops, hash_size, seed, always_transaction ? " (all within transactions)" : "");
+			}
+			exit(run_child(test_tdb, i, seed, num_loops, 0));
+		}
+	}
+
+	while (num_procs) {
+		int status, j;
+		pid_t pid;
+
+		if (error_count != 0) {
+			/* try and stop the test on any failure */
+			for (j=0;j<num_procs;j++) {
+				if (pids[j] != 0) {
+					kill(pids[j], SIGTERM);
+				}
+			}
+		}
+
+		pid = waitpid(-1, &status, kill_random ? WNOHANG : 0);
+		if (pid == 0) {
+			struct timeval tv;
+
+			/* Sleep for 1/10 second. */
+			tv.tv_sec = 0;
+			tv.tv_usec = 100000;
+			select(0, NULL, NULL, NULL, &tv);
+
+			/* Kill someone. */
+			kill(pids[random() % num_procs], SIGUSR1);
+			continue;
+		}
+
+		if (pid == -1) {
+			perror("failed to wait for child\n");
+			exit(1);
+		}
+
+		for (j=0;j<num_procs;j++) {
+			if (pids[j] == pid) break;
+		}
+		if (j == num_procs) {
+			printf("unknown child %d exited!?\n", (int)pid);
+			exit(1);
+		}
+		if (WIFSIGNALED(status)) {
+			if (WTERMSIG(status) == SIGUSR2
+			    || WTERMSIG(status) == SIGUSR1) {
+				/* SIGUSR2 means they wrote to pipe. */
+				if (WTERMSIG(status) == SIGUSR2) {
+					read(pfds[0], &done[j],
+					     sizeof(done[j]));
+				}
+				pids[j] = fork();
+				if (pids[j] == 0)
+					exit(run_child(test_tdb, j, seed,
+						       num_loops, done[j]));
+				printf("Restarting child %i for %u-%u\n",
+				       j, done[j], num_loops);
+				continue;
+			}
+			printf("child %d exited with signal %d\n",
+			       (int)pid, WTERMSIG(status));
+			error_count++;
+		} else {
+			if (WEXITSTATUS(status) != 0) {
+				printf("child %d exited with status %d\n",
+				       (int)pid, WEXITSTATUS(status));
+				error_count++;
+			}
+		}
+		memmove(&pids[j], &pids[j+1],
+			(num_procs - j - 1)*sizeof(pids[0]));
+		num_procs--;
+	}
+
+	free(pids);
+
+done:
+	if (error_count == 0) {
+		db = tdb_open_ex(test_tdb, hash_size, TDB_DEFAULT,
+				 O_RDWR, 0, &log_ctx, NULL);
+		if (!db) {
+			fatal("db open failed");
+		}
+		if (tdb_check(db, NULL, NULL) == -1) {
+			printf("db check failed");
+			exit(1);
+		}
+		tdb_close(db);
+		printf("OK\n");
+	}
+
+	free(test_tdb);
+	return error_count;
+}
diff --git a/ctdb/lib/tdb/web/index.html b/ctdb/lib/tdb/web/index.html
new file mode 100644
index 0000000000..99e8a2fe42
--- /dev/null
+++ b/ctdb/lib/tdb/web/index.html
@@ -0,0 +1,48 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
+<HTML>
+<HEAD>
+<TITLE>ldb</TITLE>
+</HEAD>
+<BODY BGCOLOR="#ffffff" TEXT="#000000" VLINK="#292555" LINK="#292555" ALINK="#cc0033">
+
+<h1>tdb</h1>
+
+TDB is a Trivial Database. In concept, it is very much like GDBM, and BSD's DB 
+except that it allows multiple simultaneous writers and uses locking 
+internally to keep writers from trampling on each other. TDB is also extremely 
+small.
+
+<h2>Download</h2>
+You can download the latest releases of tdb from the <a
+href="http://samba.org/ftp/tdb">tdb directory</a> on the samba public
+source archive.
+
+
+<h2>Discussion and bug reports</h2>
+
+tdb does not currently have its own mailing list or bug tracking
+system. For now, please use the <a
+href="https://lists.samba.org/mailman/listinfo/samba-technical">samba-technical</a>
+mailing list, and the <a href="http://bugzilla.samba.org/">Samba
+bugzilla</a> bug tracking system.
+
+<h2>Download</h2>
+
+You can download the latest code either via git or rsync.<br>
+<br>
+To fetch via git see the following guide:<br>
+<a href="http://wiki.samba.org/index.php/Using_Git_for_Samba_Development">Using Git for Samba Development</a><br>
+Once you have cloned the tree switch to the master branch and cd into the source/lib/tdb directory.<br>
+<br>
+To fetch via rsync use these commands:
+
+<pre>
+  rsync -Pavz samba.org::ftp/unpacked/standalone_projects/lib/tdb .
+  rsync -Pavz samba.org::ftp/unpacked/standalone_projects/lib/replace .
+</pre>
+
+and build in tdb. It will find the replace library in the directory
+above automatically.
+
+</BODY>
+</HTML>
diff --git a/ctdb/lib/tdb/wscript b/ctdb/lib/tdb/wscript
new file mode 100644
index 0000000000..9d309a0325
--- /dev/null
+++ b/ctdb/lib/tdb/wscript
@@ -0,0 +1,224 @@
+#!/usr/bin/env python
+
+APPNAME = 'tdb'
+VERSION = '1.2.11'
+
+blddir = 'bin'
+
+import sys, os
+
+# find the buildtools directory
+srcdir = '.'
+while not os.path.exists(srcdir+'/buildtools') and len(srcdir.split('/')) < 5:
+    srcdir = '../' + srcdir
+sys.path.insert(0, srcdir + '/buildtools/wafsamba')
+
+import wafsamba, samba_dist, Options, Logs
+
+samba_dist.DIST_DIRS('lib/tdb:. lib/replace:lib/replace buildtools:buildtools')
+
+def set_options(opt):
+    opt.BUILTIN_DEFAULT('replace')
+    opt.PRIVATE_EXTENSION_DEFAULT('tdb', noextension='tdb')
+    opt.RECURSE('lib/replace')
+    if opt.IN_LAUNCH_DIR():
+        opt.add_option('--disable-python',
+                       help=("disable the pytdb module"),
+                       action="store_true", dest='disable_python', default=False)
+
+
+def configure(conf):
+    conf.RECURSE('lib/replace')
+
+    conf.env.standalone_tdb = conf.IN_LAUNCH_DIR()
+    conf.env.building_tdb = True
+
+    if not conf.env.standalone_tdb:
+        if conf.CHECK_BUNDLED_SYSTEM_PKG('tdb', minversion=VERSION,
+                                     implied_deps='replace'):
+            conf.define('USING_SYSTEM_TDB', 1)
+            conf.env.building_tdb = False
+            if conf.CHECK_BUNDLED_SYSTEM_PYTHON('pytdb', 'tdb', minversion=VERSION):
+                conf.define('USING_SYSTEM_PYTDB', 1)
+
+    conf.env.disable_python = getattr(Options.options, 'disable_python', False)
+
+    conf.CHECK_XSLTPROC_MANPAGES()
+
+    if not conf.env.disable_python:
+        # also disable if we don't have the python libs installed
+        conf.find_program('python', var='PYTHON')
+        conf.check_tool('python')
+        conf.check_python_version((2,4,2))
+        conf.SAMBA_CHECK_PYTHON_HEADERS(mandatory=False)
+        if not conf.env.HAVE_PYTHON_H:
+            Logs.warn('Disabling pytdb as python devel libs not found')
+            conf.env.disable_python = True
+
+    conf.SAMBA_CONFIG_H()
+
+    conf.SAMBA_CHECK_UNDEFINED_SYMBOL_FLAGS()
+
+def build(bld):
+    bld.RECURSE('lib/replace')
+
+    COMMON_SRC = bld.SUBDIR('common',
+                            '''check.c error.c tdb.c traverse.c
+                            freelistcheck.c lock.c dump.c freelist.c
+                            io.c open.c transaction.c hash.c summary.c rescue.c''')
+
+    if bld.env.standalone_tdb:
+        bld.env.PKGCONFIGDIR = '${LIBDIR}/pkgconfig'
+        private_library = False
+    else:
+        private_library = True
+
+    if not bld.CONFIG_SET('USING_SYSTEM_TDB'):
+        bld.SAMBA_LIBRARY('tdb',
+                          COMMON_SRC,
+                          deps='replace',
+                          includes='include',
+                          abi_directory='ABI',
+                          abi_match='tdb_*',
+                          hide_symbols=True,
+                          vnum=VERSION,
+                          public_headers='include/tdb.h',
+                          public_headers_install=not private_library,
+                          pc_files='tdb.pc',
+                          private_library=private_library)
+
+        bld.SAMBA_BINARY('tdbtorture',
+                         'tools/tdbtorture.c',
+                         'tdb',
+                         install=False)
+
+        bld.SAMBA_BINARY('tdbrestore',
+                         'tools/tdbrestore.c',
+                         'tdb', manpages='manpages/tdbrestore.8')
+
+        bld.SAMBA_BINARY('tdbdump',
+                         'tools/tdbdump.c',
+                         'tdb', manpages='manpages/tdbdump.8')
+
+        bld.SAMBA_BINARY('tdbbackup',
+                         'tools/tdbbackup.c',
+                         'tdb',
+                         manpages='manpages/tdbbackup.8')
+
+        bld.SAMBA_BINARY('tdbtool',
+                         'tools/tdbtool.c',
+                         'tdb', manpages='manpages/tdbtool.8')
+
+        # FIXME: This hardcoded list is stupid, stupid, stupid.
+        bld.SAMBA_SUBSYSTEM('tdb-test-helpers',
+                            'test/external-agent.c test/lock-tracking.c test/logging.c',
+                            'replace',
+                            includes='include')
+
+        bld.SAMBA_BINARY('tdb1-run-3G-file', 'test/run-3G-file.c',
+                         'replace tdb-test-helpers', includes='include', install=False)
+        bld.SAMBA_BINARY('tdb1-run-bad-tdb-header', 'test/run-bad-tdb-header.c',
+                         'replace tdb-test-helpers', includes='include', install=False)
+        bld.SAMBA_BINARY('tdb1-run', 'test/run.c',
+                         'replace tdb-test-helpers', includes='include', install=False)
+        bld.SAMBA_BINARY('tdb1-run-check', 'test/run-check.c',
+                         'replace tdb-test-helpers', includes='include', install=False)
+        bld.SAMBA_BINARY('tdb1-run-corrupt', 'test/run-corrupt.c',
+                         'replace tdb-test-helpers', includes='include', install=False)
+        bld.SAMBA_BINARY('tdb1-run-die-during-transaction', 'test/run-die-during-transaction.c',
+                         'replace tdb-test-helpers', includes='include', install=False)
+        bld.SAMBA_BINARY('tdb1-run-endian', 'test/run-endian.c',
+                         'replace tdb-test-helpers', includes='include', install=False)
+        bld.SAMBA_BINARY('tdb1-run-incompatible', 'test/run-incompatible.c',
+                         'replace tdb-test-helpers', includes='include', install=False)
+        bld.SAMBA_BINARY('tdb1-run-nested-transactions', 'test/run-nested-transactions.c',
+                         'replace tdb-test-helpers', includes='include', install=False)
+        bld.SAMBA_BINARY('tdb1-run-nested-traverse', 'test/run-nested-traverse.c',
+                         'replace tdb-test-helpers', includes='include', install=False)
+        bld.SAMBA_BINARY('tdb1-run-no-lock-during-traverse', 'test/run-no-lock-during-traverse.c',
+                         'replace tdb-test-helpers', includes='include', install=False)
+        bld.SAMBA_BINARY('tdb1-run-oldhash', 'test/run-oldhash.c',
+                         'replace tdb-test-helpers', includes='include', install=False)
+        bld.SAMBA_BINARY('tdb1-run-open-during-transaction', 'test/run-open-during-transaction.c',
+                         'replace tdb-test-helpers', includes='include', install=False)
+        bld.SAMBA_BINARY('tdb1-run-readonly-check', 'test/run-readonly-check.c',
+                         'replace tdb-test-helpers', includes='include', install=False)
+        bld.SAMBA_BINARY('tdb1-run-rescue', 'test/run-rescue.c',
+                         'replace tdb-test-helpers', includes='include', install=False)
+        bld.SAMBA_BINARY('tdb1-run-rescue-find_entry', 'test/run-rescue-find_entry.c',
+                         'replace tdb-test-helpers', includes='include', install=False)
+        bld.SAMBA_BINARY('tdb1-run-rwlock-check', 'test/run-rwlock-check.c',
+                         'replace tdb-test-helpers', includes='include', install=False)
+        bld.SAMBA_BINARY('tdb1-run-summary', 'test/run-summary.c',
+                         'replace tdb-test-helpers', includes='include', install=False)
+        bld.SAMBA_BINARY('tdb1-run-transaction-expand', 'test/run-transaction-expand.c',
+                         'replace tdb-test-helpers', includes='include', install=False)
+        bld.SAMBA_BINARY('tdb1-run-traverse-in-transaction', 'test/run-traverse-in-transaction.c',
+                         'replace tdb-test-helpers', includes='include', install=False)
+        bld.SAMBA_BINARY('tdb1-run-wronghash-fail', 'test/run-wronghash-fail.c',
+                         'replace tdb-test-helpers', includes='include', install=False)
+        bld.SAMBA_BINARY('tdb1-run-zero-append', 'test/run-zero-append.c',
+                         'replace tdb-test-helpers', includes='include', install=False)
+
+    if not bld.CONFIG_SET('USING_SYSTEM_PYTDB'):
+        bld.SAMBA_PYTHON('pytdb',
+                         'pytdb.c',
+                         deps='tdb',
+                         enabled=not bld.env.disable_python,
+                         realname='tdb.so',
+                         cflags='-DPACKAGE_VERSION=\"%s\"' % VERSION)
+
+def testonly(ctx):
+    '''run tdb testsuite'''
+    import Utils, samba_utils, shutil
+    ecode = 0
+
+    test_prefix = "%s/st" % (Utils.g_module.blddir)
+    shutil.rmtree(test_prefix, ignore_errors=True)
+    os.makedirs(test_prefix)
+    os.environ['TEST_DATA_PREFIX'] = test_prefix
+
+    env = samba_utils.LOAD_ENVIRONMENT()
+    # FIXME: This is horrible :(
+    if env.building_tdb:
+        # Create scratch directory for tests.
+        testdir = os.path.join(test_prefix, 'tdb-tests')
+        samba_utils.mkdir_p(testdir)
+        # Symlink back to source dir so it can find tests in test/
+        link = os.path.join(testdir, 'test')
+        if not os.path.exists(link):
+            os.symlink(os.path.abspath(os.path.join(env.cwd, 'test')), link)
+
+        for f in 'tdb1-run-3G-file', 'tdb1-run-bad-tdb-header', 'tdb1-run', 'tdb1-run-check', 'tdb1-run-corrupt', 'tdb1-run-die-during-transaction', 'tdb1-run-endian', 'tdb1-run-incompatible', 'tdb1-run-nested-transactions', 'tdb1-run-nested-traverse', 'tdb1-run-no-lock-during-traverse', 'tdb1-run-oldhash', 'tdb1-run-open-during-transaction', 'tdb1-run-readonly-check', 'tdb1-run-rescue', 'tdb1-run-rescue-find_entry', 'tdb1-run-rwlock-check', 'tdb1-run-summary', 'tdb1-run-transaction-expand', 'tdb1-run-traverse-in-transaction', 'tdb1-run-wronghash-fail', 'tdb1-run-zero-append':
+            cmd = "cd " + testdir + " && " + os.path.abspath(os.path.join(Utils.g_module.blddir, f)) + " > test-output 2>&1"
+            print("..." + f)
+            ret = samba_utils.RUN_COMMAND(cmd)
+            if ret != 0:
+                print("%s failed:" % f)
+                samba_utils.RUN_COMMAND("cat " + os.path.join(testdir, 'test-output'))
+                ecode = ret
+                break
+
+    if ecode == 0:
+        cmd = os.path.join(Utils.g_module.blddir, 'tdbtorture')
+        ret = samba_utils.RUN_COMMAND(cmd)
+        print("testsuite returned %d" % ret)
+        if ret != 0:
+            ecode = ret
+    sys.exit(ecode)
+
+# WAF doesn't build the unit tests for this, maybe because they don't link with tdb?
+# This forces it
+def test(ctx):
+    import Scripting
+    Scripting.commands.append('build')
+    Scripting.commands.append('testonly')
+
+def dist():
+    '''makes a tarball for distribution'''
+    samba_dist.dist()
+
+def reconfigure(ctx):
+    '''reconfigure if config scripts have changed'''
+    import samba_utils
+    samba_utils.reconfigure(ctx)
diff --git a/ctdb/lib/tevent/ABI/tevent-0.9.10.sigs b/ctdb/lib/tevent/ABI/tevent-0.9.10.sigs
new file mode 100644
index 0000000000..9adaba579b
--- /dev/null
+++ b/ctdb/lib/tevent/ABI/tevent-0.9.10.sigs
@@ -0,0 +1,73 @@
+_tevent_add_fd: struct tevent_fd *(struct tevent_context *, TALLOC_CTX *, int, uint16_t, tevent_fd_handler_t, void *, const char *, const char *)
+_tevent_add_signal: struct tevent_signal *(struct tevent_context *, TALLOC_CTX *, int, int, tevent_signal_handler_t, void *, const char *, const char *)
+_tevent_add_timer: struct tevent_timer *(struct tevent_context *, TALLOC_CTX *, struct timeval, tevent_timer_handler_t, void *, const char *, const char *)
+_tevent_create_immediate: struct tevent_immediate *(TALLOC_CTX *, const char *)
+_tevent_loop_once: int (struct tevent_context *, const char *)
+_tevent_loop_until: int (struct tevent_context *, bool (*)(void *), void *, const char *)
+_tevent_loop_wait: int (struct tevent_context *, const char *)
+_tevent_queue_create: struct tevent_queue *(TALLOC_CTX *, const char *, const char *)
+_tevent_req_callback_data: void *(struct tevent_req *)
+_tevent_req_cancel: bool (struct tevent_req *, const char *)
+_tevent_req_create: struct tevent_req *(TALLOC_CTX *, void *, size_t, const char *, const char *)
+_tevent_req_data: void *(struct tevent_req *)
+_tevent_req_done: void (struct tevent_req *, const char *)
+_tevent_req_error: bool (struct tevent_req *, uint64_t, const char *)
+_tevent_req_nomem: bool (const void *, struct tevent_req *, const char *)
+_tevent_req_notify_callback: void (struct tevent_req *, const char *)
+_tevent_schedule_immediate: void (struct tevent_immediate *, struct tevent_context *, tevent_immediate_handler_t, void *, const char *, const char *)
+tevent_backend_list: const char **(TALLOC_CTX *)
+tevent_cleanup_pending_signal_handlers: void (struct tevent_signal *)
+tevent_common_add_fd: struct tevent_fd *(struct tevent_context *, TALLOC_CTX *, int, uint16_t, tevent_fd_handler_t, void *, const char *, const char *)
+tevent_common_add_signal: struct tevent_signal *(struct tevent_context *, TALLOC_CTX *, int, int, tevent_signal_handler_t, void *, const char *, const char *)
+tevent_common_add_timer: struct tevent_timer *(struct tevent_context *, TALLOC_CTX *, struct timeval, tevent_timer_handler_t, void *, const char *, const char *)
+tevent_common_check_signal: int (struct tevent_context *)
+tevent_common_context_destructor: int (struct tevent_context *)
+tevent_common_fd_destructor: int (struct tevent_fd *)
+tevent_common_fd_get_flags: uint16_t (struct tevent_fd *)
+tevent_common_fd_set_close_fn: void (struct tevent_fd *, tevent_fd_close_fn_t)
+tevent_common_fd_set_flags: void (struct tevent_fd *, uint16_t)
+tevent_common_loop_immediate: bool (struct tevent_context *)
+tevent_common_loop_timer_delay: struct timeval (struct tevent_context *)
+tevent_common_loop_wait: int (struct tevent_context *, const char *)
+tevent_common_schedule_immediate: void (struct tevent_immediate *, struct tevent_context *, tevent_immediate_handler_t, void *, const char *, const char *)
+tevent_context_init: struct tevent_context *(TALLOC_CTX *)
+tevent_context_init_byname: struct tevent_context *(TALLOC_CTX *, const char *)
+tevent_debug: void (struct tevent_context *, enum tevent_debug_level, const char *, ...)
+tevent_fd_get_flags: uint16_t (struct tevent_fd *)
+tevent_fd_set_auto_close: void (struct tevent_fd *)
+tevent_fd_set_close_fn: void (struct tevent_fd *, tevent_fd_close_fn_t)
+tevent_fd_set_flags: void (struct tevent_fd *, uint16_t)
+tevent_loop_allow_nesting: void (struct tevent_context *)
+tevent_loop_set_nesting_hook: void (struct tevent_context *, tevent_nesting_hook, void *)
+tevent_queue_add: bool (struct tevent_queue *, struct tevent_context *, struct tevent_req *, tevent_queue_trigger_fn_t, void *)
+tevent_queue_length: size_t (struct tevent_queue *)
+tevent_queue_start: void (struct tevent_queue *)
+tevent_queue_stop: void (struct tevent_queue *)
+tevent_re_initialise: int (struct tevent_context *)
+tevent_register_backend: bool (const char *, const struct tevent_ops *)
+tevent_req_default_print: char *(struct tevent_req *, TALLOC_CTX *)
+tevent_req_is_error: bool (struct tevent_req *, enum tevent_req_state *, uint64_t *)
+tevent_req_is_in_progress: bool (struct tevent_req *)
+tevent_req_poll: bool (struct tevent_req *, struct tevent_context *)
+tevent_req_post: struct tevent_req *(struct tevent_req *, struct tevent_context *)
+tevent_req_print: char *(TALLOC_CTX *, struct tevent_req *)
+tevent_req_received: void (struct tevent_req *)
+tevent_req_set_callback: void (struct tevent_req *, tevent_req_fn, void *)
+tevent_req_set_cancel_fn: void (struct tevent_req *, tevent_req_cancel_fn)
+tevent_req_set_endtime: bool (struct tevent_req *, struct tevent_context *, struct timeval)
+tevent_req_set_print_fn: void (struct tevent_req *, tevent_req_print_fn)
+tevent_set_abort_fn: void (void (*)(const char *))
+tevent_set_debug: int (struct tevent_context *, void (*)(void *, enum tevent_debug_level, const char *, va_list), void *)
+tevent_set_debug_stderr: int (struct tevent_context *)
+tevent_set_default_backend: void (const char *)
+tevent_signal_support: bool (struct tevent_context *)
+tevent_timeval_add: struct timeval (const struct timeval *, uint32_t, uint32_t)
+tevent_timeval_compare: int (const struct timeval *, const struct timeval *)
+tevent_timeval_current: struct timeval (void)
+tevent_timeval_current_ofs: struct timeval (uint32_t, uint32_t)
+tevent_timeval_is_zero: bool (const struct timeval *)
+tevent_timeval_set: struct timeval (uint32_t, uint32_t)
+tevent_timeval_until: struct timeval (const struct timeval *, const struct timeval *)
+tevent_timeval_zero: struct timeval (void)
+tevent_wakeup_recv: bool (struct tevent_req *)
+tevent_wakeup_send: struct tevent_req *(TALLOC_CTX *, struct tevent_context *, struct timeval)
diff --git a/ctdb/lib/tevent/ABI/tevent-0.9.11.sigs b/ctdb/lib/tevent/ABI/tevent-0.9.11.sigs
new file mode 100644
index 0000000000..9adaba579b
--- /dev/null
+++ b/ctdb/lib/tevent/ABI/tevent-0.9.11.sigs
@@ -0,0 +1,73 @@
+_tevent_add_fd: struct tevent_fd *(struct tevent_context *, TALLOC_CTX *, int, uint16_t, tevent_fd_handler_t, void *, const char *, const char *)
+_tevent_add_signal: struct tevent_signal *(struct tevent_context *, TALLOC_CTX *, int, int, tevent_signal_handler_t, void *, const char *, const char *)
+_tevent_add_timer: struct tevent_timer *(struct tevent_context *, TALLOC_CTX *, struct timeval, tevent_timer_handler_t, void *, const char *, const char *)
+_tevent_create_immediate: struct tevent_immediate *(TALLOC_CTX *, const char *)
+_tevent_loop_once: int (struct tevent_context *, const char *)
+_tevent_loop_until: int (struct tevent_context *, bool (*)(void *), void *, const char *)
+_tevent_loop_wait: int (struct tevent_context *, const char *)
+_tevent_queue_create: struct tevent_queue *(TALLOC_CTX *, const char *, const char *)
+_tevent_req_callback_data: void *(struct tevent_req *)
+_tevent_req_cancel: bool (struct tevent_req *, const char *)
+_tevent_req_create: struct tevent_req *(TALLOC_CTX *, void *, size_t, const char *, const char *)
+_tevent_req_data: void *(struct tevent_req *)
+_tevent_req_done: void (struct tevent_req *, const char *)
+_tevent_req_error: bool (struct tevent_req *, uint64_t, const char *)
+_tevent_req_nomem: bool (const void *, struct tevent_req *, const char *)
+_tevent_req_notify_callback: void (struct tevent_req *, const char *)
+_tevent_schedule_immediate: void (struct tevent_immediate *, struct tevent_context *, tevent_immediate_handler_t, void *, const char *, const char *)
+tevent_backend_list: const char **(TALLOC_CTX *)
+tevent_cleanup_pending_signal_handlers: void (struct tevent_signal *)
+tevent_common_add_fd: struct tevent_fd *(struct tevent_context *, TALLOC_CTX *, int, uint16_t, tevent_fd_handler_t, void *, const char *, const char *)
+tevent_common_add_signal: struct tevent_signal *(struct tevent_context *, TALLOC_CTX *, int, int, tevent_signal_handler_t, void *, const char *, const char *)
+tevent_common_add_timer: struct tevent_timer *(struct tevent_context *, TALLOC_CTX *, struct timeval, tevent_timer_handler_t, void *, const char *, const char *)
+tevent_common_check_signal: int (struct tevent_context *)
+tevent_common_context_destructor: int (struct tevent_context *)
+tevent_common_fd_destructor: int (struct tevent_fd *)
+tevent_common_fd_get_flags: uint16_t (struct tevent_fd *)
+tevent_common_fd_set_close_fn: void (struct tevent_fd *, tevent_fd_close_fn_t)
+tevent_common_fd_set_flags: void (struct tevent_fd *, uint16_t)
+tevent_common_loop_immediate: bool (struct tevent_context *)
+tevent_common_loop_timer_delay: struct timeval (struct tevent_context *)
+tevent_common_loop_wait: int (struct tevent_context *, const char *)
+tevent_common_schedule_immediate: void (struct tevent_immediate *, struct tevent_context *, tevent_immediate_handler_t, void *, const char *, const char *)
+tevent_context_init: struct tevent_context *(TALLOC_CTX *)
+tevent_context_init_byname: struct tevent_context *(TALLOC_CTX *, const char *)
+tevent_debug: void (struct tevent_context *, enum tevent_debug_level, const char *, ...)
+tevent_fd_get_flags: uint16_t (struct tevent_fd *)
+tevent_fd_set_auto_close: void (struct tevent_fd *)
+tevent_fd_set_close_fn: void (struct tevent_fd *, tevent_fd_close_fn_t)
+tevent_fd_set_flags: void (struct tevent_fd *, uint16_t)
+tevent_loop_allow_nesting: void (struct tevent_context *)
+tevent_loop_set_nesting_hook: void (struct tevent_context *, tevent_nesting_hook, void *)
+tevent_queue_add: bool (struct tevent_queue *, struct tevent_context *, struct tevent_req *, tevent_queue_trigger_fn_t, void *)
+tevent_queue_length: size_t (struct tevent_queue *)
+tevent_queue_start: void (struct tevent_queue *)
+tevent_queue_stop: void (struct tevent_queue *)
+tevent_re_initialise: int (struct tevent_context *)
+tevent_register_backend: bool (const char *, const struct tevent_ops *)
+tevent_req_default_print: char *(struct tevent_req *, TALLOC_CTX *)
+tevent_req_is_error: bool (struct tevent_req *, enum tevent_req_state *, uint64_t *)
+tevent_req_is_in_progress: bool (struct tevent_req *)
+tevent_req_poll: bool (struct tevent_req *, struct tevent_context *)
+tevent_req_post: struct tevent_req *(struct tevent_req *, struct tevent_context *)
+tevent_req_print: char *(TALLOC_CTX *, struct tevent_req *)
+tevent_req_received: void (struct tevent_req *)
+tevent_req_set_callback: void (struct tevent_req *, tevent_req_fn, void *)
+tevent_req_set_cancel_fn: void (struct tevent_req *, tevent_req_cancel_fn)
+tevent_req_set_endtime: bool (struct tevent_req *, struct tevent_context *, struct timeval)
+tevent_req_set_print_fn: void (struct tevent_req *, tevent_req_print_fn)
+tevent_set_abort_fn: void (void (*)(const char *))
+tevent_set_debug: int (struct tevent_context *, void (*)(void *, enum tevent_debug_level, const char *, va_list), void *)
+tevent_set_debug_stderr: int (struct tevent_context *)
+tevent_set_default_backend: void (const char *)
+tevent_signal_support: bool (struct tevent_context *)
+tevent_timeval_add: struct timeval (const struct timeval *, uint32_t, uint32_t)
+tevent_timeval_compare: int (const struct timeval *, const struct timeval *)
+tevent_timeval_current: struct timeval (void)
+tevent_timeval_current_ofs: struct timeval (uint32_t, uint32_t)
+tevent_timeval_is_zero: bool (const struct timeval *)
+tevent_timeval_set: struct timeval (uint32_t, uint32_t)
+tevent_timeval_until: struct timeval (const struct timeval *, const struct timeval *)
+tevent_timeval_zero: struct timeval (void)
+tevent_wakeup_recv: bool (struct tevent_req *)
+tevent_wakeup_send: struct tevent_req *(TALLOC_CTX *, struct tevent_context *, struct timeval)
diff --git a/ctdb/lib/tevent/ABI/tevent-0.9.12.sigs b/ctdb/lib/tevent/ABI/tevent-0.9.12.sigs
new file mode 100644
index 0000000000..df9b08dfd5
--- /dev/null
+++ b/ctdb/lib/tevent/ABI/tevent-0.9.12.sigs
@@ -0,0 +1,74 @@
+_tevent_add_fd: struct tevent_fd *(struct tevent_context *, TALLOC_CTX *, int, uint16_t, tevent_fd_handler_t, void *, const char *, const char *)
+_tevent_add_signal: struct tevent_signal *(struct tevent_context *, TALLOC_CTX *, int, int, tevent_signal_handler_t, void *, const char *, const char *)
+_tevent_add_timer: struct tevent_timer *(struct tevent_context *, TALLOC_CTX *, struct timeval, tevent_timer_handler_t, void *, const char *, const char *)
+_tevent_create_immediate: struct tevent_immediate *(TALLOC_CTX *, const char *)
+_tevent_loop_once: int (struct tevent_context *, const char *)
+_tevent_loop_until: int (struct tevent_context *, bool (*)(void *), void *, const char *)
+_tevent_loop_wait: int (struct tevent_context *, const char *)
+_tevent_queue_create: struct tevent_queue *(TALLOC_CTX *, const char *, const char *)
+_tevent_req_callback_data: void *(struct tevent_req *)
+_tevent_req_cancel: bool (struct tevent_req *, const char *)
+_tevent_req_create: struct tevent_req *(TALLOC_CTX *, void *, size_t, const char *, const char *)
+_tevent_req_data: void *(struct tevent_req *)
+_tevent_req_done: void (struct tevent_req *, const char *)
+_tevent_req_error: bool (struct tevent_req *, uint64_t, const char *)
+_tevent_req_nomem: bool (const void *, struct tevent_req *, const char *)
+_tevent_req_notify_callback: void (struct tevent_req *, const char *)
+_tevent_req_oom: void (struct tevent_req *, const char *)
+_tevent_schedule_immediate: void (struct tevent_immediate *, struct tevent_context *, tevent_immediate_handler_t, void *, const char *, const char *)
+tevent_backend_list: const char **(TALLOC_CTX *)
+tevent_cleanup_pending_signal_handlers: void (struct tevent_signal *)
+tevent_common_add_fd: struct tevent_fd *(struct tevent_context *, TALLOC_CTX *, int, uint16_t, tevent_fd_handler_t, void *, const char *, const char *)
+tevent_common_add_signal: struct tevent_signal *(struct tevent_context *, TALLOC_CTX *, int, int, tevent_signal_handler_t, void *, const char *, const char *)
+tevent_common_add_timer: struct tevent_timer *(struct tevent_context *, TALLOC_CTX *, struct timeval, tevent_timer_handler_t, void *, const char *, const char *)
+tevent_common_check_signal: int (struct tevent_context *)
+tevent_common_context_destructor: int (struct tevent_context *)
+tevent_common_fd_destructor: int (struct tevent_fd *)
+tevent_common_fd_get_flags: uint16_t (struct tevent_fd *)
+tevent_common_fd_set_close_fn: void (struct tevent_fd *, tevent_fd_close_fn_t)
+tevent_common_fd_set_flags: void (struct tevent_fd *, uint16_t)
+tevent_common_loop_immediate: bool (struct tevent_context *)
+tevent_common_loop_timer_delay: struct timeval (struct tevent_context *)
+tevent_common_loop_wait: int (struct tevent_context *, const char *)
+tevent_common_schedule_immediate: void (struct tevent_immediate *, struct tevent_context *, tevent_immediate_handler_t, void *, const char *, const char *)
+tevent_context_init: struct tevent_context *(TALLOC_CTX *)
+tevent_context_init_byname: struct tevent_context *(TALLOC_CTX *, const char *)
+tevent_debug: void (struct tevent_context *, enum tevent_debug_level, const char *, ...)
+tevent_fd_get_flags: uint16_t (struct tevent_fd *)
+tevent_fd_set_auto_close: void (struct tevent_fd *)
+tevent_fd_set_close_fn: void (struct tevent_fd *, tevent_fd_close_fn_t)
+tevent_fd_set_flags: void (struct tevent_fd *, uint16_t)
+tevent_loop_allow_nesting: void (struct tevent_context *)
+tevent_loop_set_nesting_hook: void (struct tevent_context *, tevent_nesting_hook, void *)
+tevent_queue_add: bool (struct tevent_queue *, struct tevent_context *, struct tevent_req *, tevent_queue_trigger_fn_t, void *)
+tevent_queue_length: size_t (struct tevent_queue *)
+tevent_queue_start: void (struct tevent_queue *)
+tevent_queue_stop: void (struct tevent_queue *)
+tevent_re_initialise: int (struct tevent_context *)
+tevent_register_backend: bool (const char *, const struct tevent_ops *)
+tevent_req_default_print: char *(struct tevent_req *, TALLOC_CTX *)
+tevent_req_is_error: bool (struct tevent_req *, enum tevent_req_state *, uint64_t *)
+tevent_req_is_in_progress: bool (struct tevent_req *)
+tevent_req_poll: bool (struct tevent_req *, struct tevent_context *)
+tevent_req_post: struct tevent_req *(struct tevent_req *, struct tevent_context *)
+tevent_req_print: char *(TALLOC_CTX *, struct tevent_req *)
+tevent_req_received: void (struct tevent_req *)
+tevent_req_set_callback: void (struct tevent_req *, tevent_req_fn, void *)
+tevent_req_set_cancel_fn: void (struct tevent_req *, tevent_req_cancel_fn)
+tevent_req_set_endtime: bool (struct tevent_req *, struct tevent_context *, struct timeval)
+tevent_req_set_print_fn: void (struct tevent_req *, tevent_req_print_fn)
+tevent_set_abort_fn: void (void (*)(const char *))
+tevent_set_debug: int (struct tevent_context *, void (*)(void *, enum tevent_debug_level, const char *, va_list), void *)
+tevent_set_debug_stderr: int (struct tevent_context *)
+tevent_set_default_backend: void (const char *)
+tevent_signal_support: bool (struct tevent_context *)
+tevent_timeval_add: struct timeval (const struct timeval *, uint32_t, uint32_t)
+tevent_timeval_compare: int (const struct timeval *, const struct timeval *)
+tevent_timeval_current: struct timeval (void)
+tevent_timeval_current_ofs: struct timeval (uint32_t, uint32_t)
+tevent_timeval_is_zero: bool (const struct timeval *)
+tevent_timeval_set: struct timeval (uint32_t, uint32_t)
+tevent_timeval_until: struct timeval (const struct timeval *, const struct timeval *)
+tevent_timeval_zero: struct timeval (void)
+tevent_wakeup_recv: bool (struct tevent_req *)
+tevent_wakeup_send: struct tevent_req *(TALLOC_CTX *, struct tevent_context *, struct timeval)
diff --git a/ctdb/lib/tevent/ABI/tevent-0.9.13.sigs b/ctdb/lib/tevent/ABI/tevent-0.9.13.sigs
new file mode 100644
index 0000000000..888ca0ee3c
--- /dev/null
+++ b/ctdb/lib/tevent/ABI/tevent-0.9.13.sigs
@@ -0,0 +1,75 @@
+_tevent_add_fd: struct tevent_fd *(struct tevent_context *, TALLOC_CTX *, int, uint16_t, tevent_fd_handler_t, void *, const char *, const char *)
+_tevent_add_signal: struct tevent_signal *(struct tevent_context *, TALLOC_CTX *, int, int, tevent_signal_handler_t, void *, const char *, const char *)
+_tevent_add_timer: struct tevent_timer *(struct tevent_context *, TALLOC_CTX *, struct timeval, tevent_timer_handler_t, void *, const char *, const char *)
+_tevent_create_immediate: struct tevent_immediate *(TALLOC_CTX *, const char *)
+_tevent_loop_once: int (struct tevent_context *, const char *)
+_tevent_loop_until: int (struct tevent_context *, bool (*)(void *), void *, const char *)
+_tevent_loop_wait: int (struct tevent_context *, const char *)
+_tevent_queue_create: struct tevent_queue *(TALLOC_CTX *, const char *, const char *)
+_tevent_req_callback_data: void *(struct tevent_req *)
+_tevent_req_cancel: bool (struct tevent_req *, const char *)
+_tevent_req_create: struct tevent_req *(TALLOC_CTX *, void *, size_t, const char *, const char *)
+_tevent_req_data: void *(struct tevent_req *)
+_tevent_req_done: void (struct tevent_req *, const char *)
+_tevent_req_error: bool (struct tevent_req *, uint64_t, const char *)
+_tevent_req_nomem: bool (const void *, struct tevent_req *, const char *)
+_tevent_req_notify_callback: void (struct tevent_req *, const char *)
+_tevent_req_oom: void (struct tevent_req *, const char *)
+_tevent_schedule_immediate: void (struct tevent_immediate *, struct tevent_context *, tevent_immediate_handler_t, void *, const char *, const char *)
+tevent_backend_list: const char **(TALLOC_CTX *)
+tevent_cleanup_pending_signal_handlers: void (struct tevent_signal *)
+tevent_common_add_fd: struct tevent_fd *(struct tevent_context *, TALLOC_CTX *, int, uint16_t, tevent_fd_handler_t, void *, const char *, const char *)
+tevent_common_add_signal: struct tevent_signal *(struct tevent_context *, TALLOC_CTX *, int, int, tevent_signal_handler_t, void *, const char *, const char *)
+tevent_common_add_timer: struct tevent_timer *(struct tevent_context *, TALLOC_CTX *, struct timeval, tevent_timer_handler_t, void *, const char *, const char *)
+tevent_common_check_signal: int (struct tevent_context *)
+tevent_common_context_destructor: int (struct tevent_context *)
+tevent_common_fd_destructor: int (struct tevent_fd *)
+tevent_common_fd_get_flags: uint16_t (struct tevent_fd *)
+tevent_common_fd_set_close_fn: void (struct tevent_fd *, tevent_fd_close_fn_t)
+tevent_common_fd_set_flags: void (struct tevent_fd *, uint16_t)
+tevent_common_loop_immediate: bool (struct tevent_context *)
+tevent_common_loop_timer_delay: struct timeval (struct tevent_context *)
+tevent_common_loop_wait: int (struct tevent_context *, const char *)
+tevent_common_schedule_immediate: void (struct tevent_immediate *, struct tevent_context *, tevent_immediate_handler_t, void *, const char *, const char *)
+tevent_context_init: struct tevent_context *(TALLOC_CTX *)
+tevent_context_init_byname: struct tevent_context *(TALLOC_CTX *, const char *)
+tevent_debug: void (struct tevent_context *, enum tevent_debug_level, const char *, ...)
+tevent_fd_get_flags: uint16_t (struct tevent_fd *)
+tevent_fd_set_auto_close: void (struct tevent_fd *)
+tevent_fd_set_close_fn: void (struct tevent_fd *, tevent_fd_close_fn_t)
+tevent_fd_set_flags: void (struct tevent_fd *, uint16_t)
+tevent_loop_allow_nesting: void (struct tevent_context *)
+tevent_loop_set_nesting_hook: void (struct tevent_context *, tevent_nesting_hook, void *)
+tevent_queue_add: bool (struct tevent_queue *, struct tevent_context *, struct tevent_req *, tevent_queue_trigger_fn_t, void *)
+tevent_queue_length: size_t (struct tevent_queue *)
+tevent_queue_start: void (struct tevent_queue *)
+tevent_queue_stop: void (struct tevent_queue *)
+tevent_re_initialise: int (struct tevent_context *)
+tevent_register_backend: bool (const char *, const struct tevent_ops *)
+tevent_req_default_print: char *(struct tevent_req *, TALLOC_CTX *)
+tevent_req_defer_callback: void (struct tevent_req *, struct tevent_context *)
+tevent_req_is_error: bool (struct tevent_req *, enum tevent_req_state *, uint64_t *)
+tevent_req_is_in_progress: bool (struct tevent_req *)
+tevent_req_poll: bool (struct tevent_req *, struct tevent_context *)
+tevent_req_post: struct tevent_req *(struct tevent_req *, struct tevent_context *)
+tevent_req_print: char *(TALLOC_CTX *, struct tevent_req *)
+tevent_req_received: void (struct tevent_req *)
+tevent_req_set_callback: void (struct tevent_req *, tevent_req_fn, void *)
+tevent_req_set_cancel_fn: void (struct tevent_req *, tevent_req_cancel_fn)
+tevent_req_set_endtime: bool (struct tevent_req *, struct tevent_context *, struct timeval)
+tevent_req_set_print_fn: void (struct tevent_req *, tevent_req_print_fn)
+tevent_set_abort_fn: void (void (*)(const char *))
+tevent_set_debug: int (struct tevent_context *, void (*)(void *, enum tevent_debug_level, const char *, va_list), void *)
+tevent_set_debug_stderr: int (struct tevent_context *)
+tevent_set_default_backend: void (const char *)
+tevent_signal_support: bool (struct tevent_context *)
+tevent_timeval_add: struct timeval (const struct timeval *, uint32_t, uint32_t)
+tevent_timeval_compare: int (const struct timeval *, const struct timeval *)
+tevent_timeval_current: struct timeval (void)
+tevent_timeval_current_ofs: struct timeval (uint32_t, uint32_t)
+tevent_timeval_is_zero: bool (const struct timeval *)
+tevent_timeval_set: struct timeval (uint32_t, uint32_t)
+tevent_timeval_until: struct timeval (const struct timeval *, const struct timeval *)
+tevent_timeval_zero: struct timeval (void)
+tevent_wakeup_recv: bool (struct tevent_req *)
+tevent_wakeup_send: struct tevent_req *(TALLOC_CTX *, struct tevent_context *, struct timeval)
diff --git a/ctdb/lib/tevent/ABI/tevent-0.9.14.sigs b/ctdb/lib/tevent/ABI/tevent-0.9.14.sigs
new file mode 100644
index 0000000000..13c461cc56
--- /dev/null
+++ b/ctdb/lib/tevent/ABI/tevent-0.9.14.sigs
@@ -0,0 +1,78 @@
+_tevent_add_fd: struct tevent_fd *(struct tevent_context *, TALLOC_CTX *, int, uint16_t, tevent_fd_handler_t, void *, const char *, const char *)
+_tevent_add_signal: struct tevent_signal *(struct tevent_context *, TALLOC_CTX *, int, int, tevent_signal_handler_t, void *, const char *, const char *)
+_tevent_add_timer: struct tevent_timer *(struct tevent_context *, TALLOC_CTX *, struct timeval, tevent_timer_handler_t, void *, const char *, const char *)
+_tevent_create_immediate: struct tevent_immediate *(TALLOC_CTX *, const char *)
+_tevent_loop_once: int (struct tevent_context *, const char *)
+_tevent_loop_until: int (struct tevent_context *, bool (*)(void *), void *, const char *)
+_tevent_loop_wait: int (struct tevent_context *, const char *)
+_tevent_queue_create: struct tevent_queue *(TALLOC_CTX *, const char *, const char *)
+_tevent_req_callback_data: void *(struct tevent_req *)
+_tevent_req_cancel: bool (struct tevent_req *, const char *)
+_tevent_req_create: struct tevent_req *(TALLOC_CTX *, void *, size_t, const char *, const char *)
+_tevent_req_data: void *(struct tevent_req *)
+_tevent_req_done: void (struct tevent_req *, const char *)
+_tevent_req_error: bool (struct tevent_req *, uint64_t, const char *)
+_tevent_req_nomem: bool (const void *, struct tevent_req *, const char *)
+_tevent_req_notify_callback: void (struct tevent_req *, const char *)
+_tevent_req_oom: void (struct tevent_req *, const char *)
+_tevent_schedule_immediate: void (struct tevent_immediate *, struct tevent_context *, tevent_immediate_handler_t, void *, const char *, const char *)
+tevent_backend_list: const char **(TALLOC_CTX *)
+tevent_cleanup_pending_signal_handlers: void (struct tevent_signal *)
+tevent_common_add_fd: struct tevent_fd *(struct tevent_context *, TALLOC_CTX *, int, uint16_t, tevent_fd_handler_t, void *, const char *, const char *)
+tevent_common_add_signal: struct tevent_signal *(struct tevent_context *, TALLOC_CTX *, int, int, tevent_signal_handler_t, void *, const char *, const char *)
+tevent_common_add_timer: struct tevent_timer *(struct tevent_context *, TALLOC_CTX *, struct timeval, tevent_timer_handler_t, void *, const char *, const char *)
+tevent_common_check_signal: int (struct tevent_context *)
+tevent_common_context_destructor: int (struct tevent_context *)
+tevent_common_fd_destructor: int (struct tevent_fd *)
+tevent_common_fd_get_flags: uint16_t (struct tevent_fd *)
+tevent_common_fd_set_close_fn: void (struct tevent_fd *, tevent_fd_close_fn_t)
+tevent_common_fd_set_flags: void (struct tevent_fd *, uint16_t)
+tevent_common_loop_immediate: bool (struct tevent_context *)
+tevent_common_loop_timer_delay: struct timeval (struct tevent_context *)
+tevent_common_loop_wait: int (struct tevent_context *, const char *)
+tevent_common_schedule_immediate: void (struct tevent_immediate *, struct tevent_context *, tevent_immediate_handler_t, void *, const char *, const char *)
+tevent_context_init: struct tevent_context *(TALLOC_CTX *)
+tevent_context_init_byname: struct tevent_context *(TALLOC_CTX *, const char *)
+tevent_debug: void (struct tevent_context *, enum tevent_debug_level, const char *, ...)
+tevent_fd_get_flags: uint16_t (struct tevent_fd *)
+tevent_fd_set_auto_close: void (struct tevent_fd *)
+tevent_fd_set_close_fn: void (struct tevent_fd *, tevent_fd_close_fn_t)
+tevent_fd_set_flags: void (struct tevent_fd *, uint16_t)
+tevent_loop_allow_nesting: void (struct tevent_context *)
+tevent_loop_set_nesting_hook: void (struct tevent_context *, tevent_nesting_hook, void *)
+tevent_queue_add: bool (struct tevent_queue *, struct tevent_context *, struct tevent_req *, tevent_queue_trigger_fn_t, void *)
+tevent_queue_add_entry: struct tevent_queue_entry *(struct tevent_queue *, struct tevent_context *, struct tevent_req *, tevent_queue_trigger_fn_t, void *)
+tevent_queue_add_optimize_empty: struct tevent_queue_entry *(struct tevent_queue *, struct tevent_context *, struct tevent_req *, tevent_queue_trigger_fn_t, void *)
+tevent_queue_length: size_t (struct tevent_queue *)
+tevent_queue_running: bool (struct tevent_queue *)
+tevent_queue_start: void (struct tevent_queue *)
+tevent_queue_stop: void (struct tevent_queue *)
+tevent_re_initialise: int (struct tevent_context *)
+tevent_register_backend: bool (const char *, const struct tevent_ops *)
+tevent_req_default_print: char *(struct tevent_req *, TALLOC_CTX *)
+tevent_req_defer_callback: void (struct tevent_req *, struct tevent_context *)
+tevent_req_is_error: bool (struct tevent_req *, enum tevent_req_state *, uint64_t *)
+tevent_req_is_in_progress: bool (struct tevent_req *)
+tevent_req_poll: bool (struct tevent_req *, struct tevent_context *)
+tevent_req_post: struct tevent_req *(struct tevent_req *, struct tevent_context *)
+tevent_req_print: char *(TALLOC_CTX *, struct tevent_req *)
+tevent_req_received: void (struct tevent_req *)
+tevent_req_set_callback: void (struct tevent_req *, tevent_req_fn, void *)
+tevent_req_set_cancel_fn: void (struct tevent_req *, tevent_req_cancel_fn)
+tevent_req_set_endtime: bool (struct tevent_req *, struct tevent_context *, struct timeval)
+tevent_req_set_print_fn: void (struct tevent_req *, tevent_req_print_fn)
+tevent_set_abort_fn: void (void (*)(const char *))
+tevent_set_debug: int (struct tevent_context *, void (*)(void *, enum tevent_debug_level, const char *, va_list), void *)
+tevent_set_debug_stderr: int (struct tevent_context *)
+tevent_set_default_backend: void (const char *)
+tevent_signal_support: bool (struct tevent_context *)
+tevent_timeval_add: struct timeval (const struct timeval *, uint32_t, uint32_t)
+tevent_timeval_compare: int (const struct timeval *, const struct timeval *)
+tevent_timeval_current: struct timeval (void)
+tevent_timeval_current_ofs: struct timeval (uint32_t, uint32_t)
+tevent_timeval_is_zero: bool (const struct timeval *)
+tevent_timeval_set: struct timeval (uint32_t, uint32_t)
+tevent_timeval_until: struct timeval (const struct timeval *, const struct timeval *)
+tevent_timeval_zero: struct timeval (void)
+tevent_wakeup_recv: bool (struct tevent_req *)
+tevent_wakeup_send: struct tevent_req *(TALLOC_CTX *, struct tevent_context *, struct timeval)
diff --git a/ctdb/lib/tevent/ABI/tevent-0.9.15.sigs b/ctdb/lib/tevent/ABI/tevent-0.9.15.sigs
new file mode 100644
index 0000000000..13c461cc56
--- /dev/null
+++ b/ctdb/lib/tevent/ABI/tevent-0.9.15.sigs
@@ -0,0 +1,78 @@
+_tevent_add_fd: struct tevent_fd *(struct tevent_context *, TALLOC_CTX *, int, uint16_t, tevent_fd_handler_t, void *, const char *, const char *)
+_tevent_add_signal: struct tevent_signal *(struct tevent_context *, TALLOC_CTX *, int, int, tevent_signal_handler_t, void *, const char *, const char *)
+_tevent_add_timer: struct tevent_timer *(struct tevent_context *, TALLOC_CTX *, struct timeval, tevent_timer_handler_t, void *, const char *, const char *)
+_tevent_create_immediate: struct tevent_immediate *(TALLOC_CTX *, const char *)
+_tevent_loop_once: int (struct tevent_context *, const char *)
+_tevent_loop_until: int (struct tevent_context *, bool (*)(void *), void *, const char *)
+_tevent_loop_wait: int (struct tevent_context *, const char *)
+_tevent_queue_create: struct tevent_queue *(TALLOC_CTX *, const char *, const char *)
+_tevent_req_callback_data: void *(struct tevent_req *)
+_tevent_req_cancel: bool (struct tevent_req *, const char *)
+_tevent_req_create: struct tevent_req *(TALLOC_CTX *, void *, size_t, const char *, const char *)
+_tevent_req_data: void *(struct tevent_req *)
+_tevent_req_done: void (struct tevent_req *, const char *)
+_tevent_req_error: bool (struct tevent_req *, uint64_t, const char *)
+_tevent_req_nomem: bool (const void *, struct tevent_req *, const char *)
+_tevent_req_notify_callback: void (struct tevent_req *, const char *)
+_tevent_req_oom: void (struct tevent_req *, const char *)
+_tevent_schedule_immediate: void (struct tevent_immediate *, struct tevent_context *, tevent_immediate_handler_t, void *, const char *, const char *)
+tevent_backend_list: const char **(TALLOC_CTX *)
+tevent_cleanup_pending_signal_handlers: void (struct tevent_signal *)
+tevent_common_add_fd: struct tevent_fd *(struct tevent_context *, TALLOC_CTX *, int, uint16_t, tevent_fd_handler_t, void *, const char *, const char *)
+tevent_common_add_signal: struct tevent_signal *(struct tevent_context *, TALLOC_CTX *, int, int, tevent_signal_handler_t, void *, const char *, const char *)
+tevent_common_add_timer: struct tevent_timer *(struct tevent_context *, TALLOC_CTX *, struct timeval, tevent_timer_handler_t, void *, const char *, const char *)
+tevent_common_check_signal: int (struct tevent_context *)
+tevent_common_context_destructor: int (struct tevent_context *)
+tevent_common_fd_destructor: int (struct tevent_fd *)
+tevent_common_fd_get_flags: uint16_t (struct tevent_fd *)
+tevent_common_fd_set_close_fn: void (struct tevent_fd *, tevent_fd_close_fn_t)
+tevent_common_fd_set_flags: void (struct tevent_fd *, uint16_t)
+tevent_common_loop_immediate: bool (struct tevent_context *)
+tevent_common_loop_timer_delay: struct timeval (struct tevent_context *)
+tevent_common_loop_wait: int (struct tevent_context *, const char *)
+tevent_common_schedule_immediate: void (struct tevent_immediate *, struct tevent_context *, tevent_immediate_handler_t, void *, const char *, const char *)
+tevent_context_init: struct tevent_context *(TALLOC_CTX *)
+tevent_context_init_byname: struct tevent_context *(TALLOC_CTX *, const char *)
+tevent_debug: void (struct tevent_context *, enum tevent_debug_level, const char *, ...)
+tevent_fd_get_flags: uint16_t (struct tevent_fd *)
+tevent_fd_set_auto_close: void (struct tevent_fd *)
+tevent_fd_set_close_fn: void (struct tevent_fd *, tevent_fd_close_fn_t)
+tevent_fd_set_flags: void (struct tevent_fd *, uint16_t)
+tevent_loop_allow_nesting: void (struct tevent_context *)
+tevent_loop_set_nesting_hook: void (struct tevent_context *, tevent_nesting_hook, void *)
+tevent_queue_add: bool (struct tevent_queue *, struct tevent_context *, struct tevent_req *, tevent_queue_trigger_fn_t, void *)
+tevent_queue_add_entry: struct tevent_queue_entry *(struct tevent_queue *, struct tevent_context *, struct tevent_req *, tevent_queue_trigger_fn_t, void *)
+tevent_queue_add_optimize_empty: struct tevent_queue_entry *(struct tevent_queue *, struct tevent_context *, struct tevent_req *, tevent_queue_trigger_fn_t, void *)
+tevent_queue_length: size_t (struct tevent_queue *)
+tevent_queue_running: bool (struct tevent_queue *)
+tevent_queue_start: void (struct tevent_queue *)
+tevent_queue_stop: void (struct tevent_queue *)
+tevent_re_initialise: int (struct tevent_context *)
+tevent_register_backend: bool (const char *, const struct tevent_ops *)
+tevent_req_default_print: char *(struct tevent_req *, TALLOC_CTX *)
+tevent_req_defer_callback: void (struct tevent_req *, struct tevent_context *)
+tevent_req_is_error: bool (struct tevent_req *, enum tevent_req_state *, uint64_t *)
+tevent_req_is_in_progress: bool (struct tevent_req *)
+tevent_req_poll: bool (struct tevent_req *, struct tevent_context *)
+tevent_req_post: struct tevent_req *(struct tevent_req *, struct tevent_context *)
+tevent_req_print: char *(TALLOC_CTX *, struct tevent_req *)
+tevent_req_received: void (struct tevent_req *)
+tevent_req_set_callback: void (struct tevent_req *, tevent_req_fn, void *)
+tevent_req_set_cancel_fn: void (struct tevent_req *, tevent_req_cancel_fn)
+tevent_req_set_endtime: bool (struct tevent_req *, struct tevent_context *, struct timeval)
+tevent_req_set_print_fn: void (struct tevent_req *, tevent_req_print_fn)
+tevent_set_abort_fn: void (void (*)(const char *))
+tevent_set_debug: int (struct tevent_context *, void (*)(void *, enum tevent_debug_level, const char *, va_list), void *)
+tevent_set_debug_stderr: int (struct tevent_context *)
+tevent_set_default_backend: void (const char *)
+tevent_signal_support: bool (struct tevent_context *)
+tevent_timeval_add: struct timeval (const struct timeval *, uint32_t, uint32_t)
+tevent_timeval_compare: int (const struct timeval *, const struct timeval *)
+tevent_timeval_current: struct timeval (void)
+tevent_timeval_current_ofs: struct timeval (uint32_t, uint32_t)
+tevent_timeval_is_zero: bool (const struct timeval *)
+tevent_timeval_set: struct timeval (uint32_t, uint32_t)
+tevent_timeval_until: struct timeval (const struct timeval *, const struct timeval *)
+tevent_timeval_zero: struct timeval (void)
+tevent_wakeup_recv: bool (struct tevent_req *)
+tevent_wakeup_send: struct tevent_req *(TALLOC_CTX *, struct tevent_context *, struct timeval)
diff --git a/ctdb/lib/tevent/ABI/tevent-0.9.16.sigs b/ctdb/lib/tevent/ABI/tevent-0.9.16.sigs
new file mode 100644
index 0000000000..ea7f9446e1
--- /dev/null
+++ b/ctdb/lib/tevent/ABI/tevent-0.9.16.sigs
@@ -0,0 +1,82 @@
+_tevent_add_fd: struct tevent_fd *(struct tevent_context *, TALLOC_CTX *, int, uint16_t, tevent_fd_handler_t, void *, const char *, const char *)
+_tevent_add_signal: struct tevent_signal *(struct tevent_context *, TALLOC_CTX *, int, int, tevent_signal_handler_t, void *, const char *, const char *)
+_tevent_add_timer: struct tevent_timer *(struct tevent_context *, TALLOC_CTX *, struct timeval, tevent_timer_handler_t, void *, const char *, const char *)
+_tevent_create_immediate: struct tevent_immediate *(TALLOC_CTX *, const char *)
+_tevent_loop_once: int (struct tevent_context *, const char *)
+_tevent_loop_until: int (struct tevent_context *, bool (*)(void *), void *, const char *)
+_tevent_loop_wait: int (struct tevent_context *, const char *)
+_tevent_queue_create: struct tevent_queue *(TALLOC_CTX *, const char *, const char *)
+_tevent_req_callback_data: void *(struct tevent_req *)
+_tevent_req_cancel: bool (struct tevent_req *, const char *)
+_tevent_req_create: struct tevent_req *(TALLOC_CTX *, void *, size_t, const char *, const char *)
+_tevent_req_data: void *(struct tevent_req *)
+_tevent_req_done: void (struct tevent_req *, const char *)
+_tevent_req_error: bool (struct tevent_req *, uint64_t, const char *)
+_tevent_req_nomem: bool (const void *, struct tevent_req *, const char *)
+_tevent_req_notify_callback: void (struct tevent_req *, const char *)
+_tevent_req_oom: void (struct tevent_req *, const char *)
+_tevent_schedule_immediate: void (struct tevent_immediate *, struct tevent_context *, tevent_immediate_handler_t, void *, const char *, const char *)
+tevent_backend_list: const char **(TALLOC_CTX *)
+tevent_cleanup_pending_signal_handlers: void (struct tevent_signal *)
+tevent_common_add_fd: struct tevent_fd *(struct tevent_context *, TALLOC_CTX *, int, uint16_t, tevent_fd_handler_t, void *, const char *, const char *)
+tevent_common_add_signal: struct tevent_signal *(struct tevent_context *, TALLOC_CTX *, int, int, tevent_signal_handler_t, void *, const char *, const char *)
+tevent_common_add_timer: struct tevent_timer *(struct tevent_context *, TALLOC_CTX *, struct timeval, tevent_timer_handler_t, void *, const char *, const char *)
+tevent_common_check_signal: int (struct tevent_context *)
+tevent_common_context_destructor: int (struct tevent_context *)
+tevent_common_fd_destructor: int (struct tevent_fd *)
+tevent_common_fd_get_flags: uint16_t (struct tevent_fd *)
+tevent_common_fd_set_close_fn: void (struct tevent_fd *, tevent_fd_close_fn_t)
+tevent_common_fd_set_flags: void (struct tevent_fd *, uint16_t)
+tevent_common_loop_immediate: bool (struct tevent_context *)
+tevent_common_loop_timer_delay: struct timeval (struct tevent_context *)
+tevent_common_loop_wait: int (struct tevent_context *, const char *)
+tevent_common_schedule_immediate: void (struct tevent_immediate *, struct tevent_context *, tevent_immediate_handler_t, void *, const char *, const char *)
+tevent_context_init: struct tevent_context *(TALLOC_CTX *)
+tevent_context_init_byname: struct tevent_context *(TALLOC_CTX *, const char *)
+tevent_context_init_ops: struct tevent_context *(TALLOC_CTX *, const struct tevent_ops *, void *)
+tevent_debug: void (struct tevent_context *, enum tevent_debug_level, const char *, ...)
+tevent_fd_get_flags: uint16_t (struct tevent_fd *)
+tevent_fd_set_auto_close: void (struct tevent_fd *)
+tevent_fd_set_close_fn: void (struct tevent_fd *, tevent_fd_close_fn_t)
+tevent_fd_set_flags: void (struct tevent_fd *, uint16_t)
+tevent_get_trace_callback: void (struct tevent_context *, tevent_trace_callback_t *, void *)
+tevent_loop_allow_nesting: void (struct tevent_context *)
+tevent_loop_set_nesting_hook: void (struct tevent_context *, tevent_nesting_hook, void *)
+tevent_queue_add: bool (struct tevent_queue *, struct tevent_context *, struct tevent_req *, tevent_queue_trigger_fn_t, void *)
+tevent_queue_add_entry: struct tevent_queue_entry *(struct tevent_queue *, struct tevent_context *, struct tevent_req *, tevent_queue_trigger_fn_t, void *)
+tevent_queue_add_optimize_empty: struct tevent_queue_entry *(struct tevent_queue *, struct tevent_context *, struct tevent_req *, tevent_queue_trigger_fn_t, void *)
+tevent_queue_length: size_t (struct tevent_queue *)
+tevent_queue_running: bool (struct tevent_queue *)
+tevent_queue_start: void (struct tevent_queue *)
+tevent_queue_stop: void (struct tevent_queue *)
+tevent_re_initialise: int (struct tevent_context *)
+tevent_register_backend: bool (const char *, const struct tevent_ops *)
+tevent_req_default_print: char *(struct tevent_req *, TALLOC_CTX *)
+tevent_req_defer_callback: void (struct tevent_req *, struct tevent_context *)
+tevent_req_is_error: bool (struct tevent_req *, enum tevent_req_state *, uint64_t *)
+tevent_req_is_in_progress: bool (struct tevent_req *)
+tevent_req_poll: bool (struct tevent_req *, struct tevent_context *)
+tevent_req_post: struct tevent_req *(struct tevent_req *, struct tevent_context *)
+tevent_req_print: char *(TALLOC_CTX *, struct tevent_req *)
+tevent_req_received: void (struct tevent_req *)
+tevent_req_set_callback: void (struct tevent_req *, tevent_req_fn, void *)
+tevent_req_set_cancel_fn: void (struct tevent_req *, tevent_req_cancel_fn)
+tevent_req_set_endtime: bool (struct tevent_req *, struct tevent_context *, struct timeval)
+tevent_req_set_print_fn: void (struct tevent_req *, tevent_req_print_fn)
+tevent_set_abort_fn: void (void (*)(const char *))
+tevent_set_debug: int (struct tevent_context *, void (*)(void *, enum tevent_debug_level, const char *, va_list), void *)
+tevent_set_debug_stderr: int (struct tevent_context *)
+tevent_set_default_backend: void (const char *)
+tevent_set_trace_callback: void (struct tevent_context *, tevent_trace_callback_t, void *)
+tevent_signal_support: bool (struct tevent_context *)
+tevent_timeval_add: struct timeval (const struct timeval *, uint32_t, uint32_t)
+tevent_timeval_compare: int (const struct timeval *, const struct timeval *)
+tevent_timeval_current: struct timeval (void)
+tevent_timeval_current_ofs: struct timeval (uint32_t, uint32_t)
+tevent_timeval_is_zero: bool (const struct timeval *)
+tevent_timeval_set: struct timeval (uint32_t, uint32_t)
+tevent_timeval_until: struct timeval (const struct timeval *, const struct timeval *)
+tevent_timeval_zero: struct timeval (void)
+tevent_trace_point_callback: void (struct tevent_context *, enum tevent_trace_point)
+tevent_wakeup_recv: bool (struct tevent_req *)
+tevent_wakeup_send: struct tevent_req *(TALLOC_CTX *, struct tevent_context *, struct timeval)
diff --git a/ctdb/lib/tevent/ABI/tevent-0.9.17.sigs b/ctdb/lib/tevent/ABI/tevent-0.9.17.sigs
new file mode 100644
index 0000000000..ea7f9446e1
--- /dev/null
+++ b/ctdb/lib/tevent/ABI/tevent-0.9.17.sigs
@@ -0,0 +1,82 @@
+_tevent_add_fd: struct tevent_fd *(struct tevent_context *, TALLOC_CTX *, int, uint16_t, tevent_fd_handler_t, void *, const char *, const char *)
+_tevent_add_signal: struct tevent_signal *(struct tevent_context *, TALLOC_CTX *, int, int, tevent_signal_handler_t, void *, const char *, const char *)
+_tevent_add_timer: struct tevent_timer *(struct tevent_context *, TALLOC_CTX *, struct timeval, tevent_timer_handler_t, void *, const char *, const char *)
+_tevent_create_immediate: struct tevent_immediate *(TALLOC_CTX *, const char *)
+_tevent_loop_once: int (struct tevent_context *, const char *)
+_tevent_loop_until: int (struct tevent_context *, bool (*)(void *), void *, const char *)
+_tevent_loop_wait: int (struct tevent_context *, const char *)
+_tevent_queue_create: struct tevent_queue *(TALLOC_CTX *, const char *, const char *)
+_tevent_req_callback_data: void *(struct tevent_req *)
+_tevent_req_cancel: bool (struct tevent_req *, const char *)
+_tevent_req_create: struct tevent_req *(TALLOC_CTX *, void *, size_t, const char *, const char *)
+_tevent_req_data: void *(struct tevent_req *)
+_tevent_req_done: void (struct tevent_req *, const char *)
+_tevent_req_error: bool (struct tevent_req *, uint64_t, const char *)
+_tevent_req_nomem: bool (const void *, struct tevent_req *, const char *)
+_tevent_req_notify_callback: void (struct tevent_req *, const char *)
+_tevent_req_oom: void (struct tevent_req *, const char *)
+_tevent_schedule_immediate: void (struct tevent_immediate *, struct tevent_context *, tevent_immediate_handler_t, void *, const char *, const char *)
+tevent_backend_list: const char **(TALLOC_CTX *)
+tevent_cleanup_pending_signal_handlers: void (struct tevent_signal *)
+tevent_common_add_fd: struct tevent_fd *(struct tevent_context *, TALLOC_CTX *, int, uint16_t, tevent_fd_handler_t, void *, const char *, const char *)
+tevent_common_add_signal: struct tevent_signal *(struct tevent_context *, TALLOC_CTX *, int, int, tevent_signal_handler_t, void *, const char *, const char *)
+tevent_common_add_timer: struct tevent_timer *(struct tevent_context *, TALLOC_CTX *, struct timeval, tevent_timer_handler_t, void *, const char *, const char *)
+tevent_common_check_signal: int (struct tevent_context *)
+tevent_common_context_destructor: int (struct tevent_context *)
+tevent_common_fd_destructor: int (struct tevent_fd *)
+tevent_common_fd_get_flags: uint16_t (struct tevent_fd *)
+tevent_common_fd_set_close_fn: void (struct tevent_fd *, tevent_fd_close_fn_t)
+tevent_common_fd_set_flags: void (struct tevent_fd *, uint16_t)
+tevent_common_loop_immediate: bool (struct tevent_context *)
+tevent_common_loop_timer_delay: struct timeval (struct tevent_context *)
+tevent_common_loop_wait: int (struct tevent_context *, const char *)
+tevent_common_schedule_immediate: void (struct tevent_immediate *, struct tevent_context *, tevent_immediate_handler_t, void *, const char *, const char *)
+tevent_context_init: struct tevent_context *(TALLOC_CTX *)
+tevent_context_init_byname: struct tevent_context *(TALLOC_CTX *, const char *)
+tevent_context_init_ops: struct tevent_context *(TALLOC_CTX *, const struct tevent_ops *, void *)
+tevent_debug: void (struct tevent_context *, enum tevent_debug_level, const char *, ...)
+tevent_fd_get_flags: uint16_t (struct tevent_fd *)
+tevent_fd_set_auto_close: void (struct tevent_fd *)
+tevent_fd_set_close_fn: void (struct tevent_fd *, tevent_fd_close_fn_t)
+tevent_fd_set_flags: void (struct tevent_fd *, uint16_t)
+tevent_get_trace_callback: void (struct tevent_context *, tevent_trace_callback_t *, void *)
+tevent_loop_allow_nesting: void (struct tevent_context *)
+tevent_loop_set_nesting_hook: void (struct tevent_context *, tevent_nesting_hook, void *)
+tevent_queue_add: bool (struct tevent_queue *, struct tevent_context *, struct tevent_req *, tevent_queue_trigger_fn_t, void *)
+tevent_queue_add_entry: struct tevent_queue_entry *(struct tevent_queue *, struct tevent_context *, struct tevent_req *, tevent_queue_trigger_fn_t, void *)
+tevent_queue_add_optimize_empty: struct tevent_queue_entry *(struct tevent_queue *, struct tevent_context *, struct tevent_req *, tevent_queue_trigger_fn_t, void *)
+tevent_queue_length: size_t (struct tevent_queue *)
+tevent_queue_running: bool (struct tevent_queue *)
+tevent_queue_start: void (struct tevent_queue *)
+tevent_queue_stop: void (struct tevent_queue *)
+tevent_re_initialise: int (struct tevent_context *)
+tevent_register_backend: bool (const char *, const struct tevent_ops *)
+tevent_req_default_print: char *(struct tevent_req *, TALLOC_CTX *)
+tevent_req_defer_callback: void (struct tevent_req *, struct tevent_context *)
+tevent_req_is_error: bool (struct tevent_req *, enum tevent_req_state *, uint64_t *)
+tevent_req_is_in_progress: bool (struct tevent_req *)
+tevent_req_poll: bool (struct tevent_req *, struct tevent_context *)
+tevent_req_post: struct tevent_req *(struct tevent_req *, struct tevent_context *)
+tevent_req_print: char *(TALLOC_CTX *, struct tevent_req *)
+tevent_req_received: void (struct tevent_req *)
+tevent_req_set_callback: void (struct tevent_req *, tevent_req_fn, void *)
+tevent_req_set_cancel_fn: void (struct tevent_req *, tevent_req_cancel_fn)
+tevent_req_set_endtime: bool (struct tevent_req *, struct tevent_context *, struct timeval)
+tevent_req_set_print_fn: void (struct tevent_req *, tevent_req_print_fn)
+tevent_set_abort_fn: void (void (*)(const char *))
+tevent_set_debug: int (struct tevent_context *, void (*)(void *, enum tevent_debug_level, const char *, va_list), void *)
+tevent_set_debug_stderr: int (struct tevent_context *)
+tevent_set_default_backend: void (const char *)
+tevent_set_trace_callback: void (struct tevent_context *, tevent_trace_callback_t, void *)
+tevent_signal_support: bool (struct tevent_context *)
+tevent_timeval_add: struct timeval (const struct timeval *, uint32_t, uint32_t)
+tevent_timeval_compare: int (const struct timeval *, const struct timeval *)
+tevent_timeval_current: struct timeval (void)
+tevent_timeval_current_ofs: struct timeval (uint32_t, uint32_t)
+tevent_timeval_is_zero: bool (const struct timeval *)
+tevent_timeval_set: struct timeval (uint32_t, uint32_t)
+tevent_timeval_until: struct timeval (const struct timeval *, const struct timeval *)
+tevent_timeval_zero: struct timeval (void)
+tevent_trace_point_callback: void (struct tevent_context *, enum tevent_trace_point)
+tevent_wakeup_recv: bool (struct tevent_req *)
+tevent_wakeup_send: struct tevent_req *(TALLOC_CTX *, struct tevent_context *, struct timeval)
diff --git a/ctdb/lib/tevent/ABI/tevent-0.9.18.sigs b/ctdb/lib/tevent/ABI/tevent-0.9.18.sigs
new file mode 100644
index 0000000000..70d20b6816
--- /dev/null
+++ b/ctdb/lib/tevent/ABI/tevent-0.9.18.sigs
@@ -0,0 +1,83 @@
+_tevent_add_fd: struct tevent_fd *(struct tevent_context *, TALLOC_CTX *, int, uint16_t, tevent_fd_handler_t, void *, const char *, const char *)
+_tevent_add_signal: struct tevent_signal *(struct tevent_context *, TALLOC_CTX *, int, int, tevent_signal_handler_t, void *, const char *, const char *)
+_tevent_add_timer: struct tevent_timer *(struct tevent_context *, TALLOC_CTX *, struct timeval, tevent_timer_handler_t, void *, const char *, const char *)
+_tevent_create_immediate: struct tevent_immediate *(TALLOC_CTX *, const char *)
+_tevent_loop_once: int (struct tevent_context *, const char *)
+_tevent_loop_until: int (struct tevent_context *, bool (*)(void *), void *, const char *)
+_tevent_loop_wait: int (struct tevent_context *, const char *)
+_tevent_queue_create: struct tevent_queue *(TALLOC_CTX *, const char *, const char *)
+_tevent_req_callback_data: void *(struct tevent_req *)
+_tevent_req_cancel: bool (struct tevent_req *, const char *)
+_tevent_req_create: struct tevent_req *(TALLOC_CTX *, void *, size_t, const char *, const char *)
+_tevent_req_data: void *(struct tevent_req *)
+_tevent_req_done: void (struct tevent_req *, const char *)
+_tevent_req_error: bool (struct tevent_req *, uint64_t, const char *)
+_tevent_req_nomem: bool (const void *, struct tevent_req *, const char *)
+_tevent_req_notify_callback: void (struct tevent_req *, const char *)
+_tevent_req_oom: void (struct tevent_req *, const char *)
+_tevent_schedule_immediate: void (struct tevent_immediate *, struct tevent_context *, tevent_immediate_handler_t, void *, const char *, const char *)
+tevent_backend_list: const char **(TALLOC_CTX *)
+tevent_cleanup_pending_signal_handlers: void (struct tevent_signal *)
+tevent_common_add_fd: struct tevent_fd *(struct tevent_context *, TALLOC_CTX *, int, uint16_t, tevent_fd_handler_t, void *, const char *, const char *)
+tevent_common_add_signal: struct tevent_signal *(struct tevent_context *, TALLOC_CTX *, int, int, tevent_signal_handler_t, void *, const char *, const char *)
+tevent_common_add_timer: struct tevent_timer *(struct tevent_context *, TALLOC_CTX *, struct timeval, tevent_timer_handler_t, void *, const char *, const char *)
+tevent_common_add_timer_v2: struct tevent_timer *(struct tevent_context *, TALLOC_CTX *, struct timeval, tevent_timer_handler_t, void *, const char *, const char *)
+tevent_common_check_signal: int (struct tevent_context *)
+tevent_common_context_destructor: int (struct tevent_context *)
+tevent_common_fd_destructor: int (struct tevent_fd *)
+tevent_common_fd_get_flags: uint16_t (struct tevent_fd *)
+tevent_common_fd_set_close_fn: void (struct tevent_fd *, tevent_fd_close_fn_t)
+tevent_common_fd_set_flags: void (struct tevent_fd *, uint16_t)
+tevent_common_loop_immediate: bool (struct tevent_context *)
+tevent_common_loop_timer_delay: struct timeval (struct tevent_context *)
+tevent_common_loop_wait: int (struct tevent_context *, const char *)
+tevent_common_schedule_immediate: void (struct tevent_immediate *, struct tevent_context *, tevent_immediate_handler_t, void *, const char *, const char *)
+tevent_context_init: struct tevent_context *(TALLOC_CTX *)
+tevent_context_init_byname: struct tevent_context *(TALLOC_CTX *, const char *)
+tevent_context_init_ops: struct tevent_context *(TALLOC_CTX *, const struct tevent_ops *, void *)
+tevent_debug: void (struct tevent_context *, enum tevent_debug_level, const char *, ...)
+tevent_fd_get_flags: uint16_t (struct tevent_fd *)
+tevent_fd_set_auto_close: void (struct tevent_fd *)
+tevent_fd_set_close_fn: void (struct tevent_fd *, tevent_fd_close_fn_t)
+tevent_fd_set_flags: void (struct tevent_fd *, uint16_t)
+tevent_get_trace_callback: void (struct tevent_context *, tevent_trace_callback_t *, void *)
+tevent_loop_allow_nesting: void (struct tevent_context *)
+tevent_loop_set_nesting_hook: void (struct tevent_context *, tevent_nesting_hook, void *)
+tevent_queue_add: bool (struct tevent_queue *, struct tevent_context *, struct tevent_req *, tevent_queue_trigger_fn_t, void *)
+tevent_queue_add_entry: struct tevent_queue_entry *(struct tevent_queue *, struct tevent_context *, struct tevent_req *, tevent_queue_trigger_fn_t, void *)
+tevent_queue_add_optimize_empty: struct tevent_queue_entry *(struct tevent_queue *, struct tevent_context *, struct tevent_req *, tevent_queue_trigger_fn_t, void *)
+tevent_queue_length: size_t (struct tevent_queue *)
+tevent_queue_running: bool (struct tevent_queue *)
+tevent_queue_start: void (struct tevent_queue *)
+tevent_queue_stop: void (struct tevent_queue *)
+tevent_re_initialise: int (struct tevent_context *)
+tevent_register_backend: bool (const char *, const struct tevent_ops *)
+tevent_req_default_print: char *(struct tevent_req *, TALLOC_CTX *)
+tevent_req_defer_callback: void (struct tevent_req *, struct tevent_context *)
+tevent_req_is_error: bool (struct tevent_req *, enum tevent_req_state *, uint64_t *)
+tevent_req_is_in_progress: bool (struct tevent_req *)
+tevent_req_poll: bool (struct tevent_req *, struct tevent_context *)
+tevent_req_post: struct tevent_req *(struct tevent_req *, struct tevent_context *)
+tevent_req_print: char *(TALLOC_CTX *, struct tevent_req *)
+tevent_req_received: void (struct tevent_req *)
+tevent_req_set_callback: void (struct tevent_req *, tevent_req_fn, void *)
+tevent_req_set_cancel_fn: void (struct tevent_req *, tevent_req_cancel_fn)
+tevent_req_set_endtime: bool (struct tevent_req *, struct tevent_context *, struct timeval)
+tevent_req_set_print_fn: void (struct tevent_req *, tevent_req_print_fn)
+tevent_set_abort_fn: void (void (*)(const char *))
+tevent_set_debug: int (struct tevent_context *, void (*)(void *, enum tevent_debug_level, const char *, va_list), void *)
+tevent_set_debug_stderr: int (struct tevent_context *)
+tevent_set_default_backend: void (const char *)
+tevent_set_trace_callback: void (struct tevent_context *, tevent_trace_callback_t, void *)
+tevent_signal_support: bool (struct tevent_context *)
+tevent_timeval_add: struct timeval (const struct timeval *, uint32_t, uint32_t)
+tevent_timeval_compare: int (const struct timeval *, const struct timeval *)
+tevent_timeval_current: struct timeval (void)
+tevent_timeval_current_ofs: struct timeval (uint32_t, uint32_t)
+tevent_timeval_is_zero: bool (const struct timeval *)
+tevent_timeval_set: struct timeval (uint32_t, uint32_t)
+tevent_timeval_until: struct timeval (const struct timeval *, const struct timeval *)
+tevent_timeval_zero: struct timeval (void)
+tevent_trace_point_callback: void (struct tevent_context *, enum tevent_trace_point)
+tevent_wakeup_recv: bool (struct tevent_req *)
+tevent_wakeup_send: struct tevent_req *(TALLOC_CTX *, struct tevent_context *, struct timeval)
diff --git a/ctdb/lib/tevent/ABI/tevent-0.9.9.sigs b/ctdb/lib/tevent/ABI/tevent-0.9.9.sigs
new file mode 100644
index 0000000000..9adaba579b
--- /dev/null
+++ b/ctdb/lib/tevent/ABI/tevent-0.9.9.sigs
@@ -0,0 +1,73 @@
+_tevent_add_fd: struct tevent_fd *(struct tevent_context *, TALLOC_CTX *, int, uint16_t, tevent_fd_handler_t, void *, const char *, const char *)
+_tevent_add_signal: struct tevent_signal *(struct tevent_context *, TALLOC_CTX *, int, int, tevent_signal_handler_t, void *, const char *, const char *)
+_tevent_add_timer: struct tevent_timer *(struct tevent_context *, TALLOC_CTX *, struct timeval, tevent_timer_handler_t, void *, const char *, const char *)
+_tevent_create_immediate: struct tevent_immediate *(TALLOC_CTX *, const char *)
+_tevent_loop_once: int (struct tevent_context *, const char *)
+_tevent_loop_until: int (struct tevent_context *, bool (*)(void *), void *, const char *)
+_tevent_loop_wait: int (struct tevent_context *, const char *)
+_tevent_queue_create: struct tevent_queue *(TALLOC_CTX *, const char *, const char *)
+_tevent_req_callback_data: void *(struct tevent_req *)
+_tevent_req_cancel: bool (struct tevent_req *, const char *)
+_tevent_req_create: struct tevent_req *(TALLOC_CTX *, void *, size_t, const char *, const char *)
+_tevent_req_data: void *(struct tevent_req *)
+_tevent_req_done: void (struct tevent_req *, const char *)
+_tevent_req_error: bool (struct tevent_req *, uint64_t, const char *)
+_tevent_req_nomem: bool (const void *, struct tevent_req *, const char *)
+_tevent_req_notify_callback: void (struct tevent_req *, const char *)
+_tevent_schedule_immediate: void (struct tevent_immediate *, struct tevent_context *, tevent_immediate_handler_t, void *, const char *, const char *)
+tevent_backend_list: const char **(TALLOC_CTX *)
+tevent_cleanup_pending_signal_handlers: void (struct tevent_signal *)
+tevent_common_add_fd: struct tevent_fd *(struct tevent_context *, TALLOC_CTX *, int, uint16_t, tevent_fd_handler_t, void *, const char *, const char *)
+tevent_common_add_signal: struct tevent_signal *(struct tevent_context *, TALLOC_CTX *, int, int, tevent_signal_handler_t, void *, const char *, const char *)
+tevent_common_add_timer: struct tevent_timer *(struct tevent_context *, TALLOC_CTX *, struct timeval, tevent_timer_handler_t, void *, const char *, const char *)
+tevent_common_check_signal: int (struct tevent_context *)
+tevent_common_context_destructor: int (struct tevent_context *)
+tevent_common_fd_destructor: int (struct tevent_fd *)
+tevent_common_fd_get_flags: uint16_t (struct tevent_fd *)
+tevent_common_fd_set_close_fn: void (struct tevent_fd *, tevent_fd_close_fn_t)
+tevent_common_fd_set_flags: void (struct tevent_fd *, uint16_t)
+tevent_common_loop_immediate: bool (struct tevent_context *)
+tevent_common_loop_timer_delay: struct timeval (struct tevent_context *)
+tevent_common_loop_wait: int (struct tevent_context *, const char *)
+tevent_common_schedule_immediate: void (struct tevent_immediate *, struct tevent_context *, tevent_immediate_handler_t, void *, const char *, const char *)
+tevent_context_init: struct tevent_context *(TALLOC_CTX *)
+tevent_context_init_byname: struct tevent_context *(TALLOC_CTX *, const char *)
+tevent_debug: void (struct tevent_context *, enum tevent_debug_level, const char *, ...)
+tevent_fd_get_flags: uint16_t (struct tevent_fd *)
+tevent_fd_set_auto_close: void (struct tevent_fd *)
+tevent_fd_set_close_fn: void (struct tevent_fd *, tevent_fd_close_fn_t)
+tevent_fd_set_flags: void (struct tevent_fd *, uint16_t)
+tevent_loop_allow_nesting: void (struct tevent_context *)
+tevent_loop_set_nesting_hook: void (struct tevent_context *, tevent_nesting_hook, void *)
+tevent_queue_add: bool (struct tevent_queue *, struct tevent_context *, struct tevent_req *, tevent_queue_trigger_fn_t, void *)
+tevent_queue_length: size_t (struct tevent_queue *)
+tevent_queue_start: void (struct tevent_queue *)
+tevent_queue_stop: void (struct tevent_queue *)
+tevent_re_initialise: int (struct tevent_context *)
+tevent_register_backend: bool (const char *, const struct tevent_ops *)
+tevent_req_default_print: char *(struct tevent_req *, TALLOC_CTX *)
+tevent_req_is_error: bool (struct tevent_req *, enum tevent_req_state *, uint64_t *)
+tevent_req_is_in_progress: bool (struct tevent_req *)
+tevent_req_poll: bool (struct tevent_req *, struct tevent_context *)
+tevent_req_post: struct tevent_req *(struct tevent_req *, struct tevent_context *)
+tevent_req_print: char *(TALLOC_CTX *, struct tevent_req *)
+tevent_req_received: void (struct tevent_req *)
+tevent_req_set_callback: void (struct tevent_req *, tevent_req_fn, void *)
+tevent_req_set_cancel_fn: void (struct tevent_req *, tevent_req_cancel_fn)
+tevent_req_set_endtime: bool (struct tevent_req *, struct tevent_context *, struct timeval)
+tevent_req_set_print_fn: void (struct tevent_req *, tevent_req_print_fn)
+tevent_set_abort_fn: void (void (*)(const char *))
+tevent_set_debug: int (struct tevent_context *, void (*)(void *, enum tevent_debug_level, const char *, va_list), void *)
+tevent_set_debug_stderr: int (struct tevent_context *)
+tevent_set_default_backend: void (const char *)
+tevent_signal_support: bool (struct tevent_context *)
+tevent_timeval_add: struct timeval (const struct timeval *, uint32_t, uint32_t)
+tevent_timeval_compare: int (const struct timeval *, const struct timeval *)
+tevent_timeval_current: struct timeval (void)
+tevent_timeval_current_ofs: struct timeval (uint32_t, uint32_t)
+tevent_timeval_is_zero: bool (const struct timeval *)
+tevent_timeval_set: struct timeval (uint32_t, uint32_t)
+tevent_timeval_until: struct timeval (const struct timeval *, const struct timeval *)
+tevent_timeval_zero: struct timeval (void)
+tevent_wakeup_recv: bool (struct tevent_req *)
+tevent_wakeup_send: struct tevent_req *(TALLOC_CTX *, struct tevent_context *, struct timeval)
diff --git a/ctdb/lib/tevent/bindings.py b/ctdb/lib/tevent/bindings.py
new file mode 100644
index 0000000000..1060caf28d
--- /dev/null
+++ b/ctdb/lib/tevent/bindings.py
@@ -0,0 +1,62 @@
+#!/usr/bin/python
+#
+#   Python integration for tevent - tests
+#
+#   Copyright (C) Jelmer Vernooij 2010
+#
+#     ** NOTE! The following LGPL license applies to the tevent
+#     ** library. This does NOT imply that all of Samba is released
+#     ** under the LGPL
+#
+#   This library is free software; you can redistribute it and/or
+#   modify it under the terms of the GNU Lesser General Public
+#   License as published by the Free Software Foundation; either
+#   version 3 of the License, or (at your option) any later version.
+#
+#   This library is distributed in the hope that it will be useful,
+#   but WITHOUT ANY WARRANTY; without even the implied warranty of
+#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+#   Lesser General Public License for more details.
+#
+#   You should have received a copy of the GNU Lesser General Public
+#   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+
+import signal
+import _tevent
+from unittest import TestCase
+
+class BackendListTests(TestCase):
+
+    def test_backend_list(self):
+        self.assertTrue(isinstance(_tevent.backend_list(), list))
+
+
+class CreateContextTests(TestCase):
+
+    def test_by_name(self):
+        ctx = _tevent.Context(_tevent.backend_list()[0])
+        self.assertTrue(ctx is not None)
+
+    def test_no_name(self):
+        ctx = _tevent.Context()
+        self.assertTrue(ctx is not None)
+
+
+class ContextTests(TestCase):
+
+    def setUp(self):
+        super(ContextTests, self).setUp()
+        self.ctx = _tevent.Context()
+
+    def test_signal_support(self):
+        self.assertTrue(type(self.ctx.signal_support) is bool)
+
+    def test_reinitialise(self):
+        self.ctx.reinitialise()
+
+    def test_loop_wait(self):
+        self.ctx.loop_wait()
+
+    def test_add_signal(self):
+        sig = self.ctx.add_signal(signal.SIGINT, 0, lambda callback: None)
+        self.assertTrue(isinstance(sig, _tevent.Signal))
diff --git a/ctdb/lib/tevent/doc/mainpage.dox b/ctdb/lib/tevent/doc/mainpage.dox
new file mode 100644
index 0000000000..e2f986e1ef
--- /dev/null
+++ b/ctdb/lib/tevent/doc/mainpage.dox
@@ -0,0 +1,42 @@
+/**
+ * @mainpage
+ *
+ * Tevent is an event system based on the talloc memory management library. It
+ * is the core event system used in Samba.
+ *
+ * The low level tevent has support for many event types, including timers,
+ * signals, and the classic file descriptor events.
+ *
+ * Tevent also provide helpers to deal with asynchronous code providing the
+ * tevent_req (tevent request) functions.
+ *
+ * @section tevent_download Download
+ *
+ * You can download the latest releases of tevent from the
+ * <a href="http://samba.org/ftp/tevent" target="_blank">tevent directory</a>
+ * on the samba public source archive.
+ *
+ * @section tevent_bugs Discussion and bug reports
+ *
+ * tevent does not currently have its own mailing list or bug tracking system.
+ * For now, please use the
+ * <a href="https://lists.samba.org/mailman/listinfo/samba-technical" target="_blank">samba-technical</a>
+ * mailing list, and the
+ * <a href="http://bugzilla.samba.org/" target="_blank">Samba bugzilla</a>
+ * bug tracking system.
+ *
+ * @section tevent_devel Development
+ * You can download the latest code either via git or rsync.
+ *
+ * To fetch via git see the following guide:
+ *
+ * <a href="http://wiki.samba.org/index.php/Using_Git_for_Samba_Development" target="_blank">Using Git for Samba Development</a>
+ *
+ * Once you have cloned the tree switch to the master branch and cd into the
+ * lib/tevent directory.
+ *
+ * To fetch via rsync use this command:
+ *
+ *   rsync -Pavz samba.org::ftp/unpacked/standalone_projects/lib/tevent .
+ *
+ */
diff --git a/ctdb/lib/tevent/doc/tutorials.dox b/ctdb/lib/tevent/doc/tutorials.dox
new file mode 100644
index 0000000000..e8beed7dc0
--- /dev/null
+++ b/ctdb/lib/tevent/doc/tutorials.dox
@@ -0,0 +1,43 @@
+/**
+ * @page tevent_queue_tutorial The tevent_queue tutorial
+ *
+ * @section Introduction
+ *
+ * A tevent_queue is used to queue up async requests that must be
+ * serialized. For example writing buffers into a socket must be
+ * serialized. Writing a large lump of data into a socket can require
+ * multiple write(2) or send(2) system calls. If more than one async
+ * request is outstanding to write large buffers into a socket, every
+ * request must individually be completed before the next one begins,
+ * even if multiple syscalls are required.
+ *
+ * To do this, every socket gets assigned a tevent_queue struct.
+ *
+ * Creating a serialized async request follows the usual convention to
+ * return a tevent_req structure with an embedded state structure. To
+ * serialize the work the requests is about to so, instead of directly
+ * starting or doing that work, tevent_queue_add must be called. When it
+ * is time for the serialized async request to do its work, the trigger
+ * callback function tevent_queue_add was given is called. In the example
+ * of writing to a socket, the trigger is called when the write request
+ * can begin accessing the socket.
+ *
+ * How does this engine work behind the scenes? When the queue is empty,
+ * tevent_queue_add schedules an immediate call to the trigger
+ * callback. The trigger callback starts its work, likely by starting
+ * other async subrequests. While these async subrequests are working,
+ * more requests can accumulate in the queue by tevent_queue_add. While
+ * there is no function to explicitly trigger the next waiter in line, it
+ * still works: When the active request in the queue is done, it will be
+ * destroyed by talloc_free. Talloc_free of an serialized async request
+ * that had been added to a queue will trigger the next request in the
+ * queue via a talloc destructor attached to a child of the serialized
+ * request. This way the queue will be kept busy when an async request
+ * finishes.
+ *
+ * @section Example
+ *
+ * @code
+ *      Metze: Please add a code example here.
+ * @endcode
+ */
diff --git a/ctdb/lib/tevent/doxy.config b/ctdb/lib/tevent/doxy.config
new file mode 100644
index 0000000000..0d67ae3592
--- /dev/null
+++ b/ctdb/lib/tevent/doxy.config
@@ -0,0 +1,1535 @@
+# Doxyfile 1.6.1
+
+# This file describes the settings to be used by the documentation system
+# doxygen (www.doxygen.org) for a project
+#
+# All text after a hash (#) is considered a comment and will be ignored
+# The format is:
+#       TAG = value [value, ...]
+# For lists items can also be appended using:
+#       TAG += value [value, ...]
+# Values that contain spaces should be placed between quotes (" ")
+
+#---------------------------------------------------------------------------
+# Project related configuration options
+#---------------------------------------------------------------------------
+
+# This tag specifies the encoding used for all characters in the config file
+# that follow. The default is UTF-8 which is also the encoding used for all
+# text before the first occurrence of this tag. Doxygen uses libiconv (or the
+# iconv built into libc) for the transcoding. See
+# http://www.gnu.org/software/libiconv for the list of possible encodings.
+
+DOXYFILE_ENCODING      = UTF-8
+
+# The PROJECT_NAME tag is a single word (or a sequence of words surrounded
+# by quotes) that should identify the project.
+
+PROJECT_NAME           = tevent
+
+# The PROJECT_NUMBER tag can be used to enter a project or revision number.
+# This could be handy for archiving the generated documentation or
+# if some version control system is used.
+
+PROJECT_NUMBER         = 0.9.8
+
+# The OUTPUT_DIRECTORY tag is used to specify the (relative or absolute)
+# base path where the generated documentation will be put.
+# If a relative path is entered, it will be relative to the location
+# where doxygen was started. If left blank the current directory will be used.
+
+OUTPUT_DIRECTORY       = doc
+
+# If the CREATE_SUBDIRS tag is set to YES, then doxygen will create
+# 4096 sub-directories (in 2 levels) under the output directory of each output
+# format and will distribute the generated files over these directories.
+# Enabling this option can be useful when feeding doxygen a huge amount of
+# source files, where putting all generated files in the same directory would
+# otherwise cause performance problems for the file system.
+
+CREATE_SUBDIRS         = NO
+
+# The OUTPUT_LANGUAGE tag is used to specify the language in which all
+# documentation generated by doxygen is written. Doxygen will use this
+# information to generate all constant output in the proper language.
+# The default language is English, other supported languages are:
+# Afrikaans, Arabic, Brazilian, Catalan, Chinese, Chinese-Traditional,
+# Croatian, Czech, Danish, Dutch, Esperanto, Farsi, Finnish, French, German,
+# Greek, Hungarian, Italian, Japanese, Japanese-en (Japanese with English
+# messages), Korean, Korean-en, Lithuanian, Norwegian, Macedonian, Persian,
+# Polish, Portuguese, Romanian, Russian, Serbian, Serbian-Cyrilic, Slovak,
+# Slovene, Spanish, Swedish, Ukrainian, and Vietnamese.
+
+OUTPUT_LANGUAGE        = English
+
+# If the BRIEF_MEMBER_DESC tag is set to YES (the default) Doxygen will
+# include brief member descriptions after the members that are listed in
+# the file and class documentation (similar to JavaDoc).
+# Set to NO to disable this.
+
+BRIEF_MEMBER_DESC      = YES
+
+# If the REPEAT_BRIEF tag is set to YES (the default) Doxygen will prepend
+# the brief description of a member or function before the detailed description.
+# Note: if both HIDE_UNDOC_MEMBERS and BRIEF_MEMBER_DESC are set to NO, the
+# brief descriptions will be completely suppressed.
+
+REPEAT_BRIEF           = YES
+
+# This tag implements a quasi-intelligent brief description abbreviator
+# that is used to form the text in various listings. Each string
+# in this list, if found as the leading text of the brief description, will be
+# stripped from the text and the result after processing the whole list, is
+# used as the annotated text. Otherwise, the brief description is used as-is.
+# If left blank, the following values are used ("$name" is automatically
+# replaced with the name of the entity): "The $name class" "The $name widget"
+# "The $name file" "is" "provides" "specifies" "contains"
+# "represents" "a" "an" "the"
+
+ABBREVIATE_BRIEF       = "The $name class" \
+                         "The $name widget" \
+                         "The $name file" \
+                         is \
+                         provides \
+                         specifies \
+                         contains \
+                         represents \
+                         a \
+                         an \
+                         the
+
+# If the ALWAYS_DETAILED_SEC and REPEAT_BRIEF tags are both set to YES then
+# Doxygen will generate a detailed section even if there is only a brief
+# description.
+
+ALWAYS_DETAILED_SEC    = NO
+
+# If the INLINE_INHERITED_MEMB tag is set to YES, doxygen will show all
+# inherited members of a class in the documentation of that class as if those
+# members were ordinary class members. Constructors, destructors and assignment
+# operators of the base classes will not be shown.
+
+INLINE_INHERITED_MEMB  = NO
+
+# If the FULL_PATH_NAMES tag is set to YES then Doxygen will prepend the full
+# path before files name in the file list and in the header files. If set
+# to NO the shortest path that makes the file name unique will be used.
+
+FULL_PATH_NAMES        = YES
+
+# If the FULL_PATH_NAMES tag is set to YES then the STRIP_FROM_PATH tag
+# can be used to strip a user-defined part of the path. Stripping is
+# only done if one of the specified strings matches the left-hand part of
+# the path. The tag can be used to show relative paths in the file list.
+# If left blank the directory from which doxygen is run is used as the
+# path to strip.
+
+STRIP_FROM_PATH        =
+
+# The STRIP_FROM_INC_PATH tag can be used to strip a user-defined part of
+# the path mentioned in the documentation of a class, which tells
+# the reader which header file to include in order to use a class.
+# If left blank only the name of the header file containing the class
+# definition is used. Otherwise one should specify the include paths that
+# are normally passed to the compiler using the -I flag.
+
+STRIP_FROM_INC_PATH    =
+
+# If the SHORT_NAMES tag is set to YES, doxygen will generate much shorter
+# (but less readable) file names. This can be useful is your file systems
+# doesn't support long names like on DOS, Mac, or CD-ROM.
+
+SHORT_NAMES            = NO
+
+# If the JAVADOC_AUTOBRIEF tag is set to YES then Doxygen
+# will interpret the first line (until the first dot) of a JavaDoc-style
+# comment as the brief description. If set to NO, the JavaDoc
+# comments will behave just like regular Qt-style comments
+# (thus requiring an explicit @brief command for a brief description.)
+
+JAVADOC_AUTOBRIEF      = YES
+
+# If the QT_AUTOBRIEF tag is set to YES then Doxygen will
+# interpret the first line (until the first dot) of a Qt-style
+# comment as the brief description. If set to NO, the comments
+# will behave just like regular Qt-style comments (thus requiring
+# an explicit \brief command for a brief description.)
+
+QT_AUTOBRIEF           = NO
+
+# The MULTILINE_CPP_IS_BRIEF tag can be set to YES to make Doxygen
+# treat a multi-line C++ special comment block (i.e. a block of //! or ///
+# comments) as a brief description. This used to be the default behaviour.
+# The new default is to treat a multi-line C++ comment block as a detailed
+# description. Set this tag to YES if you prefer the old behaviour instead.
+
+MULTILINE_CPP_IS_BRIEF = NO
+
+# If the INHERIT_DOCS tag is set to YES (the default) then an undocumented
+# member inherits the documentation from any documented member that it
+# re-implements.
+
+INHERIT_DOCS           = YES
+
+# If the SEPARATE_MEMBER_PAGES tag is set to YES, then doxygen will produce
+# a new page for each member. If set to NO, the documentation of a member will
+# be part of the file/class/namespace that contains it.
+
+SEPARATE_MEMBER_PAGES  = NO
+
+# The TAB_SIZE tag can be used to set the number of spaces in a tab.
+# Doxygen uses this value to replace tabs by spaces in code fragments.
+
+TAB_SIZE               = 8
+
+# This tag can be used to specify a number of aliases that acts
+# as commands in the documentation. An alias has the form "name=value".
+# For example adding "sideeffect=\par Side Effects:\n" will allow you to
+# put the command \sideeffect (or @sideeffect) in the documentation, which
+# will result in a user-defined paragraph with heading "Side Effects:".
+# You can put \n's in the value part of an alias to insert newlines.
+
+ALIASES                =
+
+# Set the OPTIMIZE_OUTPUT_FOR_C tag to YES if your project consists of C
+# sources only. Doxygen will then generate output that is more tailored for C.
+# For instance, some of the names that are used will be different. The list
+# of all members will be omitted, etc.
+
+OPTIMIZE_OUTPUT_FOR_C  = YES
+
+# Set the OPTIMIZE_OUTPUT_JAVA tag to YES if your project consists of Java
+# sources only. Doxygen will then generate output that is more tailored for
+# Java. For instance, namespaces will be presented as packages, qualified
+# scopes will look different, etc.
+
+OPTIMIZE_OUTPUT_JAVA   = NO
+
+# Set the OPTIMIZE_FOR_FORTRAN tag to YES if your project consists of Fortran
+# sources only. Doxygen will then generate output that is more tailored for
+# Fortran.
+
+OPTIMIZE_FOR_FORTRAN   = NO
+
+# Set the OPTIMIZE_OUTPUT_VHDL tag to YES if your project consists of VHDL
+# sources. Doxygen will then generate output that is tailored for
+# VHDL.
+
+OPTIMIZE_OUTPUT_VHDL   = NO
+
+# Doxygen selects the parser to use depending on the extension of the files it parses.
+# With this tag you can assign which parser to use for a given extension.
+# Doxygen has a built-in mapping, but you can override or extend it using this tag.
+# The format is ext=language, where ext is a file extension, and language is one of
+# the parsers supported by doxygen: IDL, Java, Javascript, C#, C, C++, D, PHP,
+# Objective-C, Python, Fortran, VHDL, C, C++. For instance to make doxygen treat
+# .inc files as Fortran files (default is PHP), and .f files as C (default is Fortran),
+# use: inc=Fortran f=C. Note that for custom extensions you also need to set FILE_PATTERNS otherwise the files are not read by doxygen.
+
+EXTENSION_MAPPING      =
+
+# If you use STL classes (i.e. std::string, std::vector, etc.) but do not want
+# to include (a tag file for) the STL sources as input, then you should
+# set this tag to YES in order to let doxygen match functions declarations and
+# definitions whose arguments contain STL classes (e.g. func(std::string); v.s.
+# func(std::string) {}). This also make the inheritance and collaboration
+# diagrams that involve STL classes more complete and accurate.
+
+BUILTIN_STL_SUPPORT    = NO
+
+# If you use Microsoft's C++/CLI language, you should set this option to YES to
+# enable parsing support.
+
+CPP_CLI_SUPPORT        = NO
+
+# Set the SIP_SUPPORT tag to YES if your project consists of sip sources only.
+# Doxygen will parse them like normal C++ but will assume all classes use public
+# instead of private inheritance when no explicit protection keyword is present.
+
+SIP_SUPPORT            = NO
+
+# For Microsoft's IDL there are propget and propput attributes to indicate getter
+# and setter methods for a property. Setting this option to YES (the default)
+# will make doxygen to replace the get and set methods by a property in the
+# documentation. This will only work if the methods are indeed getting or
+# setting a simple type. If this is not the case, or you want to show the
+# methods anyway, you should set this option to NO.
+
+IDL_PROPERTY_SUPPORT   = YES
+
+# If member grouping is used in the documentation and the DISTRIBUTE_GROUP_DOC
+# tag is set to YES, then doxygen will reuse the documentation of the first
+# member in the group (if any) for the other members of the group. By default
+# all members of a group must be documented explicitly.
+
+DISTRIBUTE_GROUP_DOC   = NO
+
+# Set the SUBGROUPING tag to YES (the default) to allow class member groups of
+# the same type (for instance a group of public functions) to be put as a
+# subgroup of that type (e.g. under the Public Functions section). Set it to
+# NO to prevent subgrouping. Alternatively, this can be done per class using
+# the \nosubgrouping command.
+
+SUBGROUPING            = YES
+
+# When TYPEDEF_HIDES_STRUCT is enabled, a typedef of a struct, union, or enum
+# is documented as struct, union, or enum with the name of the typedef. So
+# typedef struct TypeS {} TypeT, will appear in the documentation as a struct
+# with name TypeT. When disabled the typedef will appear as a member of a file,
+# namespace, or class. And the struct will be named TypeS. This can typically
+# be useful for C code in case the coding convention dictates that all compound
+# types are typedef'ed and only the typedef is referenced, never the tag name.
+
+TYPEDEF_HIDES_STRUCT   = NO
+
+# The SYMBOL_CACHE_SIZE determines the size of the internal cache use to
+# determine which symbols to keep in memory and which to flush to disk.
+# When the cache is full, less often used symbols will be written to disk.
+# For small to medium size projects (<1000 input files) the default value is
+# probably good enough. For larger projects a too small cache size can cause
+# doxygen to be busy swapping symbols to and from disk most of the time
+# causing a significant performance penality.
+# If the system has enough physical memory increasing the cache will improve the
+# performance by keeping more symbols in memory. Note that the value works on
+# a logarithmic scale so increasing the size by one will rougly double the
+# memory usage. The cache size is given by this formula:
+# 2^(16+SYMBOL_CACHE_SIZE). The valid range is 0..9, the default is 0,
+# corresponding to a cache size of 2^16 = 65536 symbols
+
+SYMBOL_CACHE_SIZE      = 0
+
+#---------------------------------------------------------------------------
+# Build related configuration options
+#---------------------------------------------------------------------------
+
+# If the EXTRACT_ALL tag is set to YES doxygen will assume all entities in
+# documentation are documented, even if no documentation was available.
+# Private class members and static file members will be hidden unless
+# the EXTRACT_PRIVATE and EXTRACT_STATIC tags are set to YES
+
+EXTRACT_ALL            = NO
+
+# If the EXTRACT_PRIVATE tag is set to YES all private members of a class
+# will be included in the documentation.
+
+EXTRACT_PRIVATE        = NO
+
+# If the EXTRACT_STATIC tag is set to YES all static members of a file
+# will be included in the documentation.
+
+EXTRACT_STATIC         = NO
+
+# If the EXTRACT_LOCAL_CLASSES tag is set to YES classes (and structs)
+# defined locally in source files will be included in the documentation.
+# If set to NO only classes defined in header files are included.
+
+EXTRACT_LOCAL_CLASSES  = NO
+
+# This flag is only useful for Objective-C code. When set to YES local
+# methods, which are defined in the implementation section but not in
+# the interface are included in the documentation.
+# If set to NO (the default) only methods in the interface are included.
+
+EXTRACT_LOCAL_METHODS  = NO
+
+# If this flag is set to YES, the members of anonymous namespaces will be
+# extracted and appear in the documentation as a namespace called
+# 'anonymous_namespace{file}', where file will be replaced with the base
+# name of the file that contains the anonymous namespace. By default
+# anonymous namespace are hidden.
+
+EXTRACT_ANON_NSPACES   = NO
+
+# If the HIDE_UNDOC_MEMBERS tag is set to YES, Doxygen will hide all
+# undocumented members of documented classes, files or namespaces.
+# If set to NO (the default) these members will be included in the
+# various overviews, but no documentation section is generated.
+# This option has no effect if EXTRACT_ALL is enabled.
+
+HIDE_UNDOC_MEMBERS     = YES
+
+# If the HIDE_UNDOC_CLASSES tag is set to YES, Doxygen will hide all
+# undocumented classes that are normally visible in the class hierarchy.
+# If set to NO (the default) these classes will be included in the various
+# overviews. This option has no effect if EXTRACT_ALL is enabled.
+
+HIDE_UNDOC_CLASSES     = YES
+
+# If the HIDE_FRIEND_COMPOUNDS tag is set to YES, Doxygen will hide all
+# friend (class|struct|union) declarations.
+# If set to NO (the default) these declarations will be included in the
+# documentation.
+
+HIDE_FRIEND_COMPOUNDS  = NO
+
+# If the HIDE_IN_BODY_DOCS tag is set to YES, Doxygen will hide any
+# documentation blocks found inside the body of a function.
+# If set to NO (the default) these blocks will be appended to the
+# function's detailed documentation block.
+
+HIDE_IN_BODY_DOCS      = NO
+
+# The INTERNAL_DOCS tag determines if documentation
+# that is typed after a \internal command is included. If the tag is set
+# to NO (the default) then the documentation will be excluded.
+# Set it to YES to include the internal documentation.
+
+INTERNAL_DOCS          = NO
+
+# If the CASE_SENSE_NAMES tag is set to NO then Doxygen will only generate
+# file names in lower-case letters. If set to YES upper-case letters are also
+# allowed. This is useful if you have classes or files whose names only differ
+# in case and if your file system supports case sensitive file names. Windows
+# and Mac users are advised to set this option to NO.
+
+CASE_SENSE_NAMES       = YES
+
+# If the HIDE_SCOPE_NAMES tag is set to NO (the default) then Doxygen
+# will show members with their full class and namespace scopes in the
+# documentation. If set to YES the scope will be hidden.
+
+HIDE_SCOPE_NAMES       = NO
+
+# If the SHOW_INCLUDE_FILES tag is set to YES (the default) then Doxygen
+# will put a list of the files that are included by a file in the documentation
+# of that file.
+
+SHOW_INCLUDE_FILES     = YES
+
+# If the INLINE_INFO tag is set to YES (the default) then a tag [inline]
+# is inserted in the documentation for inline members.
+
+INLINE_INFO            = YES
+
+# If the SORT_MEMBER_DOCS tag is set to YES (the default) then doxygen
+# will sort the (detailed) documentation of file and class members
+# alphabetically by member name. If set to NO the members will appear in
+# declaration order.
+
+SORT_MEMBER_DOCS       = YES
+
+# If the SORT_BRIEF_DOCS tag is set to YES then doxygen will sort the
+# brief documentation of file, namespace and class members alphabetically
+# by member name. If set to NO (the default) the members will appear in
+# declaration order.
+
+SORT_BRIEF_DOCS        = NO
+
+# If the SORT_MEMBERS_CTORS_1ST tag is set to YES then doxygen will sort the (brief and detailed) documentation of class members so that constructors and destructors are listed first. If set to NO (the default) the constructors will appear in the respective orders defined by SORT_MEMBER_DOCS and SORT_BRIEF_DOCS. This tag will be ignored for brief docs if SORT_BRIEF_DOCS is set to NO and ignored for detailed docs if SORT_MEMBER_DOCS is set to NO.
+
+SORT_MEMBERS_CTORS_1ST = NO
+
+# If the SORT_GROUP_NAMES tag is set to YES then doxygen will sort the
+# hierarchy of group names into alphabetical order. If set to NO (the default)
+# the group names will appear in their defined order.
+
+SORT_GROUP_NAMES       = NO
+
+# If the SORT_BY_SCOPE_NAME tag is set to YES, the class list will be
+# sorted by fully-qualified names, including namespaces. If set to
+# NO (the default), the class list will be sorted only by class name,
+# not including the namespace part.
+# Note: This option is not very useful if HIDE_SCOPE_NAMES is set to YES.
+# Note: This option applies only to the class list, not to the
+# alphabetical list.
+
+SORT_BY_SCOPE_NAME     = NO
+
+# The GENERATE_TODOLIST tag can be used to enable (YES) or
+# disable (NO) the todo list. This list is created by putting \todo
+# commands in the documentation.
+
+GENERATE_TODOLIST      = YES
+
+# The GENERATE_TESTLIST tag can be used to enable (YES) or
+# disable (NO) the test list. This list is created by putting \test
+# commands in the documentation.
+
+GENERATE_TESTLIST      = YES
+
+# The GENERATE_BUGLIST tag can be used to enable (YES) or
+# disable (NO) the bug list. This list is created by putting \bug
+# commands in the documentation.
+
+GENERATE_BUGLIST       = YES
+
+# The GENERATE_DEPRECATEDLIST tag can be used to enable (YES) or
+# disable (NO) the deprecated list. This list is created by putting
+# \deprecated commands in the documentation.
+
+GENERATE_DEPRECATEDLIST= YES
+
+# The ENABLED_SECTIONS tag can be used to enable conditional
+# documentation sections, marked by \if sectionname ... \endif.
+
+ENABLED_SECTIONS       =
+
+# The MAX_INITIALIZER_LINES tag determines the maximum number of lines
+# the initial value of a variable or define consists of for it to appear in
+# the documentation. If the initializer consists of more lines than specified
+# here it will be hidden. Use a value of 0 to hide initializers completely.
+# The appearance of the initializer of individual variables and defines in the
+# documentation can be controlled using \showinitializer or \hideinitializer
+# command in the documentation regardless of this setting.
+
+MAX_INITIALIZER_LINES  = 30
+
+# Set the SHOW_USED_FILES tag to NO to disable the list of files generated
+# at the bottom of the documentation of classes and structs. If set to YES the
+# list will mention the files that were used to generate the documentation.
+
+SHOW_USED_FILES        = YES
+
+# If the sources in your project are distributed over multiple directories
+# then setting the SHOW_DIRECTORIES tag to YES will show the directory hierarchy
+# in the documentation. The default is NO.
+
+SHOW_DIRECTORIES       = NO
+
+# Set the SHOW_FILES tag to NO to disable the generation of the Files page.
+# This will remove the Files entry from the Quick Index and from the
+# Folder Tree View (if specified). The default is YES.
+
+SHOW_FILES             = YES
+
+# Set the SHOW_NAMESPACES tag to NO to disable the generation of the
+# Namespaces page.
+# This will remove the Namespaces entry from the Quick Index
+# and from the Folder Tree View (if specified). The default is YES.
+
+SHOW_NAMESPACES        = YES
+
+# The FILE_VERSION_FILTER tag can be used to specify a program or script that
+# doxygen should invoke to get the current version for each file (typically from
+# the version control system). Doxygen will invoke the program by executing (via
+# popen()) the command <command> <input-file>, where <command> is the value of
+# the FILE_VERSION_FILTER tag, and <input-file> is the name of an input file
+# provided by doxygen. Whatever the program writes to standard output
+# is used as the file version. See the manual for examples.
+
+FILE_VERSION_FILTER    =
+
+# The LAYOUT_FILE tag can be used to specify a layout file which will be parsed by
+# doxygen. The layout file controls the global structure of the generated output files
+# in an output format independent way. The create the layout file that represents
+# doxygen's defaults, run doxygen with the -l option. You can optionally specify a
+# file name after the option, if omitted DoxygenLayout.xml will be used as the name
+# of the layout file.
+
+LAYOUT_FILE            =
+
+#---------------------------------------------------------------------------
+# configuration options related to warning and progress messages
+#---------------------------------------------------------------------------
+
+# The QUIET tag can be used to turn on/off the messages that are generated
+# by doxygen. Possible values are YES and NO. If left blank NO is used.
+
+QUIET                  = NO
+
+# The WARNINGS tag can be used to turn on/off the warning messages that are
+# generated by doxygen. Possible values are YES and NO. If left blank
+# NO is used.
+
+WARNINGS               = YES
+
+# If WARN_IF_UNDOCUMENTED is set to YES, then doxygen will generate warnings
+# for undocumented members. If EXTRACT_ALL is set to YES then this flag will
+# automatically be disabled.
+
+WARN_IF_UNDOCUMENTED   = YES
+
+# If WARN_IF_DOC_ERROR is set to YES, doxygen will generate warnings for
+# potential errors in the documentation, such as not documenting some
+# parameters in a documented function, or documenting parameters that
+# don't exist or using markup commands wrongly.
+
+WARN_IF_DOC_ERROR      = YES
+
+# This WARN_NO_PARAMDOC option can be abled to get warnings for
+# functions that are documented, but have no documentation for their parameters
+# or return value. If set to NO (the default) doxygen will only warn about
+# wrong or incomplete parameter documentation, but not about the absence of
+# documentation.
+
+WARN_NO_PARAMDOC       = NO
+
+# The WARN_FORMAT tag determines the format of the warning messages that
+# doxygen can produce. The string should contain the $file, $line, and $text
+# tags, which will be replaced by the file and line number from which the
+# warning originated and the warning text. Optionally the format may contain
+# $version, which will be replaced by the version of the file (if it could
+# be obtained via FILE_VERSION_FILTER)
+
+WARN_FORMAT            = "$file:$line: $text"
+
+# The WARN_LOGFILE tag can be used to specify a file to which warning
+# and error messages should be written. If left blank the output is written
+# to stderr.
+
+WARN_LOGFILE           =
+
+#---------------------------------------------------------------------------
+# configuration options related to the input files
+#---------------------------------------------------------------------------
+
+# The INPUT tag can be used to specify the files and/or directories that contain
+# documented source files. You may enter file names like "myfile.cpp" or
+# directories like "/usr/src/myproject". Separate the files or directories
+# with spaces.
+
+INPUT                  = . doc
+
+# This tag can be used to specify the character encoding of the source files
+# that doxygen parses. Internally doxygen uses the UTF-8 encoding, which is
+# also the default input encoding. Doxygen uses libiconv (or the iconv built
+# into libc) for the transcoding. See http://www.gnu.org/software/libiconv for
+# the list of possible encodings.
+
+INPUT_ENCODING         = UTF-8
+
+# If the value of the INPUT tag contains directories, you can use the
+# FILE_PATTERNS tag to specify one or more wildcard pattern (like *.cpp
+# and *.h) to filter out the source-files in the directories. If left
+# blank the following patterns are tested:
+# *.c *.cc *.cxx *.cpp *.c++ *.java *.ii *.ixx *.ipp *.i++ *.inl *.h *.hh *.hxx
+# *.hpp *.h++ *.idl *.odl *.cs *.php *.php3 *.inc *.m *.mm *.py *.f90
+
+FILE_PATTERNS          = *.cpp \
+                         *.cc \
+                         *.c \
+                         *.h \
+                         *.hh \
+                         *.hpp \
+                         *.dox
+
+# The RECURSIVE tag can be used to turn specify whether or not subdirectories
+# should be searched for input files as well. Possible values are YES and NO.
+# If left blank NO is used.
+
+RECURSIVE              = NO
+
+# The EXCLUDE tag can be used to specify files and/or directories that should
+# excluded from the INPUT source files. This way you can easily exclude a
+# subdirectory from a directory tree whose root is specified with the INPUT tag.
+
+EXCLUDE                =
+
+# The EXCLUDE_SYMLINKS tag can be used select whether or not files or
+# directories that are symbolic links (a Unix filesystem feature) are excluded
+# from the input.
+
+EXCLUDE_SYMLINKS       = NO
+
+# If the value of the INPUT tag contains directories, you can use the
+# EXCLUDE_PATTERNS tag to specify one or more wildcard patterns to exclude
+# certain files from those directories. Note that the wildcards are matched
+# against the file with absolute path, so to exclude all test directories
+# for example use the pattern */test/*
+
+EXCLUDE_PATTERNS       = */.git/*
+
+# The EXCLUDE_SYMBOLS tag can be used to specify one or more symbol names
+# (namespaces, classes, functions, etc.) that should be excluded from the
+# output. The symbol name can be a fully qualified name, a word, or if the
+# wildcard * is used, a substring. Examples: ANamespace, AClass,
+# AClass::ANamespace, ANamespace::*Test
+
+EXCLUDE_SYMBOLS        =
+
+# The EXAMPLE_PATH tag can be used to specify one or more files or
+# directories that contain example code fragments that are included (see
+# the \include command).
+
+EXAMPLE_PATH           =
+
+# If the value of the EXAMPLE_PATH tag contains directories, you can use the
+# EXAMPLE_PATTERNS tag to specify one or more wildcard pattern (like *.cpp
+# and *.h) to filter out the source-files in the directories. If left
+# blank all files are included.
+
+EXAMPLE_PATTERNS       =
+
+# If the EXAMPLE_RECURSIVE tag is set to YES then subdirectories will be
+# searched for input files to be used with the \include or \dontinclude
+# commands irrespective of the value of the RECURSIVE tag.
+# Possible values are YES and NO. If left blank NO is used.
+
+EXAMPLE_RECURSIVE      = NO
+
+# The IMAGE_PATH tag can be used to specify one or more files or
+# directories that contain image that are included in the documentation (see
+# the \image command).
+
+IMAGE_PATH             =
+
+# The INPUT_FILTER tag can be used to specify a program that doxygen should
+# invoke to filter for each input file. Doxygen will invoke the filter program
+# by executing (via popen()) the command <filter> <input-file>, where <filter>
+# is the value of the INPUT_FILTER tag, and <input-file> is the name of an
+# input file. Doxygen will then use the output that the filter program writes
+# to standard output.
+# If FILTER_PATTERNS is specified, this tag will be
+# ignored.
+
+INPUT_FILTER           =
+
+# The FILTER_PATTERNS tag can be used to specify filters on a per file pattern
+# basis.
+# Doxygen will compare the file name with each pattern and apply the
+# filter if there is a match.
+# The filters are a list of the form:
+# pattern=filter (like *.cpp=my_cpp_filter). See INPUT_FILTER for further
+# info on how filters are used. If FILTER_PATTERNS is empty, INPUT_FILTER
+# is applied to all files.
+
+FILTER_PATTERNS        =
+
+# If the FILTER_SOURCE_FILES tag is set to YES, the input filter (if set using
+# INPUT_FILTER) will be used to filter the input files when producing source
+# files to browse (i.e. when SOURCE_BROWSER is set to YES).
+
+FILTER_SOURCE_FILES    = NO
+
+#---------------------------------------------------------------------------
+# configuration options related to source browsing
+#---------------------------------------------------------------------------
+
+# If the SOURCE_BROWSER tag is set to YES then a list of source files will
+# be generated. Documented entities will be cross-referenced with these sources.
+# Note: To get rid of all source code in the generated output, make sure also
+# VERBATIM_HEADERS is set to NO.
+
+SOURCE_BROWSER         = NO
+
+# Setting the INLINE_SOURCES tag to YES will include the body
+# of functions and classes directly in the documentation.
+
+INLINE_SOURCES         = NO
+
+# Setting the STRIP_CODE_COMMENTS tag to YES (the default) will instruct
+# doxygen to hide any special comment blocks from generated source code
+# fragments. Normal C and C++ comments will always remain visible.
+
+STRIP_CODE_COMMENTS    = YES
+
+# If the REFERENCED_BY_RELATION tag is set to YES
+# then for each documented function all documented
+# functions referencing it will be listed.
+
+REFERENCED_BY_RELATION = NO
+
+# If the REFERENCES_RELATION tag is set to YES
+# then for each documented function all documented entities
+# called/used by that function will be listed.
+
+REFERENCES_RELATION    = NO
+
+# If the REFERENCES_LINK_SOURCE tag is set to YES (the default)
+# and SOURCE_BROWSER tag is set to YES, then the hyperlinks from
+# functions in REFERENCES_RELATION and REFERENCED_BY_RELATION lists will
+# link to the source code.
+# Otherwise they will link to the documentation.
+
+REFERENCES_LINK_SOURCE = YES
+
+# If the USE_HTAGS tag is set to YES then the references to source code
+# will point to the HTML generated by the htags(1) tool instead of doxygen
+# built-in source browser. The htags tool is part of GNU's global source
+# tagging system (see http://www.gnu.org/software/global/global.html). You
+# will need version 4.8.6 or higher.
+
+USE_HTAGS              = NO
+
+# If the VERBATIM_HEADERS tag is set to YES (the default) then Doxygen
+# will generate a verbatim copy of the header file for each class for
+# which an include is specified. Set to NO to disable this.
+
+VERBATIM_HEADERS       = YES
+
+#---------------------------------------------------------------------------
+# configuration options related to the alphabetical class index
+#---------------------------------------------------------------------------
+
+# If the ALPHABETICAL_INDEX tag is set to YES, an alphabetical index
+# of all compounds will be generated. Enable this if the project
+# contains a lot of classes, structs, unions or interfaces.
+
+ALPHABETICAL_INDEX     = NO
+
+# If the alphabetical index is enabled (see ALPHABETICAL_INDEX) then
+# the COLS_IN_ALPHA_INDEX tag can be used to specify the number of columns
+# in which this list will be split (can be a number in the range [1..20])
+
+COLS_IN_ALPHA_INDEX    = 5
+
+# In case all classes in a project start with a common prefix, all
+# classes will be put under the same header in the alphabetical index.
+# The IGNORE_PREFIX tag can be used to specify one or more prefixes that
+# should be ignored while generating the index headers.
+
+IGNORE_PREFIX          =
+
+#---------------------------------------------------------------------------
+# configuration options related to the HTML output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_HTML tag is set to YES (the default) Doxygen will
+# generate HTML output.
+
+GENERATE_HTML          = YES
+
+# If the HTML_FOOTER_DESCRIPTION tag is set to YES, Doxygen will
+# add generated date, project name and doxygen version to HTML footer.
+
+HTML_FOOTER_DESCRIPTION= NO
+
+# The HTML_OUTPUT tag is used to specify where the HTML docs will be put.
+# If a relative path is entered the value of OUTPUT_DIRECTORY will be
+# put in front of it. If left blank `html' will be used as the default path.
+
+HTML_OUTPUT            = html
+
+# The HTML_FILE_EXTENSION tag can be used to specify the file extension for
+# each generated HTML page (for example: .htm,.php,.asp). If it is left blank
+# doxygen will generate files with .html extension.
+
+HTML_FILE_EXTENSION    = .html
+
+# The HTML_HEADER tag can be used to specify a personal HTML header for
+# each generated HTML page. If it is left blank doxygen will generate a
+# standard header.
+
+HTML_HEADER            =
+
+# The HTML_FOOTER tag can be used to specify a personal HTML footer for
+# each generated HTML page. If it is left blank doxygen will generate a
+# standard footer.
+
+HTML_FOOTER            =
+
+# The HTML_STYLESHEET tag can be used to specify a user-defined cascading
+# style sheet that is used by each HTML page. It can be used to
+# fine-tune the look of the HTML output. If the tag is left blank doxygen
+# will generate a default style sheet. Note that doxygen will try to copy
+# the style sheet file to the HTML output directory, so don't put your own
+# stylesheet in the HTML output directory as well, or it will be erased!
+
+HTML_STYLESHEET        =
+
+# If the HTML_ALIGN_MEMBERS tag is set to YES, the members of classes,
+# files or namespaces will be aligned in HTML using tables. If set to
+# NO a bullet list will be used.
+
+HTML_ALIGN_MEMBERS     = YES
+
+# If the HTML_DYNAMIC_SECTIONS tag is set to YES then the generated HTML
+# documentation will contain sections that can be hidden and shown after the
+# page has loaded. For this to work a browser that supports
+# JavaScript and DHTML is required (for instance Mozilla 1.0+, Firefox
+# Netscape 6.0+, Internet explorer 5.0+, Konqueror, or Safari).
+
+HTML_DYNAMIC_SECTIONS  = NO
+
+# If the GENERATE_DOCSET tag is set to YES, additional index files
+# will be generated that can be used as input for Apple's Xcode 3
+# integrated development environment, introduced with OSX 10.5 (Leopard).
+# To create a documentation set, doxygen will generate a Makefile in the
+# HTML output directory. Running make will produce the docset in that
+# directory and running "make install" will install the docset in
+# ~/Library/Developer/Shared/Documentation/DocSets so that Xcode will find
+# it at startup.
+# See http://developer.apple.com/tools/creatingdocsetswithdoxygen.html for more information.
+
+GENERATE_DOCSET        = NO
+
+# When GENERATE_DOCSET tag is set to YES, this tag determines the name of the
+# feed. A documentation feed provides an umbrella under which multiple
+# documentation sets from a single provider (such as a company or product suite)
+# can be grouped.
+
+DOCSET_FEEDNAME        = "Doxygen generated docs"
+
+# When GENERATE_DOCSET tag is set to YES, this tag specifies a string that
+# should uniquely identify the documentation set bundle. This should be a
+# reverse domain-name style string, e.g. com.mycompany.MyDocSet. Doxygen
+# will append .docset to the name.
+
+DOCSET_BUNDLE_ID       = org.doxygen.Project
+
+# If the GENERATE_HTMLHELP tag is set to YES, additional index files
+# will be generated that can be used as input for tools like the
+# Microsoft HTML help workshop to generate a compiled HTML help file (.chm)
+# of the generated HTML documentation.
+
+GENERATE_HTMLHELP      = NO
+
+# If the GENERATE_HTMLHELP tag is set to YES, the CHM_FILE tag can
+# be used to specify the file name of the resulting .chm file. You
+# can add a path in front of the file if the result should not be
+# written to the html output directory.
+
+CHM_FILE               =
+
+# If the GENERATE_HTMLHELP tag is set to YES, the HHC_LOCATION tag can
+# be used to specify the location (absolute path including file name) of
+# the HTML help compiler (hhc.exe). If non-empty doxygen will try to run
+# the HTML help compiler on the generated index.hhp.
+
+HHC_LOCATION           =
+
+# If the GENERATE_HTMLHELP tag is set to YES, the GENERATE_CHI flag
+# controls if a separate .chi index file is generated (YES) or that
+# it should be included in the master .chm file (NO).
+
+GENERATE_CHI           = NO
+
+# If the GENERATE_HTMLHELP tag is set to YES, the CHM_INDEX_ENCODING
+# is used to encode HtmlHelp index (hhk), content (hhc) and project file
+# content.
+
+CHM_INDEX_ENCODING     =
+
+# If the GENERATE_HTMLHELP tag is set to YES, the BINARY_TOC flag
+# controls whether a binary table of contents is generated (YES) or a
+# normal table of contents (NO) in the .chm file.
+
+BINARY_TOC             = NO
+
+# The TOC_EXPAND flag can be set to YES to add extra items for group members
+# to the contents of the HTML help documentation and to the tree view.
+
+TOC_EXPAND             = NO
+
+# If the GENERATE_QHP tag is set to YES and both QHP_NAMESPACE and QHP_VIRTUAL_FOLDER
+# are set, an additional index file will be generated that can be used as input for
+# Qt's qhelpgenerator to generate a Qt Compressed Help (.qch) of the generated
+# HTML documentation.
+
+GENERATE_QHP           = NO
+
+# If the QHG_LOCATION tag is specified, the QCH_FILE tag can
+# be used to specify the file name of the resulting .qch file.
+# The path specified is relative to the HTML output folder.
+
+QCH_FILE               =
+
+# The QHP_NAMESPACE tag specifies the namespace to use when generating
+# Qt Help Project output. For more information please see
+# http://doc.trolltech.com/qthelpproject.html#namespace
+
+QHP_NAMESPACE          =
+
+# The QHP_VIRTUAL_FOLDER tag specifies the namespace to use when generating
+# Qt Help Project output. For more information please see
+# http://doc.trolltech.com/qthelpproject.html#virtual-folders
+
+QHP_VIRTUAL_FOLDER     = doc
+
+# If QHP_CUST_FILTER_NAME is set, it specifies the name of a custom filter to add.
+# For more information please see
+# http://doc.trolltech.com/qthelpproject.html#custom-filters
+
+QHP_CUST_FILTER_NAME   =
+
+# The QHP_CUST_FILT_ATTRS tag specifies the list of the attributes of the custom filter to add.For more information please see
+# <a href="http://doc.trolltech.com/qthelpproject.html#custom-filters">Qt Help Project / Custom Filters</a>.
+
+QHP_CUST_FILTER_ATTRS  =
+
+# The QHP_SECT_FILTER_ATTRS tag specifies the list of the attributes this project's
+# filter section matches.
+# <a href="http://doc.trolltech.com/qthelpproject.html#filter-attributes">Qt Help Project / Filter Attributes</a>.
+
+QHP_SECT_FILTER_ATTRS  =
+
+# If the GENERATE_QHP tag is set to YES, the QHG_LOCATION tag can
+# be used to specify the location of Qt's qhelpgenerator.
+# If non-empty doxygen will try to run qhelpgenerator on the generated
+# .qhp file.
+
+QHG_LOCATION           =
+
+# The DISABLE_INDEX tag can be used to turn on/off the condensed index at
+# top of each HTML page. The value NO (the default) enables the index and
+# the value YES disables it.
+
+DISABLE_INDEX          = NO
+
+# This tag can be used to set the number of enum values (range [1..20])
+# that doxygen will group on one line in the generated HTML documentation.
+
+ENUM_VALUES_PER_LINE   = 4
+
+# The GENERATE_TREEVIEW tag is used to specify whether a tree-like index
+# structure should be generated to display hierarchical information.
+# If the tag value is set to YES, a side panel will be generated
+# containing a tree-like index structure (just like the one that
+# is generated for HTML Help). For this to work a browser that supports
+# JavaScript, DHTML, CSS and frames is required (i.e. any modern browser).
+# Windows users are probably better off using the HTML help feature.
+
+GENERATE_TREEVIEW      = NONE
+
+# By enabling USE_INLINE_TREES, doxygen will generate the Groups, Directories,
+# and Class Hierarchy pages using a tree view instead of an ordered list.
+
+USE_INLINE_TREES       = NO
+
+# If the treeview is enabled (see GENERATE_TREEVIEW) then this tag can be
+# used to set the initial width (in pixels) of the frame in which the tree
+# is shown.
+
+TREEVIEW_WIDTH         = 250
+
+# Use this tag to change the font size of Latex formulas included
+# as images in the HTML documentation. The default is 10. Note that
+# when you change the font size after a successful doxygen run you need
+# to manually remove any form_*.png images from the HTML output directory
+# to force them to be regenerated.
+
+FORMULA_FONTSIZE       = 10
+
+# When the SEARCHENGINE tag is enable doxygen will generate a search box for the HTML output. The underlying search engine uses javascript
+# and DHTML and should work on any modern browser. Note that when using HTML help (GENERATE_HTMLHELP) or Qt help (GENERATE_QHP)
+# there is already a search function so this one should typically
+# be disabled.
+
+SEARCHENGINE           = NO
+
+#---------------------------------------------------------------------------
+# configuration options related to the LaTeX output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_LATEX tag is set to YES (the default) Doxygen will
+# generate Latex output.
+
+GENERATE_LATEX         = YES
+
+# The LATEX_OUTPUT tag is used to specify where the LaTeX docs will be put.
+# If a relative path is entered the value of OUTPUT_DIRECTORY will be
+# put in front of it. If left blank `latex' will be used as the default path.
+
+LATEX_OUTPUT           = latex
+
+# The LATEX_CMD_NAME tag can be used to specify the LaTeX command name to be
+# invoked. If left blank `latex' will be used as the default command name.
+
+LATEX_CMD_NAME         = latex
+
+# The MAKEINDEX_CMD_NAME tag can be used to specify the command name to
+# generate index for LaTeX. If left blank `makeindex' will be used as the
+# default command name.
+
+MAKEINDEX_CMD_NAME     = makeindex
+
+# If the COMPACT_LATEX tag is set to YES Doxygen generates more compact
+# LaTeX documents. This may be useful for small projects and may help to
+# save some trees in general.
+
+COMPACT_LATEX          = NO
+
+# The PAPER_TYPE tag can be used to set the paper type that is used
+# by the printer. Possible values are: a4, a4wide, letter, legal and
+# executive. If left blank a4wide will be used.
+
+PAPER_TYPE             = a4wide
+
+# The EXTRA_PACKAGES tag can be to specify one or more names of LaTeX
+# packages that should be included in the LaTeX output.
+
+EXTRA_PACKAGES         =
+
+# The LATEX_HEADER tag can be used to specify a personal LaTeX header for
+# the generated latex document. The header should contain everything until
+# the first chapter. If it is left blank doxygen will generate a
+# standard header. Notice: only use this tag if you know what you are doing!
+
+LATEX_HEADER           =
+
+# If the PDF_HYPERLINKS tag is set to YES, the LaTeX that is generated
+# is prepared for conversion to pdf (using ps2pdf). The pdf file will
+# contain links (just like the HTML output) instead of page references
+# This makes the output suitable for online browsing using a pdf viewer.
+
+PDF_HYPERLINKS         = YES
+
+# If the USE_PDFLATEX tag is set to YES, pdflatex will be used instead of
+# plain latex in the generated Makefile. Set this option to YES to get a
+# higher quality PDF documentation.
+
+USE_PDFLATEX           = YES
+
+# If the LATEX_BATCHMODE tag is set to YES, doxygen will add the \\batchmode.
+# command to the generated LaTeX files. This will instruct LaTeX to keep
+# running if errors occur, instead of asking the user for help.
+# This option is also used when generating formulas in HTML.
+
+LATEX_BATCHMODE        = NO
+
+# If LATEX_HIDE_INDICES is set to YES then doxygen will not
+# include the index chapters (such as File Index, Compound Index, etc.)
+# in the output.
+
+LATEX_HIDE_INDICES     = NO
+
+# If LATEX_SOURCE_CODE is set to YES then doxygen will include source code with syntax highlighting in the LaTeX output. Note that which sources are shown also depends on other settings such as SOURCE_BROWSER.
+
+LATEX_SOURCE_CODE      = NO
+
+#---------------------------------------------------------------------------
+# configuration options related to the RTF output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_RTF tag is set to YES Doxygen will generate RTF output
+# The RTF output is optimized for Word 97 and may not look very pretty with
+# other RTF readers or editors.
+
+GENERATE_RTF           = NO
+
+# The RTF_OUTPUT tag is used to specify where the RTF docs will be put.
+# If a relative path is entered the value of OUTPUT_DIRECTORY will be
+# put in front of it. If left blank `rtf' will be used as the default path.
+
+RTF_OUTPUT             = rtf
+
+# If the COMPACT_RTF tag is set to YES Doxygen generates more compact
+# RTF documents. This may be useful for small projects and may help to
+# save some trees in general.
+
+COMPACT_RTF            = NO
+
+# If the RTF_HYPERLINKS tag is set to YES, the RTF that is generated
+# will contain hyperlink fields. The RTF file will
+# contain links (just like the HTML output) instead of page references.
+# This makes the output suitable for online browsing using WORD or other
+# programs which support those fields.
+# Note: wordpad (write) and others do not support links.
+
+RTF_HYPERLINKS         = NO
+
+# Load stylesheet definitions from file. Syntax is similar to doxygen's
+# config file, i.e. a series of assignments. You only have to provide
+# replacements, missing definitions are set to their default value.
+
+RTF_STYLESHEET_FILE    =
+
+# Set optional variables used in the generation of an rtf document.
+# Syntax is similar to doxygen's config file.
+
+RTF_EXTENSIONS_FILE    =
+
+#---------------------------------------------------------------------------
+# configuration options related to the man page output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_MAN tag is set to YES (the default) Doxygen will
+# generate man pages
+
+GENERATE_MAN           = YES
+
+# The MAN_OUTPUT tag is used to specify where the man pages will be put.
+# If a relative path is entered the value of OUTPUT_DIRECTORY will be
+# put in front of it. If left blank `man' will be used as the default path.
+
+MAN_OUTPUT             = man
+
+# The MAN_EXTENSION tag determines the extension that is added to
+# the generated man pages (default is the subroutine's section .3)
+
+MAN_EXTENSION          = .3
+
+# If the MAN_LINKS tag is set to YES and Doxygen generates man output,
+# then it will generate one additional man file for each entity
+# documented in the real man page(s). These additional files
+# only source the real man page, but without them the man command
+# would be unable to find the correct page. The default is NO.
+
+MAN_LINKS              = NO
+
+#---------------------------------------------------------------------------
+# configuration options related to the XML output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_XML tag is set to YES Doxygen will
+# generate an XML file that captures the structure of
+# the code including all documentation.
+
+GENERATE_XML           = NO
+
+# The XML_OUTPUT tag is used to specify where the XML pages will be put.
+# If a relative path is entered the value of OUTPUT_DIRECTORY will be
+# put in front of it. If left blank `xml' will be used as the default path.
+
+XML_OUTPUT             = xml
+
+# The XML_SCHEMA tag can be used to specify an XML schema,
+# which can be used by a validating XML parser to check the
+# syntax of the XML files.
+
+XML_SCHEMA             =
+
+# The XML_DTD tag can be used to specify an XML DTD,
+# which can be used by a validating XML parser to check the
+# syntax of the XML files.
+
+XML_DTD                =
+
+# If the XML_PROGRAMLISTING tag is set to YES Doxygen will
+# dump the program listings (including syntax highlighting
+# and cross-referencing information) to the XML output. Note that
+# enabling this will significantly increase the size of the XML output.
+
+XML_PROGRAMLISTING     = YES
+
+#---------------------------------------------------------------------------
+# configuration options for the AutoGen Definitions output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_AUTOGEN_DEF tag is set to YES Doxygen will
+# generate an AutoGen Definitions (see autogen.sf.net) file
+# that captures the structure of the code including all
+# documentation. Note that this feature is still experimental
+# and incomplete at the moment.
+
+GENERATE_AUTOGEN_DEF   = NO
+
+#---------------------------------------------------------------------------
+# configuration options related to the Perl module output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_PERLMOD tag is set to YES Doxygen will
+# generate a Perl module file that captures the structure of
+# the code including all documentation. Note that this
+# feature is still experimental and incomplete at the
+# moment.
+
+GENERATE_PERLMOD       = NO
+
+# If the PERLMOD_LATEX tag is set to YES Doxygen will generate
+# the necessary Makefile rules, Perl scripts and LaTeX code to be able
+# to generate PDF and DVI output from the Perl module output.
+
+PERLMOD_LATEX          = NO
+
+# If the PERLMOD_PRETTY tag is set to YES the Perl module output will be
+# nicely formatted so it can be parsed by a human reader.
+# This is useful
+# if you want to understand what is going on.
+# On the other hand, if this
+# tag is set to NO the size of the Perl module output will be much smaller
+# and Perl will parse it just the same.
+
+PERLMOD_PRETTY         = YES
+
+# The names of the make variables in the generated doxyrules.make file
+# are prefixed with the string contained in PERLMOD_MAKEVAR_PREFIX.
+# This is useful so different doxyrules.make files included by the same
+# Makefile don't overwrite each other's variables.
+
+PERLMOD_MAKEVAR_PREFIX =
+
+#---------------------------------------------------------------------------
+# Configuration options related to the preprocessor
+#---------------------------------------------------------------------------
+
+# If the ENABLE_PREPROCESSING tag is set to YES (the default) Doxygen will
+# evaluate all C-preprocessor directives found in the sources and include
+# files.
+
+ENABLE_PREPROCESSING   = YES
+
+# If the MACRO_EXPANSION tag is set to YES Doxygen will expand all macro
+# names in the source code. If set to NO (the default) only conditional
+# compilation will be performed. Macro expansion can be done in a controlled
+# way by setting EXPAND_ONLY_PREDEF to YES.
+
+MACRO_EXPANSION        = YES
+
+# If the EXPAND_ONLY_PREDEF and MACRO_EXPANSION tags are both set to YES
+# then the macro expansion is limited to the macros specified with the
+# PREDEFINED and EXPAND_AS_DEFINED tags.
+
+EXPAND_ONLY_PREDEF     = YES
+
+# If the SEARCH_INCLUDES tag is set to YES (the default) the includes files
+# in the INCLUDE_PATH (see below) will be search if a #include is found.
+
+SEARCH_INCLUDES        = YES
+
+# The INCLUDE_PATH tag can be used to specify one or more directories that
+# contain include files that are not input files but should be processed by
+# the preprocessor.
+
+INCLUDE_PATH           =
+
+# You can use the INCLUDE_FILE_PATTERNS tag to specify one or more wildcard
+# patterns (like *.h and *.hpp) to filter out the header-files in the
+# directories. If left blank, the patterns specified with FILE_PATTERNS will
+# be used.
+
+INCLUDE_FILE_PATTERNS  =
+
+# The PREDEFINED tag can be used to specify one or more macro names that
+# are defined before the preprocessor is started (similar to the -D option of
+# gcc). The argument of the tag is a list of macros of the form: name
+# or name=definition (no spaces). If the definition and the = are
+# omitted =1 is assumed. To prevent a macro definition from being
+# undefined via #undef or recursively expanded use the := operator
+# instead of the = operator.
+
+PREDEFINED             = DOXYGEN PRINTF_ATTRIBUTE(x,y)=
+
+# If the MACRO_EXPANSION and EXPAND_ONLY_PREDEF tags are set to YES then
+# this tag can be used to specify a list of macro names that should be expanded.
+# The macro definition that is found in the sources will be used.
+# Use the PREDEFINED tag if you want to use a different macro definition.
+
+EXPAND_AS_DEFINED      =
+
+# If the SKIP_FUNCTION_MACROS tag is set to YES (the default) then
+# doxygen's preprocessor will remove all function-like macros that are alone
+# on a line, have an all uppercase name, and do not end with a semicolon. Such
+# function macros are typically used for boiler-plate code, and will confuse
+# the parser if not removed.
+
+SKIP_FUNCTION_MACROS   = YES
+
+#---------------------------------------------------------------------------
+# Configuration::additions related to external references
+#---------------------------------------------------------------------------
+
+# The TAGFILES option can be used to specify one or more tagfiles.
+# Optionally an initial location of the external documentation
+# can be added for each tagfile. The format of a tag file without
+# this location is as follows:
+#
+# TAGFILES = file1 file2 ...
+# Adding location for the tag files is done as follows:
+#
+# TAGFILES = file1=loc1 "file2 = loc2" ...
+# where "loc1" and "loc2" can be relative or absolute paths or
+# URLs. If a location is present for each tag, the installdox tool
+# does not have to be run to correct the links.
+# Note that each tag file must have a unique name
+# (where the name does NOT include the path)
+# If a tag file is not located in the directory in which doxygen
+# is run, you must also specify the path to the tagfile here.
+
+TAGFILES               =
+
+# When a file name is specified after GENERATE_TAGFILE, doxygen will create
+# a tag file that is based on the input files it reads.
+
+GENERATE_TAGFILE       =
+
+# If the ALLEXTERNALS tag is set to YES all external classes will be listed
+# in the class index. If set to NO only the inherited external classes
+# will be listed.
+
+ALLEXTERNALS           = NO
+
+# If the EXTERNAL_GROUPS tag is set to YES all external groups will be listed
+# in the modules index. If set to NO, only the current project's groups will
+# be listed.
+
+EXTERNAL_GROUPS        = YES
+
+# The PERL_PATH should be the absolute path and name of the perl script
+# interpreter (i.e. the result of `which perl').
+
+PERL_PATH              = /usr/bin/perl
+
+#---------------------------------------------------------------------------
+# Configuration options related to the dot tool
+#---------------------------------------------------------------------------
+
+# If the CLASS_DIAGRAMS tag is set to YES (the default) Doxygen will
+# generate a inheritance diagram (in HTML, RTF and LaTeX) for classes with base
+# or super classes. Setting the tag to NO turns the diagrams off. Note that
+# this option is superseded by the HAVE_DOT option below. This is only a
+# fallback. It is recommended to install and use dot, since it yields more
+# powerful graphs.
+
+CLASS_DIAGRAMS         = YES
+
+# You can define message sequence charts within doxygen comments using the \msc
+# command. Doxygen will then run the mscgen tool (see
+# http://www.mcternan.me.uk/mscgen/) to produce the chart and insert it in the
+# documentation. The MSCGEN_PATH tag allows you to specify the directory where
+# the mscgen tool resides. If left empty the tool is assumed to be found in the
+# default search path.
+
+MSCGEN_PATH            =
+
+# If set to YES, the inheritance and collaboration graphs will hide
+# inheritance and usage relations if the target is undocumented
+# or is not a class.
+
+HIDE_UNDOC_RELATIONS   = YES
+
+# If you set the HAVE_DOT tag to YES then doxygen will assume the dot tool is
+# available from the path. This tool is part of Graphviz, a graph visualization
+# toolkit from AT&T and Lucent Bell Labs. The other options in this section
+# have no effect if this option is set to NO (the default)
+
+HAVE_DOT               = NO
+
+# By default doxygen will write a font called FreeSans.ttf to the output
+# directory and reference it in all dot files that doxygen generates. This
+# font does not include all possible unicode characters however, so when you need
+# these (or just want a differently looking font) you can specify the font name
+# using DOT_FONTNAME. You need need to make sure dot is able to find the font,
+# which can be done by putting it in a standard location or by setting the
+# DOTFONTPATH environment variable or by setting DOT_FONTPATH to the directory
+# containing the font.
+
+DOT_FONTNAME           = FreeSans
+
+# The DOT_FONTSIZE tag can be used to set the size of the font of dot graphs.
+# The default size is 10pt.
+
+DOT_FONTSIZE           = 10
+
+# By default doxygen will tell dot to use the output directory to look for the
+# FreeSans.ttf font (which doxygen will put there itself). If you specify a
+# different font using DOT_FONTNAME you can set the path where dot
+# can find it using this tag.
+
+DOT_FONTPATH           =
+
+# If the CLASS_GRAPH and HAVE_DOT tags are set to YES then doxygen
+# will generate a graph for each documented class showing the direct and
+# indirect inheritance relations. Setting this tag to YES will force the
+# the CLASS_DIAGRAMS tag to NO.
+
+CLASS_GRAPH            = YES
+
+# If the COLLABORATION_GRAPH and HAVE_DOT tags are set to YES then doxygen
+# will generate a graph for each documented class showing the direct and
+# indirect implementation dependencies (inheritance, containment, and
+# class references variables) of the class with other documented classes.
+
+COLLABORATION_GRAPH    = YES
+
+# If the GROUP_GRAPHS and HAVE_DOT tags are set to YES then doxygen
+# will generate a graph for groups, showing the direct groups dependencies
+
+GROUP_GRAPHS           = YES
+
+# If the UML_LOOK tag is set to YES doxygen will generate inheritance and
+# collaboration diagrams in a style similar to the OMG's Unified Modeling
+# Language.
+
+UML_LOOK               = NO
+
+# If set to YES, the inheritance and collaboration graphs will show the
+# relations between templates and their instances.
+
+TEMPLATE_RELATIONS     = NO
+
+# If the ENABLE_PREPROCESSING, SEARCH_INCLUDES, INCLUDE_GRAPH, and HAVE_DOT
+# tags are set to YES then doxygen will generate a graph for each documented
+# file showing the direct and indirect include dependencies of the file with
+# other documented files.
+
+INCLUDE_GRAPH          = YES
+
+# If the ENABLE_PREPROCESSING, SEARCH_INCLUDES, INCLUDED_BY_GRAPH, and
+# HAVE_DOT tags are set to YES then doxygen will generate a graph for each
+# documented header file showing the documented files that directly or
+# indirectly include this file.
+
+INCLUDED_BY_GRAPH      = YES
+
+# If the CALL_GRAPH and HAVE_DOT options are set to YES then
+# doxygen will generate a call dependency graph for every global function
+# or class method. Note that enabling this option will significantly increase
+# the time of a run. So in most cases it will be better to enable call graphs
+# for selected functions only using the \callgraph command.
+
+CALL_GRAPH             = NO
+
+# If the CALLER_GRAPH and HAVE_DOT tags are set to YES then
+# doxygen will generate a caller dependency graph for every global function
+# or class method. Note that enabling this option will significantly increase
+# the time of a run. So in most cases it will be better to enable caller
+# graphs for selected functions only using the \callergraph command.
+
+CALLER_GRAPH           = NO
+
+# If the GRAPHICAL_HIERARCHY and HAVE_DOT tags are set to YES then doxygen
+# will graphical hierarchy of all classes instead of a textual one.
+
+GRAPHICAL_HIERARCHY    = YES
+
+# If the DIRECTORY_GRAPH, SHOW_DIRECTORIES and HAVE_DOT tags are set to YES
+# then doxygen will show the dependencies a directory has on other directories
+# in a graphical way. The dependency relations are determined by the #include
+# relations between the files in the directories.
+
+DIRECTORY_GRAPH        = YES
+
+# The DOT_IMAGE_FORMAT tag can be used to set the image format of the images
+# generated by dot. Possible values are png, jpg, or gif
+# If left blank png will be used.
+
+DOT_IMAGE_FORMAT       = png
+
+# The tag DOT_PATH can be used to specify the path where the dot tool can be
+# found. If left blank, it is assumed the dot tool can be found in the path.
+
+DOT_PATH               =
+
+# The DOTFILE_DIRS tag can be used to specify one or more directories that
+# contain dot files that are included in the documentation (see the
+# \dotfile command).
+
+DOTFILE_DIRS           =
+
+# The DOT_GRAPH_MAX_NODES tag can be used to set the maximum number of
+# nodes that will be shown in the graph. If the number of nodes in a graph
+# becomes larger than this value, doxygen will truncate the graph, which is
+# visualized by representing a node as a red box. Note that doxygen if the
+# number of direct children of the root node in a graph is already larger than
+# DOT_GRAPH_MAX_NODES then the graph will not be shown at all. Also note
+# that the size of a graph can be further restricted by MAX_DOT_GRAPH_DEPTH.
+
+DOT_GRAPH_MAX_NODES    = 50
+
+# The MAX_DOT_GRAPH_DEPTH tag can be used to set the maximum depth of the
+# graphs generated by dot. A depth value of 3 means that only nodes reachable
+# from the root by following a path via at most 3 edges will be shown. Nodes
+# that lay further from the root node will be omitted. Note that setting this
+# option to 1 or 2 may greatly reduce the computation time needed for large
+# code bases. Also note that the size of a graph can be further restricted by
+# DOT_GRAPH_MAX_NODES. Using a depth of 0 means no depth restriction.
+
+MAX_DOT_GRAPH_DEPTH    = 0
+
+# Set the DOT_TRANSPARENT tag to YES to generate images with a transparent
+# background. This is disabled by default, because dot on Windows does not
+# seem to support this out of the box. Warning: Depending on the platform used,
+# enabling this option may lead to badly anti-aliased labels on the edges of
+# a graph (i.e. they become hard to read).
+
+DOT_TRANSPARENT        = YES
+
+# Set the DOT_MULTI_TARGETS tag to YES allow dot to generate multiple output
+# files in one run (i.e. multiple -o and -T options on the command line). This
+# makes dot run faster, but since only newer versions of dot (>1.8.10)
+# support this, this feature is disabled by default.
+
+DOT_MULTI_TARGETS      = NO
+
+# If the GENERATE_LEGEND tag is set to YES (the default) Doxygen will
+# generate a legend page explaining the meaning of the various boxes and
+# arrows in the dot generated graphs.
+
+GENERATE_LEGEND        = YES
+
+# If the DOT_CLEANUP tag is set to YES (the default) Doxygen will
+# remove the intermediate dot files that are used to generate
+# the various graphs.
+
+DOT_CLEANUP            = YES
diff --git a/ctdb/lib/tevent/libtevent.m4 b/ctdb/lib/tevent/libtevent.m4
new file mode 100644
index 0000000000..0682f5b315
--- /dev/null
+++ b/ctdb/lib/tevent/libtevent.m4
@@ -0,0 +1,94 @@
+dnl Check to see if we should use the included tevent
+
+INCLUDED_TEVENT=auto
+AC_ARG_WITH(included-tevent,
+    [AC_HELP_STRING([--with-included-tevent], [use bundled tevent library, not from system])],
+    [ INCLUDED_TEVENT=$withval ])
+
+AC_SUBST(TEVENT_LIBS)
+AC_SUBST(TEVENT_CFLAGS)
+
+if test x"$INCLUDED_TEVENT" != x"yes" ; then
+    AC_CHECK_HEADERS(tevent.h)
+    AC_CHECK_LIB(tevent, tevent_context_init, [ TEVENT_LIBS="-ltevent" ])
+    AC_CHECK_DECLS([TEVENT_TRACE_BEFORE_WAIT],,, [[#include <tevent.h>]])
+    if test x"$ac_cv_header_tevent_h" = x"no" -o \
+	x"$ac_cv_lib_tevent_tevent_context_init" = x"no" -o \
+	x"$ac_cv_have_decl_TEVENT_TRACE_BEFORE_WAIT" = x"no" ; then
+        INCLUDED_TEVENT=yes
+        TEVENT_CFLAGS=""
+    else
+        INCLUDED_TEVENT=no
+    fi
+fi
+
+AC_MSG_CHECKING(whether to use included tevent)
+AC_MSG_RESULT($INCLUDED_TEVENT)
+if test x"$INCLUDED_TEVENT" != x"no" ; then
+    dnl find the tevent sources. This is meant to work both for
+    dnl standalone builds, and builds of packages using libtevent
+	teventdir=""
+	teventpaths="$srcdir $srcdir/lib/tevent $srcdir/tevent $srcdir/../tevent"
+	for d in $teventpaths; do
+		if test -f "$d/tevent.c"; then
+			teventdir="$d"
+            AC_SUBST(teventdir)
+			break
+		fi
+	done
+	if test x"$teventdir" = "x"; then
+	   AC_MSG_ERROR([cannot find tevent source in $teventpaths])
+	fi
+    TEVENT_OBJ="tevent.o tevent_debug.o tevent_util.o"
+    TEVENT_OBJ="$TEVENT_OBJ tevent_fd.o tevent_timed.o tevent_immediate.o tevent_signal.o"
+    TEVENT_OBJ="$TEVENT_OBJ tevent_req.o tevent_wakeup.o tevent_queue.o"
+    TEVENT_OBJ="$TEVENT_OBJ tevent_standard.o tevent_select.o tevent_poll.o"
+    AC_SUBST(TEVENT_OBJ)
+
+    TEVENT_CFLAGS="-I$teventdir"
+    AC_SUBST(TEVENT_CFLAGS)
+
+    TEVENT_LIBS=""
+    AC_SUBST(TEVENT_LIBS)
+
+    AC_CHECK_HEADERS(sys/epoll.h)
+    AC_CHECK_FUNCS(epoll_create)
+    if test x"$ac_cv_header_sys_epoll_h" = x"yes" -a x"$ac_cv_func_epoll_create" = x"yes"; then
+        TEVENT_OBJ="$TEVENT_OBJ tevent_epoll.o"
+        AC_DEFINE(HAVE_EPOLL, 1, [Whether epoll available])
+    fi
+
+    tevent_num_signals_includes="$ac_includes_default
+    #include <signal.h>
+    "
+    tevent_num_signals=64
+    AC_CHECK_VALUEOF(NSIG, [$tevent_num_signals_includes])
+    v=$ac_cv_valueof_NSIG
+    test -n "$v" && test "$v" -gt "$tevent_num_signals" && {
+            tevent_num_signals=$v
+    }
+    AC_CHECK_VALUEOF(_NSIG, [$tevent_num_signals_includes])
+    v=$ac_cv_valueof__NSIG
+    test -n "$v" && test "$v" -gt "$tevent_num_signals" && {
+            tevent_num_signals=$v
+    }
+    AC_CHECK_VALUEOF(SIGRTMAX, [$tevent_num_signals_includes])
+    v=$ac_cv_valueof_SIGRTMAX
+    test -n "$v" && test "$v" -gt "$tevent_num_signals" && {
+            tevent_num_signals=$v
+    }
+    AC_CHECK_VALUEOF(SIGRTMIN, [$tevent_num_signals_includes])
+    v=$ac_cv_valueof_SIGRTMIN
+    test -n "$v" && {
+            v=`expr $v + $v`
+    }
+    test -n "$v" && test "$v" -gt "$tevent_num_signals" && {
+            tevent_num_signals=$v
+    }
+    AC_DEFINE_UNQUOTED(TEVENT_NUM_SIGNALS, $tevent_num_signals, [Max signal number value])
+
+    if test x"$VERSIONSCRIPT" != "x"; then
+        EXPORTSFILE=tevent.exports
+        AC_SUBST(EXPORTSFILE)
+    fi
+fi
diff --git a/ctdb/lib/tevent/pytevent.c b/ctdb/lib/tevent/pytevent.c
new file mode 100644
index 0000000000..870f5aad61
--- /dev/null
+++ b/ctdb/lib/tevent/pytevent.c
@@ -0,0 +1,766 @@
+/*
+   Unix SMB/CIFS implementation.
+   Python bindings for tevent
+
+   Copyright (C) Jelmer Vernooij 2010
+
+     ** NOTE! The following LGPL license applies to the tevent
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include <Python.h>
+#include <tevent.h>
+
+void init_tevent(void);
+
+typedef struct {
+	PyObject_HEAD
+	struct tevent_context *ev;
+} TeventContext_Object;
+
+typedef struct {
+	PyObject_HEAD
+	struct tevent_queue *queue;
+} TeventQueue_Object;
+
+typedef struct {
+	PyObject_HEAD
+	struct tevent_req *req;
+} TeventReq_Object;
+
+typedef struct {
+	PyObject_HEAD
+	struct tevent_signal *signal;
+} TeventSignal_Object;
+
+typedef struct {
+	PyObject_HEAD
+	struct tevent_timer *timer;
+} TeventTimer_Object;
+
+typedef struct {
+	PyObject_HEAD
+	struct tevent_fd *fd;
+} TeventFd_Object;
+
+staticforward PyTypeObject TeventContext_Type;
+staticforward PyTypeObject TeventReq_Type;
+staticforward PyTypeObject TeventQueue_Type;
+staticforward PyTypeObject TeventSignal_Type;
+staticforward PyTypeObject TeventTimer_Type;
+staticforward PyTypeObject TeventFd_Type;
+
+static int py_context_init(struct tevent_context *ev)
+{
+	/* FIXME */
+	return 0;
+}
+
+static struct tevent_fd *py_add_fd(struct tevent_context *ev,
+				    TALLOC_CTX *mem_ctx,
+				    int fd, uint16_t flags,
+				    tevent_fd_handler_t handler,
+				    void *private_data,
+				    const char *handler_name,
+				    const char *location)
+{
+	/* FIXME */
+	return NULL;
+}
+
+static void py_set_fd_close_fn(struct tevent_fd *fde,
+				tevent_fd_close_fn_t close_fn)
+{
+	/* FIXME */
+}
+
+static uint16_t py_get_fd_flags(struct tevent_fd *fde)
+{
+	/* FIXME */
+	return 0;
+}
+
+static void py_set_fd_flags(struct tevent_fd *fde, uint16_t flags)
+{
+	/* FIXME */
+}
+
+/* timed_event functions */
+static struct tevent_timer *py_add_timer(struct tevent_context *ev,
+					  TALLOC_CTX *mem_ctx,
+					  struct timeval next_event,
+					  tevent_timer_handler_t handler,
+					  void *private_data,
+					  const char *handler_name,
+					  const char *location)
+{
+	/* FIXME */
+	return NULL;
+}
+
+/* immediate event functions */
+static void py_schedule_immediate(struct tevent_immediate *im,
+				   struct tevent_context *ev,
+				   tevent_immediate_handler_t handler,
+				   void *private_data,
+				   const char *handler_name,
+				   const char *location)
+{
+	/* FIXME */
+}
+
+/* signal functions */
+static struct tevent_signal *py_add_signal(struct tevent_context *ev,
+					    TALLOC_CTX *mem_ctx,
+					    int signum, int sa_flags,
+					    tevent_signal_handler_t handler,
+					    void *private_data,
+					    const char *handler_name,
+					    const char *location)
+{
+	/* FIXME */
+	return NULL;
+}
+
+/* loop functions */
+static int py_loop_once(struct tevent_context *ev, const char *location)
+{
+	/* FIXME */
+	return 0;
+}
+
+static int py_loop_wait(struct tevent_context *ev, const char *location)
+{
+	/* FIXME */
+	return 0;
+}
+
+const static struct tevent_ops py_tevent_ops = {
+	.context_init = py_context_init,
+	.add_fd = py_add_fd,
+	.set_fd_close_fn = py_set_fd_close_fn,
+	.get_fd_flags = py_get_fd_flags,
+	.set_fd_flags = py_set_fd_flags,
+	.add_timer = py_add_timer,
+	.schedule_immediate = py_schedule_immediate,
+	.add_signal = py_add_signal,
+	.loop_wait = py_loop_wait,
+	.loop_once = py_loop_once,
+};
+
+static PyObject *py_register_backend(PyObject *self, PyObject *args)
+{
+	PyObject *name, *py_backend;
+
+	if (!PyArg_ParseTuple(args, "O", &py_backend))
+		return NULL;
+
+	name = PyObject_GetAttrString(py_backend, "name");
+	if (name == NULL) {
+		PyErr_SetNone(PyExc_AttributeError);
+		return NULL;
+	}
+
+	if (!PyString_Check(name)) {
+		PyErr_SetNone(PyExc_TypeError);
+		return NULL;
+	}
+
+	if (!tevent_register_backend(PyString_AsString(name), &py_tevent_ops)) { /* FIXME: What to do with backend */
+		PyErr_SetNone(PyExc_RuntimeError);
+		return NULL;
+	}
+
+	Py_RETURN_NONE;
+}
+
+static PyObject *py_tevent_context_reinitialise(TeventContext_Object *self)
+{
+	int ret = tevent_re_initialise(self->ev);
+	if (ret != 0) {
+		PyErr_SetNone(PyExc_RuntimeError);
+		return NULL;
+	}
+	Py_RETURN_NONE;
+}
+
+static PyObject *py_tevent_queue_stop(TeventQueue_Object *self)
+{
+	tevent_queue_stop(self->queue);
+	Py_RETURN_NONE;
+}
+
+static PyObject *py_tevent_queue_start(TeventQueue_Object *self)
+{
+	tevent_queue_start(self->queue);
+	Py_RETURN_NONE;
+}
+
+static void py_queue_trigger(struct tevent_req *req, void *private_data)
+{
+	PyObject *callback = private_data, *ret;
+
+	ret = PyObject_CallFunction(callback, "");
+	Py_XDECREF(ret);
+}
+
+static PyObject *py_tevent_queue_add(TeventQueue_Object *self, PyObject *args)
+{
+	TeventContext_Object *py_ev;
+	TeventReq_Object *py_req;
+	PyObject *trigger;
+	bool ret;
+
+	if (!PyArg_ParseTuple(args, "O!O!O", 
+						  &TeventContext_Type, &py_ev,
+						  &TeventReq_Type, &py_req,
+						  &trigger))
+		return NULL;
+
+	Py_INCREF(trigger);
+
+	ret = tevent_queue_add(self->queue, py_ev->ev, py_req->req,
+						   py_queue_trigger, trigger);
+	if (!ret) {
+		PyErr_SetString(PyExc_RuntimeError, "queue add failed");
+		Py_DECREF(trigger);
+		return NULL;
+	}
+
+	Py_RETURN_NONE;
+}
+
+static PyMethodDef py_tevent_queue_methods[] = {
+	{ "stop", (PyCFunction)py_tevent_queue_stop, METH_NOARGS,
+		"S.stop()" },
+	{ "start", (PyCFunction)py_tevent_queue_start, METH_NOARGS,
+		"S.start()" },
+	{ "add", (PyCFunction)py_tevent_queue_add, METH_VARARGS,
+		"S.add(ctx, req, trigger, baton)" },
+	{ NULL },
+};
+
+static PyObject *py_tevent_context_wakeup_send(PyObject *self, PyObject *args)
+{
+	/* FIXME */
+
+	Py_RETURN_NONE;
+}
+
+static PyObject *py_tevent_context_loop_wait(TeventContext_Object *self)
+{
+	if (tevent_loop_wait(self->ev) != 0) {
+		PyErr_SetNone(PyExc_RuntimeError);
+		return NULL;
+	}
+	Py_RETURN_NONE;
+}
+
+static PyObject *py_tevent_context_loop_once(TeventContext_Object *self)
+{
+	if (tevent_loop_once(self->ev) != 0) {
+		PyErr_SetNone(PyExc_RuntimeError);
+		return NULL;
+	}
+	Py_RETURN_NONE;
+}
+
+#ifdef TEVENT_DEPRECATED
+static bool py_tevent_finished(PyObject *callback)
+{
+	PyObject *py_ret;
+	bool ret;
+
+	py_ret = PyObject_CallFunction(callback, "");
+	if (py_ret == NULL)
+		return true;
+	ret = PyObject_IsTrue(py_ret);
+	Py_DECREF(py_ret);
+	return ret;
+}
+
+static PyObject *py_tevent_context_loop_until(TeventContext_Object *self, PyObject *args)
+{
+	PyObject *callback;
+	if (!PyArg_ParseTuple(args, "O", &callback))
+		return NULL;
+
+	if (tevent_loop_until(self->ev, py_tevent_finished, callback) != 0) {
+		PyErr_SetNone(PyExc_RuntimeError);
+		return NULL;
+	}
+
+	if (PyErr_Occurred())
+		return NULL;
+
+	Py_RETURN_NONE;
+}
+#endif
+
+static void py_tevent_signal_handler(struct tevent_context *ev,
+					struct tevent_signal *se,
+					int signum,
+					int count,
+					void *siginfo,
+					void *private_data)
+{
+	PyObject *callback = (PyObject *)private_data, *ret;
+
+	ret = PyObject_CallFunction(callback, "ii", signum, count);
+	Py_XDECREF(ret);
+}
+
+static void py_tevent_signal_dealloc(TeventSignal_Object *self)
+{
+	talloc_free(self->signal);
+	PyObject_Del(self);
+}
+
+static PyTypeObject TeventSignal_Type = {
+	.tp_name = "tevent.Signal",
+	.tp_basicsize = sizeof(TeventSignal_Object),
+	.tp_dealloc = (destructor)py_tevent_signal_dealloc,
+	.tp_flags = Py_TPFLAGS_DEFAULT,
+};
+
+static PyObject *py_tevent_context_add_signal(TeventContext_Object *self, PyObject *args)
+{
+	int signum, sa_flags;
+	PyObject *handler;
+	struct tevent_signal *sig;
+	TeventSignal_Object *ret;
+
+	if (!PyArg_ParseTuple(args, "iiO", &signum, &sa_flags, &handler))
+		return NULL;
+
+	Py_INCREF(handler);
+	sig = tevent_add_signal(self->ev, NULL, signum, sa_flags,
+							py_tevent_signal_handler, handler);
+
+	ret = PyObject_New(TeventSignal_Object, &TeventSignal_Type);
+	if (ret == NULL) {
+		PyErr_NoMemory();
+		talloc_free(sig);
+		return NULL;
+	}
+
+	ret->signal = sig;
+
+	return (PyObject *)ret;
+}
+
+static void py_timer_handler(struct tevent_context *ev,
+				       struct tevent_timer *te,
+				       struct timeval current_time,
+				       void *private_data)
+{
+	PyObject *callback = private_data, *ret;
+	ret = PyObject_CallFunction(callback, "l", te);
+	Py_XDECREF(ret);
+}
+
+static PyObject *py_tevent_context_add_timer(TeventContext_Object *self, PyObject *args)
+{
+	TeventTimer_Object *ret;
+	struct timeval next_event;
+	struct tevent_timer *timer;
+	PyObject *handler;
+	if (!PyArg_ParseTuple(args, "lO", &next_event, &handler))
+		return NULL;
+
+	timer = tevent_add_timer(self->ev, NULL, next_event, py_timer_handler,
+							 handler);
+	if (timer == NULL) {
+		PyErr_SetNone(PyExc_RuntimeError);
+		return NULL;
+	}
+
+	ret = PyObject_New(TeventTimer_Object, &TeventTimer_Type);
+	if (ret == NULL) {
+		PyErr_NoMemory();
+		talloc_free(timer);
+		return NULL;
+	}
+	ret->timer = timer;
+
+	return (PyObject *)ret;
+}
+
+static void py_fd_handler(struct tevent_context *ev,
+				    struct tevent_fd *fde,
+				    uint16_t flags,
+				    void *private_data)
+{
+	PyObject *callback = private_data, *ret;
+
+	ret = PyObject_CallFunction(callback, "i", flags);
+	Py_XDECREF(ret);
+}
+
+static PyObject *py_tevent_context_add_fd(TeventContext_Object *self, PyObject *args)
+{
+	int fd, flags;
+	PyObject *handler;
+	struct tevent_fd *tfd;
+	TeventFd_Object *ret;
+
+	if (!PyArg_ParseTuple(args, "iiO", &fd, &flags, &handler))
+		return NULL;
+
+	tfd = tevent_add_fd(self->ev, NULL, fd, flags, py_fd_handler, handler);
+	if (tfd == NULL) {
+		PyErr_SetNone(PyExc_RuntimeError);
+		return NULL;
+	}
+
+	ret = PyObject_New(TeventFd_Object, &TeventFd_Type);
+	if (ret == NULL) {
+		talloc_free(tfd);
+		return NULL;
+	}
+	ret->fd = tfd;
+
+	return (PyObject *)ret;
+}
+
+#ifdef TEVENT_DEPRECATED
+static PyObject *py_tevent_context_set_allow_nesting(TeventContext_Object *self)
+{
+	tevent_loop_allow_nesting(self->ev);
+	Py_RETURN_NONE;
+}
+#endif
+
+static PyMethodDef py_tevent_context_methods[] = {
+	{ "reinitialise", (PyCFunction)py_tevent_context_reinitialise, METH_NOARGS,
+		"S.reinitialise()" },
+	{ "wakeup_send", (PyCFunction)py_tevent_context_wakeup_send, 
+		METH_VARARGS, "S.wakeup_send(wakeup_time) -> req" },
+	{ "loop_wait", (PyCFunction)py_tevent_context_loop_wait,
+		METH_NOARGS, "S.loop_wait()" },
+	{ "loop_once", (PyCFunction)py_tevent_context_loop_once,
+		METH_NOARGS, "S.loop_once()" },
+#ifdef TEVENT_DEPRECATED
+	{ "loop_until", (PyCFunction)py_tevent_context_loop_until,
+		METH_VARARGS, "S.loop_until(callback)" },
+#endif
+	{ "add_signal", (PyCFunction)py_tevent_context_add_signal,
+		METH_VARARGS, "S.add_signal(signum, sa_flags, handler) -> signal" },
+	{ "add_timer", (PyCFunction)py_tevent_context_add_timer,
+		METH_VARARGS, "S.add_timer(next_event, handler) -> timer" },
+	{ "add_fd", (PyCFunction)py_tevent_context_add_fd, 
+		METH_VARARGS, "S.add_fd(fd, flags, handler) -> fd" },
+#ifdef TEVENT_DEPRECATED
+	{ "allow_nesting", (PyCFunction)py_tevent_context_set_allow_nesting, 
+		METH_NOARGS, "Whether to allow nested tevent loops." },
+#endif
+	{ NULL },
+};
+
+static PyObject *py_tevent_req_wakeup_recv(PyObject *self)
+{
+	/* FIXME */
+	Py_RETURN_NONE;
+}
+
+static PyObject *py_tevent_req_received(PyObject *self)
+{
+	/* FIXME */
+	Py_RETURN_NONE;
+}
+
+static PyObject *py_tevent_req_is_error(PyObject *self)
+{
+	/* FIXME */
+	Py_RETURN_NONE;
+}
+
+static PyObject *py_tevent_req_poll(PyObject *self)
+{
+	/* FIXME */
+	Py_RETURN_NONE;
+}
+
+static PyObject *py_tevent_req_is_in_progress(PyObject *self)
+{
+	/* FIXME */
+	Py_RETURN_NONE;
+}
+
+static PyGetSetDef py_tevent_req_getsetters[] = {
+	{ "in_progress", (getter)py_tevent_req_is_in_progress, NULL,
+		"Whether the request is in progress" },
+	{ NULL }
+};
+
+static PyObject *py_tevent_req_post(PyObject *self, PyObject *args)
+{
+	/* FIXME */
+	Py_RETURN_NONE;
+}
+
+static PyObject *py_tevent_req_set_error(PyObject *self, PyObject *args)
+{
+	/* FIXME */
+	Py_RETURN_NONE;
+}
+
+static PyObject *py_tevent_req_done(PyObject *self)
+{
+	/* FIXME */
+	Py_RETURN_NONE;
+}
+
+static PyObject *py_tevent_req_notify_callback(PyObject *self)
+{
+	/* FIXME */
+	Py_RETURN_NONE;
+}
+
+static PyObject *py_tevent_req_set_endtime(PyObject *self, PyObject *args)
+{
+	/* FIXME */
+	Py_RETURN_NONE;
+}
+
+static PyObject *py_tevent_req_cancel(TeventReq_Object *self)
+{
+	if (!tevent_req_cancel(self->req)) {
+		PyErr_SetNone(PyExc_RuntimeError);
+		return NULL;
+	}
+	Py_RETURN_NONE;
+}
+
+static PyMethodDef py_tevent_req_methods[] = {
+	{ "wakeup_recv", (PyCFunction)py_tevent_req_wakeup_recv, METH_NOARGS,
+		"Wakeup received" },
+	{ "received", (PyCFunction)py_tevent_req_received, METH_NOARGS,
+		"Receive finished" },
+	{ "is_error", (PyCFunction)py_tevent_req_is_error, METH_NOARGS,
+		"is_error() -> (error, state)" },
+	{ "poll", (PyCFunction)py_tevent_req_poll, METH_VARARGS,
+		"poll(ctx)" },
+	{ "post", (PyCFunction)py_tevent_req_post, METH_VARARGS,
+		"post(ctx) -> req" },
+	{ "set_error", (PyCFunction)py_tevent_req_set_error, METH_VARARGS,
+		"set_error(error)" },
+	{ "done", (PyCFunction)py_tevent_req_done, METH_NOARGS,
+		"done()" },
+	{ "notify_callback", (PyCFunction)py_tevent_req_notify_callback,
+		METH_NOARGS, "notify_callback()" },
+	{ "set_endtime", (PyCFunction)py_tevent_req_set_endtime,
+		METH_VARARGS, "set_endtime(ctx, endtime)" },
+	{ "cancel", (PyCFunction)py_tevent_req_cancel,
+		METH_NOARGS, "cancel()" },
+	{ NULL }
+};
+
+static void py_tevent_req_dealloc(TeventReq_Object *self)
+{
+	talloc_free(self->req);
+	PyObject_DEL(self);
+}
+
+static PyTypeObject TeventReq_Type = {
+	.tp_name = "tevent.Request",
+	.tp_basicsize = sizeof(TeventReq_Object),
+	.tp_methods = py_tevent_req_methods,
+	.tp_dealloc = (destructor)py_tevent_req_dealloc,
+	.tp_getset = py_tevent_req_getsetters,
+	/* FIXME: .tp_new = py_tevent_req_new, */
+};
+
+static PyObject *py_tevent_queue_get_length(TeventQueue_Object *self)
+{
+	return PyInt_FromLong(tevent_queue_length(self->queue));
+}
+
+static PyGetSetDef py_tevent_queue_getsetters[] = {
+	{ "length", (getter)py_tevent_queue_get_length,
+		NULL, "The number of elements in the queue." },
+	{ NULL },
+};
+
+static void py_tevent_queue_dealloc(TeventQueue_Object *self)
+{
+	talloc_free(self->queue);
+	PyObject_Del(self);
+}
+
+static PyTypeObject TeventQueue_Type = {
+	.tp_name = "tevent.Queue",
+	.tp_basicsize = sizeof(TeventQueue_Object),
+	.tp_dealloc = (destructor)py_tevent_queue_dealloc,
+	.tp_flags = Py_TPFLAGS_DEFAULT,
+	.tp_getset = py_tevent_queue_getsetters,
+	.tp_methods = py_tevent_queue_methods,
+};
+
+static PyObject *py_tevent_context_signal_support(PyObject *_self)
+{
+	TeventContext_Object *self = (TeventContext_Object *)_self;
+	return PyBool_FromLong(tevent_signal_support(self->ev));
+}
+
+static PyGetSetDef py_tevent_context_getsetters[] = {
+	{ "signal_support", (getter)py_tevent_context_signal_support,
+		NULL, "if this platform and tevent context support signal handling" },
+	{ NULL }
+};
+
+static void py_tevent_context_dealloc(TeventContext_Object *self)
+{
+	talloc_free(self->ev);
+	PyObject_Del(self);
+}
+
+static PyObject *py_tevent_context_new(PyTypeObject *type, PyObject *args, PyObject *kwargs)
+{
+	const char * const kwnames[] = { "name", NULL };
+	char *name = NULL;
+	struct tevent_context *ev;
+	TeventContext_Object *ret;
+
+	if (!PyArg_ParseTupleAndKeywords(args, kwargs, "|s", kwnames, &name))
+		return NULL;
+
+	if (name == NULL) {
+		ev = tevent_context_init(NULL);
+	} else {
+		ev = tevent_context_init_byname(NULL, name);
+	}
+
+	if (ev == NULL) {
+		PyErr_SetNone(PyExc_RuntimeError);
+		return NULL;
+	}
+
+	ret = PyObject_New(TeventContext_Object, type);
+	if (ret == NULL) {
+		PyErr_NoMemory();
+		talloc_free(ev);
+		return NULL;
+	}
+
+	ret->ev = ev;
+	return (PyObject *)ret;
+}
+
+static PyTypeObject TeventContext_Type = {
+	.tp_name = "tevent.Context",
+	.tp_new = py_tevent_context_new,
+	.tp_basicsize = sizeof(TeventContext_Object),
+	.tp_dealloc = (destructor)py_tevent_context_dealloc,
+	.tp_methods = py_tevent_context_methods,
+	.tp_getset = py_tevent_context_getsetters,
+	.tp_flags = Py_TPFLAGS_DEFAULT,
+};
+
+static PyObject *py_set_default_backend(PyObject *self, PyObject *args)
+{
+	char *backend_name;
+	if (!PyArg_ParseTuple(args, "s", &backend_name))
+		return NULL;
+
+	tevent_set_default_backend(backend_name);
+
+	Py_RETURN_NONE;
+}
+
+static PyObject *py_backend_list(PyObject *self)
+{
+	PyObject *ret;
+	int i;
+	const char **backends;
+
+	ret = PyList_New(0);
+	if (ret == NULL) {
+		return NULL;
+	}
+
+	backends = tevent_backend_list(NULL);
+	if (backends == NULL) {
+		PyErr_SetNone(PyExc_RuntimeError);
+		Py_DECREF(ret);
+		return NULL;
+	}
+	for (i = 0; backends[i]; i++) {
+		PyList_Append(ret, PyString_FromString(backends[i]));
+	}
+
+	talloc_free(backends);
+
+	return ret;
+}
+
+static PyMethodDef tevent_methods[] = {
+	{ "register_backend", (PyCFunction)py_register_backend, METH_VARARGS,
+		"register_backend(backend)" },
+	{ "set_default_backend", (PyCFunction)py_set_default_backend, 
+		METH_VARARGS, "set_default_backend(backend)" },
+	{ "backend_list", (PyCFunction)py_backend_list, 
+		METH_NOARGS, "backend_list() -> list" },
+	{ NULL },
+};
+
+void init_tevent(void)
+{
+	PyObject *m;
+
+	if (PyType_Ready(&TeventContext_Type) < 0)
+		return;
+
+	if (PyType_Ready(&TeventQueue_Type) < 0)
+		return;
+
+	if (PyType_Ready(&TeventReq_Type) < 0)
+		return;
+
+	if (PyType_Ready(&TeventSignal_Type) < 0)
+		return;
+
+	if (PyType_Ready(&TeventTimer_Type) < 0)
+		return;
+
+	if (PyType_Ready(&TeventFd_Type) < 0)
+		return;
+
+	m = Py_InitModule3("_tevent", tevent_methods, "Tevent integration for twisted.");
+	if (m == NULL)
+		return;
+
+	Py_INCREF(&TeventContext_Type);
+	PyModule_AddObject(m, "Context", (PyObject *)&TeventContext_Type);
+
+	Py_INCREF(&TeventQueue_Type);
+	PyModule_AddObject(m, "Queue", (PyObject *)&TeventQueue_Type);
+
+	Py_INCREF(&TeventReq_Type);
+	PyModule_AddObject(m, "Request", (PyObject *)&TeventReq_Type);
+
+	Py_INCREF(&TeventSignal_Type);
+	PyModule_AddObject(m, "Signal", (PyObject *)&TeventSignal_Type);
+
+	Py_INCREF(&TeventTimer_Type);
+	PyModule_AddObject(m, "Timer", (PyObject *)&TeventTimer_Type);
+
+	Py_INCREF(&TeventFd_Type);
+	PyModule_AddObject(m, "Fd", (PyObject *)&TeventFd_Type);
+
+	PyModule_AddObject(m, "__version__", PyString_FromString(PACKAGE_VERSION));
+}
diff --git a/ctdb/lib/tevent/release-script.sh b/ctdb/lib/tevent/release-script.sh
new file mode 100755
index 0000000000..077f562967
--- /dev/null
+++ b/ctdb/lib/tevent/release-script.sh
@@ -0,0 +1,48 @@
+#!/bin/bash
+
+if [ "$1" = "" ]; then
+    echo "Please provide version string, eg: 1.2.0"
+    exit 1
+fi
+
+if [ ! -d "lib/tevent" ]; then
+    echo "Run this script from the samba base directory."
+    exit 1
+fi
+
+git clean -f -x -d lib/tevent
+git clean -f -x -d lib/replace
+
+curbranch=`git-branch |grep "^*" | tr -d "* "`
+
+version=$1
+strver=`echo ${version} | tr "." "-"`
+
+# Checkout the release tag
+git branch -f tevent-release-script-${strver} tevent-${strver}
+if [ ! "$?" = "0" ];  then
+    echo "Unable to checkout tevent-${strver} release"
+    exit 1
+fi
+
+git checkout tevent-release-script-${strver}
+
+# Test configure agrees with us
+confver=`grep "^AC_INIT" lib/tevent/configure.ac | tr -d "AC_INIT(tevent, " | tr -d ")"`
+if [ ! "$confver" = "$version" ]; then
+    echo "Wrong version, requested release for ${version}, found ${confver}"
+    exit 1
+fi
+
+# Now build tarball
+cp -a lib/tevent tevent-${version}
+cp -a lib/replace tevent-${version}/libreplace
+pushd tevent-${version}
+./autogen.sh
+popd
+tar cvzf tevent-${version}.tar.gz tevent-${version}
+rm -fr tevent-${version}
+
+#Clean up
+git checkout $curbranch
+git branch -d tevent-release-script-${strver}
diff --git a/ctdb/lib/tevent/testsuite.c b/ctdb/lib/tevent/testsuite.c
new file mode 100644
index 0000000000..8e3f4af51c
--- /dev/null
+++ b/ctdb/lib/tevent/testsuite.c
@@ -0,0 +1,833 @@
+/* 
+   Unix SMB/CIFS implementation.
+
+   testing of the events subsystem
+
+   Copyright (C) Stefan Metzmacher 2006-2009
+   Copyright (C) Jeremy Allison    2013
+
+     ** NOTE! The following LGPL license applies to the tevent
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "includes.h"
+#include "lib/tevent/tevent.h"
+#include "system/filesys.h"
+#include "system/select.h"
+#include "system/network.h"
+#include "torture/torture.h"
+#ifdef HAVE_PTHREAD
+#include <pthread.h>
+#include <assert.h>
+#endif
+
+static int fde_count;
+
+static void fde_handler_read(struct tevent_context *ev_ctx, struct tevent_fd *f,
+			uint16_t flags, void *private_data)
+{
+	int *fd = (int *)private_data;
+	char c;
+#ifdef SA_SIGINFO
+	kill(getpid(), SIGUSR1);
+#endif
+	kill(getpid(), SIGALRM);
+
+	read(fd[0], &c, 1);
+	fde_count++;
+}
+
+static void fde_handler_write(struct tevent_context *ev_ctx, struct tevent_fd *f,
+			uint16_t flags, void *private_data)
+{
+	int *fd = (int *)private_data;
+	char c = 0;
+	write(fd[1], &c, 1);
+}
+
+
+/* These should never fire... */
+static void fde_handler_read_1(struct tevent_context *ev_ctx, struct tevent_fd *f,
+			uint16_t flags, void *private_data)
+{
+	struct torture_context *test = (struct torture_context *)private_data;
+	torture_comment(test, "fde_handler_read_1 should never fire !\n");
+	abort();
+}
+
+/* These should never fire... */
+static void fde_handler_write_1(struct tevent_context *ev_ctx, struct tevent_fd *f,
+			uint16_t flags, void *private_data)
+{
+	struct torture_context *test = (struct torture_context *)private_data;
+	torture_comment(test, "fde_handler_write_1 should never fire !\n");
+	abort();
+}
+
+static void finished_handler(struct tevent_context *ev_ctx, struct tevent_timer *te,
+			     struct timeval tval, void *private_data)
+{
+	int *finished = (int *)private_data;
+	(*finished) = 1;
+}
+
+static void count_handler(struct tevent_context *ev_ctx, struct tevent_signal *te,
+			  int signum, int count, void *info, void *private_data)
+{
+	int *countp = (int *)private_data;
+	(*countp) += count;
+}
+
+static bool test_event_context(struct torture_context *test,
+			       const void *test_data)
+{
+	struct tevent_context *ev_ctx;
+	int fd[2] = { -1, -1 };
+	const char *backend = (const char *)test_data;
+	int alarm_count=0, info_count=0;
+	struct tevent_fd *fde_read;
+	struct tevent_fd *fde_read_1;
+	struct tevent_fd *fde_write;
+	struct tevent_fd *fde_write_1;
+#ifdef SA_RESTART
+	struct tevent_signal *se1 = NULL;
+#endif
+#ifdef SA_RESETHAND
+	struct tevent_signal *se2 = NULL;
+#endif
+#ifdef SA_SIGINFO
+	struct tevent_signal *se3 = NULL;
+#endif
+	int finished=0;
+	struct timeval t;
+
+	ev_ctx = tevent_context_init_byname(test, backend);
+	if (ev_ctx == NULL) {
+		torture_comment(test, "event backend '%s' not supported\n", backend);
+		return true;
+	}
+
+	torture_comment(test, "backend '%s' - %s\n",
+			backend, __FUNCTION__);
+
+	/* reset globals */
+	fde_count = 0;
+
+	/* create a pipe */
+	pipe(fd);
+
+	fde_read = tevent_add_fd(ev_ctx, ev_ctx, fd[0], TEVENT_FD_READ,
+			    fde_handler_read, fd);
+	fde_write_1 = tevent_add_fd(ev_ctx, ev_ctx, fd[0], TEVENT_FD_WRITE,
+			    fde_handler_write_1, test);
+
+	fde_write = tevent_add_fd(ev_ctx, ev_ctx, fd[1], TEVENT_FD_WRITE,
+			    fde_handler_write, fd);
+	fde_read_1 = tevent_add_fd(ev_ctx, ev_ctx, fd[1], TEVENT_FD_READ,
+			    fde_handler_read_1, test);
+
+	tevent_fd_set_auto_close(fde_read);
+	tevent_fd_set_auto_close(fde_write);
+
+	tevent_add_timer(ev_ctx, ev_ctx, timeval_current_ofs(2,0),
+			 finished_handler, &finished);
+
+#ifdef SA_RESTART
+	se1 = tevent_add_signal(ev_ctx, ev_ctx, SIGALRM, SA_RESTART, count_handler, &alarm_count);
+	torture_assert(test, se1 != NULL, "failed to setup se1");
+#endif
+#ifdef SA_RESETHAND
+	se2 = tevent_add_signal(ev_ctx, ev_ctx, SIGALRM, SA_RESETHAND, count_handler, &alarm_count);
+	torture_assert(test, se2 != NULL, "failed to setup se2");
+#endif
+#ifdef SA_SIGINFO
+	se3 = tevent_add_signal(ev_ctx, ev_ctx, SIGUSR1, SA_SIGINFO, count_handler, &info_count);
+	torture_assert(test, se3 != NULL, "failed to setup se3");
+#endif
+
+	t = timeval_current();
+	while (!finished) {
+		errno = 0;
+		if (tevent_loop_once(ev_ctx) == -1) {
+			talloc_free(ev_ctx);
+			torture_fail(test, talloc_asprintf(test, "Failed event loop %s\n", strerror(errno)));
+		}
+	}
+
+	talloc_free(fde_read);
+	talloc_free(fde_write);
+	talloc_free(fde_read_1);
+	talloc_free(fde_write_1);
+
+	while (alarm_count < fde_count+1) {
+		if (tevent_loop_once(ev_ctx) == -1) {
+			break;
+		}
+	}
+
+	torture_comment(test, "Got %.2f pipe events/sec\n", fde_count/timeval_elapsed(&t));
+
+#ifdef SA_RESTART
+	talloc_free(se1);
+#endif
+
+	torture_assert_int_equal(test, alarm_count, 1+fde_count, "alarm count mismatch");
+
+#ifdef SA_RESETHAND
+	/*
+	 * we do not call talloc_free(se2)
+	 * because it is already gone,
+	 * after triggering the event handler.
+	 */
+#endif
+
+#ifdef SA_SIGINFO
+	talloc_free(se3);
+	torture_assert_int_equal(test, info_count, fde_count, "info count mismatch");
+#endif
+
+	talloc_free(ev_ctx);
+
+	return true;
+}
+
+struct test_event_fd1_state {
+	struct torture_context *tctx;
+	const char *backend;
+	struct tevent_context *ev;
+	int sock[2];
+	struct tevent_timer *te;
+	struct tevent_fd *fde0;
+	struct tevent_fd *fde1;
+	bool got_write;
+	bool got_read;
+	bool drain;
+	bool drain_done;
+	unsigned loop_count;
+	bool finished;
+	const char *error;
+};
+
+static void test_event_fd1_fde_handler(struct tevent_context *ev_ctx,
+				       struct tevent_fd *fde,
+				       uint16_t flags,
+				       void *private_data)
+{
+	struct test_event_fd1_state *state =
+		(struct test_event_fd1_state *)private_data;
+
+	if (state->drain_done) {
+		state->finished = true;
+		state->error = __location__;
+		return;
+	}
+
+	if (state->drain) {
+		ssize_t ret;
+		uint8_t c = 0;
+
+		if (!(flags & TEVENT_FD_READ)) {
+			state->finished = true;
+			state->error = __location__;
+			return;
+		}
+
+		ret = read(state->sock[0], &c, 1);
+		if (ret == 1) {
+			return;
+		}
+
+		/*
+		 * end of test...
+		 */
+		tevent_fd_set_flags(fde, 0);
+		state->drain_done = true;
+		return;
+	}
+
+	if (!state->got_write) {
+		uint8_t c = 0;
+
+		if (flags != TEVENT_FD_WRITE) {
+			state->finished = true;
+			state->error = __location__;
+			return;
+		}
+		state->got_write = true;
+
+		/*
+		 * we write to the other socket...
+		 */
+		write(state->sock[1], &c, 1);
+		TEVENT_FD_NOT_WRITEABLE(fde);
+		TEVENT_FD_READABLE(fde);
+		return;
+	}
+
+	if (!state->got_read) {
+		if (flags != TEVENT_FD_READ) {
+			state->finished = true;
+			state->error = __location__;
+			return;
+		}
+		state->got_read = true;
+
+		TEVENT_FD_NOT_READABLE(fde);
+		return;
+	}
+
+	state->finished = true;
+	state->error = __location__;
+	return;
+}
+
+static void test_event_fd1_finished(struct tevent_context *ev_ctx,
+				    struct tevent_timer *te,
+				    struct timeval tval,
+				    void *private_data)
+{
+	struct test_event_fd1_state *state =
+		(struct test_event_fd1_state *)private_data;
+
+	if (state->drain_done) {
+		state->finished = true;
+		return;
+	}
+
+	if (!state->got_write) {
+		state->finished = true;
+		state->error = __location__;
+		return;
+	}
+
+	if (!state->got_read) {
+		state->finished = true;
+		state->error = __location__;
+		return;
+	}
+
+	state->loop_count++;
+	if (state->loop_count > 3) {
+		state->finished = true;
+		state->error = __location__;
+		return;
+	}
+
+	state->got_write = false;
+	state->got_read = false;
+
+	tevent_fd_set_flags(state->fde0, TEVENT_FD_WRITE);
+
+	if (state->loop_count > 2) {
+		state->drain = true;
+		TALLOC_FREE(state->fde1);
+		TEVENT_FD_READABLE(state->fde0);
+	}
+
+	state->te = tevent_add_timer(state->ev, state->ev,
+				    timeval_current_ofs(0,2000),
+				    test_event_fd1_finished, state);
+}
+
+static bool test_event_fd1(struct torture_context *tctx,
+			   const void *test_data)
+{
+	struct test_event_fd1_state state;
+
+	ZERO_STRUCT(state);
+	state.tctx = tctx;
+	state.backend = (const char *)test_data;
+
+	state.ev = tevent_context_init_byname(tctx, state.backend);
+	if (state.ev == NULL) {
+		torture_skip(tctx, talloc_asprintf(tctx,
+			     "event backend '%s' not supported\n",
+			     state.backend));
+		return true;
+	}
+
+	tevent_set_debug_stderr(state.ev);
+	torture_comment(tctx, "backend '%s' - %s\n",
+			state.backend, __FUNCTION__);
+
+	/*
+	 * This tests the following:
+	 *
+	 * It monitors the state of state.sock[0]
+	 * with tevent_fd, but we never read/write on state.sock[0]
+	 * while state.sock[1] * is only used to write a few bytes.
+	 *
+	 * We have a loop:
+	 *   - we wait only for TEVENT_FD_WRITE on state.sock[0]
+	 *   - we write 1 byte to state.sock[1]
+	 *   - we wait only for TEVENT_FD_READ on state.sock[0]
+	 *   - we disable events on state.sock[0]
+	 *   - the timer event restarts the loop
+	 * Then we close state.sock[1]
+	 * We have a loop:
+	 *   - we wait for TEVENT_FD_READ/WRITE on state.sock[0]
+	 *   - we try to read 1 byte
+	 *   - if the read gets an error of returns 0
+	 *     we disable the event handler
+	 *   - the timer finishes the test
+	 */
+	state.sock[0] = -1;
+	state.sock[1] = -1;
+	socketpair(AF_UNIX, SOCK_STREAM, 0, state.sock);
+
+	state.te = tevent_add_timer(state.ev, state.ev,
+				    timeval_current_ofs(0,1000),
+				    test_event_fd1_finished, &state);
+	state.fde0 = tevent_add_fd(state.ev, state.ev,
+				   state.sock[0], TEVENT_FD_WRITE,
+				   test_event_fd1_fde_handler, &state);
+	/* state.fde1 is only used to auto close */
+	state.fde1 = tevent_add_fd(state.ev, state.ev,
+				   state.sock[1], 0,
+				   test_event_fd1_fde_handler, &state);
+
+	tevent_fd_set_auto_close(state.fde0);
+	tevent_fd_set_auto_close(state.fde1);
+
+	while (!state.finished) {
+		errno = 0;
+		if (tevent_loop_once(state.ev) == -1) {
+			talloc_free(state.ev);
+			torture_fail(tctx, talloc_asprintf(tctx,
+				     "Failed event loop %s\n",
+				     strerror(errno)));
+		}
+	}
+
+	talloc_free(state.ev);
+
+	torture_assert(tctx, state.error == NULL, talloc_asprintf(tctx,
+		       "%s", state.error));
+
+	return true;
+}
+
+struct test_event_fd2_state {
+	struct torture_context *tctx;
+	const char *backend;
+	struct tevent_context *ev;
+	struct tevent_timer *te;
+	struct test_event_fd2_sock {
+		struct test_event_fd2_state *state;
+		int fd;
+		struct tevent_fd *fde;
+		size_t num_written;
+		size_t num_read;
+		bool got_full;
+	} sock0, sock1;
+	bool finished;
+	const char *error;
+};
+
+static void test_event_fd2_sock_handler(struct tevent_context *ev_ctx,
+					struct tevent_fd *fde,
+					uint16_t flags,
+					void *private_data)
+{
+	struct test_event_fd2_sock *cur_sock =
+		(struct test_event_fd2_sock *)private_data;
+	struct test_event_fd2_state *state = cur_sock->state;
+	struct test_event_fd2_sock *oth_sock = NULL;
+	uint8_t v = 0, c;
+	ssize_t ret;
+
+	if (cur_sock == &state->sock0) {
+		oth_sock = &state->sock1;
+	} else {
+		oth_sock = &state->sock0;
+	}
+
+	if (oth_sock->num_written == 1) {
+		if (flags != (TEVENT_FD_READ | TEVENT_FD_WRITE)) {
+			state->finished = true;
+			state->error = __location__;
+			return;
+		}
+	}
+
+	if (cur_sock->num_read == oth_sock->num_written) {
+		state->finished = true;
+		state->error = __location__;
+		return;
+	}
+
+	if (!(flags & TEVENT_FD_READ)) {
+		state->finished = true;
+		state->error = __location__;
+		return;
+	}
+
+	if (oth_sock->num_read > 0) {
+		/*
+		 * There should be room to write a byte again
+		 */
+		if (!(flags & TEVENT_FD_WRITE)) {
+			state->finished = true;
+			state->error = __location__;
+			return;
+		}
+	}
+
+	if ((flags & TEVENT_FD_WRITE) && !cur_sock->got_full) {
+		v = (uint8_t)cur_sock->num_written;
+		ret = write(cur_sock->fd, &v, 1);
+		if (ret != 1) {
+			state->finished = true;
+			state->error = __location__;
+			return;
+		}
+		cur_sock->num_written++;
+		if (cur_sock->num_written > 0x80000000) {
+			state->finished = true;
+			state->error = __location__;
+			return;
+		}
+		return;
+	}
+
+	if (!cur_sock->got_full) {
+		cur_sock->got_full = true;
+
+		if (!oth_sock->got_full) {
+			/*
+			 * cur_sock is full,
+			 * lets wait for oth_sock
+			 * to be filled
+			 */
+			tevent_fd_set_flags(cur_sock->fde, 0);
+			return;
+		}
+
+		/*
+		 * oth_sock waited for cur_sock,
+		 * lets restart it
+		 */
+		tevent_fd_set_flags(oth_sock->fde,
+				    TEVENT_FD_READ|TEVENT_FD_WRITE);
+	}
+
+	ret = read(cur_sock->fd, &v, 1);
+	if (ret != 1) {
+		state->finished = true;
+		state->error = __location__;
+		return;
+	}
+	c = (uint8_t)cur_sock->num_read;
+	if (c != v) {
+		state->finished = true;
+		state->error = __location__;
+		return;
+	}
+	cur_sock->num_read++;
+
+	if (cur_sock->num_read < oth_sock->num_written) {
+		/* there is more to read */
+		return;
+	}
+	/*
+	 * we read everything, we need to remove TEVENT_FD_WRITE
+	 * to avoid spinning
+	 */
+	TEVENT_FD_NOT_WRITEABLE(cur_sock->fde);
+
+	if (oth_sock->num_read == cur_sock->num_written) {
+		/*
+		 * both directions are finished
+		 */
+		state->finished = true;
+	}
+
+	return;
+}
+
+static void test_event_fd2_finished(struct tevent_context *ev_ctx,
+				    struct tevent_timer *te,
+				    struct timeval tval,
+				    void *private_data)
+{
+	struct test_event_fd2_state *state =
+		(struct test_event_fd2_state *)private_data;
+
+	/*
+	 * this should never be triggered
+	 */
+	state->finished = true;
+	state->error = __location__;
+}
+
+static bool test_event_fd2(struct torture_context *tctx,
+			   const void *test_data)
+{
+	struct test_event_fd2_state state;
+	int sock[2];
+	uint8_t c = 0;
+
+	ZERO_STRUCT(state);
+	state.tctx = tctx;
+	state.backend = (const char *)test_data;
+
+	state.ev = tevent_context_init_byname(tctx, state.backend);
+	if (state.ev == NULL) {
+		torture_skip(tctx, talloc_asprintf(tctx,
+			     "event backend '%s' not supported\n",
+			     state.backend));
+		return true;
+	}
+
+	tevent_set_debug_stderr(state.ev);
+	torture_comment(tctx, "backend '%s' - %s\n",
+			state.backend, __FUNCTION__);
+
+	/*
+	 * This tests the following
+	 *
+	 * - We write 1 byte to each socket
+	 * - We wait for TEVENT_FD_READ/WRITE on both sockets
+	 * - When we get TEVENT_FD_WRITE we write 1 byte
+	 *   until both socket buffers are full, which
+	 *   means both sockets only get TEVENT_FD_READ.
+	 * - Then we read 1 byte until we have consumed
+	 *   all bytes the other end has written.
+	 */
+	sock[0] = -1;
+	sock[1] = -1;
+	socketpair(AF_UNIX, SOCK_STREAM, 0, sock);
+
+	/*
+	 * the timer should never expire
+	 */
+	state.te = tevent_add_timer(state.ev, state.ev,
+				    timeval_current_ofs(600, 0),
+				    test_event_fd2_finished, &state);
+	state.sock0.state = &state;
+	state.sock0.fd = sock[0];
+	state.sock0.fde = tevent_add_fd(state.ev, state.ev,
+					state.sock0.fd,
+					TEVENT_FD_READ | TEVENT_FD_WRITE,
+					test_event_fd2_sock_handler,
+					&state.sock0);
+	state.sock1.state = &state;
+	state.sock1.fd = sock[1];
+	state.sock1.fde = tevent_add_fd(state.ev, state.ev,
+					state.sock1.fd,
+					TEVENT_FD_READ | TEVENT_FD_WRITE,
+					test_event_fd2_sock_handler,
+					&state.sock1);
+
+	tevent_fd_set_auto_close(state.sock0.fde);
+	tevent_fd_set_auto_close(state.sock1.fde);
+
+	write(state.sock0.fd, &c, 1);
+	state.sock0.num_written++;
+	write(state.sock1.fd, &c, 1);
+	state.sock1.num_written++;
+
+	while (!state.finished) {
+		errno = 0;
+		if (tevent_loop_once(state.ev) == -1) {
+			talloc_free(state.ev);
+			torture_fail(tctx, talloc_asprintf(tctx,
+				     "Failed event loop %s\n",
+				     strerror(errno)));
+		}
+	}
+
+	talloc_free(state.ev);
+
+	torture_assert(tctx, state.error == NULL, talloc_asprintf(tctx,
+		       "%s", state.error));
+
+	return true;
+}
+
+#ifdef HAVE_PTHREAD
+
+static pthread_mutex_t threaded_mutex = PTHREAD_MUTEX_INITIALIZER;
+static bool do_shutdown = false;
+
+static void test_event_threaded_lock(void)
+{
+	int ret;
+	ret = pthread_mutex_lock(&threaded_mutex);
+	assert(ret == 0);
+}
+
+static void test_event_threaded_unlock(void)
+{
+	int ret;
+	ret = pthread_mutex_unlock(&threaded_mutex);
+	assert(ret == 0);
+}
+
+static void test_event_threaded_trace(enum tevent_trace_point point,
+				      void *private_data)
+{
+	switch (point) {
+	case TEVENT_TRACE_BEFORE_WAIT:
+		test_event_threaded_unlock();
+		break;
+	case TEVENT_TRACE_AFTER_WAIT:
+		test_event_threaded_lock();
+		break;
+	case TEVENT_TRACE_BEFORE_LOOP_ONCE:
+	case TEVENT_TRACE_AFTER_LOOP_ONCE:
+		break;
+	}
+}
+
+static void test_event_threaded_timer(struct tevent_context *ev,
+				      struct tevent_timer *te,
+				      struct timeval current_time,
+				      void *private_data)
+{
+	return;
+}
+
+static void *test_event_poll_thread(void *private_data)
+{
+	struct tevent_context *ev = (struct tevent_context *)private_data;
+
+	test_event_threaded_lock();
+
+	while (true) {
+		int ret;
+		ret = tevent_loop_once(ev);
+		assert(ret == 0);
+		if (do_shutdown) {
+			test_event_threaded_unlock();
+			return NULL;
+		}
+	}
+
+}
+
+static void test_event_threaded_read_handler(struct tevent_context *ev,
+					     struct tevent_fd *fde,
+					     uint16_t flags,
+					     void *private_data)
+{
+	int *pfd = (int *)private_data;
+	char c;
+	ssize_t nread;
+
+	if ((flags & TEVENT_FD_READ) == 0) {
+		return;
+	}
+
+	do {
+		nread = read(*pfd, &c, 1);
+	} while ((nread == -1) && (errno == EINTR));
+
+	assert(nread == 1);
+}
+
+static bool test_event_context_threaded(struct torture_context *test,
+					const void *test_data)
+{
+	struct tevent_context *ev;
+	struct tevent_timer *te;
+	struct tevent_fd *fde;
+	pthread_t poll_thread;
+	int fds[2];
+	int ret;
+	char c = 0;
+
+	ev = tevent_context_init_byname(test, "poll_mt");
+	torture_assert(test, ev != NULL, "poll_mt not supported");
+
+	tevent_set_trace_callback(ev, test_event_threaded_trace, NULL);
+
+	te = tevent_add_timer(ev, ev, timeval_current_ofs(5, 0),
+			      test_event_threaded_timer, NULL);
+	torture_assert(test, te != NULL, "Could not add timer");
+
+	ret = pthread_create(&poll_thread, NULL, test_event_poll_thread, ev);
+	torture_assert(test, ret == 0, "Could not create poll thread");
+
+	ret = pipe(fds);
+	torture_assert(test, ret == 0, "Could not create pipe");
+
+	poll(NULL, 0, 100);
+
+	test_event_threaded_lock();
+
+	fde = tevent_add_fd(ev, ev, fds[0], TEVENT_FD_READ,
+			    test_event_threaded_read_handler, &fds[0]);
+	torture_assert(test, fde != NULL, "Could not add fd event");
+
+	test_event_threaded_unlock();
+
+	poll(NULL, 0, 100);
+
+	write(fds[1], &c, 1);
+
+	poll(NULL, 0, 100);
+
+	test_event_threaded_lock();
+	do_shutdown = true;
+	test_event_threaded_unlock();
+
+	write(fds[1], &c, 1);
+
+	ret = pthread_join(poll_thread, NULL);
+	torture_assert(test, ret == 0, "pthread_join failed");
+
+	return true;
+}
+
+#endif
+
+struct torture_suite *torture_local_event(TALLOC_CTX *mem_ctx)
+{
+	struct torture_suite *suite = torture_suite_create(mem_ctx, "event");
+	const char **list = tevent_backend_list(suite);
+	int i;
+
+	for (i=0;list && list[i];i++) {
+		struct torture_suite *backend_suite;
+
+		backend_suite = torture_suite_create(mem_ctx, list[i]);
+
+		torture_suite_add_simple_tcase_const(backend_suite,
+					       "context",
+					       test_event_context,
+					       (const void *)list[i]);
+		torture_suite_add_simple_tcase_const(backend_suite,
+					       "fd1",
+					       test_event_fd1,
+					       (const void *)list[i]);
+		torture_suite_add_simple_tcase_const(backend_suite,
+					       "fd2",
+					       test_event_fd2,
+					       (const void *)list[i]);
+
+		torture_suite_add_suite(suite, backend_suite);
+	}
+
+#ifdef HAVE_PTHREAD
+	torture_suite_add_simple_tcase_const(suite, "threaded_poll_mt",
+					     test_event_context_threaded,
+					     NULL);
+#endif
+
+	return suite;
+}
diff --git a/ctdb/lib/tevent/tevent.c b/ctdb/lib/tevent/tevent.c
new file mode 100644
index 0000000000..be0afd453b
--- /dev/null
+++ b/ctdb/lib/tevent/tevent.c
@@ -0,0 +1,669 @@
+/* 
+   Unix SMB/CIFS implementation.
+   main select loop and event handling
+   Copyright (C) Andrew Tridgell 2003
+   Copyright (C) Stefan Metzmacher 2009
+
+     ** NOTE! The following LGPL license applies to the tevent
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+/*
+  PLEASE READ THIS BEFORE MODIFYING!
+
+  This module is a general abstraction for the main select loop and
+  event handling. Do not ever put any localised hacks in here, instead
+  register one of the possible event types and implement that event
+  somewhere else.
+
+  There are 2 types of event handling that are handled in this module:
+
+  1) a file descriptor becoming readable or writeable. This is mostly
+     used for network sockets, but can be used for any type of file
+     descriptor. You may only register one handler for each file
+     descriptor/io combination or you will get unpredictable results
+     (this means that you can have a handler for read events, and a
+     separate handler for write events, but not two handlers that are
+     both handling read events)
+
+  2) a timed event. You can register an event that happens at a
+     specific time.  You can register as many of these as you
+     like. They are single shot - add a new timed event in the event
+     handler to get another event.
+
+  To setup a set of events you first need to create a event_context
+  structure using the function tevent_context_init(); This returns a
+  'struct tevent_context' that you use in all subsequent calls.
+
+  After that you can add/remove events that you are interested in
+  using tevent_add_*() and talloc_free()
+
+  Finally, you call tevent_loop_wait_once() to block waiting for one of the
+  events to occor or tevent_loop_wait() which will loop
+  forever.
+
+*/
+#include "replace.h"
+#include "system/filesys.h"
+#define TEVENT_DEPRECATED 1
+#include "tevent.h"
+#include "tevent_internal.h"
+#include "tevent_util.h"
+
+struct tevent_ops_list {
+	struct tevent_ops_list *next, *prev;
+	const char *name;
+	const struct tevent_ops *ops;
+};
+
+/* list of registered event backends */
+static struct tevent_ops_list *tevent_backends = NULL;
+static char *tevent_default_backend = NULL;
+
+/*
+  register an events backend
+*/
+bool tevent_register_backend(const char *name, const struct tevent_ops *ops)
+{
+	struct tevent_ops_list *e;
+
+	for (e = tevent_backends; e != NULL; e = e->next) {
+		if (0 == strcmp(e->name, name)) {
+			/* already registered, skip it */
+			return true;
+		}
+	}
+
+	e = talloc(NULL, struct tevent_ops_list);
+	if (e == NULL) return false;
+
+	e->name = name;
+	e->ops = ops;
+	DLIST_ADD(tevent_backends, e);
+
+	return true;
+}
+
+/*
+  set the default event backend
+ */
+void tevent_set_default_backend(const char *backend)
+{
+	talloc_free(tevent_default_backend);
+	tevent_default_backend = talloc_strdup(NULL, backend);
+}
+
+/*
+  initialise backends if not already done
+*/
+static void tevent_backend_init(void)
+{
+	static bool done;
+
+	if (done) {
+		return;
+	}
+
+	done = true;
+
+	tevent_select_init();
+	tevent_poll_init();
+	tevent_poll_mt_init();
+#ifdef HAVE_EPOLL
+	tevent_epoll_init();
+#endif
+	tevent_standard_init();
+}
+
+_PRIVATE_ const struct tevent_ops *tevent_find_ops_byname(const char *name)
+{
+	struct tevent_ops_list *e;
+
+	tevent_backend_init();
+
+	if (name == NULL) {
+		name = tevent_default_backend;
+	}
+	if (name == NULL) {
+		name = "standard";
+	}
+
+	for (e = tevent_backends; e != NULL; e = e->next) {
+		if (0 == strcmp(e->name, name)) {
+			return e->ops;
+		}
+	}
+
+	return NULL;
+}
+
+/*
+  list available backends
+*/
+const char **tevent_backend_list(TALLOC_CTX *mem_ctx)
+{
+	const char **list = NULL;
+	struct tevent_ops_list *e;
+
+	tevent_backend_init();
+
+	for (e=tevent_backends;e;e=e->next) {
+		list = ev_str_list_add(list, e->name);
+	}
+
+	talloc_steal(mem_ctx, list);
+
+	return list;
+}
+
+int tevent_common_context_destructor(struct tevent_context *ev)
+{
+	struct tevent_fd *fd, *fn;
+	struct tevent_timer *te, *tn;
+	struct tevent_immediate *ie, *in;
+	struct tevent_signal *se, *sn;
+
+	if (ev->pipe_fde) {
+		talloc_free(ev->pipe_fde);
+		close(ev->pipe_fds[0]);
+		close(ev->pipe_fds[1]);
+		ev->pipe_fde = NULL;
+	}
+
+	for (fd = ev->fd_events; fd; fd = fn) {
+		fn = fd->next;
+		fd->event_ctx = NULL;
+		DLIST_REMOVE(ev->fd_events, fd);
+	}
+
+	ev->last_zero_timer = NULL;
+	for (te = ev->timer_events; te; te = tn) {
+		tn = te->next;
+		te->event_ctx = NULL;
+		DLIST_REMOVE(ev->timer_events, te);
+	}
+
+	for (ie = ev->immediate_events; ie; ie = in) {
+		in = ie->next;
+		ie->event_ctx = NULL;
+		ie->cancel_fn = NULL;
+		DLIST_REMOVE(ev->immediate_events, ie);
+	}
+
+	for (se = ev->signal_events; se; se = sn) {
+		sn = se->next;
+		se->event_ctx = NULL;
+		DLIST_REMOVE(ev->signal_events, se);
+		/*
+		 * This is important, Otherwise signals
+		 * are handled twice in child. eg, SIGHUP.
+		 * one added in parent, and another one in
+		 * the child. -- BoYang
+		 */
+		tevent_cleanup_pending_signal_handlers(se);
+	}
+
+	/* removing nesting hook or we get an abort when nesting is
+	 * not allowed. -- SSS
+	 * Note that we need to leave the allowed flag at its current
+	 * value, otherwise the use in tevent_re_initialise() will
+	 * leave the event context with allowed forced to false, which
+	 * will break users that expect nesting to be allowed
+	 */
+	ev->nesting.level = 0;
+	ev->nesting.hook_fn = NULL;
+	ev->nesting.hook_private = NULL;
+
+	return 0;
+}
+
+/*
+  create a event_context structure for a specific implemementation.
+  This must be the first events call, and all subsequent calls pass
+  this event_context as the first element. Event handlers also
+  receive this as their first argument.
+
+  This function is for allowing third-party-applications to hook in gluecode
+  to their own event loop code, so that they can make async usage of our client libs
+
+  NOTE: use tevent_context_init() inside of samba!
+*/
+struct tevent_context *tevent_context_init_ops(TALLOC_CTX *mem_ctx,
+					       const struct tevent_ops *ops,
+					       void *additional_data)
+{
+	struct tevent_context *ev;
+	int ret;
+
+	ev = talloc_zero(mem_ctx, struct tevent_context);
+	if (!ev) return NULL;
+
+	talloc_set_destructor(ev, tevent_common_context_destructor);
+
+	ev->ops = ops;
+	ev->additional_data = additional_data;
+
+	ret = ev->ops->context_init(ev);
+	if (ret != 0) {
+		talloc_free(ev);
+		return NULL;
+	}
+
+	return ev;
+}
+
+/*
+  create a event_context structure. This must be the first events
+  call, and all subsequent calls pass this event_context as the first
+  element. Event handlers also receive this as their first argument.
+*/
+struct tevent_context *tevent_context_init_byname(TALLOC_CTX *mem_ctx,
+						  const char *name)
+{
+	const struct tevent_ops *ops;
+
+	ops = tevent_find_ops_byname(name);
+	if (ops == NULL) {
+		return NULL;
+	}
+
+	return tevent_context_init_ops(mem_ctx, ops, NULL);
+}
+
+
+/*
+  create a event_context structure. This must be the first events
+  call, and all subsequent calls pass this event_context as the first
+  element. Event handlers also receive this as their first argument.
+*/
+struct tevent_context *tevent_context_init(TALLOC_CTX *mem_ctx)
+{
+	return tevent_context_init_byname(mem_ctx, NULL);
+}
+
+/*
+  add a fd based event
+  return NULL on failure (memory allocation error)
+*/
+struct tevent_fd *_tevent_add_fd(struct tevent_context *ev,
+				 TALLOC_CTX *mem_ctx,
+				 int fd,
+				 uint16_t flags,
+				 tevent_fd_handler_t handler,
+				 void *private_data,
+				 const char *handler_name,
+				 const char *location)
+{
+	return ev->ops->add_fd(ev, mem_ctx, fd, flags, handler, private_data,
+			       handler_name, location);
+}
+
+/*
+  set a close function on the fd event
+*/
+void tevent_fd_set_close_fn(struct tevent_fd *fde,
+			    tevent_fd_close_fn_t close_fn)
+{
+	if (!fde) return;
+	if (!fde->event_ctx) return;
+	fde->event_ctx->ops->set_fd_close_fn(fde, close_fn);
+}
+
+static void tevent_fd_auto_close_fn(struct tevent_context *ev,
+				    struct tevent_fd *fde,
+				    int fd,
+				    void *private_data)
+{
+	close(fd);
+}
+
+void tevent_fd_set_auto_close(struct tevent_fd *fde)
+{
+	tevent_fd_set_close_fn(fde, tevent_fd_auto_close_fn);
+}
+
+/*
+  return the fd event flags
+*/
+uint16_t tevent_fd_get_flags(struct tevent_fd *fde)
+{
+	if (!fde) return 0;
+	if (!fde->event_ctx) return 0;
+	return fde->event_ctx->ops->get_fd_flags(fde);
+}
+
+/*
+  set the fd event flags
+*/
+void tevent_fd_set_flags(struct tevent_fd *fde, uint16_t flags)
+{
+	if (!fde) return;
+	if (!fde->event_ctx) return;
+	fde->event_ctx->ops->set_fd_flags(fde, flags);
+}
+
+bool tevent_signal_support(struct tevent_context *ev)
+{
+	if (ev->ops->add_signal) {
+		return true;
+	}
+	return false;
+}
+
+static void (*tevent_abort_fn)(const char *reason);
+
+void tevent_set_abort_fn(void (*abort_fn)(const char *reason))
+{
+	tevent_abort_fn = abort_fn;
+}
+
+static void tevent_abort(struct tevent_context *ev, const char *reason)
+{
+	tevent_debug(ev, TEVENT_DEBUG_FATAL,
+		     "abort: %s\n", reason);
+
+	if (!tevent_abort_fn) {
+		abort();
+	}
+
+	tevent_abort_fn(reason);
+}
+
+/*
+  add a timer event
+  return NULL on failure
+*/
+struct tevent_timer *_tevent_add_timer(struct tevent_context *ev,
+				       TALLOC_CTX *mem_ctx,
+				       struct timeval next_event,
+				       tevent_timer_handler_t handler,
+				       void *private_data,
+				       const char *handler_name,
+				       const char *location)
+{
+	return ev->ops->add_timer(ev, mem_ctx, next_event, handler, private_data,
+				  handler_name, location);
+}
+
+/*
+  allocate an immediate event
+  return NULL on failure (memory allocation error)
+*/
+struct tevent_immediate *_tevent_create_immediate(TALLOC_CTX *mem_ctx,
+						  const char *location)
+{
+	struct tevent_immediate *im;
+
+	im = talloc(mem_ctx, struct tevent_immediate);
+	if (im == NULL) return NULL;
+
+	im->prev		= NULL;
+	im->next		= NULL;
+	im->event_ctx		= NULL;
+	im->create_location	= location;
+	im->handler		= NULL;
+	im->private_data	= NULL;
+	im->handler_name	= NULL;
+	im->schedule_location	= NULL;
+	im->cancel_fn		= NULL;
+	im->additional_data	= NULL;
+
+	return im;
+}
+
+/*
+  schedule an immediate event
+*/
+void _tevent_schedule_immediate(struct tevent_immediate *im,
+				struct tevent_context *ev,
+				tevent_immediate_handler_t handler,
+				void *private_data,
+				const char *handler_name,
+				const char *location)
+{
+	ev->ops->schedule_immediate(im, ev, handler, private_data,
+				    handler_name, location);
+}
+
+/*
+  add a signal event
+
+  sa_flags are flags to sigaction(2)
+
+  return NULL on failure
+*/
+struct tevent_signal *_tevent_add_signal(struct tevent_context *ev,
+					 TALLOC_CTX *mem_ctx,
+					 int signum,
+					 int sa_flags,
+					 tevent_signal_handler_t handler,
+					 void *private_data,
+					 const char *handler_name,
+					 const char *location)
+{
+	return ev->ops->add_signal(ev, mem_ctx, signum, sa_flags, handler, private_data,
+				   handler_name, location);
+}
+
+void tevent_loop_allow_nesting(struct tevent_context *ev)
+{
+	ev->nesting.allowed = true;
+}
+
+void tevent_loop_set_nesting_hook(struct tevent_context *ev,
+				  tevent_nesting_hook hook,
+				  void *private_data)
+{
+	if (ev->nesting.hook_fn && 
+	    (ev->nesting.hook_fn != hook ||
+	     ev->nesting.hook_private != private_data)) {
+		/* the way the nesting hook code is currently written
+		   we cannot support two different nesting hooks at the
+		   same time. */
+		tevent_abort(ev, "tevent: Violation of nesting hook rules\n");
+	}
+	ev->nesting.hook_fn = hook;
+	ev->nesting.hook_private = private_data;
+}
+
+static void tevent_abort_nesting(struct tevent_context *ev, const char *location)
+{
+	const char *reason;
+
+	reason = talloc_asprintf(NULL, "tevent_loop_once() nesting at %s",
+				 location);
+	if (!reason) {
+		reason = "tevent_loop_once() nesting";
+	}
+
+	tevent_abort(ev, reason);
+}
+
+/*
+  do a single event loop using the events defined in ev 
+*/
+int _tevent_loop_once(struct tevent_context *ev, const char *location)
+{
+	int ret;
+	void *nesting_stack_ptr = NULL;
+
+	ev->nesting.level++;
+
+	if (ev->nesting.level > 1) {
+		if (!ev->nesting.allowed) {
+			tevent_abort_nesting(ev, location);
+			errno = ELOOP;
+			return -1;
+		}
+	}
+	if (ev->nesting.level > 0) {
+		if (ev->nesting.hook_fn) {
+			int ret2;
+			ret2 = ev->nesting.hook_fn(ev,
+						   ev->nesting.hook_private,
+						   ev->nesting.level,
+						   true,
+						   (void *)&nesting_stack_ptr,
+						   location);
+			if (ret2 != 0) {
+				ret = ret2;
+				goto done;
+			}
+		}
+	}
+
+	tevent_trace_point_callback(ev, TEVENT_TRACE_BEFORE_LOOP_ONCE);
+	ret = ev->ops->loop_once(ev, location);
+	tevent_trace_point_callback(ev, TEVENT_TRACE_AFTER_LOOP_ONCE);
+
+	if (ev->nesting.level > 0) {
+		if (ev->nesting.hook_fn) {
+			int ret2;
+			ret2 = ev->nesting.hook_fn(ev,
+						   ev->nesting.hook_private,
+						   ev->nesting.level,
+						   false,
+						   (void *)&nesting_stack_ptr,
+						   location);
+			if (ret2 != 0) {
+				ret = ret2;
+				goto done;
+			}
+		}
+	}
+
+done:
+	ev->nesting.level--;
+	return ret;
+}
+
+/*
+  this is a performance optimization for the samba4 nested event loop problems
+*/
+int _tevent_loop_until(struct tevent_context *ev,
+		       bool (*finished)(void *private_data),
+		       void *private_data,
+		       const char *location)
+{
+	int ret = 0;
+	void *nesting_stack_ptr = NULL;
+
+	ev->nesting.level++;
+
+	if (ev->nesting.level > 1) {
+		if (!ev->nesting.allowed) {
+			tevent_abort_nesting(ev, location);
+			errno = ELOOP;
+			return -1;
+		}
+	}
+	if (ev->nesting.level > 0) {
+		if (ev->nesting.hook_fn) {
+			int ret2;
+			ret2 = ev->nesting.hook_fn(ev,
+						   ev->nesting.hook_private,
+						   ev->nesting.level,
+						   true,
+						   (void *)&nesting_stack_ptr,
+						   location);
+			if (ret2 != 0) {
+				ret = ret2;
+				goto done;
+			}
+		}
+	}
+
+	while (!finished(private_data)) {
+		tevent_trace_point_callback(ev, TEVENT_TRACE_BEFORE_LOOP_ONCE);
+		ret = ev->ops->loop_once(ev, location);
+		tevent_trace_point_callback(ev, TEVENT_TRACE_AFTER_LOOP_ONCE);
+		if (ret != 0) {
+			break;
+		}
+	}
+
+	if (ev->nesting.level > 0) {
+		if (ev->nesting.hook_fn) {
+			int ret2;
+			ret2 = ev->nesting.hook_fn(ev,
+						   ev->nesting.hook_private,
+						   ev->nesting.level,
+						   false,
+						   (void *)&nesting_stack_ptr,
+						   location);
+			if (ret2 != 0) {
+				ret = ret2;
+				goto done;
+			}
+		}
+	}
+
+done:
+	ev->nesting.level--;
+	return ret;
+}
+
+/*
+  return on failure or (with 0) if all fd events are removed
+*/
+int tevent_common_loop_wait(struct tevent_context *ev,
+			    const char *location)
+{
+	/*
+	 * loop as long as we have events pending
+	 */
+	while (ev->fd_events ||
+	       ev->timer_events ||
+	       ev->immediate_events ||
+	       ev->signal_events) {
+		int ret;
+		ret = _tevent_loop_once(ev, location);
+		if (ret != 0) {
+			tevent_debug(ev, TEVENT_DEBUG_FATAL,
+				     "_tevent_loop_once() failed: %d - %s\n",
+				     ret, strerror(errno));
+			return ret;
+		}
+	}
+
+	tevent_debug(ev, TEVENT_DEBUG_WARNING,
+		     "tevent_common_loop_wait() out of events\n");
+	return 0;
+}
+
+/*
+  return on failure or (with 0) if all fd events are removed
+*/
+int _tevent_loop_wait(struct tevent_context *ev, const char *location)
+{
+	return ev->ops->loop_wait(ev, location);
+}
+
+
+/*
+  re-initialise a tevent context. This leaves you with the same
+  event context, but all events are wiped and the structure is
+  re-initialised. This is most useful after a fork()  
+
+  zero is returned on success, non-zero on failure
+*/
+int tevent_re_initialise(struct tevent_context *ev)
+{
+	tevent_common_context_destructor(ev);
+
+	return ev->ops->context_init(ev);
+}
diff --git a/ctdb/lib/tevent/tevent.h b/ctdb/lib/tevent/tevent.h
new file mode 100644
index 0000000000..6b4d371e4c
--- /dev/null
+++ b/ctdb/lib/tevent/tevent.h
@@ -0,0 +1,1776 @@
+/* 
+   Unix SMB/CIFS implementation.
+
+   generalised event loop handling
+
+   Copyright (C) Andrew Tridgell 2005
+   Copyright (C) Stefan Metzmacher 2005-2009
+   Copyright (C) Volker Lendecke 2008
+
+     ** NOTE! The following LGPL license applies to the tevent
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#ifndef __TEVENT_H__
+#define __TEVENT_H__
+
+#include <stdint.h>
+#include <talloc.h>
+#include <sys/time.h>
+#include <stdbool.h>
+
+struct tevent_context;
+struct tevent_ops;
+struct tevent_fd;
+struct tevent_timer;
+struct tevent_immediate;
+struct tevent_signal;
+
+/**
+ * @defgroup tevent The tevent API
+ *
+ * The tevent low-level API
+ *
+ * This API provides the public interface to manage events in the tevent
+ * mainloop. Functions are provided for managing low-level events such
+ * as timer events, fd events and signal handling.
+ *
+ * @{
+ */
+
+/* event handler types */
+/**
+ * Called when a file descriptor monitored by tevent has
+ * data to be read or written on it.
+ */
+typedef void (*tevent_fd_handler_t)(struct tevent_context *ev,
+				    struct tevent_fd *fde,
+				    uint16_t flags,
+				    void *private_data);
+
+/**
+ * Called when tevent is ceasing the monitoring of a file descriptor.
+ */
+typedef void (*tevent_fd_close_fn_t)(struct tevent_context *ev,
+				     struct tevent_fd *fde,
+				     int fd,
+				     void *private_data);
+
+/**
+ * Called when a tevent timer has fired.
+ */
+typedef void (*tevent_timer_handler_t)(struct tevent_context *ev,
+				       struct tevent_timer *te,
+				       struct timeval current_time,
+				       void *private_data);
+
+/**
+ * Called when a tevent immediate event is invoked.
+ */
+typedef void (*tevent_immediate_handler_t)(struct tevent_context *ctx,
+					   struct tevent_immediate *im,
+					   void *private_data);
+
+/**
+ * Called after tevent detects the specified signal.
+ */
+typedef void (*tevent_signal_handler_t)(struct tevent_context *ev,
+					struct tevent_signal *se,
+					int signum,
+					int count,
+					void *siginfo,
+					void *private_data);
+
+/**
+ * @brief Create a event_context structure.
+ *
+ * This must be the first events call, and all subsequent calls pass this
+ * event_context as the first element. Event handlers also receive this as
+ * their first argument.
+ *
+ * @param[in]  mem_ctx  The memory context to use.
+ *
+ * @return              An allocated tevent context, NULL on error.
+ *
+ * @see tevent_context_init()
+ */
+struct tevent_context *tevent_context_init(TALLOC_CTX *mem_ctx);
+
+/**
+ * @brief Create a event_context structure and select a specific backend.
+ *
+ * This must be the first events call, and all subsequent calls pass this
+ * event_context as the first element. Event handlers also receive this as
+ * their first argument.
+ *
+ * @param[in]  mem_ctx  The memory context to use.
+ *
+ * @param[in]  name     The name of the backend to use.
+ *
+ * @return              An allocated tevent context, NULL on error.
+ */
+struct tevent_context *tevent_context_init_byname(TALLOC_CTX *mem_ctx, const char *name);
+
+/**
+ * @brief Create a custom event context
+ *
+ * @param[in]  mem_ctx  The memory context to use.
+ * @param[in]  ops      The function pointer table of the backend.
+ * @param[in]  additional_data  The additional/private data to this instance
+ *
+ * @return              An allocated tevent context, NULL on error.
+ *
+ */
+struct tevent_context *tevent_context_init_ops(TALLOC_CTX *mem_ctx,
+					       const struct tevent_ops *ops,
+					       void *additional_data);
+
+/**
+ * @brief List available backends.
+ *
+ * @param[in]  mem_ctx  The memory context to use.
+ *
+ * @return              A string vector with a terminating NULL element, NULL
+ *                      on error.
+ */
+const char **tevent_backend_list(TALLOC_CTX *mem_ctx);
+
+/**
+ * @brief Set the default tevent backend.
+ *
+ * @param[in]  backend  The name of the backend to set.
+ */
+void tevent_set_default_backend(const char *backend);
+
+#ifdef DOXYGEN
+/**
+ * @brief Add a file descriptor based event.
+ *
+ * @param[in]  ev       The event context to work on.
+ *
+ * @param[in]  mem_ctx  The talloc memory context to use.
+ *
+ * @param[in]  fd       The file descriptor to base the event on.
+ *
+ * @param[in]  flags    #TEVENT_FD_READ or #TEVENT_FD_WRITE
+ *
+ * @param[in]  handler  The callback handler for the event.
+ *
+ * @param[in]  private_data  The private data passed to the callback handler.
+ *
+ * @return              The file descriptor based event, NULL on error.
+ *
+ * @note To cancel the monitoring of a file descriptor, call talloc_free()
+ * on the object returned by this function.
+ */
+struct tevent_fd *tevent_add_fd(struct tevent_context *ev,
+				TALLOC_CTX *mem_ctx,
+				int fd,
+				uint16_t flags,
+				tevent_fd_handler_t handler,
+				void *private_data);
+#else
+struct tevent_fd *_tevent_add_fd(struct tevent_context *ev,
+				 TALLOC_CTX *mem_ctx,
+				 int fd,
+				 uint16_t flags,
+				 tevent_fd_handler_t handler,
+				 void *private_data,
+				 const char *handler_name,
+				 const char *location);
+#define tevent_add_fd(ev, mem_ctx, fd, flags, handler, private_data) \
+	_tevent_add_fd(ev, mem_ctx, fd, flags, handler, private_data, \
+		       #handler, __location__)
+#endif
+
+#ifdef DOXYGEN
+/**
+ * @brief Add a timed event
+ *
+ * @param[in]  ev       The event context to work on.
+ *
+ * @param[in]  mem_ctx  The talloc memory context to use.
+ *
+ * @param[in]  next_event  Timeval specifying the absolute time to fire this
+ * event. This is not an offset.
+ *
+ * @param[in]  handler  The callback handler for the event.
+ *
+ * @param[in]  private_data  The private data passed to the callback handler.
+ *
+ * @return The newly-created timer event, or NULL on error.
+ *
+ * @note To cancel a timer event before it fires, call talloc_free() on the
+ * event returned from this function. This event is automatically
+ * talloc_free()-ed after its event handler files, if it hasn't been freed yet.
+ *
+ * @note Unlike some mainloops, tevent timers are one-time events. To set up
+ * a recurring event, it is necessary to call tevent_add_timer() again during
+ * the handler processing.
+ *
+ * @note Due to the internal mainloop processing, a timer set to run
+ * immediately will do so after any other pending timers fire, but before
+ * any further file descriptor or signal handling events fire. Callers should
+ * not rely on this behavior!
+ */
+struct tevent_timer *tevent_add_timer(struct tevent_context *ev,
+                                      TALLOC_CTX *mem_ctx,
+                                      struct timeval next_event,
+                                      tevent_timer_handler_t handler,
+                                      void *private_data);
+#else
+struct tevent_timer *_tevent_add_timer(struct tevent_context *ev,
+				       TALLOC_CTX *mem_ctx,
+				       struct timeval next_event,
+				       tevent_timer_handler_t handler,
+				       void *private_data,
+				       const char *handler_name,
+				       const char *location);
+#define tevent_add_timer(ev, mem_ctx, next_event, handler, private_data) \
+	_tevent_add_timer(ev, mem_ctx, next_event, handler, private_data, \
+			  #handler, __location__)
+#endif
+
+#ifdef DOXYGEN
+/**
+ * Initialize an immediate event object
+ *
+ * This object can be used to trigger an event to occur immediately after
+ * returning from the current event (before any other event occurs)
+ *
+ * @param[in] mem_ctx  The talloc memory context to use as the parent
+ *
+ * @return An empty tevent_immediate object. Use tevent_schedule_immediate
+ * to populate and use it.
+ *
+ * @note Available as of tevent 0.9.8
+ */
+struct tevent_immediate *tevent_create_immediate(TALLOC_CTX *mem_ctx);
+#else
+struct tevent_immediate *_tevent_create_immediate(TALLOC_CTX *mem_ctx,
+						  const char *location);
+#define tevent_create_immediate(mem_ctx) \
+	_tevent_create_immediate(mem_ctx, __location__)
+#endif
+
+#ifdef DOXYGEN
+
+/**
+ * Schedule an event for immediate execution. This event will occur
+ * immediately after returning from the current event (before any other
+ * event occurs)
+ *
+ * @param[in] im       The tevent_immediate object to populate and use
+ * @param[in] ctx      The tevent_context to run this event
+ * @param[in] handler  The event handler to run when this event fires
+ * @param[in] private_data  Data to pass to the event handler
+ */
+void tevent_schedule_immediate(struct tevent_immediate *im,
+                struct tevent_context *ctx,
+                tevent_immediate_handler_t handler,
+                void *private_data);
+#else
+void _tevent_schedule_immediate(struct tevent_immediate *im,
+				struct tevent_context *ctx,
+				tevent_immediate_handler_t handler,
+				void *private_data,
+				const char *handler_name,
+				const char *location);
+#define tevent_schedule_immediate(im, ctx, handler, private_data) \
+	_tevent_schedule_immediate(im, ctx, handler, private_data, \
+				   #handler, __location__);
+#endif
+
+#ifdef DOXYGEN
+/**
+ * @brief Add a tevent signal handler
+ *
+ * tevent_add_signal() creates a new event for handling a signal the next
+ * time through the mainloop. It implements a very simple traditional signal
+ * handler whose only purpose is to add the handler event into the mainloop.
+ *
+ * @param[in]  ev       The event context to work on.
+ *
+ * @param[in]  mem_ctx  The talloc memory context to use.
+ *
+ * @param[in]  signum   The signal to trap
+ *
+ * @param[in]  handler  The callback handler for the signal.
+ *
+ * @param[in]  sa_flags sigaction flags for this signal handler.
+ *
+ * @param[in]  private_data  The private data passed to the callback handler.
+ *
+ * @return The newly-created signal handler event, or NULL on error.
+ *
+ * @note To cancel a signal handler, call talloc_free() on the event returned
+ * from this function.
+ */
+struct tevent_signal *tevent_add_signal(struct tevent_context *ev,
+                     TALLOC_CTX *mem_ctx,
+                     int signum,
+                     int sa_flags,
+                     tevent_signal_handler_t handler,
+                     void *private_data);
+#else
+struct tevent_signal *_tevent_add_signal(struct tevent_context *ev,
+					 TALLOC_CTX *mem_ctx,
+					 int signum,
+					 int sa_flags,
+					 tevent_signal_handler_t handler,
+					 void *private_data,
+					 const char *handler_name,
+					 const char *location);
+#define tevent_add_signal(ev, mem_ctx, signum, sa_flags, handler, private_data) \
+	_tevent_add_signal(ev, mem_ctx, signum, sa_flags, handler, private_data, \
+			   #handler, __location__)
+#endif
+
+#ifdef DOXYGEN
+/**
+ * @brief Pass a single time through the mainloop
+ *
+ * This will process any appropriate signal, immediate, fd and timer events
+ *
+ * @param[in]  ev The event context to process
+ *
+ * @return Zero on success, nonzero if an internal error occurred
+ */
+int tevent_loop_once(struct tevent_context *ev);
+#else
+int _tevent_loop_once(struct tevent_context *ev, const char *location);
+#define tevent_loop_once(ev) \
+	_tevent_loop_once(ev, __location__)
+#endif
+
+#ifdef DOXYGEN
+/**
+ * @brief Run the mainloop
+ *
+ * The mainloop will run until there are no events remaining to be processed
+ *
+ * @param[in]  ev The event context to process
+ *
+ * @return Zero if all events have been processed. Nonzero if an internal
+ * error occurred.
+ */
+int tevent_loop_wait(struct tevent_context *ev);
+#else
+int _tevent_loop_wait(struct tevent_context *ev, const char *location);
+#define tevent_loop_wait(ev) \
+	_tevent_loop_wait(ev, __location__)
+#endif
+
+
+/**
+ * Assign a function to run when a tevent_fd is freed
+ *
+ * This function is a destructor for the tevent_fd. It does not automatically
+ * close the file descriptor. If this is the desired behavior, then it must be
+ * performed by the close_fn.
+ *
+ * @param[in] fde       File descriptor event on which to set the destructor
+ * @param[in] close_fn  Destructor to execute when fde is freed
+ */
+void tevent_fd_set_close_fn(struct tevent_fd *fde,
+			    tevent_fd_close_fn_t close_fn);
+
+/**
+ * Automatically close the file descriptor when the tevent_fd is freed
+ *
+ * This function calls close(fd) internally.
+ *
+ * @param[in] fde  File descriptor event to auto-close
+ */
+void tevent_fd_set_auto_close(struct tevent_fd *fde);
+
+/**
+ * Return the flags set on this file descriptor event
+ *
+ * @param[in] fde  File descriptor event to query
+ *
+ * @return The flags set on the event. See #TEVENT_FD_READ and
+ * #TEVENT_FD_WRITE
+ */
+uint16_t tevent_fd_get_flags(struct tevent_fd *fde);
+
+/**
+ * Set flags on a file descriptor event
+ *
+ * @param[in] fde    File descriptor event to set
+ * @param[in] flags  Flags to set on the event. See #TEVENT_FD_READ and
+ * #TEVENT_FD_WRITE
+ */
+void tevent_fd_set_flags(struct tevent_fd *fde, uint16_t flags);
+
+/**
+ * Query whether tevent supports signal handling
+ *
+ * @param[in] ev  An initialized tevent context
+ *
+ * @return True if this platform and tevent context support signal handling
+ */
+bool tevent_signal_support(struct tevent_context *ev);
+
+void tevent_set_abort_fn(void (*abort_fn)(const char *reason));
+
+/* bits for file descriptor event flags */
+
+/**
+ * Monitor a file descriptor for write availability
+ */
+#define TEVENT_FD_READ 1
+/**
+ * Monitor a file descriptor for data to be read
+ */
+#define TEVENT_FD_WRITE 2
+
+/**
+ * Convenience function for declaring a tevent_fd writable
+ */
+#define TEVENT_FD_WRITEABLE(fde) \
+	tevent_fd_set_flags(fde, tevent_fd_get_flags(fde) | TEVENT_FD_WRITE)
+
+/**
+ * Convenience function for declaring a tevent_fd readable
+ */
+#define TEVENT_FD_READABLE(fde) \
+	tevent_fd_set_flags(fde, tevent_fd_get_flags(fde) | TEVENT_FD_READ)
+
+/**
+ * Convenience function for declaring a tevent_fd non-writable
+ */
+#define TEVENT_FD_NOT_WRITEABLE(fde) \
+	tevent_fd_set_flags(fde, tevent_fd_get_flags(fde) & ~TEVENT_FD_WRITE)
+
+/**
+ * Convenience function for declaring a tevent_fd non-readable
+ */
+#define TEVENT_FD_NOT_READABLE(fde) \
+	tevent_fd_set_flags(fde, tevent_fd_get_flags(fde) & ~TEVENT_FD_READ)
+
+/**
+ * Debug level of tevent
+ */
+enum tevent_debug_level {
+	TEVENT_DEBUG_FATAL,
+	TEVENT_DEBUG_ERROR,
+	TEVENT_DEBUG_WARNING,
+	TEVENT_DEBUG_TRACE
+};
+
+/**
+ * @brief The tevent debug callbac.
+ *
+ * @param[in]  context  The memory context to use.
+ *
+ * @param[in]  level    The debug level.
+ *
+ * @param[in]  fmt      The format string.
+ *
+ * @param[in]  ap       The arguments for the format string.
+ */
+typedef void (*tevent_debug_fn)(void *context,
+				enum tevent_debug_level level,
+				const char *fmt,
+				va_list ap) PRINTF_ATTRIBUTE(3,0);
+
+/**
+ * Set destination for tevent debug messages
+ *
+ * @param[in] ev        Event context to debug
+ * @param[in] debug     Function to handle output printing
+ * @param[in] context   The context to pass to the debug function.
+ *
+ * @return Always returns 0 as of version 0.9.8
+ *
+ * @note Default is to emit no debug messages
+ */
+int tevent_set_debug(struct tevent_context *ev,
+		     tevent_debug_fn debug,
+		     void *context);
+
+/**
+ * Designate stderr for debug message output
+ *
+ * @param[in] ev     Event context to debug
+ *
+ * @note This function will only output TEVENT_DEBUG_FATAL, TEVENT_DEBUG_ERROR
+ * and TEVENT_DEBUG_WARNING messages. For TEVENT_DEBUG_TRACE, please define a
+ * function for tevent_set_debug()
+ */
+int tevent_set_debug_stderr(struct tevent_context *ev);
+
+enum tevent_trace_point {
+	/**
+	 * Corresponds to a trace point just before waiting
+	 */
+	TEVENT_TRACE_BEFORE_WAIT,
+	/**
+	 * Corresponds to a trace point just after waiting
+	 */
+	TEVENT_TRACE_AFTER_WAIT,
+#define TEVENT_HAS_LOOP_ONCE_TRACE_POINTS 1
+	/**
+	 * Corresponds to a trace point just before calling
+	 * the loop_once() backend function.
+	 */
+	TEVENT_TRACE_BEFORE_LOOP_ONCE,
+	/**
+	 * Corresponds to a trace point right after the
+	 * loop_once() backend function has returned.
+	 */
+	TEVENT_TRACE_AFTER_LOOP_ONCE,
+};
+
+typedef void (*tevent_trace_callback_t)(enum tevent_trace_point,
+					void *private_data);
+
+/**
+ * Register a callback to be called at certain trace points
+ *
+ * @param[in] ev             Event context
+ * @param[in] cb             Trace callback
+ * @param[in] private_data   Data to be passed to callback
+ *
+ * @note The callback will be called at trace points defined by
+ * tevent_trace_point.  Call with NULL to reset.
+ */
+void tevent_set_trace_callback(struct tevent_context *ev,
+			       tevent_trace_callback_t cb,
+			       void *private_data);
+
+/**
+ * Retrieve the current trace callback
+ *
+ * @param[in] ev             Event context
+ * @param[out] cb            Registered trace callback
+ * @param[out] private_data  Registered data to be passed to callback
+ *
+ * @note This can be used to allow one component that wants to
+ * register a callback to respect the callback that another component
+ * has already registered.
+ */
+void tevent_get_trace_callback(struct tevent_context *ev,
+			       tevent_trace_callback_t *cb,
+			       void *private_data);
+
+/**
+ * @}
+ */
+
+/**
+ * @defgroup tevent_request The tevent request functions.
+ * @ingroup tevent
+ *
+ * A tevent_req represents an asynchronous computation.
+ *
+ * The tevent_req group of API calls is the recommended way of
+ * programming async computations within tevent. In particular the
+ * file descriptor (tevent_add_fd) and timer (tevent_add_timed) events
+ * are considered too low-level to be used in larger computations. To
+ * read and write from and to sockets, Samba provides two calls on top
+ * of tevent_add_fd: read_packet_send/recv and writev_send/recv. These
+ * requests are much easier to compose than the low-level event
+ * handlers called from tevent_add_fd.
+ *
+ * A lot of the simplicity tevent_req has brought to the notoriously
+ * hairy async programming came via a set of conventions that every
+ * async computation programmed should follow. One central piece of
+ * these conventions is the naming of routines and variables.
+ *
+ * Every async computation needs a name (sensibly called "computation"
+ * down from here). From this name quite a few naming conventions are
+ * derived.
+ *
+ * Every computation that requires local state needs a
+ * @code
+ * struct computation_state {
+ *     int local_var;
+ * };
+ * @endcode
+ * Even if no local variables are required, such a state struct should
+ * be created containing a dummy variable. Quite a few helper
+ * functions and macros (for example tevent_req_create()) assume such
+ * a state struct.
+ *
+ * An async computation is started by a computation_send
+ * function. When it is finished, its result can be received by a
+ * computation_recv function. For an example how to set up an async
+ * computation, see the code example in the documentation for
+ * tevent_req_create() and tevent_req_post(). The prototypes for _send
+ * and _recv functions should follow some conventions:
+ *
+ * @code
+ * struct tevent_req *computation_send(TALLOC_CTX *mem_ctx,
+ *                                     struct tevent_req *ev,
+ *                                     ... further args);
+ * int computation_recv(struct tevent_req *req, ... further output args);
+ * @endcode
+ *
+ * The "int" result of computation_recv() depends on the result the
+ * sync version of the function would have, "int" is just an example
+ * here.
+ *
+ * Another important piece of the conventions is that the program flow
+ * is interrupted as little as possible. Because a blocking
+ * sub-computation requires that the flow needs to continue in a
+ * separate function that is the logical sequel of some computation,
+ * it should lexically follow sending off the blocking
+ * sub-computation. Setting the callback function via
+ * tevent_req_set_callback() requires referencing a function lexically
+ * below the call to tevent_req_set_callback(), forward declarations
+ * are required. A lot of the async computations thus begin with a
+ * sequence of declarations such as
+ *
+ * @code
+ * static void computation_step1_done(struct tevent_req *subreq);
+ * static void computation_step2_done(struct tevent_req *subreq);
+ * static void computation_step3_done(struct tevent_req *subreq);
+ * @endcode
+ *
+ * It really helps readability a lot to do these forward declarations,
+ * because the lexically sequential program flow makes the async
+ * computations almost as clear to read as a normal, sync program
+ * flow.
+ *
+ * It is up to the user of the async computation to talloc_free it
+ * after it has finished. If an async computation should be aborted,
+ * the tevent_req structure can be talloc_free'ed. After it has
+ * finished, it should talloc_free'ed by the API user.
+ *
+ * @{
+ */
+
+/**
+ * An async request moves from TEVENT_REQ_INIT to
+ * TEVENT_REQ_IN_PROGRESS. All other states are valid after a request
+ * has finished.
+ */
+enum tevent_req_state {
+	/**
+	 * We are creating the request
+	 */
+	TEVENT_REQ_INIT,
+	/**
+	 * We are waiting the request to complete
+	 */
+	TEVENT_REQ_IN_PROGRESS,
+	/**
+	 * The request is finished successfully
+	 */
+	TEVENT_REQ_DONE,
+	/**
+	 * A user error has occurred. The user error has been
+	 * indicated by tevent_req_error(), it can be retrieved via
+	 * tevent_req_is_error().
+	 */
+	TEVENT_REQ_USER_ERROR,
+	/**
+	 * Request timed out after the timeout set by tevent_req_set_endtime.
+	 */
+	TEVENT_REQ_TIMED_OUT,
+	/**
+	 * An internal allocation has failed, or tevent_req_nomem has
+	 * been given a NULL pointer as the first argument.
+	 */
+	TEVENT_REQ_NO_MEMORY,
+	/**
+	 * The request has been received by the caller. No further
+	 * action is valid.
+	 */
+	TEVENT_REQ_RECEIVED
+};
+
+/**
+ * @brief An async request
+ */
+struct tevent_req;
+
+/**
+ * @brief A tevent request callback function.
+ *
+ * @param[in]  req      The tevent async request which executed this callback.
+ */
+typedef void (*tevent_req_fn)(struct tevent_req *req);
+
+/**
+ * @brief Set an async request callback.
+ *
+ * See the documentation of tevent_req_post() for an example how this
+ * is supposed to be used.
+ *
+ * @param[in]  req      The async request to set the callback.
+ *
+ * @param[in]  fn       The callback function to set.
+ *
+ * @param[in]  pvt      A pointer to private data to pass to the async request
+ *                      callback.
+ */
+void tevent_req_set_callback(struct tevent_req *req, tevent_req_fn fn, void *pvt);
+
+#ifdef DOXYGEN
+/**
+ * @brief Get the private data cast to the given type for a callback from
+ *        a tevent request structure.
+ *
+ * @code
+ * static void computation_done(struct tevent_req *subreq) {
+ *     struct tevent_req *req = tevent_req_callback_data(subreq, struct tevent_req);
+ *     struct computation_state *state = tevent_req_data(req, struct computation_state);
+ *     .... more things, eventually maybe call tevent_req_done(req);
+ * }
+ * @endcode
+ *
+ * @param[in]  req      The structure to get the callback data from.
+ *
+ * @param[in]  type     The type of the private callback data to get.
+ *
+ * @return              The type casted private data set NULL if not set.
+ */
+void *tevent_req_callback_data(struct tevent_req *req, #type);
+#else
+void *_tevent_req_callback_data(struct tevent_req *req);
+#define tevent_req_callback_data(_req, _type) \
+	talloc_get_type_abort(_tevent_req_callback_data(_req), _type)
+#endif
+
+#ifdef DOXYGEN
+/**
+ * @brief Get the private data for a callback from a tevent request structure.
+ *
+ * @param[in]  req      The structure to get the callback data from.
+ *
+ * @param[in]  req      The structure to get the data from.
+ *
+ * @return              The private data or NULL if not set.
+ */
+void *tevent_req_callback_data_void(struct tevent_req *req);
+#else
+#define tevent_req_callback_data_void(_req) \
+	_tevent_req_callback_data(_req)
+#endif
+
+#ifdef DOXYGEN
+/**
+ * @brief Get the private data from a tevent request structure.
+ *
+ * When the tevent_req has been created by tevent_req_create, the
+ * result of tevent_req_data() is the state variable created by
+ * tevent_req_create() as a child of the req.
+ *
+ * @param[in]  req      The structure to get the private data from.
+ *
+ * @param[in]  type	The type of the private data
+ *
+ * @return              The private data or NULL if not set.
+ */
+void *tevent_req_data(struct tevent_req *req, #type);
+#else
+void *_tevent_req_data(struct tevent_req *req);
+#define tevent_req_data(_req, _type) \
+	talloc_get_type_abort(_tevent_req_data(_req), _type)
+#endif
+
+/**
+ * @brief The print function which can be set for a tevent async request.
+ *
+ * @param[in]  req      The tevent async request.
+ *
+ * @param[in]  ctx      A talloc memory context which can be uses to allocate
+ *                      memory.
+ *
+ * @return              An allocated string buffer to print.
+ *
+ * Example:
+ * @code
+ *   static char *my_print(struct tevent_req *req, TALLOC_CTX *mem_ctx)
+ *   {
+ *     struct my_data *data = tevent_req_data(req, struct my_data);
+ *     char *result;
+ *
+ *     result = tevent_req_default_print(mem_ctx, req);
+ *     if (result == NULL) {
+ *       return NULL;
+ *     }
+ *
+ *     return talloc_asprintf_append_buffer(result, "foo=%d, bar=%d",
+ *       data->foo, data->bar);
+ *   }
+ * @endcode
+ */
+typedef char *(*tevent_req_print_fn)(struct tevent_req *req, TALLOC_CTX *ctx);
+
+/**
+ * @brief This function sets a print function for the given request.
+ *
+ * This function can be used to setup a print function for the given request.
+ * This will be triggered if the tevent_req_print() function was
+ * called on the given request.
+ *
+ * @param[in]  req      The request to use.
+ *
+ * @param[in]  fn       A pointer to the print function
+ *
+ * @note This function should only be used for debugging.
+ */
+void tevent_req_set_print_fn(struct tevent_req *req, tevent_req_print_fn fn);
+
+/**
+ * @brief The default print function for creating debug messages.
+ *
+ * The function should not be used by users of the async API,
+ * but custom print function can use it and append custom text
+ * to the string.
+ *
+ * @param[in]  req      The request to be printed.
+ *
+ * @param[in]  mem_ctx  The memory context for the result.
+ *
+ * @return              Text representation of request.
+ *
+ */
+char *tevent_req_default_print(struct tevent_req *req, TALLOC_CTX *mem_ctx);
+
+/**
+ * @brief Print an tevent_req structure in debug messages.
+ *
+ * This function should be used by callers of the async API.
+ *
+ * @param[in]  mem_ctx  The memory context for the result.
+ *
+ * @param[in] req       The request to be printed.
+ *
+ * @return              Text representation of request.
+ */
+char *tevent_req_print(TALLOC_CTX *mem_ctx, struct tevent_req *req);
+
+/**
+ * @brief A typedef for a cancel function for a tevent request.
+ *
+ * @param[in]  req      The tevent request calling this function.
+ *
+ * @return              True if the request could be canceled, false if not.
+ */
+typedef bool (*tevent_req_cancel_fn)(struct tevent_req *req);
+
+/**
+ * @brief This function sets a cancel function for the given tevent request.
+ *
+ * This function can be used to setup a cancel function for the given request.
+ * This will be triggered if the tevent_req_cancel() function was
+ * called on the given request.
+ *
+ * @param[in]  req      The request to use.
+ *
+ * @param[in]  fn       A pointer to the cancel function.
+ */
+void tevent_req_set_cancel_fn(struct tevent_req *req, tevent_req_cancel_fn fn);
+
+#ifdef DOXYGEN
+/**
+ * @brief Try to cancel the given tevent request.
+ *
+ * This function can be used to cancel the given request.
+ *
+ * It is only possible to cancel a request when the implementation
+ * has registered a cancel function via the tevent_req_set_cancel_fn().
+ *
+ * @param[in]  req      The request to use.
+ *
+ * @return              This function returns true is the request is cancelable,
+ *                      othererwise false is returned.
+ *
+ * @note Even if the function returns true, the caller need to wait
+ *       for the function to complete normally.
+ *       Only the _recv() function of the given request indicates
+ *       if the request was really canceled.
+ */
+bool tevent_req_cancel(struct tevent_req *req);
+#else
+bool _tevent_req_cancel(struct tevent_req *req, const char *location);
+#define tevent_req_cancel(req) \
+	_tevent_req_cancel(req, __location__)
+#endif
+
+#ifdef DOXYGEN
+/**
+ * @brief Create an async tevent request.
+ *
+ * The new async request will be initialized in state TEVENT_REQ_IN_PROGRESS.
+ *
+ * @code
+ * struct tevent_req *req;
+ * struct computation_state *state;
+ * req = tevent_req_create(mem_ctx, &state, struct computation_state);
+ * @endcode
+ *
+ * Tevent_req_create() creates the state variable as a talloc child of
+ * its result. The state variable should be used as the talloc parent
+ * for all temporary variables that are allocated during the async
+ * computation. This way, when the user of the async computation frees
+ * the request, the state as a talloc child will be free'd along with
+ * all the temporary variables hanging off the state.
+ *
+ * @param[in] mem_ctx   The memory context for the result.
+ * @param[in] pstate    Pointer to the private request state.
+ * @param[in] type      The name of the request.
+ *
+ * @return              A new async request. NULL on error.
+ */
+struct tevent_req *tevent_req_create(TALLOC_CTX *mem_ctx,
+				     void **pstate, #type);
+#else
+struct tevent_req *_tevent_req_create(TALLOC_CTX *mem_ctx,
+				      void *pstate,
+				      size_t state_size,
+				      const char *type,
+				      const char *location);
+
+#define tevent_req_create(_mem_ctx, _pstate, _type) \
+	_tevent_req_create((_mem_ctx), (_pstate), sizeof(_type), \
+			   #_type, __location__)
+#endif
+
+/**
+ * @brief Set a timeout for an async request.
+ *
+ * @param[in]  req      The request to set the timeout for.
+ *
+ * @param[in]  ev       The event context to use for the timer.
+ *
+ * @param[in]  endtime  The endtime of the request.
+ *
+ * @return              True if succeeded, false if not.
+ */
+bool tevent_req_set_endtime(struct tevent_req *req,
+			    struct tevent_context *ev,
+			    struct timeval endtime);
+
+#ifdef DOXYGEN
+/**
+ * @brief Call the notify callback of the given tevent request manually.
+ *
+ * @param[in]  req      The tevent request to call the notify function from.
+ *
+ * @see tevent_req_set_callback()
+ */
+void tevent_req_notify_callback(struct tevent_req *req);
+#else
+void _tevent_req_notify_callback(struct tevent_req *req, const char *location);
+#define tevent_req_notify_callback(req)		\
+	_tevent_req_notify_callback(req, __location__)
+#endif
+
+#ifdef DOXYGEN
+/**
+ * @brief An async request has successfully finished.
+ *
+ * This function is to be used by implementors of async requests. When a
+ * request is successfully finished, this function calls the user's completion
+ * function.
+ *
+ * @param[in]  req       The finished request.
+ */
+void tevent_req_done(struct tevent_req *req);
+#else
+void _tevent_req_done(struct tevent_req *req,
+		      const char *location);
+#define tevent_req_done(req) \
+	_tevent_req_done(req, __location__)
+#endif
+
+#ifdef DOXYGEN
+/**
+ * @brief An async request has seen an error.
+ *
+ * This function is to be used by implementors of async requests. When a
+ * request can not successfully completed, the implementation should call this
+ * function with the appropriate status code.
+ *
+ * If error is 0 the function returns false and does nothing more.
+ *
+ * @param[in]  req      The request with an error.
+ *
+ * @param[in]  error    The error code.
+ *
+ * @return              On success true is returned, false if error is 0.
+ *
+ * @code
+ * int error = first_function();
+ * if (tevent_req_error(req, error)) {
+ *      return;
+ * }
+ *
+ * error = second_function();
+ * if (tevent_req_error(req, error)) {
+ *      return;
+ * }
+ *
+ * tevent_req_done(req);
+ * return;
+ * @endcode
+ */
+bool tevent_req_error(struct tevent_req *req,
+		      uint64_t error);
+#else
+bool _tevent_req_error(struct tevent_req *req,
+		       uint64_t error,
+		       const char *location);
+#define tevent_req_error(req, error) \
+	_tevent_req_error(req, error, __location__)
+#endif
+
+#ifdef DOXYGEN
+/**
+ * @brief Helper function for nomem check.
+ *
+ * Convenience helper to easily check alloc failure within a callback
+ * implementing the next step of an async request.
+ *
+ * @param[in]  p        The pointer to be checked.
+ *
+ * @param[in]  req      The request being processed.
+ *
+ * @code
+ * p = talloc(mem_ctx, bla);
+ * if (tevent_req_nomem(p, req)) {
+ *      return;
+ * }
+ * @endcode
+ */
+bool tevent_req_nomem(const void *p,
+		      struct tevent_req *req);
+#else
+bool _tevent_req_nomem(const void *p,
+		       struct tevent_req *req,
+		       const char *location);
+#define tevent_req_nomem(p, req) \
+	_tevent_req_nomem(p, req, __location__)
+#endif
+
+#ifdef DOXYGEN
+/**
+ * @brief Indicate out of memory to a request
+ *
+ * @param[in]  req      The request being processed.
+ */
+void tevent_req_oom(struct tevent_req *req);
+#else
+void _tevent_req_oom(struct tevent_req *req,
+		     const char *location);
+#define tevent_req_oom(req) \
+	_tevent_req_oom(req, __location__)
+#endif
+
+/**
+ * @brief Finish a request before the caller had the change to set the callback.
+ *
+ * An implementation of an async request might find that it can either finish
+ * the request without waiting for an external event, or it can not even start
+ * the engine. To present the illusion of a callback to the user of the API,
+ * the implementation can call this helper function which triggers an
+ * immediate event. This way the caller can use the same calling
+ * conventions, independent of whether the request was actually deferred.
+ *
+ * @code
+ * struct tevent_req *computation_send(TALLOC_CTX *mem_ctx,
+ *                                     struct tevent_context *ev)
+ * {
+ *     struct tevent_req *req, *subreq;
+ *     struct computation_state *state;
+ *     req = tevent_req_create(mem_ctx, &state, struct computation_state);
+ *     if (req == NULL) {
+ *         return NULL;
+ *     }
+ *     subreq = subcomputation_send(state, ev);
+ *     if (tevent_req_nomem(subreq, req)) {
+ *         return tevent_req_post(req, ev);
+ *     }
+ *     tevent_req_set_callback(subreq, computation_done, req);
+ *     return req;
+ * }
+ * @endcode
+ *
+ * @param[in]  req      The finished request.
+ *
+ * @param[in]  ev       The tevent_context for the immediate event.
+ *
+ * @return              The given request will be returned.
+ */
+struct tevent_req *tevent_req_post(struct tevent_req *req,
+				   struct tevent_context *ev);
+
+/**
+ * @brief Finish multiple requests within one function
+ *
+ * Normally tevent_req_notify_callback() and all wrappers
+ * (e.g. tevent_req_done() and tevent_req_error())
+ * need to be the last thing an event handler should call.
+ * This is because the callback is likely to destroy the
+ * context of the current function.
+ *
+ * If a function wants to notify more than one caller,
+ * it is dangerous if it just triggers multiple callbacks
+ * in a row. With tevent_req_defer_callback() it is possible
+ * to set an event context that will be used to defer the callback
+ * via an immediate event (similar to tevent_req_post()).
+ *
+ * @code
+ * struct complete_state {
+ *       struct tevent_context *ev;
+ *
+ *       struct tevent_req **reqs;
+ * };
+ *
+ * void complete(struct complete_state *state)
+ * {
+ *       size_t i, c = talloc_array_length(state->reqs);
+ *
+ *       for (i=0; i < c; i++) {
+ *            tevent_req_defer_callback(state->reqs[i], state->ev);
+ *            tevent_req_done(state->reqs[i]);
+ *       }
+ * }
+ * @endcode
+ *
+ * @param[in]  req      The finished request.
+ *
+ * @param[in]  ev       The tevent_context for the immediate event.
+ *
+ * @return              The given request will be returned.
+ */
+void tevent_req_defer_callback(struct tevent_req *req,
+			       struct tevent_context *ev);
+
+/**
+ * @brief Check if the given request is still in progress.
+ *
+ * It is typically used by sync wrapper functions.
+ *
+ * @param[in]  req      The request to poll.
+ *
+ * @return              The boolean form of "is in progress".
+ */
+bool tevent_req_is_in_progress(struct tevent_req *req);
+
+/**
+ * @brief Actively poll for the given request to finish.
+ *
+ * This function is typically used by sync wrapper functions.
+ *
+ * @param[in]  req      The request to poll.
+ *
+ * @param[in]  ev       The tevent_context to be used.
+ *
+ * @return              On success true is returned. If a critical error has
+ *                      happened in the tevent loop layer false is returned.
+ *                      This is not the return value of the given request!
+ *
+ * @note This should only be used if the given tevent context was created by the
+ * caller, to avoid event loop nesting.
+ *
+ * @code
+ * req = tstream_writev_queue_send(mem_ctx,
+ *                                 ev_ctx,
+ *                                 tstream,
+ *                                 send_queue,
+ *                                 iov, 2);
+ * ok = tevent_req_poll(req, tctx->ev);
+ * rc = tstream_writev_queue_recv(req, &sys_errno);
+ * TALLOC_FREE(req);
+ * @endcode
+ */
+bool tevent_req_poll(struct tevent_req *req,
+		     struct tevent_context *ev);
+
+/**
+ * @brief Get the tevent request state and the actual error set by
+ * tevent_req_error.
+ *
+ * @code
+ * int computation_recv(struct tevent_req *req, uint64_t *perr)
+ * {
+ *     enum tevent_req_state state;
+ *     uint64_t err;
+ *     if (tevent_req_is_error(req, &state, &err)) {
+ *         *perr = err;
+ *         return -1;
+ *     }
+ *     return 0;
+ * }
+ * @endcode
+ *
+ * @param[in]  req      The tevent request to get the error from.
+ *
+ * @param[out] state    A pointer to store the tevent request error state.
+ *
+ * @param[out] error    A pointer to store the error set by tevent_req_error().
+ *
+ * @return              True if the function could set error and state, false
+ *                      otherwise.
+ *
+ * @see tevent_req_error()
+ */
+bool tevent_req_is_error(struct tevent_req *req,
+			 enum tevent_req_state *state,
+			 uint64_t *error);
+
+/**
+ * @brief Use as the last action of a _recv() function.
+ *
+ * This function destroys the attached private data.
+ *
+ * @param[in]  req      The finished request.
+ */
+void tevent_req_received(struct tevent_req *req);
+
+/**
+ * @brief Create a tevent subrequest at a given time.
+ *
+ * The idea is that always the same syntax for tevent requests.
+ *
+ * @param[in]  mem_ctx  The talloc memory context to use.
+ *
+ * @param[in]  ev       The event handle to setup the request.
+ *
+ * @param[in]  wakeup_time The time to wakeup and execute the request.
+ *
+ * @return              The new subrequest, NULL on error.
+ *
+ * Example:
+ * @code
+ *   static void my_callback_wakeup_done(tevent_req *subreq)
+ *   {
+ *     struct tevent_req *req = tevent_req_callback_data(subreq,
+ *                              struct tevent_req);
+ *     bool ok;
+ *
+ *     ok = tevent_wakeup_recv(subreq);
+ *     TALLOC_FREE(subreq);
+ *     if (!ok) {
+ *         tevent_req_error(req, -1);
+ *         return;
+ *     }
+ *     ...
+ *   }
+ * @endcode
+ *
+ * @code
+ *   subreq = tevent_wakeup_send(mem_ctx, ev, wakeup_time);
+ *   if (tevent_req_nomem(subreq, req)) {
+ *     return false;
+ *   }
+ *   tevent_set_callback(subreq, my_callback_wakeup_done, req);
+ * @endcode
+ *
+ * @see tevent_wakeup_recv()
+ */
+struct tevent_req *tevent_wakeup_send(TALLOC_CTX *mem_ctx,
+				      struct tevent_context *ev,
+				      struct timeval wakeup_time);
+
+/**
+ * @brief Check if the wakeup has been correctly executed.
+ *
+ * This function needs to be called in the callback function set after calling
+ * tevent_wakeup_send().
+ *
+ * @param[in]  req      The tevent request to check.
+ *
+ * @return              True on success, false otherwise.
+ *
+ * @see tevent_wakeup_recv()
+ */
+bool tevent_wakeup_recv(struct tevent_req *req);
+
+/* @} */
+
+/**
+ * @defgroup tevent_helpers The tevent helper functiions
+ * @ingroup tevent
+ *
+ * @todo description
+ *
+ * @{
+ */
+
+/**
+ * @brief Compare two timeval values.
+ *
+ * @param[in]  tv1      The first timeval value to compare.
+ *
+ * @param[in]  tv2      The second timeval value to compare.
+ *
+ * @return              0 if they are equal.
+ *                      1 if the first time is greater than the second.
+ *                      -1 if the first time is smaller than the second.
+ */
+int tevent_timeval_compare(const struct timeval *tv1,
+			   const struct timeval *tv2);
+
+/**
+ * @brief Get a zero timval value.
+ *
+ * @return              A zero timval value.
+ */
+struct timeval tevent_timeval_zero(void);
+
+/**
+ * @brief Get a timeval value for the current time.
+ *
+ * @return              A timval value with the current time.
+ */
+struct timeval tevent_timeval_current(void);
+
+/**
+ * @brief Get a timeval structure with the given values.
+ *
+ * @param[in]  secs     The seconds to set.
+ *
+ * @param[in]  usecs    The microseconds to set.
+ *
+ * @return              A timeval structure with the given values.
+ */
+struct timeval tevent_timeval_set(uint32_t secs, uint32_t usecs);
+
+/**
+ * @brief Get the difference between two timeval values.
+ *
+ * @param[in]  tv1      The first timeval.
+ *
+ * @param[in]  tv2      The second timeval.
+ *
+ * @return              A timeval structure with the difference between the
+ *                      first and the second value.
+ */
+struct timeval tevent_timeval_until(const struct timeval *tv1,
+				    const struct timeval *tv2);
+
+/**
+ * @brief Check if a given timeval structure is zero.
+ *
+ * @param[in]  tv       The timeval to check if it is zero.
+ *
+ * @return              True if it is zero, false otherwise.
+ */
+bool tevent_timeval_is_zero(const struct timeval *tv);
+
+/**
+ * @brief Add the given amount of time to a timeval structure.
+ *
+ * @param[in]  tv        The timeval structure to add the time.
+ *
+ * @param[in]  secs      The seconds to add to the timeval.
+ *
+ * @param[in]  usecs     The microseconds to add to the timeval.
+ *
+ * @return               The timeval structure with the new time.
+ */
+struct timeval tevent_timeval_add(const struct timeval *tv, uint32_t secs,
+				  uint32_t usecs);
+
+/**
+ * @brief Get a timeval in the future with a specified offset from now.
+ *
+ * @param[in]  secs     The seconds of the offset from now.
+ *
+ * @param[in]  usecs    The microseconds of the offset from now.
+ *
+ * @return              A timval with the given offset in the future.
+ */
+struct timeval tevent_timeval_current_ofs(uint32_t secs, uint32_t usecs);
+
+/* @} */
+
+
+/**
+ * @defgroup tevent_queue The tevent queue functions
+ * @ingroup tevent
+ *
+ * A tevent_queue is used to queue up async requests that must be
+ * serialized. For example writing buffers into a socket must be
+ * serialized. Writing a large lump of data into a socket can require
+ * multiple write(2) or send(2) system calls. If more than one async
+ * request is outstanding to write large buffers into a socket, every
+ * request must individually be completed before the next one begins,
+ * even if multiple syscalls are required.
+ *
+ * Take a look at @ref tevent_queue_tutorial for more details.
+ * @{
+ */
+
+struct tevent_queue;
+struct tevent_queue_entry;
+
+#ifdef DOXYGEN
+/**
+ * @brief Create and start a tevent queue.
+ *
+ * @param[in]  mem_ctx  The talloc memory context to allocate the queue.
+ *
+ * @param[in]  name     The name to use to identify the queue.
+ *
+ * @return              An allocated tevent queue on success, NULL on error.
+ *
+ * @see tevent_queue_start()
+ * @see tevent_queue_stop()
+ */
+struct tevent_queue *tevent_queue_create(TALLOC_CTX *mem_ctx,
+					 const char *name);
+#else
+struct tevent_queue *_tevent_queue_create(TALLOC_CTX *mem_ctx,
+					  const char *name,
+					  const char *location);
+
+#define tevent_queue_create(_mem_ctx, _name) \
+	_tevent_queue_create((_mem_ctx), (_name), __location__)
+#endif
+
+/**
+ * @brief A callback trigger function run by the queue.
+ *
+ * @param[in]  req      The tevent request the trigger function is executed on.
+ *
+ * @param[in]  private_data The private data pointer specified by
+ *                          tevent_queue_add().
+ *
+ * @see tevent_queue_add()
+ * @see tevent_queue_add_entry()
+ * @see tevent_queue_add_optimize_empty()
+ */
+typedef void (*tevent_queue_trigger_fn_t)(struct tevent_req *req,
+					  void *private_data);
+
+/**
+ * @brief Add a tevent request to the queue.
+ *
+ * @param[in]  queue    The queue to add the request.
+ *
+ * @param[in]  ev       The event handle to use for the request.
+ *
+ * @param[in]  req      The tevent request to add to the queue.
+ *
+ * @param[in]  trigger  The function triggered by the queue when the request
+ *                      is called. Since tevent 0.9.14 it's possible to
+ *                      pass NULL, in order to just add a "blocker" to the
+ *                      queue.
+ *
+ * @param[in]  private_data The private data passed to the trigger function.
+ *
+ * @return              True if the request has been successfully added, false
+ *                      otherwise.
+ */
+bool tevent_queue_add(struct tevent_queue *queue,
+		      struct tevent_context *ev,
+		      struct tevent_req *req,
+		      tevent_queue_trigger_fn_t trigger,
+		      void *private_data);
+
+/**
+ * @brief Add a tevent request to the queue.
+ *
+ * The request can be removed from the queue by calling talloc_free()
+ * (or a similar function) on the returned queue entry. This
+ * is the only difference to tevent_queue_add().
+ *
+ * @param[in]  queue    The queue to add the request.
+ *
+ * @param[in]  ev       The event handle to use for the request.
+ *
+ * @param[in]  req      The tevent request to add to the queue.
+ *
+ * @param[in]  trigger  The function triggered by the queue when the request
+ *                      is called. Since tevent 0.9.14 it's possible to
+ *                      pass NULL, in order to just add a "blocker" to the
+ *                      queue.
+ *
+ * @param[in]  private_data The private data passed to the trigger function.
+ *
+ * @return              a pointer to the tevent_queue_entry if the request
+ *                      has been successfully added, NULL otherwise.
+ *
+ * @see tevent_queue_add()
+ * @see tevent_queue_add_optimize_empty()
+ */
+struct tevent_queue_entry *tevent_queue_add_entry(
+					struct tevent_queue *queue,
+					struct tevent_context *ev,
+					struct tevent_req *req,
+					tevent_queue_trigger_fn_t trigger,
+					void *private_data);
+
+/**
+ * @brief Add a tevent request to the queue using a possible optimization.
+ *
+ * This tries to optimize for the empty queue case and may calls
+ * the trigger function directly. This is the only difference compared
+ * to tevent_queue_add_entry().
+ *
+ * The caller needs to be prepared that the trigger function has
+ * already called tevent_req_notify_callback(), tevent_req_error(),
+ * tevent_req_done() or a similar function.
+ *
+ * The request can be removed from the queue by calling talloc_free()
+ * (or a similar function) on the returned queue entry.
+ *
+ * @param[in]  queue    The queue to add the request.
+ *
+ * @param[in]  ev       The event handle to use for the request.
+ *
+ * @param[in]  req      The tevent request to add to the queue.
+ *
+ * @param[in]  trigger  The function triggered by the queue when the request
+ *                      is called. Since tevent 0.9.14 it's possible to
+ *                      pass NULL, in order to just add a "blocker" to the
+ *                      queue.
+ *
+ * @param[in]  private_data The private data passed to the trigger function.
+ *
+ * @return              a pointer to the tevent_queue_entry if the request
+ *                      has been successfully added, NULL otherwise.
+ *
+ * @see tevent_queue_add()
+ * @see tevent_queue_add_entry()
+ */
+struct tevent_queue_entry *tevent_queue_add_optimize_empty(
+					struct tevent_queue *queue,
+					struct tevent_context *ev,
+					struct tevent_req *req,
+					tevent_queue_trigger_fn_t trigger,
+					void *private_data);
+
+/**
+ * @brief Start a tevent queue.
+ *
+ * The queue is started by default.
+ *
+ * @param[in]  queue    The queue to start.
+ */
+void tevent_queue_start(struct tevent_queue *queue);
+
+/**
+ * @brief Stop a tevent queue.
+ *
+ * The queue is started by default.
+ *
+ * @param[in]  queue    The queue to stop.
+ */
+void tevent_queue_stop(struct tevent_queue *queue);
+
+/**
+ * @brief Get the length of the queue.
+ *
+ * @param[in]  queue    The queue to get the length from.
+ *
+ * @return              The number of elements.
+ */
+size_t tevent_queue_length(struct tevent_queue *queue);
+
+/**
+ * @brief Is the tevent queue running.
+ *
+ * The queue is started by default.
+ *
+ * @param[in]  queue    The queue.
+ *
+ * @return              Wether the queue is running or not..
+ */
+bool tevent_queue_running(struct tevent_queue *queue);
+
+typedef int (*tevent_nesting_hook)(struct tevent_context *ev,
+				   void *private_data,
+				   uint32_t level,
+				   bool begin,
+				   void *stack_ptr,
+				   const char *location);
+#ifdef TEVENT_DEPRECATED
+#ifndef _DEPRECATED_
+#if (__GNUC__ >= 3) && (__GNUC_MINOR__ >= 1 )
+#define _DEPRECATED_ __attribute__ ((deprecated))
+#else
+#define _DEPRECATED_
+#endif
+#endif
+void tevent_loop_allow_nesting(struct tevent_context *ev) _DEPRECATED_;
+void tevent_loop_set_nesting_hook(struct tevent_context *ev,
+				  tevent_nesting_hook hook,
+				  void *private_data) _DEPRECATED_;
+int _tevent_loop_until(struct tevent_context *ev,
+		       bool (*finished)(void *private_data),
+		       void *private_data,
+		       const char *location) _DEPRECATED_;
+#define tevent_loop_until(ev, finished, private_data) \
+	_tevent_loop_until(ev, finished, private_data, __location__)
+#endif
+
+int tevent_re_initialise(struct tevent_context *ev);
+
+/* @} */
+
+/**
+ * @defgroup tevent_ops The tevent operation functions
+ * @ingroup tevent
+ *
+ * The following structure and registration functions are exclusively
+ * needed for people writing and pluggin a different event engine.
+ * There is nothing useful for normal tevent user in here.
+ * @{
+ */
+
+struct tevent_ops {
+	/* context init */
+	int (*context_init)(struct tevent_context *ev);
+
+	/* fd_event functions */
+	struct tevent_fd *(*add_fd)(struct tevent_context *ev,
+				    TALLOC_CTX *mem_ctx,
+				    int fd, uint16_t flags,
+				    tevent_fd_handler_t handler,
+				    void *private_data,
+				    const char *handler_name,
+				    const char *location);
+	void (*set_fd_close_fn)(struct tevent_fd *fde,
+				tevent_fd_close_fn_t close_fn);
+	uint16_t (*get_fd_flags)(struct tevent_fd *fde);
+	void (*set_fd_flags)(struct tevent_fd *fde, uint16_t flags);
+
+	/* timed_event functions */
+	struct tevent_timer *(*add_timer)(struct tevent_context *ev,
+					  TALLOC_CTX *mem_ctx,
+					  struct timeval next_event,
+					  tevent_timer_handler_t handler,
+					  void *private_data,
+					  const char *handler_name,
+					  const char *location);
+
+	/* immediate event functions */
+	void (*schedule_immediate)(struct tevent_immediate *im,
+				   struct tevent_context *ev,
+				   tevent_immediate_handler_t handler,
+				   void *private_data,
+				   const char *handler_name,
+				   const char *location);
+
+	/* signal functions */
+	struct tevent_signal *(*add_signal)(struct tevent_context *ev,
+					    TALLOC_CTX *mem_ctx,
+					    int signum, int sa_flags,
+					    tevent_signal_handler_t handler,
+					    void *private_data,
+					    const char *handler_name,
+					    const char *location);
+
+	/* loop functions */
+	int (*loop_once)(struct tevent_context *ev, const char *location);
+	int (*loop_wait)(struct tevent_context *ev, const char *location);
+};
+
+bool tevent_register_backend(const char *name, const struct tevent_ops *ops);
+
+/* @} */
+
+/**
+ * @defgroup tevent_compat The tevent compatibility functions
+ * @ingroup tevent
+ *
+ * The following definitions are usueful only for compatibility with the
+ * implementation originally developed within the samba4 code and will be
+ * soon removed. Please NEVER use in new code.
+ *
+ * @todo Ignore it?
+ *
+ * @{
+ */
+
+#ifdef TEVENT_COMPAT_DEFINES
+
+#define event_context	tevent_context
+#define event_ops	tevent_ops
+#define fd_event	tevent_fd
+#define timed_event	tevent_timer
+#define signal_event	tevent_signal
+
+#define event_fd_handler_t	tevent_fd_handler_t
+#define event_timed_handler_t	tevent_timer_handler_t
+#define event_signal_handler_t	tevent_signal_handler_t
+
+#define event_context_init(mem_ctx) \
+	tevent_context_init(mem_ctx)
+
+#define event_context_init_byname(mem_ctx, name) \
+	tevent_context_init_byname(mem_ctx, name)
+
+#define event_backend_list(mem_ctx) \
+	tevent_backend_list(mem_ctx)
+
+#define event_set_default_backend(backend) \
+	tevent_set_default_backend(backend)
+
+#define event_add_fd(ev, mem_ctx, fd, flags, handler, private_data) \
+	tevent_add_fd(ev, mem_ctx, fd, flags, handler, private_data)
+
+#define event_add_timed(ev, mem_ctx, next_event, handler, private_data) \
+	tevent_add_timer(ev, mem_ctx, next_event, handler, private_data)
+
+#define event_add_signal(ev, mem_ctx, signum, sa_flags, handler, private_data) \
+	tevent_add_signal(ev, mem_ctx, signum, sa_flags, handler, private_data)
+
+#define event_loop_once(ev) \
+	tevent_loop_once(ev)
+
+#define event_loop_wait(ev) \
+	tevent_loop_wait(ev)
+
+#define event_get_fd_flags(fde) \
+	tevent_fd_get_flags(fde)
+
+#define event_set_fd_flags(fde, flags) \
+	tevent_fd_set_flags(fde, flags)
+
+#define EVENT_FD_READ		TEVENT_FD_READ
+#define EVENT_FD_WRITE		TEVENT_FD_WRITE
+
+#define EVENT_FD_WRITEABLE(fde) \
+	TEVENT_FD_WRITEABLE(fde)
+
+#define EVENT_FD_READABLE(fde) \
+	TEVENT_FD_READABLE(fde)
+
+#define EVENT_FD_NOT_WRITEABLE(fde) \
+	TEVENT_FD_NOT_WRITEABLE(fde)
+
+#define EVENT_FD_NOT_READABLE(fde) \
+	TEVENT_FD_NOT_READABLE(fde)
+
+#define ev_debug_level		tevent_debug_level
+
+#define EV_DEBUG_FATAL		TEVENT_DEBUG_FATAL
+#define EV_DEBUG_ERROR		TEVENT_DEBUG_ERROR
+#define EV_DEBUG_WARNING	TEVENT_DEBUG_WARNING
+#define EV_DEBUG_TRACE		TEVENT_DEBUG_TRACE
+
+#define ev_set_debug(ev, debug, context) \
+	tevent_set_debug(ev, debug, context)
+
+#define ev_set_debug_stderr(_ev) tevent_set_debug_stderr(ev)
+
+#endif /* TEVENT_COMPAT_DEFINES */
+
+/* @} */
+
+#endif /* __TEVENT_H__ */
diff --git a/ctdb/lib/tevent/tevent.pc.in b/ctdb/lib/tevent/tevent.pc.in
new file mode 100644
index 0000000000..1091ff00f1
--- /dev/null
+++ b/ctdb/lib/tevent/tevent.pc.in
@@ -0,0 +1,12 @@
+prefix=@prefix@
+exec_prefix=@exec_prefix@
+libdir=@libdir@
+includedir=@includedir@
+
+Name: tevent
+Description: An event system library
+Version: @PACKAGE_VERSION@
+Requires: talloc
+Libs: -L${libdir} -ltevent
+Cflags: @LIB_RPATH@ -I${includedir}
+URL: http://samba.org/
diff --git a/ctdb/lib/tevent/tevent.py b/ctdb/lib/tevent/tevent.py
new file mode 100644
index 0000000000..c296544eb4
--- /dev/null
+++ b/ctdb/lib/tevent/tevent.py
@@ -0,0 +1,29 @@
+#!/usr/bin/python
+#
+#   Python integration for tevent
+#
+#   Copyright (C) Jelmer Vernooij 2011
+#
+#     ** NOTE! The following LGPL license applies to the tevent
+#     ** library. This does NOT imply that all of Samba is released
+#     ** under the LGPL
+#
+#   This library is free software; you can redistribute it and/or
+#   modify it under the terms of the GNU Lesser General Public
+#   License as published by the Free Software Foundation; either
+#   version 3 of the License, or (at your option) any later version.
+#
+#   This library is distributed in the hope that it will be useful,
+#   but WITHOUT ANY WARRANTY; without even the implied warranty of
+#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+#   Lesser General Public License for more details.
+#
+#   You should have received a copy of the GNU Lesser General Public
+#   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+
+from _tevent import (
+    __version__,
+    backend_list,
+    Context,
+    Signal,
+    )
diff --git a/ctdb/lib/tevent/tevent_debug.c b/ctdb/lib/tevent/tevent_debug.c
new file mode 100644
index 0000000000..31da7b9683
--- /dev/null
+++ b/ctdb/lib/tevent/tevent_debug.c
@@ -0,0 +1,119 @@
+/*
+   Unix SMB/CIFS implementation.
+
+   Copyright (C) Andrew Tridgell 2005
+   Copyright (C) Jelmer Vernooij 2005
+
+     ** NOTE! The following LGPL license applies to the tevent
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "replace.h"
+#include "tevent.h"
+#include "tevent_internal.h"
+
+/********************************************************************
+ * Debug wrapper functions, modeled (with lot's of code copied as is)
+ * after the ev debug wrapper functions
+ ********************************************************************/
+
+/*
+  this allows the user to choose their own debug function
+*/
+int tevent_set_debug(struct tevent_context *ev,
+		     void (*debug)(void *context,
+				   enum tevent_debug_level level,
+				   const char *fmt,
+				   va_list ap) PRINTF_ATTRIBUTE(3,0),
+		     void *context)
+{
+	ev->debug_ops.debug = debug;
+	ev->debug_ops.context = context;
+	return 0;
+}
+
+/*
+  debug function for ev_set_debug_stderr
+*/
+static void tevent_debug_stderr(void *private_data,
+				enum tevent_debug_level level,
+				const char *fmt,
+				va_list ap) PRINTF_ATTRIBUTE(3,0);
+static void tevent_debug_stderr(void *private_data,
+				enum tevent_debug_level level,
+				const char *fmt, va_list ap)
+{
+	if (level <= TEVENT_DEBUG_WARNING) {
+		vfprintf(stderr, fmt, ap);
+	}
+}
+
+/*
+  convenience function to setup debug messages on stderr
+  messages of level TEVENT_DEBUG_WARNING and higher are printed
+*/
+int tevent_set_debug_stderr(struct tevent_context *ev)
+{
+	return tevent_set_debug(ev, tevent_debug_stderr, ev);
+}
+
+/*
+ * log a message
+ *
+ * The default debug action is to ignore debugging messages.
+ * This is the most appropriate action for a library.
+ * Applications using the library must decide where to
+ * redirect debugging messages
+*/
+void tevent_debug(struct tevent_context *ev, enum tevent_debug_level level,
+		  const char *fmt, ...)
+{
+	va_list ap;
+	if (!ev) {
+		return;
+	}
+	if (ev->debug_ops.debug == NULL) {
+		return;
+	}
+	va_start(ap, fmt);
+	ev->debug_ops.debug(ev->debug_ops.context, level, fmt, ap);
+	va_end(ap);
+}
+
+void tevent_set_trace_callback(struct tevent_context *ev,
+			       tevent_trace_callback_t cb,
+			       void *private_data)
+{
+	ev->tracing.callback = cb;
+	ev->tracing.private_data = private_data;
+}
+
+void tevent_get_trace_callback(struct tevent_context *ev,
+			       tevent_trace_callback_t *cb,
+			       void *private_data)
+{
+	*cb = ev->tracing.callback;
+	*(void**)private_data = ev->tracing.private_data;
+}
+
+void tevent_trace_point_callback(struct tevent_context *ev,
+				 enum tevent_trace_point tp)
+{
+	if (ev->tracing.callback != NULL) {
+		ev->tracing.callback(tp, ev->tracing.private_data);
+	}
+}
diff --git a/ctdb/lib/tevent/tevent_epoll.c b/ctdb/lib/tevent/tevent_epoll.c
new file mode 100644
index 0000000000..599c190658
--- /dev/null
+++ b/ctdb/lib/tevent/tevent_epoll.c
@@ -0,0 +1,950 @@
+/* 
+   Unix SMB/CIFS implementation.
+
+   main select loop and event handling - epoll implementation
+
+   Copyright (C) Andrew Tridgell	2003-2005
+   Copyright (C) Stefan Metzmacher	2005-2013
+   Copyright (C) Jeremy Allison		2013
+
+     ** NOTE! The following LGPL license applies to the tevent
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "replace.h"
+#include "system/filesys.h"
+#include "system/select.h"
+#include "tevent.h"
+#include "tevent_internal.h"
+#include "tevent_util.h"
+
+struct epoll_event_context {
+	/* a pointer back to the generic event_context */
+	struct tevent_context *ev;
+
+	/* when using epoll this is the handle from epoll_create */
+	int epoll_fd;
+
+	pid_t pid;
+
+	bool panic_force_replay;
+	bool *panic_state;
+	bool (*panic_fallback)(struct tevent_context *ev, bool replay);
+};
+
+#define EPOLL_ADDITIONAL_FD_FLAG_HAS_EVENT	(1<<0)
+#define EPOLL_ADDITIONAL_FD_FLAG_REPORT_ERROR	(1<<1)
+#define EPOLL_ADDITIONAL_FD_FLAG_GOT_ERROR	(1<<2)
+#define EPOLL_ADDITIONAL_FD_FLAG_HAS_MPX	(1<<3)
+
+#ifdef TEST_PANIC_FALLBACK
+
+static int epoll_create_panic_fallback(struct epoll_event_context *epoll_ev,
+				       int size)
+{
+	if (epoll_ev->panic_fallback == NULL) {
+		return epoll_create(size);
+	}
+
+	/* 50% of the time, fail... */
+	if ((random() % 2) == 0) {
+		errno = EINVAL;
+		return -1;
+	}
+
+	return epoll_create(size);
+}
+
+static int epoll_ctl_panic_fallback(struct epoll_event_context *epoll_ev,
+				    int epfd, int op, int fd,
+				    struct epoll_event *event)
+{
+	if (epoll_ev->panic_fallback == NULL) {
+		return epoll_ctl(epfd, op, fd, event);
+	}
+
+	/* 50% of the time, fail... */
+	if ((random() % 2) == 0) {
+		errno = EINVAL;
+		return -1;
+	}
+
+	return epoll_ctl(epfd, op, fd, event);
+}
+
+static int epoll_wait_panic_fallback(struct epoll_event_context *epoll_ev,
+				     int epfd,
+				     struct epoll_event *events,
+				     int maxevents,
+				     int timeout)
+{
+	if (epoll_ev->panic_fallback == NULL) {
+		return epoll_wait(epfd, events, maxevents, timeout);
+	}
+
+	/* 50% of the time, fail... */
+	if ((random() % 2) == 0) {
+		errno = EINVAL;
+		return -1;
+	}
+
+	return epoll_wait(epfd, events, maxevents, timeout);
+}
+
+#define epoll_create(_size) \
+	epoll_create_panic_fallback(epoll_ev, _size)
+#define epoll_ctl(_epfd, _op, _fd, _event) \
+	epoll_ctl_panic_fallback(epoll_ev,_epfd, _op, _fd, _event)
+#define epoll_wait(_epfd, _events, _maxevents, _timeout) \
+	epoll_wait_panic_fallback(epoll_ev, _epfd, _events, _maxevents, _timeout)
+#endif
+
+/*
+  called to set the panic fallback function.
+*/
+_PRIVATE_ bool tevent_epoll_set_panic_fallback(struct tevent_context *ev,
+				bool (*panic_fallback)(struct tevent_context *ev,
+						       bool replay))
+{
+	struct epoll_event_context *epoll_ev;
+
+	if (ev->additional_data == NULL) {
+		return false;
+	}
+
+	epoll_ev = talloc_get_type(ev->additional_data,
+				struct epoll_event_context);
+	if (epoll_ev == NULL) {
+		return false;
+	}
+	epoll_ev->panic_fallback = panic_fallback;
+	return true;
+}
+
+/*
+  called when a epoll call fails
+*/
+static void epoll_panic(struct epoll_event_context *epoll_ev,
+			const char *reason, bool replay)
+{
+	struct tevent_context *ev = epoll_ev->ev;
+	bool (*panic_fallback)(struct tevent_context *ev, bool replay);
+
+	panic_fallback = epoll_ev->panic_fallback;
+
+	if (epoll_ev->panic_state != NULL) {
+		*epoll_ev->panic_state = true;
+	}
+
+	if (epoll_ev->panic_force_replay) {
+		replay = true;
+	}
+
+	TALLOC_FREE(ev->additional_data);
+
+	if (panic_fallback == NULL) {
+		tevent_debug(ev, TEVENT_DEBUG_FATAL,
+			"%s (%s) replay[%u] - calling abort()\n",
+			reason, strerror(errno), (unsigned)replay);
+		abort();
+	}
+
+	tevent_debug(ev, TEVENT_DEBUG_ERROR,
+		     "%s (%s) replay[%u] - calling panic_fallback\n",
+		     reason, strerror(errno), (unsigned)replay);
+
+	if (!panic_fallback(ev, replay)) {
+		/* Fallback failed. */
+		tevent_debug(ev, TEVENT_DEBUG_FATAL,
+			"%s (%s) replay[%u] - calling abort()\n",
+			reason, strerror(errno), (unsigned)replay);
+		abort();
+	}
+}
+
+/*
+  map from TEVENT_FD_* to EPOLLIN/EPOLLOUT
+*/
+static uint32_t epoll_map_flags(uint16_t flags)
+{
+	uint32_t ret = 0;
+	if (flags & TEVENT_FD_READ) ret |= (EPOLLIN | EPOLLERR | EPOLLHUP);
+	if (flags & TEVENT_FD_WRITE) ret |= (EPOLLOUT | EPOLLERR | EPOLLHUP);
+	return ret;
+}
+
+/*
+ free the epoll fd
+*/
+static int epoll_ctx_destructor(struct epoll_event_context *epoll_ev)
+{
+	close(epoll_ev->epoll_fd);
+	epoll_ev->epoll_fd = -1;
+	return 0;
+}
+
+/*
+ init the epoll fd
+*/
+static int epoll_init_ctx(struct epoll_event_context *epoll_ev)
+{
+	epoll_ev->epoll_fd = epoll_create(64);
+	if (epoll_ev->epoll_fd == -1) {
+		tevent_debug(epoll_ev->ev, TEVENT_DEBUG_FATAL,
+			     "Failed to create epoll handle.\n");
+		return -1;
+	}
+
+	if (!ev_set_close_on_exec(epoll_ev->epoll_fd)) {
+		tevent_debug(epoll_ev->ev, TEVENT_DEBUG_WARNING,
+			     "Failed to set close-on-exec, file descriptor may be leaked to children.\n");
+	}
+
+	epoll_ev->pid = getpid();
+	talloc_set_destructor(epoll_ev, epoll_ctx_destructor);
+
+	return 0;
+}
+
+static void epoll_update_event(struct epoll_event_context *epoll_ev, struct tevent_fd *fde);
+
+/*
+  reopen the epoll handle when our pid changes
+  see http://junkcode.samba.org/ftp/unpacked/junkcode/epoll_fork.c for an 
+  demonstration of why this is needed
+ */
+static void epoll_check_reopen(struct epoll_event_context *epoll_ev)
+{
+	struct tevent_fd *fde;
+	bool *caller_panic_state = epoll_ev->panic_state;
+	bool panic_triggered = false;
+
+	if (epoll_ev->pid == getpid()) {
+		return;
+	}
+
+	close(epoll_ev->epoll_fd);
+	epoll_ev->epoll_fd = epoll_create(64);
+	if (epoll_ev->epoll_fd == -1) {
+		epoll_panic(epoll_ev, "epoll_create() failed", false);
+		return;
+	}
+
+	if (!ev_set_close_on_exec(epoll_ev->epoll_fd)) {
+		tevent_debug(epoll_ev->ev, TEVENT_DEBUG_WARNING,
+			     "Failed to set close-on-exec, file descriptor may be leaked to children.\n");
+	}
+
+	epoll_ev->pid = getpid();
+	epoll_ev->panic_state = &panic_triggered;
+	for (fde=epoll_ev->ev->fd_events;fde;fde=fde->next) {
+		fde->additional_flags &= ~EPOLL_ADDITIONAL_FD_FLAG_HAS_EVENT;
+		epoll_update_event(epoll_ev, fde);
+
+		if (panic_triggered) {
+			if (caller_panic_state != NULL) {
+				*caller_panic_state = true;
+			}
+			return;
+		}
+	}
+	epoll_ev->panic_state = NULL;
+}
+
+/*
+ epoll cannot add the same file descriptor twice, once
+ with read, once with write which is allowed by the
+ tevent backend. Multiplex the existing fde, flag it
+ as such so we can search for the correct fde on
+ event triggering.
+*/
+
+static int epoll_add_multiplex_fd(struct epoll_event_context *epoll_ev,
+				  struct tevent_fd *add_fde)
+{
+	struct epoll_event event;
+	struct tevent_fd *mpx_fde;
+	int ret;
+
+	/* Find the existing fde that caused the EEXIST error. */
+	for (mpx_fde = epoll_ev->ev->fd_events; mpx_fde; mpx_fde = mpx_fde->next) {
+		if (mpx_fde->fd != add_fde->fd) {
+			continue;
+		}
+
+		if (mpx_fde == add_fde) {
+			continue;
+		}
+
+		break;
+	}
+	if (mpx_fde == NULL) {
+		tevent_debug(epoll_ev->ev, TEVENT_DEBUG_FATAL,
+			     "can't find multiplex fde for fd[%d]",
+			     add_fde->fd);
+		return -1;
+	}
+
+	if (mpx_fde->additional_flags & EPOLL_ADDITIONAL_FD_FLAG_HAS_MPX) {
+		/* Logic error. Can't have more than 2 multiplexed fde's. */
+		tevent_debug(epoll_ev->ev, TEVENT_DEBUG_FATAL,
+			     "multiplex fde for fd[%d] is already multiplexed\n",
+			     mpx_fde->fd);
+		return -1;
+	}
+
+	/*
+	 * The multiplex fde must have the same fd, and also
+	 * already have an epoll event attached.
+	 */
+	if (!(mpx_fde->additional_flags & EPOLL_ADDITIONAL_FD_FLAG_HAS_EVENT)) {
+		/* Logic error. Can't have more than 2 multiplexed fde's. */
+		tevent_debug(epoll_ev->ev, TEVENT_DEBUG_FATAL,
+			     "multiplex fde for fd[%d] has no event\n",
+			     mpx_fde->fd);
+		return -1;
+	}
+
+	/* Modify the mpx_fde to add in the new flags. */
+	ZERO_STRUCT(event);
+	event.events = epoll_map_flags(mpx_fde->flags);
+	event.events |= epoll_map_flags(add_fde->flags);
+	event.data.ptr = mpx_fde;
+	ret = epoll_ctl(epoll_ev->epoll_fd, EPOLL_CTL_MOD, mpx_fde->fd, &event);
+	if (ret != 0 && errno == EBADF) {
+		tevent_debug(epoll_ev->ev, TEVENT_DEBUG_ERROR,
+			     "EPOLL_CTL_MOD EBADF for "
+			     "add_fde[%p] mpx_fde[%p] fd[%d] - disabling\n",
+			     add_fde, mpx_fde, add_fde->fd);
+		DLIST_REMOVE(epoll_ev->ev->fd_events, mpx_fde);
+		mpx_fde->event_ctx = NULL;
+		DLIST_REMOVE(epoll_ev->ev->fd_events, add_fde);
+		add_fde->event_ctx = NULL;
+		return 0;
+	} else if (ret != 0) {
+		return ret;
+	}
+
+	/*
+	 * Make each fde->additional_data pointers point at each other
+	 * so we can look them up from each other. They are now paired.
+	 */
+	mpx_fde->additional_data = (struct tevent_fd *)add_fde;
+	add_fde->additional_data = (struct tevent_fd *)mpx_fde;
+
+	/* Now flag both fde's as being multiplexed. */
+	mpx_fde->additional_flags |= EPOLL_ADDITIONAL_FD_FLAG_HAS_MPX;
+	add_fde->additional_flags |= EPOLL_ADDITIONAL_FD_FLAG_HAS_MPX;
+
+	/* we need to keep the GOT_ERROR flag */
+	if (mpx_fde->additional_flags & EPOLL_ADDITIONAL_FD_FLAG_GOT_ERROR) {
+		add_fde->additional_flags |= EPOLL_ADDITIONAL_FD_FLAG_GOT_ERROR;
+	}
+
+	return 0;
+}
+
+/*
+ add the epoll event to the given fd_event
+*/
+static void epoll_add_event(struct epoll_event_context *epoll_ev, struct tevent_fd *fde)
+{
+	struct epoll_event event;
+	int ret;
+	struct tevent_fd *mpx_fde = NULL;
+
+	fde->additional_flags &= ~EPOLL_ADDITIONAL_FD_FLAG_HAS_EVENT;
+	fde->additional_flags &= ~EPOLL_ADDITIONAL_FD_FLAG_REPORT_ERROR;
+
+	if (fde->additional_flags & EPOLL_ADDITIONAL_FD_FLAG_HAS_MPX) {
+		/*
+		 * This is a multiplexed fde, we need to include both
+		 * flags in the modified event.
+		 */
+		mpx_fde = talloc_get_type_abort(fde->additional_data,
+						struct tevent_fd);
+
+		mpx_fde->additional_flags &= ~EPOLL_ADDITIONAL_FD_FLAG_HAS_EVENT;
+		mpx_fde->additional_flags &= ~EPOLL_ADDITIONAL_FD_FLAG_REPORT_ERROR;
+	}
+
+	ZERO_STRUCT(event);
+	event.events = epoll_map_flags(fde->flags);
+	if (mpx_fde != NULL) {
+		event.events |= epoll_map_flags(mpx_fde->flags);
+	}
+	event.data.ptr = fde;
+	ret = epoll_ctl(epoll_ev->epoll_fd, EPOLL_CTL_ADD, fde->fd, &event);
+	if (ret != 0 && errno == EBADF) {
+		tevent_debug(epoll_ev->ev, TEVENT_DEBUG_ERROR,
+			     "EPOLL_CTL_ADD EBADF for "
+			     "fde[%p] mpx_fde[%p] fd[%d] - disabling\n",
+			     fde, mpx_fde, fde->fd);
+		DLIST_REMOVE(epoll_ev->ev->fd_events, fde);
+		fde->event_ctx = NULL;
+		if (mpx_fde != NULL) {
+			DLIST_REMOVE(epoll_ev->ev->fd_events, mpx_fde);
+			mpx_fde->event_ctx = NULL;
+		}
+		return;
+	} else if (ret != 0 && errno == EEXIST && mpx_fde == NULL) {
+		ret = epoll_add_multiplex_fd(epoll_ev, fde);
+		if (ret != 0) {
+			epoll_panic(epoll_ev, "epoll_add_multiplex_fd failed",
+				    false);
+			return;
+		}
+	} else if (ret != 0) {
+		epoll_panic(epoll_ev, "EPOLL_CTL_ADD failed", false);
+		return;
+	}
+
+	fde->additional_flags |= EPOLL_ADDITIONAL_FD_FLAG_HAS_EVENT;
+	/* only if we want to read we want to tell the event handler about errors */
+	if (fde->flags & TEVENT_FD_READ) {
+		fde->additional_flags |= EPOLL_ADDITIONAL_FD_FLAG_REPORT_ERROR;
+	}
+
+	if (mpx_fde == NULL) {
+		return;
+	}
+
+	mpx_fde->additional_flags |= EPOLL_ADDITIONAL_FD_FLAG_HAS_EVENT;
+	/* only if we want to read we want to tell the event handler about errors */
+	if (mpx_fde->flags & TEVENT_FD_READ) {
+		mpx_fde->additional_flags |= EPOLL_ADDITIONAL_FD_FLAG_REPORT_ERROR;
+	}
+}
+
+/*
+ delete the epoll event for given fd_event
+*/
+static void epoll_del_event(struct epoll_event_context *epoll_ev, struct tevent_fd *fde)
+{
+	struct epoll_event event;
+	int ret;
+	struct tevent_fd *mpx_fde = NULL;
+
+	fde->additional_flags &= ~EPOLL_ADDITIONAL_FD_FLAG_HAS_EVENT;
+	fde->additional_flags &= ~EPOLL_ADDITIONAL_FD_FLAG_REPORT_ERROR;
+
+	if (fde->additional_flags & EPOLL_ADDITIONAL_FD_FLAG_HAS_MPX) {
+		/*
+		 * This is a multiplexed fde, we need to modify both events.
+		 */
+		mpx_fde = talloc_get_type_abort(fde->additional_data,
+						struct tevent_fd);
+
+		mpx_fde->additional_flags &= ~EPOLL_ADDITIONAL_FD_FLAG_HAS_EVENT;
+		mpx_fde->additional_flags &= ~EPOLL_ADDITIONAL_FD_FLAG_REPORT_ERROR;
+	}
+
+	ZERO_STRUCT(event);
+	ret = epoll_ctl(epoll_ev->epoll_fd, EPOLL_CTL_DEL, fde->fd, &event);
+	if (ret != 0 && errno == ENOENT) {
+		/*
+		 * This can happen after a epoll_check_reopen
+		 * within epoll_event_fd_destructor.
+		 */
+		tevent_debug(epoll_ev->ev, TEVENT_DEBUG_TRACE,
+			     "EPOLL_CTL_DEL ignoring ENOENT for fd[%d]\n",
+			     fde->fd);
+		return;
+	} else if (ret != 0 && errno == EBADF) {
+		tevent_debug(epoll_ev->ev, TEVENT_DEBUG_WARNING,
+			     "EPOLL_CTL_DEL EBADF for "
+			     "fde[%p] mpx_fde[%p] fd[%d] - disabling\n",
+			     fde, mpx_fde, fde->fd);
+		DLIST_REMOVE(epoll_ev->ev->fd_events, fde);
+		fde->event_ctx = NULL;
+		if (mpx_fde != NULL) {
+			DLIST_REMOVE(epoll_ev->ev->fd_events, mpx_fde);
+			mpx_fde->event_ctx = NULL;
+		}
+		return;
+	} else if (ret != 0) {
+		epoll_panic(epoll_ev, "EPOLL_CTL_DEL failed", false);
+		return;
+	}
+}
+
+/*
+ change the epoll event to the given fd_event
+*/
+static void epoll_mod_event(struct epoll_event_context *epoll_ev, struct tevent_fd *fde)
+{
+	struct tevent_fd *mpx_fde = NULL;
+	struct epoll_event event;
+	int ret;
+
+	fde->additional_flags &= ~EPOLL_ADDITIONAL_FD_FLAG_HAS_EVENT;
+	fde->additional_flags &= ~EPOLL_ADDITIONAL_FD_FLAG_REPORT_ERROR;
+
+	if (fde->additional_flags & EPOLL_ADDITIONAL_FD_FLAG_HAS_MPX) {
+		/*
+		 * This is a multiplexed fde, we need to include both
+		 * flags in the modified event.
+		 */
+		mpx_fde = talloc_get_type_abort(fde->additional_data,
+						struct tevent_fd);
+
+		mpx_fde->additional_flags &= ~EPOLL_ADDITIONAL_FD_FLAG_HAS_EVENT;
+		mpx_fde->additional_flags &= ~EPOLL_ADDITIONAL_FD_FLAG_REPORT_ERROR;
+	}
+
+	ZERO_STRUCT(event);
+	event.events = epoll_map_flags(fde->flags);
+	if (mpx_fde != NULL) {
+		event.events |= epoll_map_flags(mpx_fde->flags);
+	}
+	event.data.ptr = fde;
+	ret = epoll_ctl(epoll_ev->epoll_fd, EPOLL_CTL_MOD, fde->fd, &event);
+	if (ret != 0 && errno == EBADF) {
+		tevent_debug(epoll_ev->ev, TEVENT_DEBUG_ERROR,
+			     "EPOLL_CTL_MOD EBADF for "
+			     "fde[%p] mpx_fde[%p] fd[%d] - disabling\n",
+			     fde, mpx_fde, fde->fd);
+		DLIST_REMOVE(epoll_ev->ev->fd_events, fde);
+		fde->event_ctx = NULL;
+		if (mpx_fde != NULL) {
+			DLIST_REMOVE(epoll_ev->ev->fd_events, mpx_fde);
+			mpx_fde->event_ctx = NULL;
+		}
+		return;
+	} else if (ret != 0) {
+		epoll_panic(epoll_ev, "EPOLL_CTL_MOD failed", false);
+		return;
+	}
+
+	fde->additional_flags |= EPOLL_ADDITIONAL_FD_FLAG_HAS_EVENT;
+	/* only if we want to read we want to tell the event handler about errors */
+	if (fde->flags & TEVENT_FD_READ) {
+		fde->additional_flags |= EPOLL_ADDITIONAL_FD_FLAG_REPORT_ERROR;
+	}
+
+	if (mpx_fde == NULL) {
+		return;
+	}
+
+	mpx_fde->additional_flags |= EPOLL_ADDITIONAL_FD_FLAG_HAS_EVENT;
+	/* only if we want to read we want to tell the event handler about errors */
+	if (mpx_fde->flags & TEVENT_FD_READ) {
+		mpx_fde->additional_flags |= EPOLL_ADDITIONAL_FD_FLAG_REPORT_ERROR;
+	}
+}
+
+static void epoll_update_event(struct epoll_event_context *epoll_ev, struct tevent_fd *fde)
+{
+	bool got_error = (fde->additional_flags & EPOLL_ADDITIONAL_FD_FLAG_GOT_ERROR);
+	bool want_read = (fde->flags & TEVENT_FD_READ);
+	bool want_write= (fde->flags & TEVENT_FD_WRITE);
+	struct tevent_fd *mpx_fde = NULL;
+
+	if (fde->additional_flags & EPOLL_ADDITIONAL_FD_FLAG_HAS_MPX) {
+		/*
+		 * work out what the multiplexed fde wants.
+		 */
+		mpx_fde = talloc_get_type_abort(fde->additional_data,
+						struct tevent_fd);
+
+		if (mpx_fde->flags & TEVENT_FD_READ) {
+			want_read = true;
+		}
+
+		if (mpx_fde->flags & TEVENT_FD_WRITE) {
+			want_write = true;
+		}
+	}
+
+	/* there's already an event */
+	if (fde->additional_flags & EPOLL_ADDITIONAL_FD_FLAG_HAS_EVENT) {
+		if (want_read || (want_write && !got_error)) {
+			epoll_mod_event(epoll_ev, fde);
+			return;
+		}
+		/* 
+		 * if we want to match the select behavior, we need to remove the epoll_event
+		 * when the caller isn't interested in events.
+		 *
+		 * this is because epoll reports EPOLLERR and EPOLLHUP, even without asking for them
+		 */
+		epoll_del_event(epoll_ev, fde);
+		return;
+	}
+
+	/* there's no epoll_event attached to the fde */
+	if (want_read || (want_write && !got_error)) {
+		epoll_add_event(epoll_ev, fde);
+		return;
+	}
+}
+
+/*
+  Cope with epoll returning EPOLLHUP|EPOLLERR on an event.
+  Return true if there's nothing else to do, false if
+  this event needs further handling.
+*/
+static bool epoll_handle_hup_or_err(struct epoll_event_context *epoll_ev,
+				struct tevent_fd *fde)
+{
+	if (fde == NULL) {
+		/* Nothing to do if no event. */
+		return true;
+	}
+
+	fde->additional_flags |= EPOLL_ADDITIONAL_FD_FLAG_GOT_ERROR;
+	/*
+	 * if we only wait for TEVENT_FD_WRITE, we should not tell the
+	 * event handler about it, and remove the epoll_event,
+	 * as we only report errors when waiting for read events,
+	 * to match the select() behavior
+	 */
+	if (!(fde->additional_flags & EPOLL_ADDITIONAL_FD_FLAG_REPORT_ERROR)) {
+		/*
+		 * Do the same as the poll backend and
+		 * remove the writeable flag.
+		 */
+		fde->flags &= ~TEVENT_FD_WRITE;
+		return true;
+	}
+	/* This has TEVENT_FD_READ set, we're not finished. */
+	return false;
+}
+
+/*
+  event loop handling using epoll
+*/
+static int epoll_event_loop(struct epoll_event_context *epoll_ev, struct timeval *tvalp)
+{
+	int ret, i;
+#define MAXEVENTS 1
+	struct epoll_event events[MAXEVENTS];
+	int timeout = -1;
+	int wait_errno;
+
+	if (tvalp) {
+		/* it's better to trigger timed events a bit later than too early */
+		timeout = ((tvalp->tv_usec+999) / 1000) + (tvalp->tv_sec*1000);
+	}
+
+	if (epoll_ev->ev->signal_events &&
+	    tevent_common_check_signal(epoll_ev->ev)) {
+		return 0;
+	}
+
+	tevent_trace_point_callback(epoll_ev->ev, TEVENT_TRACE_BEFORE_WAIT);
+	ret = epoll_wait(epoll_ev->epoll_fd, events, MAXEVENTS, timeout);
+	wait_errno = errno;
+	tevent_trace_point_callback(epoll_ev->ev, TEVENT_TRACE_AFTER_WAIT);
+
+	if (ret == -1 && wait_errno == EINTR && epoll_ev->ev->signal_events) {
+		if (tevent_common_check_signal(epoll_ev->ev)) {
+			return 0;
+		}
+	}
+
+	if (ret == -1 && wait_errno != EINTR) {
+		epoll_panic(epoll_ev, "epoll_wait() failed", true);
+		return -1;
+	}
+
+	if (ret == 0 && tvalp) {
+		/* we don't care about a possible delay here */
+		tevent_common_loop_timer_delay(epoll_ev->ev);
+		return 0;
+	}
+
+	for (i=0;i<ret;i++) {
+		struct tevent_fd *fde = talloc_get_type(events[i].data.ptr, 
+						       struct tevent_fd);
+		uint16_t flags = 0;
+		struct tevent_fd *mpx_fde = NULL;
+
+		if (fde == NULL) {
+			epoll_panic(epoll_ev, "epoll_wait() gave bad data", true);
+			return -1;
+		}
+		if (fde->additional_flags & EPOLL_ADDITIONAL_FD_FLAG_HAS_MPX) {
+			/*
+			 * Save off the multiplexed event in case we need
+			 * to use it to call the handler function.
+			 */
+			mpx_fde = talloc_get_type_abort(fde->additional_data,
+							struct tevent_fd);
+		}
+		if (events[i].events & (EPOLLHUP|EPOLLERR)) {
+			bool handled_fde = epoll_handle_hup_or_err(epoll_ev, fde);
+			bool handled_mpx = epoll_handle_hup_or_err(epoll_ev, mpx_fde);
+
+			if (handled_fde && handled_mpx) {
+				epoll_update_event(epoll_ev, fde);
+				continue;
+			}
+
+			if (!handled_mpx) {
+				/*
+				 * If the mpx event was the one that needs
+				 * further handling, it's the TEVENT_FD_READ
+				 * event so switch over and call that handler.
+				 */
+				fde = mpx_fde;
+				mpx_fde = NULL;
+			}
+			flags |= TEVENT_FD_READ;
+		}
+		if (events[i].events & EPOLLIN) flags |= TEVENT_FD_READ;
+		if (events[i].events & EPOLLOUT) flags |= TEVENT_FD_WRITE;
+
+		if (flags & TEVENT_FD_WRITE) {
+			if (fde->flags & TEVENT_FD_WRITE) {
+				mpx_fde = NULL;
+			}
+			if (mpx_fde && mpx_fde->flags & TEVENT_FD_WRITE) {
+				fde = mpx_fde;
+				mpx_fde = NULL;
+			}
+		}
+
+		if (mpx_fde) {
+			/* Ensure we got the right fde. */
+			if ((flags & fde->flags) == 0) {
+				fde = mpx_fde;
+				mpx_fde = NULL;
+			}
+		}
+
+		/*
+		 * make sure we only pass the flags
+		 * the handler is expecting.
+		 */
+		flags &= fde->flags;
+		if (flags) {
+			fde->handler(epoll_ev->ev, fde, flags, fde->private_data);
+			break;
+		}
+	}
+
+	return 0;
+}
+
+/*
+  create a epoll_event_context structure.
+*/
+static int epoll_event_context_init(struct tevent_context *ev)
+{
+	int ret;
+	struct epoll_event_context *epoll_ev;
+
+	/*
+	 * We might be called during tevent_re_initialise()
+	 * which means we need to free our old additional_data.
+	 */
+	TALLOC_FREE(ev->additional_data);
+
+	epoll_ev = talloc_zero(ev, struct epoll_event_context);
+	if (!epoll_ev) return -1;
+	epoll_ev->ev = ev;
+	epoll_ev->epoll_fd = -1;
+
+	ret = epoll_init_ctx(epoll_ev);
+	if (ret != 0) {
+		talloc_free(epoll_ev);
+		return ret;
+	}
+
+	ev->additional_data = epoll_ev;
+	return 0;
+}
+
+/*
+  destroy an fd_event
+*/
+static int epoll_event_fd_destructor(struct tevent_fd *fde)
+{
+	struct tevent_context *ev = fde->event_ctx;
+	struct epoll_event_context *epoll_ev = NULL;
+	bool panic_triggered = false;
+	struct tevent_fd *mpx_fde = NULL;
+	int flags = fde->flags;
+
+	if (ev == NULL) {
+		return tevent_common_fd_destructor(fde);
+	}
+
+	epoll_ev = talloc_get_type_abort(ev->additional_data,
+					 struct epoll_event_context);
+
+	/*
+	 * we must remove the event from the list
+	 * otherwise a panic fallback handler may
+	 * reuse invalid memory
+	 */
+	DLIST_REMOVE(ev->fd_events, fde);
+
+	if (fde->additional_flags & EPOLL_ADDITIONAL_FD_FLAG_HAS_MPX) {
+		mpx_fde = talloc_get_type_abort(fde->additional_data,
+						struct tevent_fd);
+
+		fde->additional_flags &= ~EPOLL_ADDITIONAL_FD_FLAG_HAS_MPX;
+		mpx_fde->additional_flags &= ~EPOLL_ADDITIONAL_FD_FLAG_HAS_MPX;
+
+		fde->additional_data = NULL;
+		mpx_fde->additional_data = NULL;
+
+		fde->additional_flags &= ~EPOLL_ADDITIONAL_FD_FLAG_HAS_EVENT;
+	}
+
+	epoll_ev->panic_state = &panic_triggered;
+	epoll_check_reopen(epoll_ev);
+	if (panic_triggered) {
+		return tevent_common_fd_destructor(fde);
+	}
+
+	if (mpx_fde != NULL) {
+		epoll_update_event(epoll_ev, mpx_fde);
+		if (panic_triggered) {
+			return tevent_common_fd_destructor(fde);
+		}
+	}
+
+	fde->flags = 0;
+	epoll_update_event(epoll_ev, fde);
+	fde->flags = flags;
+	if (panic_triggered) {
+		return tevent_common_fd_destructor(fde);
+	}
+	epoll_ev->panic_state = NULL;
+
+	return tevent_common_fd_destructor(fde);
+}
+
+/*
+  add a fd based event
+  return NULL on failure (memory allocation error)
+*/
+static struct tevent_fd *epoll_event_add_fd(struct tevent_context *ev, TALLOC_CTX *mem_ctx,
+					    int fd, uint16_t flags,
+					    tevent_fd_handler_t handler,
+					    void *private_data,
+					    const char *handler_name,
+					    const char *location)
+{
+	struct epoll_event_context *epoll_ev = talloc_get_type(ev->additional_data,
+							   struct epoll_event_context);
+	struct tevent_fd *fde;
+	bool panic_triggered = false;
+
+	fde = tevent_common_add_fd(ev, mem_ctx, fd, flags,
+				   handler, private_data,
+				   handler_name, location);
+	if (!fde) return NULL;
+
+	talloc_set_destructor(fde, epoll_event_fd_destructor);
+
+	epoll_ev->panic_state = &panic_triggered;
+	epoll_check_reopen(epoll_ev);
+	if (panic_triggered) {
+		return fde;
+	}
+	epoll_ev->panic_state = NULL;
+
+	epoll_update_event(epoll_ev, fde);
+
+	return fde;
+}
+
+/*
+  set the fd event flags
+*/
+static void epoll_event_set_fd_flags(struct tevent_fd *fde, uint16_t flags)
+{
+	struct tevent_context *ev;
+	struct epoll_event_context *epoll_ev;
+	bool panic_triggered = false;
+
+	if (fde->flags == flags) return;
+
+	ev = fde->event_ctx;
+	epoll_ev = talloc_get_type(ev->additional_data, struct epoll_event_context);
+
+	fde->flags = flags;
+
+	epoll_ev->panic_state = &panic_triggered;
+	epoll_check_reopen(epoll_ev);
+	if (panic_triggered) {
+		return;
+	}
+	epoll_ev->panic_state = NULL;
+
+	epoll_update_event(epoll_ev, fde);
+}
+
+/*
+  do a single event loop using the events defined in ev 
+*/
+static int epoll_event_loop_once(struct tevent_context *ev, const char *location)
+{
+	struct epoll_event_context *epoll_ev = talloc_get_type(ev->additional_data,
+		 					   struct epoll_event_context);
+	struct timeval tval;
+	bool panic_triggered = false;
+
+	if (ev->signal_events &&
+	    tevent_common_check_signal(ev)) {
+		return 0;
+	}
+
+	if (ev->immediate_events &&
+	    tevent_common_loop_immediate(ev)) {
+		return 0;
+	}
+
+	tval = tevent_common_loop_timer_delay(ev);
+	if (tevent_timeval_is_zero(&tval)) {
+		return 0;
+	}
+
+	epoll_ev->panic_state = &panic_triggered;
+	epoll_ev->panic_force_replay = true;
+	epoll_check_reopen(epoll_ev);
+	if (panic_triggered) {
+		errno = EINVAL;
+		return -1;
+	}
+	epoll_ev->panic_force_replay = false;
+	epoll_ev->panic_state = NULL;
+
+	return epoll_event_loop(epoll_ev, &tval);
+}
+
+static const struct tevent_ops epoll_event_ops = {
+	.context_init		= epoll_event_context_init,
+	.add_fd			= epoll_event_add_fd,
+	.set_fd_close_fn	= tevent_common_fd_set_close_fn,
+	.get_fd_flags		= tevent_common_fd_get_flags,
+	.set_fd_flags		= epoll_event_set_fd_flags,
+	.add_timer		= tevent_common_add_timer_v2,
+	.schedule_immediate	= tevent_common_schedule_immediate,
+	.add_signal		= tevent_common_add_signal,
+	.loop_once		= epoll_event_loop_once,
+	.loop_wait		= tevent_common_loop_wait,
+};
+
+_PRIVATE_ bool tevent_epoll_init(void)
+{
+	return tevent_register_backend("epoll", &epoll_event_ops);
+}
diff --git a/ctdb/lib/tevent/tevent_fd.c b/ctdb/lib/tevent/tevent_fd.c
new file mode 100644
index 0000000000..455961b67c
--- /dev/null
+++ b/ctdb/lib/tevent/tevent_fd.c
@@ -0,0 +1,95 @@
+/*
+   Unix SMB/CIFS implementation.
+
+   common events code for fd events
+
+   Copyright (C) Stefan Metzmacher 2009
+
+     ** NOTE! The following LGPL license applies to the tevent
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "replace.h"
+#include "tevent.h"
+#include "tevent_internal.h"
+#include "tevent_util.h"
+
+int tevent_common_fd_destructor(struct tevent_fd *fde)
+{
+	if (fde->event_ctx) {
+		DLIST_REMOVE(fde->event_ctx->fd_events, fde);
+	}
+
+	if (fde->close_fn) {
+		fde->close_fn(fde->event_ctx, fde, fde->fd, fde->private_data);
+		fde->fd = -1;
+	}
+
+	return 0;
+}
+
+struct tevent_fd *tevent_common_add_fd(struct tevent_context *ev, TALLOC_CTX *mem_ctx,
+				       int fd, uint16_t flags,
+				       tevent_fd_handler_t handler,
+				       void *private_data,
+				       const char *handler_name,
+				       const char *location)
+{
+	struct tevent_fd *fde;
+
+	/* tevent will crash later on select() if we save
+	 * a negative file descriptor. Better to fail here
+	 * so that consumers will be able to debug it
+	 */
+	if (fd < 0) return NULL;
+
+	fde = talloc(mem_ctx?mem_ctx:ev, struct tevent_fd);
+	if (!fde) return NULL;
+
+	fde->event_ctx		= ev;
+	fde->fd			= fd;
+	fde->flags		= flags;
+	fde->handler		= handler;
+	fde->close_fn		= NULL;
+	fde->private_data	= private_data;
+	fde->handler_name	= handler_name;
+	fde->location		= location;
+	fde->additional_flags	= 0;
+	fde->additional_data	= NULL;
+
+	DLIST_ADD(ev->fd_events, fde);
+
+	talloc_set_destructor(fde, tevent_common_fd_destructor);
+
+	return fde;
+}
+uint16_t tevent_common_fd_get_flags(struct tevent_fd *fde)
+{
+	return fde->flags;
+}
+
+void tevent_common_fd_set_flags(struct tevent_fd *fde, uint16_t flags)
+{
+	if (fde->flags == flags) return;
+	fde->flags = flags;
+}
+
+void tevent_common_fd_set_close_fn(struct tevent_fd *fde,
+				   tevent_fd_close_fn_t close_fn)
+{
+	fde->close_fn = close_fn;
+}
diff --git a/ctdb/lib/tevent/tevent_immediate.c b/ctdb/lib/tevent/tevent_immediate.c
new file mode 100644
index 0000000000..1ac293e175
--- /dev/null
+++ b/ctdb/lib/tevent/tevent_immediate.c
@@ -0,0 +1,139 @@
+/*
+   Unix SMB/CIFS implementation.
+
+   common events code for immediate events
+
+   Copyright (C) Stefan Metzmacher 2009
+
+     ** NOTE! The following LGPL license applies to the tevent
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "replace.h"
+#include "tevent.h"
+#include "tevent_internal.h"
+#include "tevent_util.h"
+
+static void tevent_common_immediate_cancel(struct tevent_immediate *im)
+{
+	if (!im->event_ctx) {
+		return;
+	}
+
+	tevent_debug(im->event_ctx, TEVENT_DEBUG_TRACE,
+		     "Cancel immediate event %p \"%s\"\n",
+		     im, im->handler_name);
+
+	/* let the backend free im->additional_data */
+	if (im->cancel_fn) {
+		im->cancel_fn(im);
+	}
+
+	DLIST_REMOVE(im->event_ctx->immediate_events, im);
+	im->event_ctx		= NULL;
+	im->handler		= NULL;
+	im->private_data	= NULL;
+	im->handler_name	= NULL;
+	im->schedule_location	= NULL;
+	im->cancel_fn		= NULL;
+	im->additional_data	= NULL;
+
+	talloc_set_destructor(im, NULL);
+}
+
+/*
+  destroy an immediate event
+*/
+static int tevent_common_immediate_destructor(struct tevent_immediate *im)
+{
+	tevent_common_immediate_cancel(im);
+	return 0;
+}
+
+/*
+ * schedule an immediate event on
+ */
+void tevent_common_schedule_immediate(struct tevent_immediate *im,
+				      struct tevent_context *ev,
+				      tevent_immediate_handler_t handler,
+				      void *private_data,
+				      const char *handler_name,
+				      const char *location)
+{
+	tevent_common_immediate_cancel(im);
+
+	if (!handler) {
+		return;
+	}
+
+	im->event_ctx		= ev;
+	im->handler		= handler;
+	im->private_data	= private_data;
+	im->handler_name	= handler_name;
+	im->schedule_location	= location;
+	im->cancel_fn		= NULL;
+	im->additional_data	= NULL;
+
+	DLIST_ADD_END(ev->immediate_events, im, struct tevent_immediate *);
+	talloc_set_destructor(im, tevent_common_immediate_destructor);
+
+	tevent_debug(ev, TEVENT_DEBUG_TRACE,
+		     "Schedule immediate event \"%s\": %p\n",
+		     handler_name, im);
+}
+
+/*
+  trigger the first immediate event and return true
+  if no event was triggered return false
+*/
+bool tevent_common_loop_immediate(struct tevent_context *ev)
+{
+	struct tevent_immediate *im = ev->immediate_events;
+	tevent_immediate_handler_t handler;
+	void *private_data;
+
+	if (!im) {
+		return false;
+	}
+
+	tevent_debug(ev, TEVENT_DEBUG_TRACE,
+		     "Run immediate event \"%s\": %p\n",
+		     im->handler_name, im);
+
+	/*
+	 * remember the handler and then clear the event
+	 * the handler might reschedule the event
+	 */
+	handler = im->handler;
+	private_data = im->private_data;
+
+	DLIST_REMOVE(im->event_ctx->immediate_events, im);
+	im->event_ctx		= NULL;
+	im->handler		= NULL;
+	im->private_data	= NULL;
+	im->handler_name	= NULL;
+	im->schedule_location	= NULL;
+	im->cancel_fn		= NULL;
+	im->additional_data	= NULL;
+
+	talloc_set_destructor(im, NULL);
+
+	handler(ev, im, private_data);
+
+	return true;
+}
+
diff --git a/ctdb/lib/tevent/tevent_internal.h b/ctdb/lib/tevent/tevent_internal.h
new file mode 100644
index 0000000000..b239e7403a
--- /dev/null
+++ b/ctdb/lib/tevent/tevent_internal.h
@@ -0,0 +1,345 @@
+/* 
+   Unix SMB/CIFS implementation.
+
+   generalised event loop handling
+
+   INTERNAL STRUCTS. THERE ARE NO API GUARANTEES.
+   External users should only ever have to include this header when 
+   implementing new tevent backends.
+
+   Copyright (C) Stefan Metzmacher 2005-2009
+
+     ** NOTE! The following LGPL license applies to the tevent
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+struct tevent_req {
+	/**
+	 * @brief What to do on completion
+	 *
+	 * This is used for the user of an async request, fn is called when
+	 * the request completes, either successfully or with an error.
+	 */
+	struct {
+		/**
+		 * @brief Completion function
+		 * Completion function, to be filled by the API user
+		 */
+		tevent_req_fn fn;
+		/**
+		 * @brief Private data for the completion function
+		 */
+		void *private_data;
+	} async;
+
+	/**
+	 * @brief Private state pointer for the actual implementation
+	 *
+	 * The implementation doing the work for the async request needs to
+	 * keep around current data like for example a fd event. The user of
+	 * an async request should not touch this.
+	 */
+	void *data;
+
+	/**
+	 * @brief A function to overwrite the default print function
+	 *
+	 * The implementation doing the work may want to implement a
+	 * custom function to print the text representation of the async
+	 * request.
+	 */
+	tevent_req_print_fn private_print;
+
+	/**
+	 * @brief A function to cancel the request
+	 *
+	 * The implementation might want to set a function
+	 * that is called when the tevent_req_cancel() function
+	 * was called.
+	 */
+	tevent_req_cancel_fn private_cancel;
+
+	/**
+	 * @brief Internal state of the request
+	 *
+	 * Callers should only access this via functions and never directly.
+	 */
+	struct {
+		/**
+		 * @brief The talloc type of the data pointer
+		 *
+		 * This is filled by the tevent_req_create() macro.
+		 *
+		 * This for debugging only.
+		 */
+		const char *private_type;
+
+		/**
+		 * @brief The location where the request was created
+		 *
+		 * This uses the __location__ macro via the tevent_req_create()
+		 * macro.
+		 *
+		 * This for debugging only.
+		 */
+		const char *create_location;
+
+		/**
+		 * @brief The location where the request was finished
+		 *
+		 * This uses the __location__ macro via the tevent_req_done(),
+		 * tevent_req_error() or tevent_req_nomem() macro.
+		 *
+		 * This for debugging only.
+		 */
+		const char *finish_location;
+
+		/**
+		 * @brief The location where the request was canceled
+		 *
+		 * This uses the __location__ macro via the
+		 * tevent_req_cancel() macro.
+		 *
+		 * This for debugging only.
+		 */
+		const char *cancel_location;
+
+		/**
+		 * @brief The external state - will be queried by the caller
+		 *
+		 * While the async request is being processed, state will remain in
+		 * TEVENT_REQ_IN_PROGRESS. A request is finished if
+		 * req->state>=TEVENT_REQ_DONE.
+		 */
+		enum tevent_req_state state;
+
+		/**
+		 * @brief status code when finished
+		 *
+		 * This status can be queried in the async completion function. It
+		 * will be set to 0 when everything went fine.
+		 */
+		uint64_t error;
+
+		/**
+		 * @brief the immediate event used by tevent_req_post
+		 *
+		 */
+		struct tevent_immediate *trigger;
+
+		/**
+		 * @brief An event context which will be used to
+		 *        defer the _tevent_req_notify_callback().
+		 */
+		struct tevent_context *defer_callback_ev;
+
+		/**
+		 * @brief the timer event if tevent_req_set_endtime was used
+		 *
+		 */
+		struct tevent_timer *timer;
+	} internal;
+};
+
+struct tevent_fd {
+	struct tevent_fd *prev, *next;
+	struct tevent_context *event_ctx;
+	int fd;
+	uint16_t flags; /* see TEVENT_FD_* flags */
+	tevent_fd_handler_t handler;
+	tevent_fd_close_fn_t close_fn;
+	/* this is private for the specific handler */
+	void *private_data;
+	/* this is for debugging only! */
+	const char *handler_name;
+	const char *location;
+	/* this is private for the events_ops implementation */
+	uint64_t additional_flags;
+	void *additional_data;
+};
+
+struct tevent_timer {
+	struct tevent_timer *prev, *next;
+	struct tevent_context *event_ctx;
+	struct timeval next_event;
+	tevent_timer_handler_t handler;
+	/* this is private for the specific handler */
+	void *private_data;
+	/* this is for debugging only! */
+	const char *handler_name;
+	const char *location;
+	/* this is private for the events_ops implementation */
+	void *additional_data;
+};
+
+struct tevent_immediate {
+	struct tevent_immediate *prev, *next;
+	struct tevent_context *event_ctx;
+	tevent_immediate_handler_t handler;
+	/* this is private for the specific handler */
+	void *private_data;
+	/* this is for debugging only! */
+	const char *handler_name;
+	const char *create_location;
+	const char *schedule_location;
+	/* this is private for the events_ops implementation */
+	void (*cancel_fn)(struct tevent_immediate *im);
+	void *additional_data;
+};
+
+struct tevent_signal {
+	struct tevent_signal *prev, *next;
+	struct tevent_context *event_ctx;
+	int signum;
+	int sa_flags;
+	tevent_signal_handler_t handler;
+	/* this is private for the specific handler */
+	void *private_data;
+	/* this is for debugging only! */
+	const char *handler_name;
+	const char *location;
+	/* this is private for the events_ops implementation */
+	void *additional_data;
+};
+
+struct tevent_debug_ops {
+	void (*debug)(void *context, enum tevent_debug_level level,
+		      const char *fmt, va_list ap) PRINTF_ATTRIBUTE(3,0);
+	void *context;
+};
+
+void tevent_debug(struct tevent_context *ev, enum tevent_debug_level level,
+		  const char *fmt, ...) PRINTF_ATTRIBUTE(3,4);
+
+struct tevent_context {
+	/* the specific events implementation */
+	const struct tevent_ops *ops;
+
+	/* list of fd events - used by common code */
+	struct tevent_fd *fd_events;
+
+	/* list of timed events - used by common code */
+	struct tevent_timer *timer_events;
+
+	/* list of immediate events - used by common code */
+	struct tevent_immediate *immediate_events;
+
+	/* list of signal events - used by common code */
+	struct tevent_signal *signal_events;
+
+	/* this is private for the events_ops implementation */
+	void *additional_data;
+
+	/* pipe hack used with signal handlers */
+	struct tevent_fd *pipe_fde;
+	int pipe_fds[2];
+
+	/* debugging operations */
+	struct tevent_debug_ops debug_ops;
+
+	/* info about the nesting status */
+	struct {
+		bool allowed;
+		uint32_t level;
+		tevent_nesting_hook hook_fn;
+		void *hook_private;
+	} nesting;
+
+	struct {
+		tevent_trace_callback_t callback;
+		void *private_data;
+	} tracing;
+
+	/*
+	 * an optimization pointer into timer_events
+	 * used by used by common code via
+	 * tevent_common_add_timer_v2()
+	 */
+	struct tevent_timer *last_zero_timer;
+};
+
+const struct tevent_ops *tevent_find_ops_byname(const char *name);
+
+int tevent_common_context_destructor(struct tevent_context *ev);
+int tevent_common_loop_wait(struct tevent_context *ev,
+			    const char *location);
+
+int tevent_common_fd_destructor(struct tevent_fd *fde);
+struct tevent_fd *tevent_common_add_fd(struct tevent_context *ev,
+				       TALLOC_CTX *mem_ctx,
+				       int fd,
+				       uint16_t flags,
+				       tevent_fd_handler_t handler,
+				       void *private_data,
+				       const char *handler_name,
+				       const char *location);
+void tevent_common_fd_set_close_fn(struct tevent_fd *fde,
+				   tevent_fd_close_fn_t close_fn);
+uint16_t tevent_common_fd_get_flags(struct tevent_fd *fde);
+void tevent_common_fd_set_flags(struct tevent_fd *fde, uint16_t flags);
+
+struct tevent_timer *tevent_common_add_timer(struct tevent_context *ev,
+					     TALLOC_CTX *mem_ctx,
+					     struct timeval next_event,
+					     tevent_timer_handler_t handler,
+					     void *private_data,
+					     const char *handler_name,
+					     const char *location);
+struct tevent_timer *tevent_common_add_timer_v2(struct tevent_context *ev,
+						TALLOC_CTX *mem_ctx,
+					        struct timeval next_event,
+					        tevent_timer_handler_t handler,
+					        void *private_data,
+					        const char *handler_name,
+					        const char *location);
+struct timeval tevent_common_loop_timer_delay(struct tevent_context *);
+
+void tevent_common_schedule_immediate(struct tevent_immediate *im,
+				      struct tevent_context *ev,
+				      tevent_immediate_handler_t handler,
+				      void *private_data,
+				      const char *handler_name,
+				      const char *location);
+bool tevent_common_loop_immediate(struct tevent_context *ev);
+
+struct tevent_signal *tevent_common_add_signal(struct tevent_context *ev,
+					       TALLOC_CTX *mem_ctx,
+					       int signum,
+					       int sa_flags,
+					       tevent_signal_handler_t handler,
+					       void *private_data,
+					       const char *handler_name,
+					       const char *location);
+int tevent_common_check_signal(struct tevent_context *ev);
+void tevent_cleanup_pending_signal_handlers(struct tevent_signal *se);
+
+bool tevent_standard_init(void);
+bool tevent_select_init(void);
+bool tevent_poll_init(void);
+void tevent_poll_event_add_fd_internal(struct tevent_context *ev,
+				       struct tevent_fd *fde);
+bool tevent_poll_mt_init(void);
+#ifdef HAVE_EPOLL
+bool tevent_epoll_init(void);
+bool tevent_epoll_set_panic_fallback(struct tevent_context *ev,
+			bool (*panic_fallback)(struct tevent_context *ev,
+					       bool replay));
+#endif
+
+
+void tevent_trace_point_callback(struct tevent_context *ev,
+				 enum tevent_trace_point);
diff --git a/ctdb/lib/tevent/tevent_liboop.c b/ctdb/lib/tevent/tevent_liboop.c
new file mode 100644
index 0000000000..68be76baca
--- /dev/null
+++ b/ctdb/lib/tevent/tevent_liboop.c
@@ -0,0 +1,292 @@
+/* 
+   Unix SMB/CIFS implementation.
+   main select loop and event handling
+   wrapper for http://git.lysator.liu.se/liboop/
+
+   Copyright (C) Stefan Metzmacher 2005
+
+     ** NOTE! The following LGPL license applies to the tevent
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "events.h"
+#include "events_internal.h"
+
+#include <oop.h>
+
+/*
+ !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 
+
+ NOTE: this code compiles fine, but is completely *UNTESTED*
+       and is only committed as an example
+
+ !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!	 
+*/
+
+static int oop_event_context_destructor(struct tevent_context *ev)
+{
+	oop_source_sys *oop_sys = ev->additional_data;
+
+	oop_sys_delete(oop_sys);
+
+	return 0;
+}
+
+/*
+  create a oop_event_context structure.
+*/
+static int oop_event_context_init(struct tevent_context *ev, void *private_data)
+{
+	oop_source_sys *oop_sys = private_data;
+
+	if (!oop_sys) {
+		oop_sys = oop_sys_new();
+		if (!oop_sys) {
+			return -1;
+		}
+
+		talloc_set_destructor(ev, oop_event_context_destructor);
+	}
+
+	ev->additional_data = oop_sys;
+
+	return 0;
+}
+
+static void *oop_event_fd_handler(oop_source *oop, int fd, oop_event oop_type, void *ptr)
+{
+	struct tevent_fd *fde = ptr;
+
+	if (fd != fde->fd) return OOP_ERROR;
+
+	switch(oop_type) {
+		case OOP_READ:
+			fde->handler(fde->event_ctx, fde, EVENT_FD_READ, fde->private_data);
+			return OOP_CONTINUE;
+		case OOP_WRITE:
+			fde->handler(fde->event_ctx, fde, EVENT_FD_WRITE, fde->private_data);
+			return OOP_CONTINUE;			
+		case OOP_EXCEPTION:
+			return OOP_ERROR;
+		case OOP_NUM_EVENTS:
+			return OOP_ERROR;
+	}
+
+	return OOP_ERROR;
+}
+
+/*
+  destroy an fd_event
+*/
+static int oop_event_fd_destructor(struct tevent_fd *fde)
+{
+	struct tevent_context *ev = fde->event_ctx;
+	oop_source_sys *oop_sys = ev->additional_data;
+	oop_source *oop = oop_sys_source(oop_sys);
+
+	if (fde->flags & EVENT_FD_READ)
+		oop->cancel_fd(oop, fde->fd, OOP_READ);
+	if (fde->flags & EVENT_FD_WRITE)
+		oop->cancel_fd(oop, fde->fd, OOP_WRITE);
+
+	if (fde->flags & EVENT_FD_AUTOCLOSE) {
+		close(fde->fd);
+		fde->fd = -1;
+	}
+
+	return 0;
+}
+
+/*
+  add a fd based event
+  return NULL on failure (memory allocation error)
+*/
+static struct tevent_fd *oop_event_add_fd(struct tevent_context *ev, TALLOC_CTX *mem_ctx,
+					 int fd, uint16_t flags,
+					 event_fd_handler_t handler,
+					 void *private_data)
+{
+	struct tevent_fd *fde;
+	oop_source_sys *oop_sys = ev->additional_data;
+	oop_source *oop = oop_sys_source(oop_sys);
+	
+	fde = talloc(mem_ctx?mem_ctx:ev, struct tevent_fd);
+	if (!fde) return NULL;
+
+	fde->event_ctx		= ev;
+	fde->fd			= fd;
+	fde->flags		= flags;
+	fde->handler		= handler;
+	fde->private_data	= private_data;
+	fde->additional_flags	= 0;
+	fde->additional_data	= NULL;
+
+	if (fde->flags & EVENT_FD_READ)
+		oop->on_fd(oop, fde->fd, OOP_READ, oop_event_fd_handler, fde);
+	if (fde->flags & EVENT_FD_WRITE)
+		oop->on_fd(oop, fde->fd, OOP_WRITE, oop_event_fd_handler, fde);
+
+	talloc_set_destructor(fde, oop_event_fd_destructor);
+
+	return fde;
+}
+
+/*
+  return the fd event flags
+*/
+static uint16_t oop_event_get_fd_flags(struct tevent_fd *fde)
+{
+	return fde->flags;
+}
+
+/*
+  set the fd event flags
+*/
+static void oop_event_set_fd_flags(struct tevent_fd *fde, uint16_t flags)
+{
+	oop_source_sys *oop_sys;
+	oop_source *oop;
+
+	oop_sys = fde->event_ctx->additional_data;
+	oop = oop_sys_source(oop_sys);
+
+	if ((fde->flags & EVENT_FD_READ)&&(!(flags & EVENT_FD_READ)))
+		oop->cancel_fd(oop, fde->fd, OOP_READ);
+
+	if ((!(fde->flags & EVENT_FD_READ))&&(flags & EVENT_FD_READ))
+		oop->on_fd(oop, fde->fd, OOP_READ, oop_event_fd_handler, fde);
+
+	if ((fde->flags & EVENT_FD_WRITE)&&(!(flags & EVENT_FD_WRITE)))
+		oop->cancel_fd(oop, fde->fd, OOP_WRITE);
+
+	if ((!(fde->flags & EVENT_FD_WRITE))&&(flags & EVENT_FD_WRITE))
+		oop->on_fd(oop, fde->fd, OOP_WRITE, oop_event_fd_handler, fde);
+
+	fde->flags = flags;
+}
+
+static int oop_event_timed_destructor(struct tevent_timer *te);
+
+static int oop_event_timed_deny_destructor(struct tevent_timer *te)
+{
+	return -1;
+}
+
+static void *oop_event_timed_handler(oop_source *oop, struct timeval t, void *ptr)
+{
+	struct tevent_timer *te = ptr;
+
+	/* deny the handler to free the event */
+	talloc_set_destructor(te, oop_event_timed_deny_destructor);
+	te->handler(te->event_ctx, te, t, te->private_data);
+
+	talloc_set_destructor(te, oop_event_timed_destructor);
+	talloc_free(te);
+
+	return OOP_CONTINUE;
+}
+
+/*
+  destroy a timed event
+*/
+static int oop_event_timed_destructor(struct tevent_timer *te)
+{
+	struct tevent_context *ev = te->event_ctx;
+	oop_source_sys *oop_sys = ev->additional_data;
+	oop_source *oop = oop_sys_source(oop_sys);
+
+	oop->cancel_time(oop, te->next_event, oop_event_timed_handler, te);
+
+	return 0;
+}
+
+/*
+  add a timed event
+  return NULL on failure (memory allocation error)
+*/
+static struct tevent_timer *oop_event_add_timed(struct tevent_context *ev, TALLOC_CTX *mem_ctx,
+					       struct timeval next_event, 
+					       event_timed_handler_t handler, 
+					       void *private_data) 
+{
+	oop_source_sys *oop_sys = ev->additional_data;
+	oop_source *oop = oop_sys_source(oop_sys);
+	struct tevent_timer *te;
+
+	te = talloc(mem_ctx?mem_ctx:ev, struct tevent_timer);
+	if (te == NULL) return NULL;
+
+	te->event_ctx		= ev;
+	te->next_event		= next_event;
+	te->handler		= handler;
+	te->private_data	= private_data;
+	te->additional_data	= NULL;
+
+	oop->on_time(oop, te->next_event, oop_event_timed_handler, te);
+
+	talloc_set_destructor(te, oop_event_timed_destructor);
+
+	return te;
+}
+
+/*
+  do a single event loop using the events defined in ev 
+*/
+static int oop_event_loop_once(struct tevent_context *ev)
+{
+	void *oop_ret;
+	oop_source_sys *oop_sys = ev->additional_data;
+
+	oop_ret = oop_sys_run_once(oop_sys);
+	if (oop_ret == OOP_CONTINUE) {
+		return 0;
+	}
+
+	return -1;
+}
+
+/*
+  return on failure or (with 0) if all fd events are removed
+*/
+static int oop_event_loop_wait(struct tevent_context *ev)
+{
+	void *oop_ret;
+	oop_source_sys *oop_sys = ev->additional_data;
+
+	oop_ret = oop_sys_run(oop_sys);
+	if (oop_ret == OOP_CONTINUE) {
+		return 0;
+	}
+
+	return -1;
+}
+
+static const struct event_ops event_oop_ops = {
+	.context_init	= oop_event_context_init,
+	.add_fd		= oop_event_add_fd,
+	.get_fd_flags	= oop_event_get_fd_flags,
+	.set_fd_flags	= oop_event_set_fd_flags,
+	.add_timer	= oop_event_add_timed,
+	.add_signal	= common_event_add_signal,
+	.loop_once	= oop_event_loop_once,
+	.loop_wait	= oop_event_loop_wait,
+};
+
+const struct event_ops *event_liboop_get_ops(void)
+{
+	return &event_oop_ops;
+}
diff --git a/ctdb/lib/tevent/tevent_poll.c b/ctdb/lib/tevent/tevent_poll.c
new file mode 100644
index 0000000000..92fcc441ac
--- /dev/null
+++ b/ctdb/lib/tevent/tevent_poll.c
@@ -0,0 +1,725 @@
+/*
+   Unix SMB/CIFS implementation.
+   main select loop and event handling
+   Copyright (C) Andrew Tridgell	2003-2005
+   Copyright (C) Stefan Metzmacher	2005-2009
+
+     ** NOTE! The following LGPL license applies to the tevent
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "replace.h"
+#include "system/filesys.h"
+#include "system/select.h"
+#include "tevent.h"
+#include "tevent_util.h"
+#include "tevent_internal.h"
+
+struct poll_event_context {
+	/* a pointer back to the generic event_context */
+	struct tevent_context *ev;
+
+	/*
+	 * A DLIST for fresh fde's added by poll_event_add_fd but not
+	 * picked up yet by poll_event_loop_once
+	 */
+	struct tevent_fd *fresh;
+	/*
+	 * A DLIST for disabled fde's.
+	 */
+	struct tevent_fd *disabled;
+	/*
+	 * one or more events were deleted or disabled
+	 */
+	bool deleted;
+
+	/*
+	 * These two arrays are maintained together.
+	 */
+	struct pollfd *fds;
+	struct tevent_fd **fdes;
+	unsigned num_fds;
+
+	/*
+	 * Signal fd to wake the poll() thread
+	 */
+	int signal_fd;
+
+	/* information for exiting from the event loop */
+	int exit_code;
+};
+
+static int poll_event_context_destructor(struct poll_event_context *poll_ev)
+{
+	struct tevent_fd *fd, *fn;
+
+	for (fd = poll_ev->fresh; fd; fd = fn) {
+		fn = fd->next;
+		fd->event_ctx = NULL;
+		DLIST_REMOVE(poll_ev->fresh, fd);
+	}
+
+	for (fd = poll_ev->disabled; fd; fd = fn) {
+		fn = fd->next;
+		fd->event_ctx = NULL;
+		DLIST_REMOVE(poll_ev->disabled, fd);
+	}
+
+	if (poll_ev->signal_fd == -1) {
+		/*
+		 * Non-threaded, no signal pipe
+		 */
+		return 0;
+	}
+
+	close(poll_ev->signal_fd);
+	poll_ev->signal_fd = -1;
+
+	if (poll_ev->num_fds == 0) {
+		return 0;
+	}
+	if (poll_ev->fds[0].fd != -1) {
+		close(poll_ev->fds[0].fd);
+		poll_ev->fds[0].fd = -1;
+	}
+	return 0;
+}
+
+/*
+  create a poll_event_context structure.
+*/
+static int poll_event_context_init(struct tevent_context *ev)
+{
+	struct poll_event_context *poll_ev;
+
+	/*
+	 * we might be called during tevent_re_initialise()
+	 * which means we need to free our old additional_data
+	 * in order to detach old fd events from the
+	 * poll_ev->fresh list
+	 */
+	TALLOC_FREE(ev->additional_data);
+
+	poll_ev = talloc_zero(ev, struct poll_event_context);
+	if (poll_ev == NULL) {
+		return -1;
+	}
+	poll_ev->ev = ev;
+	poll_ev->signal_fd = -1;
+	ev->additional_data = poll_ev;
+	talloc_set_destructor(poll_ev, poll_event_context_destructor);
+	return 0;
+}
+
+static bool set_nonblock(int fd)
+{
+	int val;
+
+	val = fcntl(fd, F_GETFL, 0);
+	if (val == -1) {
+		return false;
+	}
+	val |= O_NONBLOCK;
+
+	return (fcntl(fd, F_SETFL, val) != -1);
+}
+
+static int poll_event_context_init_mt(struct tevent_context *ev)
+{
+	struct poll_event_context *poll_ev;
+	struct pollfd *pfd;
+	int fds[2];
+	int ret;
+
+	ret = poll_event_context_init(ev);
+	if (ret == -1) {
+		return ret;
+	}
+
+	poll_ev = talloc_get_type_abort(
+		ev->additional_data, struct poll_event_context);
+
+	poll_ev->fds = talloc_zero(poll_ev, struct pollfd);
+	if (poll_ev->fds == NULL) {
+		return -1;
+	}
+
+	ret = pipe(fds);
+	if (ret == -1) {
+		return -1;
+	}
+
+	if (!set_nonblock(fds[0]) || !set_nonblock(fds[1])) {
+		close(fds[0]);
+		close(fds[1]);
+		return -1;
+	}
+
+	poll_ev->signal_fd = fds[1];
+
+	pfd = &poll_ev->fds[0];
+	pfd->fd = fds[0];
+	pfd->events = (POLLIN|POLLHUP);
+
+	poll_ev->num_fds = 1;
+
+	talloc_set_destructor(poll_ev, poll_event_context_destructor);
+
+	return 0;
+}
+
+static void poll_event_wake_pollthread(struct poll_event_context *poll_ev)
+{
+	char c;
+	ssize_t ret;
+
+	if (poll_ev->signal_fd == -1) {
+		return;
+	}
+	c = 0;
+	do {
+		ret = write(poll_ev->signal_fd, &c, sizeof(c));
+	} while ((ret == -1) && (errno == EINTR));
+}
+
+static void poll_event_drain_signal_fd(struct poll_event_context *poll_ev)
+{
+	char buf[16];
+	ssize_t ret;
+	int fd;
+
+	if (poll_ev->signal_fd == -1) {
+		return;
+	}
+
+	if (poll_ev->num_fds < 1) {
+		return;
+	}
+	fd = poll_ev->fds[0].fd;
+
+	do {
+		ret = read(fd, buf, sizeof(buf));
+	} while (ret == sizeof(buf));
+}
+
+/*
+  destroy an fd_event
+*/
+static int poll_event_fd_destructor(struct tevent_fd *fde)
+{
+	struct tevent_context *ev = fde->event_ctx;
+	struct poll_event_context *poll_ev;
+	uint64_t del_idx = fde->additional_flags;
+
+	if (ev == NULL) {
+		goto done;
+	}
+
+	poll_ev = talloc_get_type_abort(
+		ev->additional_data, struct poll_event_context);
+
+	if (del_idx == UINT64_MAX) {
+		struct tevent_fd **listp =
+			(struct tevent_fd **)fde->additional_data;
+
+		DLIST_REMOVE((*listp), fde);
+		goto done;
+	}
+
+	poll_ev->fdes[del_idx] = NULL;
+	poll_ev->deleted = true;
+	poll_event_wake_pollthread(poll_ev);
+done:
+	return tevent_common_fd_destructor(fde);
+}
+
+static void poll_event_schedule_immediate(struct tevent_immediate *im,
+					  struct tevent_context *ev,
+					  tevent_immediate_handler_t handler,
+					  void *private_data,
+					  const char *handler_name,
+					  const char *location)
+{
+	struct poll_event_context *poll_ev = talloc_get_type_abort(
+		ev->additional_data, struct poll_event_context);
+
+	tevent_common_schedule_immediate(im, ev, handler, private_data,
+					 handler_name, location);
+	poll_event_wake_pollthread(poll_ev);
+}
+
+/*
+  Private function called by "standard" backend fallback.
+  Note this only allows fallback to "poll" backend, not "poll-mt".
+*/
+_PRIVATE_ void tevent_poll_event_add_fd_internal(struct tevent_context *ev,
+						 struct tevent_fd *fde)
+{
+	struct poll_event_context *poll_ev = talloc_get_type_abort(
+		ev->additional_data, struct poll_event_context);
+	struct tevent_fd **listp;
+
+	if (fde->flags != 0) {
+		listp = &poll_ev->fresh;
+	} else {
+		listp = &poll_ev->disabled;
+	}
+
+	fde->additional_flags	= UINT64_MAX;
+	fde->additional_data	= listp;
+
+	DLIST_ADD((*listp), fde);
+	talloc_set_destructor(fde, poll_event_fd_destructor);
+}
+
+/*
+  add a fd based event
+  return NULL on failure (memory allocation error)
+*/
+static struct tevent_fd *poll_event_add_fd(struct tevent_context *ev,
+					   TALLOC_CTX *mem_ctx,
+					   int fd, uint16_t flags,
+					   tevent_fd_handler_t handler,
+					   void *private_data,
+					   const char *handler_name,
+					   const char *location)
+{
+	struct poll_event_context *poll_ev = talloc_get_type_abort(
+		ev->additional_data, struct poll_event_context);
+	struct tevent_fd *fde;
+
+	if (fd < 0) {
+		return NULL;
+	}
+
+	fde = talloc(mem_ctx ? mem_ctx : ev, struct tevent_fd);
+	if (fde == NULL) {
+		return NULL;
+	}
+	fde->event_ctx		= ev;
+	fde->fd			= fd;
+	fde->flags		= flags;
+	fde->handler		= handler;
+	fde->close_fn		= NULL;
+	fde->private_data	= private_data;
+	fde->handler_name	= handler_name;
+	fde->location		= location;
+	fde->additional_flags	= UINT64_MAX;
+	fde->additional_data	= NULL;
+
+	tevent_poll_event_add_fd_internal(ev, fde);
+	poll_event_wake_pollthread(poll_ev);
+
+	/*
+	 * poll_event_loop_poll will take care of the rest in
+	 * poll_event_setup_fresh
+	 */
+	return fde;
+}
+
+/*
+  set the fd event flags
+*/
+static void poll_event_set_fd_flags(struct tevent_fd *fde, uint16_t flags)
+{
+	struct tevent_context *ev = fde->event_ctx;
+	struct poll_event_context *poll_ev;
+	uint64_t idx = fde->additional_flags;
+	uint16_t pollflags;
+
+	if (ev == NULL) {
+		return;
+	}
+	poll_ev = talloc_get_type_abort(
+		ev->additional_data, struct poll_event_context);
+
+	fde->flags = flags;
+
+	if (idx == UINT64_MAX) {
+		struct tevent_fd **listp =
+			(struct tevent_fd **)fde->additional_data;
+
+		/*
+		 * We move it between the fresh and disabled lists.
+		 */
+		DLIST_REMOVE((*listp), fde);
+		tevent_poll_event_add_fd_internal(ev, fde);
+		poll_event_wake_pollthread(poll_ev);
+		return;
+	}
+
+	if (fde->flags == 0) {
+		/*
+		 * We need to remove it from the array
+		 * and move it to the disabled list.
+		 */
+		poll_ev->fdes[idx] = NULL;
+		poll_ev->deleted = true;
+		DLIST_REMOVE(ev->fd_events, fde);
+		tevent_poll_event_add_fd_internal(ev, fde);
+		poll_event_wake_pollthread(poll_ev);
+		return;
+	}
+
+	pollflags = 0;
+
+	if (flags & TEVENT_FD_READ) {
+		pollflags |= (POLLIN|POLLHUP);
+	}
+	if (flags & TEVENT_FD_WRITE) {
+		pollflags |= (POLLOUT);
+	}
+	poll_ev->fds[idx].events = pollflags;
+
+	poll_event_wake_pollthread(poll_ev);
+}
+
+static bool poll_event_setup_fresh(struct tevent_context *ev,
+				   struct poll_event_context *poll_ev)
+{
+	struct tevent_fd *fde, *next;
+	unsigned num_fresh, num_fds;
+
+	if (poll_ev->deleted) {
+		unsigned first_fd = (poll_ev->signal_fd != -1) ? 1 : 0;
+		unsigned i;
+
+		for (i=first_fd; i < poll_ev->num_fds;) {
+			fde = poll_ev->fdes[i];
+			if (fde != NULL) {
+				i++;
+				continue;
+			}
+
+			/*
+			 * This fde was talloc_free()'ed. Delete it
+			 * from the arrays
+			 */
+			poll_ev->num_fds -= 1;
+			if (poll_ev->num_fds == i) {
+				break;
+			}
+			poll_ev->fds[i] = poll_ev->fds[poll_ev->num_fds];
+			poll_ev->fdes[i] = poll_ev->fdes[poll_ev->num_fds];
+			if (poll_ev->fdes[i] != NULL) {
+				poll_ev->fdes[i]->additional_flags = i;
+			}
+		}
+	}
+	poll_ev->deleted = false;
+
+	if (poll_ev->fresh == NULL) {
+		return true;
+	}
+
+	num_fresh = 0;
+	for (fde = poll_ev->fresh; fde; fde = fde->next) {
+		num_fresh += 1;
+	}
+	num_fds = poll_ev->num_fds + num_fresh;
+
+	/*
+	 * We check the length of fdes here. It is the last one
+	 * enlarged, so if the realloc for poll_fd->fdes fails,
+	 * poll_fd->fds will have at least the size of poll_fd->fdes
+	 */
+
+	if (num_fds >= talloc_array_length(poll_ev->fdes)) {
+		struct pollfd *tmp_fds;
+		struct tevent_fd **tmp_fdes;
+		unsigned array_length;
+
+		array_length = (num_fds + 15) & ~15; /* round up to 16 */
+
+		tmp_fds = talloc_realloc(
+			poll_ev, poll_ev->fds, struct pollfd, array_length);
+		if (tmp_fds == NULL) {
+			return false;
+		}
+		poll_ev->fds = tmp_fds;
+
+		tmp_fdes = talloc_realloc(
+			poll_ev, poll_ev->fdes, struct tevent_fd *,
+			array_length);
+		if (tmp_fdes == NULL) {
+			return false;
+		}
+		poll_ev->fdes = tmp_fdes;
+	}
+
+	for (fde = poll_ev->fresh; fde; fde = next) {
+		struct pollfd *pfd;
+
+		pfd = &poll_ev->fds[poll_ev->num_fds];
+
+		pfd->fd = fde->fd;
+		pfd->events = 0;
+		pfd->revents = 0;
+
+		if (fde->flags & TEVENT_FD_READ) {
+			pfd->events |= (POLLIN|POLLHUP);
+		}
+		if (fde->flags & TEVENT_FD_WRITE) {
+			pfd->events |= (POLLOUT);
+		}
+
+		fde->additional_flags = poll_ev->num_fds;
+		poll_ev->fdes[poll_ev->num_fds] = fde;
+
+		next = fde->next;
+		DLIST_REMOVE(poll_ev->fresh, fde);
+		DLIST_ADD(ev->fd_events, fde);
+
+		poll_ev->num_fds += 1;
+	}
+	return true;
+}
+
+/*
+  event loop handling using poll()
+*/
+static int poll_event_loop_poll(struct tevent_context *ev,
+				struct timeval *tvalp)
+{
+	struct poll_event_context *poll_ev = talloc_get_type_abort(
+		ev->additional_data, struct poll_event_context);
+	int pollrtn;
+	int timeout = -1;
+	int poll_errno;
+	struct tevent_fd *fde = NULL;
+	unsigned i;
+
+	if (ev->signal_events && tevent_common_check_signal(ev)) {
+		return 0;
+	}
+
+	if (tvalp != NULL) {
+		timeout = tvalp->tv_sec * 1000;
+		timeout += (tvalp->tv_usec + 999) / 1000;
+	}
+
+	poll_event_drain_signal_fd(poll_ev);
+
+	if (!poll_event_setup_fresh(ev, poll_ev)) {
+		return -1;
+	}
+
+	tevent_trace_point_callback(poll_ev->ev, TEVENT_TRACE_BEFORE_WAIT);
+	pollrtn = poll(poll_ev->fds, poll_ev->num_fds, timeout);
+	poll_errno = errno;
+	tevent_trace_point_callback(poll_ev->ev, TEVENT_TRACE_AFTER_WAIT);
+
+	if (pollrtn == -1 && poll_errno == EINTR && ev->signal_events) {
+		tevent_common_check_signal(ev);
+		return 0;
+	}
+
+	if (pollrtn == 0 && tvalp) {
+		/* we don't care about a possible delay here */
+		tevent_common_loop_timer_delay(ev);
+		return 0;
+	}
+
+	if (pollrtn <= 0) {
+		/*
+		 * No fd's ready
+		 */
+		return 0;
+	}
+
+	/* at least one file descriptor is ready - check
+	   which ones and call the handler, being careful to allow
+	   the handler to remove itself when called */
+
+	for (fde = ev->fd_events; fde; fde = fde->next) {
+		unsigned idx = fde->additional_flags;
+		struct pollfd *pfd;
+		uint16_t flags = 0;
+
+		if (idx == UINT64_MAX) {
+			continue;
+		}
+
+		pfd = &poll_ev->fds[idx];
+
+		if (pfd->revents & POLLNVAL) {
+			/*
+			 * the socket is dead! this should never
+			 * happen as the socket should have first been
+			 * made readable and that should have removed
+			 * the event, so this must be a bug.
+			 *
+			 * We ignore it here to match the epoll
+			 * behavior.
+			 */
+			tevent_debug(ev, TEVENT_DEBUG_ERROR,
+				     "POLLNVAL on fde[%p] fd[%d] - disabling\n",
+				     fde, pfd->fd);
+			poll_ev->fdes[idx] = NULL;
+			poll_ev->deleted = true;
+			DLIST_REMOVE(ev->fd_events, fde);
+			fde->event_ctx = NULL;
+			continue;
+		}
+
+		if (pfd->revents & (POLLHUP|POLLERR)) {
+			/* If we only wait for TEVENT_FD_WRITE, we
+			   should not tell the event handler about it,
+			   and remove the writable flag, as we only
+			   report errors when waiting for read events
+			   to match the select behavior. */
+			if (!(fde->flags & TEVENT_FD_READ)) {
+				TEVENT_FD_NOT_WRITEABLE(fde);
+				continue;
+			}
+			flags |= TEVENT_FD_READ;
+		}
+		if (pfd->revents & POLLIN) {
+			flags |= TEVENT_FD_READ;
+		}
+		if (pfd->revents & POLLOUT) {
+			flags |= TEVENT_FD_WRITE;
+		}
+		/*
+		 * Note that fde->flags could be changed when using
+		 * the poll_mt backend together with threads,
+		 * that why we need to check pfd->revents and fde->flags
+		 */
+		flags &= fde->flags;
+		if (flags != 0) {
+			DLIST_DEMOTE(ev->fd_events, fde, struct tevent_fd);
+			fde->handler(ev, fde, flags, fde->private_data);
+			return 0;
+		}
+	}
+
+	for (i = 0; i < poll_ev->num_fds; i++) {
+		if (poll_ev->fds[i].revents & POLLNVAL) {
+			/*
+			 * the socket is dead! this should never
+			 * happen as the socket should have first been
+			 * made readable and that should have removed
+			 * the event, so this must be a bug or
+			 * a race in the poll_mt usage.
+			 */
+			fde = poll_ev->fdes[i];
+			tevent_debug(ev, TEVENT_DEBUG_WARNING,
+				     "POLLNVAL on dangling fd[%d] fde[%p] - disabling\n",
+				     poll_ev->fds[i].fd, fde);
+			poll_ev->fdes[i] = NULL;
+			poll_ev->deleted = true;
+			if (fde != NULL) {
+				DLIST_REMOVE(ev->fd_events, fde);
+				fde->event_ctx = NULL;
+			}
+		}
+	}
+
+	return 0;
+}
+
+/*
+  do a single event loop using the events defined in ev
+*/
+static int poll_event_loop_once(struct tevent_context *ev,
+				const char *location)
+{
+	struct timeval tval;
+
+	if (ev->signal_events &&
+	    tevent_common_check_signal(ev)) {
+		return 0;
+	}
+
+	if (ev->immediate_events &&
+	    tevent_common_loop_immediate(ev)) {
+		return 0;
+	}
+
+	tval = tevent_common_loop_timer_delay(ev);
+	if (tevent_timeval_is_zero(&tval)) {
+		return 0;
+	}
+
+	return poll_event_loop_poll(ev, &tval);
+}
+
+static int poll_event_loop_wait(struct tevent_context *ev,
+				const char *location)
+{
+	struct poll_event_context *poll_ev = talloc_get_type_abort(
+		ev->additional_data, struct poll_event_context);
+
+	/*
+	 * loop as long as we have events pending
+	 */
+	while (ev->fd_events ||
+	       ev->timer_events ||
+	       ev->immediate_events ||
+	       ev->signal_events ||
+	       poll_ev->fresh ||
+	       poll_ev->disabled) {
+		int ret;
+		ret = _tevent_loop_once(ev, location);
+		if (ret != 0) {
+			tevent_debug(ev, TEVENT_DEBUG_FATAL,
+				     "_tevent_loop_once() failed: %d - %s\n",
+				     ret, strerror(errno));
+			return ret;
+		}
+	}
+
+	tevent_debug(ev, TEVENT_DEBUG_WARNING,
+		     "poll_event_loop_wait() out of events\n");
+	return 0;
+}
+
+static const struct tevent_ops poll_event_ops = {
+	.context_init		= poll_event_context_init,
+	.add_fd			= poll_event_add_fd,
+	.set_fd_close_fn	= tevent_common_fd_set_close_fn,
+	.get_fd_flags		= tevent_common_fd_get_flags,
+	.set_fd_flags		= poll_event_set_fd_flags,
+	.add_timer		= tevent_common_add_timer_v2,
+	.schedule_immediate	= tevent_common_schedule_immediate,
+	.add_signal		= tevent_common_add_signal,
+	.loop_once		= poll_event_loop_once,
+	.loop_wait		= poll_event_loop_wait,
+};
+
+_PRIVATE_ bool tevent_poll_init(void)
+{
+	return tevent_register_backend("poll", &poll_event_ops);
+}
+
+static const struct tevent_ops poll_event_mt_ops = {
+	.context_init		= poll_event_context_init_mt,
+	.add_fd			= poll_event_add_fd,
+	.set_fd_close_fn	= tevent_common_fd_set_close_fn,
+	.get_fd_flags		= tevent_common_fd_get_flags,
+	.set_fd_flags		= poll_event_set_fd_flags,
+	.add_timer		= tevent_common_add_timer_v2,
+	.schedule_immediate	= poll_event_schedule_immediate,
+	.add_signal		= tevent_common_add_signal,
+	.loop_once		= poll_event_loop_once,
+	.loop_wait		= poll_event_loop_wait,
+};
+
+_PRIVATE_ bool tevent_poll_mt_init(void)
+{
+	return tevent_register_backend("poll_mt", &poll_event_mt_ops);
+}
diff --git a/ctdb/lib/tevent/tevent_queue.c b/ctdb/lib/tevent/tevent_queue.c
new file mode 100644
index 0000000000..4750675802
--- /dev/null
+++ b/ctdb/lib/tevent/tevent_queue.c
@@ -0,0 +1,300 @@
+/*
+   Unix SMB/CIFS implementation.
+   Infrastructure for async requests
+   Copyright (C) Volker Lendecke 2008
+   Copyright (C) Stefan Metzmacher 2009
+
+     ** NOTE! The following LGPL license applies to the tevent
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "replace.h"
+#include "tevent.h"
+#include "tevent_internal.h"
+#include "tevent_util.h"
+
+struct tevent_queue_entry {
+	struct tevent_queue_entry *prev, *next;
+	struct tevent_queue *queue;
+
+	bool triggered;
+
+	struct tevent_req *req;
+	struct tevent_context *ev;
+
+	tevent_queue_trigger_fn_t trigger;
+	void *private_data;
+};
+
+struct tevent_queue {
+	const char *name;
+	const char *location;
+
+	bool running;
+	struct tevent_immediate *immediate;
+
+	size_t length;
+	struct tevent_queue_entry *list;
+};
+
+static void tevent_queue_immediate_trigger(struct tevent_context *ev,
+					   struct tevent_immediate *im,
+					   void *private_data);
+
+static int tevent_queue_entry_destructor(struct tevent_queue_entry *e)
+{
+	struct tevent_queue *q = e->queue;
+
+	if (!q) {
+		return 0;
+	}
+
+	DLIST_REMOVE(q->list, e);
+	q->length--;
+
+	if (!q->running) {
+		return 0;
+	}
+
+	if (!q->list) {
+		return 0;
+	}
+
+	if (q->list->triggered) {
+		return 0;
+	}
+
+	tevent_schedule_immediate(q->immediate,
+				  q->list->ev,
+				  tevent_queue_immediate_trigger,
+				  q);
+
+	return 0;
+}
+
+static int tevent_queue_destructor(struct tevent_queue *q)
+{
+	q->running = false;
+
+	while (q->list) {
+		struct tevent_queue_entry *e = q->list;
+		talloc_free(e);
+	}
+
+	return 0;
+}
+
+struct tevent_queue *_tevent_queue_create(TALLOC_CTX *mem_ctx,
+					  const char *name,
+					  const char *location)
+{
+	struct tevent_queue *queue;
+
+	queue = talloc_zero(mem_ctx, struct tevent_queue);
+	if (!queue) {
+		return NULL;
+	}
+
+	queue->name = talloc_strdup(queue, name);
+	if (!queue->name) {
+		talloc_free(queue);
+		return NULL;
+	}
+	queue->immediate = tevent_create_immediate(queue);
+	if (!queue->immediate) {
+		talloc_free(queue);
+		return NULL;
+	}
+
+	queue->location = location;
+
+	/* queue is running by default */
+	queue->running = true;
+
+	talloc_set_destructor(queue, tevent_queue_destructor);
+	return queue;
+}
+
+static void tevent_queue_immediate_trigger(struct tevent_context *ev,
+					   struct tevent_immediate *im,
+					   void *private_data)
+{
+	struct tevent_queue *q = talloc_get_type(private_data,
+				  struct tevent_queue);
+
+	if (!q->running) {
+		return;
+	}
+
+	q->list->triggered = true;
+	q->list->trigger(q->list->req, q->list->private_data);
+}
+
+static struct tevent_queue_entry *tevent_queue_add_internal(
+					struct tevent_queue *queue,
+					struct tevent_context *ev,
+					struct tevent_req *req,
+					tevent_queue_trigger_fn_t trigger,
+					void *private_data,
+					bool allow_direct)
+{
+	struct tevent_queue_entry *e;
+
+	e = talloc_zero(req, struct tevent_queue_entry);
+	if (e == NULL) {
+		return NULL;
+	}
+
+	e->queue = queue;
+	e->req = req;
+	e->ev = ev;
+	e->trigger = trigger;
+	e->private_data = private_data;
+
+	/*
+	 * if there is no trigger, it is just a blocker
+	 */
+	if (trigger == NULL) {
+		e->triggered = true;
+	}
+
+	if (queue->length > 0) {
+		/*
+		 * if there are already entries in the
+		 * queue do not optimize.
+		 */
+		allow_direct = false;
+	}
+
+	if (req->async.fn != NULL) {
+		/*
+		 * If the callers wants to optimize for the
+		 * empty queue case, call the trigger only
+		 * if there is no callback defined for the
+		 * request yet.
+		 */
+		allow_direct = false;
+	}
+
+	DLIST_ADD_END(queue->list, e, struct tevent_queue_entry *);
+	queue->length++;
+	talloc_set_destructor(e, tevent_queue_entry_destructor);
+
+	if (!queue->running) {
+		return e;
+	}
+
+	if (queue->list->triggered) {
+		return e;
+	}
+
+	/*
+	 * If allowed we directly call the trigger
+	 * avoiding possible delays caused by
+	 * an immediate event.
+	 */
+	if (allow_direct) {
+		queue->list->triggered = true;
+		queue->list->trigger(queue->list->req,
+				     queue->list->private_data);
+		return e;
+	}
+
+	tevent_schedule_immediate(queue->immediate,
+				  queue->list->ev,
+				  tevent_queue_immediate_trigger,
+				  queue);
+
+	return e;
+}
+
+bool tevent_queue_add(struct tevent_queue *queue,
+		      struct tevent_context *ev,
+		      struct tevent_req *req,
+		      tevent_queue_trigger_fn_t trigger,
+		      void *private_data)
+{
+	struct tevent_queue_entry *e;
+
+	e = tevent_queue_add_internal(queue, ev, req,
+				      trigger, private_data, false);
+	if (e == NULL) {
+		return false;
+	}
+
+	return true;
+}
+
+struct tevent_queue_entry *tevent_queue_add_entry(
+					struct tevent_queue *queue,
+					struct tevent_context *ev,
+					struct tevent_req *req,
+					tevent_queue_trigger_fn_t trigger,
+					void *private_data)
+{
+	return tevent_queue_add_internal(queue, ev, req,
+					 trigger, private_data, false);
+}
+
+struct tevent_queue_entry *tevent_queue_add_optimize_empty(
+					struct tevent_queue *queue,
+					struct tevent_context *ev,
+					struct tevent_req *req,
+					tevent_queue_trigger_fn_t trigger,
+					void *private_data)
+{
+	return tevent_queue_add_internal(queue, ev, req,
+					 trigger, private_data, true);
+}
+
+void tevent_queue_start(struct tevent_queue *queue)
+{
+	if (queue->running) {
+		/* already started */
+		return;
+	}
+
+	queue->running = true;
+
+	if (!queue->list) {
+		return;
+	}
+
+	if (queue->list->triggered) {
+		return;
+	}
+
+	tevent_schedule_immediate(queue->immediate,
+				  queue->list->ev,
+				  tevent_queue_immediate_trigger,
+				  queue);
+}
+
+void tevent_queue_stop(struct tevent_queue *queue)
+{
+	queue->running = false;
+}
+
+size_t tevent_queue_length(struct tevent_queue *queue)
+{
+	return queue->length;
+}
+
+bool tevent_queue_running(struct tevent_queue *queue)
+{
+	return queue->running;
+}
diff --git a/ctdb/lib/tevent/tevent_req.c b/ctdb/lib/tevent/tevent_req.c
new file mode 100644
index 0000000000..d8d0c5f564
--- /dev/null
+++ b/ctdb/lib/tevent/tevent_req.c
@@ -0,0 +1,294 @@
+/*
+   Unix SMB/CIFS implementation.
+   Infrastructure for async requests
+   Copyright (C) Volker Lendecke 2008
+   Copyright (C) Stefan Metzmacher 2009
+
+     ** NOTE! The following LGPL license applies to the tevent
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "replace.h"
+#include "tevent.h"
+#include "tevent_internal.h"
+#include "tevent_util.h"
+
+char *tevent_req_default_print(struct tevent_req *req, TALLOC_CTX *mem_ctx)
+{
+	return talloc_asprintf(mem_ctx,
+			       "tevent_req[%p/%s]: state[%d] error[%lld (0x%llX)] "
+			       " state[%s (%p)] timer[%p]",
+			       req, req->internal.create_location,
+			       req->internal.state,
+			       (unsigned long long)req->internal.error,
+			       (unsigned long long)req->internal.error,
+			       talloc_get_name(req->data),
+			       req->data,
+			       req->internal.timer
+			       );
+}
+
+char *tevent_req_print(TALLOC_CTX *mem_ctx, struct tevent_req *req)
+{
+	if (!req->private_print) {
+		return tevent_req_default_print(req, mem_ctx);
+	}
+
+	return req->private_print(req, mem_ctx);
+}
+
+struct tevent_req *_tevent_req_create(TALLOC_CTX *mem_ctx,
+				    void *pdata,
+				    size_t data_size,
+				    const char *type,
+				    const char *location)
+{
+	struct tevent_req *req;
+	void **ppdata = (void **)pdata;
+	void *data;
+
+	req = talloc_zero(mem_ctx, struct tevent_req);
+	if (req == NULL) {
+		return NULL;
+	}
+	req->internal.private_type	= type;
+	req->internal.create_location	= location;
+	req->internal.finish_location	= NULL;
+	req->internal.state		= TEVENT_REQ_IN_PROGRESS;
+	req->internal.trigger		= tevent_create_immediate(req);
+	if (!req->internal.trigger) {
+		talloc_free(req);
+		return NULL;
+	}
+	req->internal.defer_callback_ev	= NULL;
+
+	data = talloc_zero_size(req, data_size);
+	if (data == NULL) {
+		talloc_free(req);
+		return NULL;
+	}
+	talloc_set_name_const(data, type);
+
+	req->data = data;
+
+	*ppdata = data;
+	return req;
+}
+
+void _tevent_req_notify_callback(struct tevent_req *req, const char *location)
+{
+	req->internal.finish_location = location;
+	if (req->internal.defer_callback_ev) {
+		(void)tevent_req_post(req, req->internal.defer_callback_ev);
+		req->internal.defer_callback_ev = NULL;
+		return;
+	}
+	if (req->async.fn != NULL) {
+		req->async.fn(req);
+	}
+}
+
+static void tevent_req_finish(struct tevent_req *req,
+			      enum tevent_req_state state,
+			      const char *location)
+{
+	req->internal.state = state;
+	_tevent_req_notify_callback(req, location);
+}
+
+void _tevent_req_done(struct tevent_req *req,
+		      const char *location)
+{
+	tevent_req_finish(req, TEVENT_REQ_DONE, location);
+}
+
+bool _tevent_req_error(struct tevent_req *req,
+		       uint64_t error,
+		       const char *location)
+{
+	if (error == 0) {
+		return false;
+	}
+
+	req->internal.error = error;
+	tevent_req_finish(req, TEVENT_REQ_USER_ERROR, location);
+	return true;
+}
+
+void _tevent_req_oom(struct tevent_req *req, const char *location)
+{
+	tevent_req_finish(req, TEVENT_REQ_NO_MEMORY, location);
+}
+
+bool _tevent_req_nomem(const void *p,
+		       struct tevent_req *req,
+		       const char *location)
+{
+	if (p != NULL) {
+		return false;
+	}
+	_tevent_req_oom(req, location);
+	return true;
+}
+
+/**
+ * @internal
+ *
+ * @brief Immediate event callback.
+ *
+ * @param[in]  ev       The event context to use.
+ *
+ * @param[in]  im       The immediate event.
+ *
+ * @param[in]  priv     The async request to be finished.
+ */
+static void tevent_req_trigger(struct tevent_context *ev,
+			       struct tevent_immediate *im,
+			       void *private_data)
+{
+	struct tevent_req *req = talloc_get_type(private_data,
+				 struct tevent_req);
+
+	tevent_req_finish(req, req->internal.state,
+			  req->internal.finish_location);
+}
+
+struct tevent_req *tevent_req_post(struct tevent_req *req,
+				   struct tevent_context *ev)
+{
+	tevent_schedule_immediate(req->internal.trigger,
+				  ev, tevent_req_trigger, req);
+	return req;
+}
+
+void tevent_req_defer_callback(struct tevent_req *req,
+			       struct tevent_context *ev)
+{
+	req->internal.defer_callback_ev = ev;
+}
+
+bool tevent_req_is_in_progress(struct tevent_req *req)
+{
+	if (req->internal.state == TEVENT_REQ_IN_PROGRESS) {
+		return true;
+	}
+
+	return false;
+}
+
+void tevent_req_received(struct tevent_req *req)
+{
+	TALLOC_FREE(req->data);
+	req->private_print = NULL;
+
+	TALLOC_FREE(req->internal.trigger);
+	TALLOC_FREE(req->internal.timer);
+
+	req->internal.state = TEVENT_REQ_RECEIVED;
+}
+
+bool tevent_req_poll(struct tevent_req *req,
+		     struct tevent_context *ev)
+{
+	while (tevent_req_is_in_progress(req)) {
+		int ret;
+
+		ret = tevent_loop_once(ev);
+		if (ret != 0) {
+			return false;
+		}
+	}
+
+	return true;
+}
+
+bool tevent_req_is_error(struct tevent_req *req, enum tevent_req_state *state,
+			uint64_t *error)
+{
+	if (req->internal.state == TEVENT_REQ_DONE) {
+		return false;
+	}
+	if (req->internal.state == TEVENT_REQ_USER_ERROR) {
+		*error = req->internal.error;
+	}
+	*state = req->internal.state;
+	return true;
+}
+
+static void tevent_req_timedout(struct tevent_context *ev,
+			       struct tevent_timer *te,
+			       struct timeval now,
+			       void *private_data)
+{
+	struct tevent_req *req = talloc_get_type(private_data,
+				 struct tevent_req);
+
+	TALLOC_FREE(req->internal.timer);
+
+	tevent_req_finish(req, TEVENT_REQ_TIMED_OUT, __FUNCTION__);
+}
+
+bool tevent_req_set_endtime(struct tevent_req *req,
+			    struct tevent_context *ev,
+			    struct timeval endtime)
+{
+	TALLOC_FREE(req->internal.timer);
+
+	req->internal.timer = tevent_add_timer(ev, req, endtime,
+					       tevent_req_timedout,
+					       req);
+	if (tevent_req_nomem(req->internal.timer, req)) {
+		return false;
+	}
+
+	return true;
+}
+
+void tevent_req_set_callback(struct tevent_req *req, tevent_req_fn fn, void *pvt)
+{
+	req->async.fn = fn;
+	req->async.private_data = pvt;
+}
+
+void *_tevent_req_callback_data(struct tevent_req *req)
+{
+	return req->async.private_data;
+}
+
+void *_tevent_req_data(struct tevent_req *req)
+{
+	return req->data;
+}
+
+void tevent_req_set_print_fn(struct tevent_req *req, tevent_req_print_fn fn)
+{
+	req->private_print = fn;
+}
+
+void tevent_req_set_cancel_fn(struct tevent_req *req, tevent_req_cancel_fn fn)
+{
+	req->private_cancel = fn;
+}
+
+bool _tevent_req_cancel(struct tevent_req *req, const char *location)
+{
+	if (req->private_cancel == NULL) {
+		return false;
+	}
+
+	return req->private_cancel(req);
+}
diff --git a/ctdb/lib/tevent/tevent_select.c b/ctdb/lib/tevent/tevent_select.c
new file mode 100644
index 0000000000..bfce246e22
--- /dev/null
+++ b/ctdb/lib/tevent/tevent_select.c
@@ -0,0 +1,277 @@
+/* 
+   Unix SMB/CIFS implementation.
+   main select loop and event handling
+   Copyright (C) Andrew Tridgell	2003-2005
+   Copyright (C) Stefan Metzmacher	2005-2009
+
+     ** NOTE! The following LGPL license applies to the tevent
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "replace.h"
+#include "system/filesys.h"
+#include "system/select.h"
+#include "tevent.h"
+#include "tevent_util.h"
+#include "tevent_internal.h"
+
+struct select_event_context {
+	/* a pointer back to the generic event_context */
+	struct tevent_context *ev;
+
+	/* the maximum file descriptor number in fd_events */
+	int maxfd;
+
+	/* information for exiting from the event loop */
+	int exit_code;
+};
+
+/*
+  create a select_event_context structure.
+*/
+static int select_event_context_init(struct tevent_context *ev)
+{
+	struct select_event_context *select_ev;
+
+	/*
+	 * We might be called during tevent_re_initialise()
+	 * which means we need to free our old additional_data.
+	 */
+	TALLOC_FREE(ev->additional_data);
+
+	select_ev = talloc_zero(ev, struct select_event_context);
+	if (!select_ev) return -1;
+	select_ev->ev = ev;
+
+	ev->additional_data = select_ev;
+	return 0;
+}
+
+/*
+  recalculate the maxfd
+*/
+static void calc_maxfd(struct select_event_context *select_ev)
+{
+	struct tevent_fd *fde;
+
+	select_ev->maxfd = 0;
+	for (fde = select_ev->ev->fd_events; fde; fde = fde->next) {
+		if (fde->fd > select_ev->maxfd) {
+			select_ev->maxfd = fde->fd;
+		}
+	}
+}
+
+
+/* to mark the ev->maxfd invalid
+ * this means we need to recalculate it
+ */
+#define EVENT_INVALID_MAXFD (-1)
+
+/*
+  destroy an fd_event
+*/
+static int select_event_fd_destructor(struct tevent_fd *fde)
+{
+	struct tevent_context *ev = fde->event_ctx;
+	struct select_event_context *select_ev = NULL;
+
+	if (ev) {
+		select_ev = talloc_get_type(ev->additional_data,
+					    struct select_event_context);
+
+		if (select_ev->maxfd == fde->fd) {
+			select_ev->maxfd = EVENT_INVALID_MAXFD;
+		}
+	}
+
+	return tevent_common_fd_destructor(fde);
+}
+
+/*
+  add a fd based event
+  return NULL on failure (memory allocation error)
+*/
+static struct tevent_fd *select_event_add_fd(struct tevent_context *ev, TALLOC_CTX *mem_ctx,
+					     int fd, uint16_t flags,
+					     tevent_fd_handler_t handler,
+					     void *private_data,
+					     const char *handler_name,
+					     const char *location)
+{
+	struct select_event_context *select_ev = talloc_get_type(ev->additional_data,
+							   struct select_event_context);
+	struct tevent_fd *fde;
+
+	if (fd < 0 || fd >= FD_SETSIZE) {
+		errno = EBADF;
+		return NULL;
+	}
+
+	fde = tevent_common_add_fd(ev, mem_ctx, fd, flags,
+				   handler, private_data,
+				   handler_name, location);
+	if (!fde) return NULL;
+
+	if ((select_ev->maxfd != EVENT_INVALID_MAXFD)
+	    && (fde->fd > select_ev->maxfd)) {
+		select_ev->maxfd = fde->fd;
+	}
+	talloc_set_destructor(fde, select_event_fd_destructor);
+
+	return fde;
+}
+
+/*
+  event loop handling using select()
+*/
+static int select_event_loop_select(struct select_event_context *select_ev, struct timeval *tvalp)
+{
+	fd_set r_fds, w_fds;
+	struct tevent_fd *fde;
+	int selrtn;
+	int select_errno;
+
+	/* we maybe need to recalculate the maxfd */
+	if (select_ev->maxfd == EVENT_INVALID_MAXFD) {
+		calc_maxfd(select_ev);
+	}
+
+	FD_ZERO(&r_fds);
+	FD_ZERO(&w_fds);
+
+	/* setup any fd events */
+	for (fde = select_ev->ev->fd_events; fde; fde = fde->next) {
+		if (fde->fd < 0 || fde->fd >= FD_SETSIZE) {
+			tevent_debug(select_ev->ev, TEVENT_DEBUG_FATAL,
+				     "ERROR: EBADF fd[%d] >= %d "
+				     "select_event_loop_once\n",
+				     fde->fd, FD_SETSIZE);
+			errno = EBADF;
+			return -1;
+		}
+
+		if (fde->flags & TEVENT_FD_READ) {
+			FD_SET(fde->fd, &r_fds);
+		}
+		if (fde->flags & TEVENT_FD_WRITE) {
+			FD_SET(fde->fd, &w_fds);
+		}
+	}
+
+	if (select_ev->ev->signal_events &&
+	    tevent_common_check_signal(select_ev->ev)) {
+		return 0;
+	}
+
+	tevent_trace_point_callback(select_ev->ev, TEVENT_TRACE_BEFORE_WAIT);
+	selrtn = select(select_ev->maxfd+1, &r_fds, &w_fds, NULL, tvalp);
+	select_errno = errno;
+	tevent_trace_point_callback(select_ev->ev, TEVENT_TRACE_AFTER_WAIT);
+
+	if (selrtn == -1 && select_errno == EINTR &&
+	    select_ev->ev->signal_events) {
+		tevent_common_check_signal(select_ev->ev);
+		return 0;
+	}
+
+	if (selrtn == -1 && select_errno == EBADF) {
+		/* the socket is dead! this should never
+		   happen as the socket should have first been
+		   made readable and that should have removed
+		   the event, so this must be a bug. This is a
+		   fatal error. */
+		tevent_debug(select_ev->ev, TEVENT_DEBUG_FATAL,
+			     "ERROR: EBADF on select_event_loop_once\n");
+		select_ev->exit_code = EBADF;
+		return -1;
+	}
+
+	if (selrtn == 0 && tvalp) {
+		/* we don't care about a possible delay here */
+		tevent_common_loop_timer_delay(select_ev->ev);
+		return 0;
+	}
+
+	if (selrtn > 0) {
+		/* at least one file descriptor is ready - check
+		   which ones and call the handler, being careful to allow
+		   the handler to remove itself when called */
+		for (fde = select_ev->ev->fd_events; fde; fde = fde->next) {
+			uint16_t flags = 0;
+
+			if (FD_ISSET(fde->fd, &r_fds) && (fde->flags & TEVENT_FD_READ)) {
+				flags |= TEVENT_FD_READ;
+			}
+			if (FD_ISSET(fde->fd, &w_fds) && (fde->flags & TEVENT_FD_WRITE)) {
+				flags |= TEVENT_FD_WRITE;
+			}
+			if (flags) {
+				DLIST_DEMOTE(select_ev->ev->fd_events, fde, struct tevent_fd);
+				fde->handler(select_ev->ev, fde, flags, fde->private_data);
+				break;
+			}
+		}
+	}
+
+	return 0;
+}
+
+/*
+  do a single event loop using the events defined in ev 
+*/
+static int select_event_loop_once(struct tevent_context *ev, const char *location)
+{
+	struct select_event_context *select_ev = talloc_get_type(ev->additional_data,
+		 					   struct select_event_context);
+	struct timeval tval;
+
+	if (ev->signal_events &&
+	    tevent_common_check_signal(ev)) {
+		return 0;
+	}
+
+	if (ev->immediate_events &&
+	    tevent_common_loop_immediate(ev)) {
+		return 0;
+	}
+
+	tval = tevent_common_loop_timer_delay(ev);
+	if (tevent_timeval_is_zero(&tval)) {
+		return 0;
+	}
+
+	return select_event_loop_select(select_ev, &tval);
+}
+
+static const struct tevent_ops select_event_ops = {
+	.context_init		= select_event_context_init,
+	.add_fd			= select_event_add_fd,
+	.set_fd_close_fn	= tevent_common_fd_set_close_fn,
+	.get_fd_flags		= tevent_common_fd_get_flags,
+	.set_fd_flags		= tevent_common_fd_set_flags,
+	.add_timer		= tevent_common_add_timer_v2,
+	.schedule_immediate	= tevent_common_schedule_immediate,
+	.add_signal		= tevent_common_add_signal,
+	.loop_once		= select_event_loop_once,
+	.loop_wait		= tevent_common_loop_wait,
+};
+
+_PRIVATE_ bool tevent_select_init(void)
+{
+	return tevent_register_backend("select", &select_event_ops);
+}
diff --git a/ctdb/lib/tevent/tevent_signal.c b/ctdb/lib/tevent/tevent_signal.c
new file mode 100644
index 0000000000..b5a56ef03a
--- /dev/null
+++ b/ctdb/lib/tevent/tevent_signal.c
@@ -0,0 +1,484 @@
+/* 
+   Unix SMB/CIFS implementation.
+
+   common events code for signal events
+
+   Copyright (C) Andrew Tridgell	2007
+
+     ** NOTE! The following LGPL license applies to the tevent
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "replace.h"
+#include "system/filesys.h"
+#include "system/wait.h"
+#include "tevent.h"
+#include "tevent_internal.h"
+#include "tevent_util.h"
+
+/* maximum number of SA_SIGINFO signals to hold in the queue.
+  NB. This *MUST* be a power of 2, in order for the ring buffer
+  wrap to work correctly. Thanks to Petr Vandrovec <petr@vandrovec.name>
+  for this. */
+
+#define TEVENT_SA_INFO_QUEUE_COUNT 64
+
+struct tevent_sigcounter {
+	uint32_t count;
+	uint32_t seen;
+};
+
+#define TEVENT_SIG_INCREMENT(s) (s).count++
+#define TEVENT_SIG_SEEN(s, n) (s).seen += (n)
+#define TEVENT_SIG_PENDING(s) ((s).seen != (s).count)
+
+struct tevent_common_signal_list {
+	struct tevent_common_signal_list *prev, *next;
+	struct tevent_signal *se;
+};
+
+/*
+  the poor design of signals means that this table must be static global
+*/
+static struct tevent_sig_state {
+	struct tevent_common_signal_list *sig_handlers[TEVENT_NUM_SIGNALS+1];
+	struct sigaction *oldact[TEVENT_NUM_SIGNALS+1];
+	struct tevent_sigcounter signal_count[TEVENT_NUM_SIGNALS+1];
+	struct tevent_sigcounter got_signal;
+#ifdef SA_SIGINFO
+	/* with SA_SIGINFO we get quite a lot of info per signal */
+	siginfo_t *sig_info[TEVENT_NUM_SIGNALS+1];
+	struct tevent_sigcounter sig_blocked[TEVENT_NUM_SIGNALS+1];
+#endif
+} *sig_state;
+
+/*
+  return number of sigcounter events not processed yet
+*/
+static uint32_t tevent_sig_count(struct tevent_sigcounter s)
+{
+	return s.count - s.seen;
+}
+
+/*
+  signal handler - redirects to registered signals
+*/
+static void tevent_common_signal_handler(int signum)
+{
+	char c = 0;
+	struct tevent_common_signal_list *sl;
+	struct tevent_context *ev = NULL;
+	int saved_errno = errno;
+
+	TEVENT_SIG_INCREMENT(sig_state->signal_count[signum]);
+	TEVENT_SIG_INCREMENT(sig_state->got_signal);
+
+	/* Write to each unique event context. */
+	for (sl = sig_state->sig_handlers[signum]; sl; sl = sl->next) {
+		if (sl->se->event_ctx && sl->se->event_ctx != ev) {
+			ev = sl->se->event_ctx;
+			/* doesn't matter if this pipe overflows */
+			(void) write(ev->pipe_fds[1], &c, 1);
+		}
+	}
+
+	errno = saved_errno;
+}
+
+#ifdef SA_SIGINFO
+/*
+  signal handler with SA_SIGINFO - redirects to registered signals
+*/
+static void tevent_common_signal_handler_info(int signum, siginfo_t *info,
+					      void *uctx)
+{
+	uint32_t count = tevent_sig_count(sig_state->signal_count[signum]);
+	/* sig_state->signal_count[signum].seen % TEVENT_SA_INFO_QUEUE_COUNT
+	 * is the base of the unprocessed signals in the ringbuffer. */
+	uint32_t ofs = (sig_state->signal_count[signum].seen + count) %
+				TEVENT_SA_INFO_QUEUE_COUNT;
+	sig_state->sig_info[signum][ofs] = *info;
+
+	tevent_common_signal_handler(signum);
+
+	/* handle SA_SIGINFO */
+	if (count+1 == TEVENT_SA_INFO_QUEUE_COUNT) {
+		/* we've filled the info array - block this signal until
+		   these ones are delivered */
+#ifdef HAVE_UCONTEXT_T
+		/*
+		 * This is the only way for this to work.
+		 * By default signum is blocked inside this
+		 * signal handler using a temporary mask,
+		 * but what we really need to do now is
+		 * block it in the callers mask, so it
+		 * stays blocked when the temporary signal
+		 * handler mask is replaced when we return
+		 * from here. The callers mask can be found
+		 * in the ucontext_t passed in as the
+		 * void *uctx argument.
+		 */
+		ucontext_t *ucp = (ucontext_t *)uctx;
+		sigaddset(&ucp->uc_sigmask, signum);
+#else
+		/*
+		 * WARNING !!! WARNING !!!!
+		 *
+		 * This code doesn't work.
+		 * By default signum is blocked inside this
+		 * signal handler, but calling sigprocmask
+		 * modifies the temporary signal mask being
+		 * used *inside* this handler, which will be
+		 * replaced by the callers signal mask once
+		 * we return from here. See Samba
+		 * bug #9550 for details.
+		 */
+		sigset_t set;
+		sigemptyset(&set);
+		sigaddset(&set, signum);
+		sigprocmask(SIG_BLOCK, &set, NULL);
+#endif
+		TEVENT_SIG_INCREMENT(sig_state->sig_blocked[signum]);
+	}
+}
+#endif
+
+static int tevent_common_signal_list_destructor(struct tevent_common_signal_list *sl)
+{
+	if (sig_state->sig_handlers[sl->se->signum]) {
+		DLIST_REMOVE(sig_state->sig_handlers[sl->se->signum], sl);
+	}
+	return 0;
+}
+
+/*
+  destroy a signal event
+*/
+static int tevent_signal_destructor(struct tevent_signal *se)
+{
+	struct tevent_common_signal_list *sl;
+	sl = talloc_get_type(se->additional_data,
+			     struct tevent_common_signal_list);
+
+	if (se->event_ctx) {
+		DLIST_REMOVE(se->event_ctx->signal_events, se);
+	}
+
+	talloc_free(sl);
+
+	if (sig_state->sig_handlers[se->signum] == NULL) {
+		/* restore old handler, if any */
+		if (sig_state->oldact[se->signum]) {
+			sigaction(se->signum, sig_state->oldact[se->signum], NULL);
+			sig_state->oldact[se->signum] = NULL;
+		}
+#ifdef SA_SIGINFO
+		if (se->sa_flags & SA_SIGINFO) {
+			if (sig_state->sig_info[se->signum]) {
+				talloc_free(sig_state->sig_info[se->signum]);
+				sig_state->sig_info[se->signum] = NULL;
+			}
+		}
+#endif
+	}
+
+	return 0;
+}
+
+/*
+  this is part of the pipe hack needed to avoid the signal race condition
+*/
+static void signal_pipe_handler(struct tevent_context *ev, struct tevent_fd *fde, 
+				uint16_t flags, void *_private)
+{
+	char c[16];
+	/* its non-blocking, doesn't matter if we read too much */
+	(void) read(fde->fd, c, sizeof(c));
+}
+
+/*
+  add a signal event
+  return NULL on failure (memory allocation error)
+*/
+struct tevent_signal *tevent_common_add_signal(struct tevent_context *ev,
+					       TALLOC_CTX *mem_ctx,
+					       int signum,
+					       int sa_flags,
+					       tevent_signal_handler_t handler,
+					       void *private_data,
+					       const char *handler_name,
+					       const char *location)
+{
+	struct tevent_signal *se;
+	struct tevent_common_signal_list *sl;
+	sigset_t set, oldset;
+
+	if (signum >= TEVENT_NUM_SIGNALS) {
+		errno = EINVAL;
+		return NULL;
+	}
+
+	/* the sig_state needs to be on a global context as it can last across
+	   multiple event contexts */
+	if (sig_state == NULL) {
+		sig_state = talloc_zero(NULL, struct tevent_sig_state);
+		if (sig_state == NULL) {
+			return NULL;
+		}
+	}
+
+	se = talloc(mem_ctx?mem_ctx:ev, struct tevent_signal);
+	if (se == NULL) return NULL;
+
+	se->event_ctx		= ev;
+	se->signum              = signum;
+	se->sa_flags            = sa_flags;
+	se->handler		= handler;
+	se->private_data	= private_data;
+	se->handler_name	= handler_name;
+	se->location		= location;
+	se->additional_data	= NULL;
+
+	sl = talloc(se, struct tevent_common_signal_list);
+	if (!sl) {
+		talloc_free(se);
+		return NULL;
+	}
+	sl->se = se;
+	se->additional_data	= sl;
+
+	/* Ensure, no matter the destruction order, that we always have a handle on the global sig_state */
+	if (!talloc_reference(se, sig_state)) {
+		talloc_free(se);
+		return NULL;
+	}
+
+	/* we need to setup the pipe hack handler if not already
+	   setup */
+	if (ev->pipe_fde == NULL) {
+		if (pipe(ev->pipe_fds) == -1) {
+			talloc_free(se);
+			return NULL;
+		}
+		ev_set_blocking(ev->pipe_fds[0], false);
+		ev_set_blocking(ev->pipe_fds[1], false);
+		ev->pipe_fde = tevent_add_fd(ev, ev, ev->pipe_fds[0],
+					     TEVENT_FD_READ,
+					     signal_pipe_handler, NULL);
+		if (!ev->pipe_fde) {
+			close(ev->pipe_fds[0]);
+			close(ev->pipe_fds[1]);
+			talloc_free(se);
+			return NULL;
+		}
+	}
+
+	/* only install a signal handler if not already installed */
+	if (sig_state->sig_handlers[signum] == NULL) {
+		struct sigaction act;
+		ZERO_STRUCT(act);
+		act.sa_handler = tevent_common_signal_handler;
+		act.sa_flags = sa_flags;
+#ifdef SA_SIGINFO
+		if (sa_flags & SA_SIGINFO) {
+			act.sa_handler   = NULL;
+			act.sa_sigaction = tevent_common_signal_handler_info;
+			if (sig_state->sig_info[signum] == NULL) {
+				sig_state->sig_info[signum] =
+					talloc_zero_array(sig_state, siginfo_t,
+							  TEVENT_SA_INFO_QUEUE_COUNT);
+				if (sig_state->sig_info[signum] == NULL) {
+					talloc_free(se);
+					return NULL;
+				}
+			}
+		}
+#endif
+		sig_state->oldact[signum] = talloc(sig_state, struct sigaction);
+		if (sig_state->oldact[signum] == NULL) {
+			talloc_free(se);
+			return NULL;			
+		}
+		if (sigaction(signum, &act, sig_state->oldact[signum]) == -1) {
+			talloc_free(se);
+			return NULL;
+		}
+	}
+
+	DLIST_ADD(se->event_ctx->signal_events, se);
+
+	/* Make sure the signal doesn't come in while we're mangling list. */
+	sigemptyset(&set);
+	sigaddset(&set, signum);
+	sigprocmask(SIG_BLOCK, &set, &oldset);
+	DLIST_ADD(sig_state->sig_handlers[signum], sl);
+	sigprocmask(SIG_SETMASK, &oldset, NULL);
+
+	talloc_set_destructor(se, tevent_signal_destructor);
+	talloc_set_destructor(sl, tevent_common_signal_list_destructor);
+
+	return se;
+}
+
+struct tevent_se_exists {
+	struct tevent_se_exists **myself;
+};
+
+static int tevent_se_exists_destructor(struct tevent_se_exists *s)
+{
+	*s->myself = NULL;
+	return 0;
+}
+
+/*
+  check if a signal is pending
+  return != 0 if a signal was pending
+*/
+int tevent_common_check_signal(struct tevent_context *ev)
+{
+	int i;
+
+	if (!sig_state || !TEVENT_SIG_PENDING(sig_state->got_signal)) {
+		return 0;
+	}
+	
+	for (i=0;i<TEVENT_NUM_SIGNALS+1;i++) {
+		struct tevent_common_signal_list *sl, *next;
+		struct tevent_sigcounter counter = sig_state->signal_count[i];
+		uint32_t count = tevent_sig_count(counter);
+#ifdef SA_SIGINFO
+		/* Ensure we null out any stored siginfo_t entries
+		 * after processing for debugging purposes. */
+		bool clear_processed_siginfo = false;
+#endif
+
+		if (count == 0) {
+			continue;
+		}
+		for (sl=sig_state->sig_handlers[i];sl;sl=next) {
+			struct tevent_signal *se = sl->se;
+			struct tevent_se_exists *exists;
+
+			next = sl->next;
+
+			/*
+			 * We have to be careful to not touch "se"
+			 * after it was deleted in its handler. Thus
+			 * we allocate a child whose destructor will
+			 * tell by nulling out itself that its parent
+			 * is gone.
+			 */
+			exists = talloc(se, struct tevent_se_exists);
+			if (exists == NULL) {
+				continue;
+			}
+			exists->myself = &exists;
+			talloc_set_destructor(
+				exists, tevent_se_exists_destructor);
+
+#ifdef SA_SIGINFO
+			if (se->sa_flags & SA_SIGINFO) {
+				uint32_t j;
+
+				clear_processed_siginfo = true;
+
+				for (j=0;j<count;j++) {
+					/* sig_state->signal_count[i].seen
+					 * % TEVENT_SA_INFO_QUEUE_COUNT is
+					 * the base position of the unprocessed
+					 * signals in the ringbuffer. */
+					uint32_t ofs = (counter.seen + j)
+						% TEVENT_SA_INFO_QUEUE_COUNT;
+					se->handler(ev, se, i, 1,
+						    (void*)&sig_state->sig_info[i][ofs], 
+						    se->private_data);
+					if (!exists) {
+						break;
+					}
+				}
+#ifdef SA_RESETHAND
+				if (exists && (se->sa_flags & SA_RESETHAND)) {
+					talloc_free(se);
+				}
+#endif
+				talloc_free(exists);
+				continue;
+			}
+#endif
+			se->handler(ev, se, i, count, NULL, se->private_data);
+#ifdef SA_RESETHAND
+			if (exists && (se->sa_flags & SA_RESETHAND)) {
+				talloc_free(se);
+			}
+#endif
+			talloc_free(exists);
+		}
+
+#ifdef SA_SIGINFO
+		if (clear_processed_siginfo) {
+			uint32_t j;
+			for (j=0;j<count;j++) {
+				uint32_t ofs = (counter.seen + j)
+					% TEVENT_SA_INFO_QUEUE_COUNT;
+				memset((void*)&sig_state->sig_info[i][ofs],
+					'\0',
+					sizeof(siginfo_t));
+			}
+		}
+#endif
+
+		TEVENT_SIG_SEEN(sig_state->signal_count[i], count);
+		TEVENT_SIG_SEEN(sig_state->got_signal, count);
+
+#ifdef SA_SIGINFO
+		if (TEVENT_SIG_PENDING(sig_state->sig_blocked[i])) {
+			/* We'd filled the queue, unblock the
+			   signal now the queue is empty again.
+			   Note we MUST do this after the
+			   TEVENT_SIG_SEEN(sig_state->signal_count[i], count)
+			   call to prevent a new signal running
+			   out of room in the sig_state->sig_info[i][]
+			   ring buffer. */
+			sigset_t set;
+			sigemptyset(&set);
+			sigaddset(&set, i);
+			TEVENT_SIG_SEEN(sig_state->sig_blocked[i],
+				 tevent_sig_count(sig_state->sig_blocked[i]));
+			sigprocmask(SIG_UNBLOCK, &set, NULL);
+		}
+#endif
+	}
+
+	return 1;
+}
+
+void tevent_cleanup_pending_signal_handlers(struct tevent_signal *se)
+{
+	struct tevent_common_signal_list *sl;
+	sl = talloc_get_type(se->additional_data,
+			     struct tevent_common_signal_list);
+
+	tevent_common_signal_list_destructor(sl);
+
+	if (sig_state->sig_handlers[se->signum] == NULL) {
+		if (sig_state->oldact[se->signum]) {
+			sigaction(se->signum, sig_state->oldact[se->signum], NULL);
+			sig_state->oldact[se->signum] = NULL;
+		}
+	}
+	return;
+}
diff --git a/ctdb/lib/tevent/tevent_standard.c b/ctdb/lib/tevent/tevent_standard.c
new file mode 100644
index 0000000000..258499484a
--- /dev/null
+++ b/ctdb/lib/tevent/tevent_standard.c
@@ -0,0 +1,233 @@
+/* 
+   Unix SMB/CIFS implementation.
+   main select loop and event handling
+   Copyright (C) Stefan Metzmacher      2013
+   Copyright (C) Jeremy Allison         2013
+
+     ** NOTE! The following LGPL license applies to the tevent
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+/*
+  This is SAMBA's default event loop code
+
+  - we try to use epoll if configure detected support for it
+    otherwise we use poll()
+  - if epoll is broken on the system or the kernel doesn't support it
+    at runtime we fallback to poll()
+*/
+
+#include "replace.h"
+#include "tevent.h"
+#include "tevent_util.h"
+#include "tevent_internal.h"
+
+struct std_event_glue {
+	const struct tevent_ops *epoll_ops;
+	const struct tevent_ops *poll_ops;
+	struct tevent_ops *glue_ops;
+	bool fallback_replay;
+};
+
+static int std_event_context_init(struct tevent_context *ev);
+
+static const struct tevent_ops std_event_ops = {
+	.context_init           = std_event_context_init,
+};
+
+/*
+  If this function gets called. epoll failed at runtime.
+  Move us to using poll instead. If we return false here,
+  caller should abort().
+*/
+static bool std_fallback_to_poll(struct tevent_context *ev, bool replay)
+{
+	void *glue_ptr = talloc_parent(ev->ops);
+	struct std_event_glue *glue =
+		talloc_get_type_abort(glue_ptr,
+		struct std_event_glue);
+	int ret;
+	struct tevent_fd *fde;
+	struct tevent_fd *fde_next;
+
+	glue->fallback_replay = replay;
+
+	/* First switch all the ops to poll. */
+	glue->epoll_ops = NULL;
+
+	/*
+	 * Set custom_ops the same as poll.
+	 */
+	*glue->glue_ops = *glue->poll_ops;
+	glue->glue_ops->context_init = std_event_context_init;
+
+	/* Next initialize the poll backend. */
+	ret = glue->poll_ops->context_init(ev);
+	if (ret != 0) {
+		return false;
+	}
+
+	/*
+	 * Now we have to change all the existing file descriptor
+	 * events from the epoll backend to the poll backend.
+	 */
+	for (fde = ev->fd_events; fde; fde = fde_next) {
+		/*
+		 * We must remove this fde off the ev->fd_events list.
+		 */
+		fde_next = fde->next;
+
+		/* Remove from the ev->fd_events list. */
+		DLIST_REMOVE(ev->fd_events, fde);
+
+		/* Re-add this event as a poll backend event. */
+		tevent_poll_event_add_fd_internal(ev, fde);
+	}
+
+	return true;
+}
+
+static int std_event_loop_once(struct tevent_context *ev, const char *location)
+{
+	void *glue_ptr = talloc_parent(ev->ops);
+	struct std_event_glue *glue =
+		talloc_get_type_abort(glue_ptr,
+		struct std_event_glue);
+	int ret;
+
+	ret = glue->epoll_ops->loop_once(ev, location);
+	if (glue->epoll_ops != NULL) {
+		/* No fallback */
+		return ret;
+	}
+
+	if (!glue->fallback_replay) {
+		/*
+		 * The problem happened while modifying an event.
+		 * An event handler was triggered in this case
+		 * and there is no need to call loop_once() again.
+		 */
+		return ret;
+	}
+
+	return glue->poll_ops->loop_once(ev, location);
+}
+
+static int std_event_loop_wait(struct tevent_context *ev, const char *location)
+{
+	void *glue_ptr = talloc_parent(ev->ops);
+	struct std_event_glue *glue =
+		talloc_get_type_abort(glue_ptr,
+		struct std_event_glue);
+	int ret;
+
+	ret = glue->epoll_ops->loop_wait(ev, location);
+	if (glue->epoll_ops != NULL) {
+		/* No fallback */
+		return ret;
+	}
+
+	return glue->poll_ops->loop_wait(ev, location);
+}
+/*
+  Initialize the epoll backend and allow it to call a
+  switch function if epoll fails at runtime.
+*/
+static int std_event_context_init(struct tevent_context *ev)
+{
+	struct std_event_glue *glue;
+	int ret;
+
+	/*
+	 * If this is the first initialization
+	 * we need to set up the allocated ops
+	 * pointers.
+	 */
+
+	if (ev->ops == &std_event_ops) {
+		glue = talloc_zero(ev, struct std_event_glue);
+		if (glue == NULL) {
+			return -1;
+		}
+
+		glue->epoll_ops = tevent_find_ops_byname("epoll");
+
+		glue->poll_ops = tevent_find_ops_byname("poll");
+		if (glue->poll_ops == NULL) {
+			return -1;
+		}
+
+		/*
+		 * Allocate space for our custom ops.
+		 * Allocate as a child of our epoll_ops pointer
+		 * so we can easily get to it using talloc_parent.
+		 */
+		glue->glue_ops = talloc_zero(glue, struct tevent_ops);
+		if (glue->glue_ops == NULL) {
+			talloc_free(glue);
+			return -1;
+		}
+
+		ev->ops = glue->glue_ops;
+	} else {
+		void *glue_ptr = talloc_parent(ev->ops);
+		glue = talloc_get_type_abort(glue_ptr, struct std_event_glue);
+	}
+
+	if (glue->epoll_ops != NULL) {
+		/*
+		 * Set custom_ops the same as epoll,
+		 * except re-init using std_event_context_init()
+		 * and use std_event_loop_once() to add the
+		 * ability to fallback to a poll backend on
+		 * epoll runtime error.
+		 */
+		*glue->glue_ops = *glue->epoll_ops;
+		glue->glue_ops->context_init = std_event_context_init;
+		glue->glue_ops->loop_once = std_event_loop_once;
+		glue->glue_ops->loop_wait = std_event_loop_wait;
+
+		ret = glue->epoll_ops->context_init(ev);
+		if (ret == -1) {
+			goto fallback;
+		}
+#ifdef HAVE_EPOLL
+		if (!tevent_epoll_set_panic_fallback(ev, std_fallback_to_poll)) {
+			TALLOC_FREE(ev->additional_data);
+			goto fallback;
+		}
+#endif
+
+		return ret;
+	}
+
+fallback:
+	glue->epoll_ops = NULL;
+
+	/*
+	 * Set custom_ops the same as poll.
+	 */
+	*glue->glue_ops = *glue->poll_ops;
+	glue->glue_ops->context_init = std_event_context_init;
+
+	return glue->poll_ops->context_init(ev);
+}
+
+_PRIVATE_ bool tevent_standard_init(void)
+{
+	return tevent_register_backend("standard", &std_event_ops);
+}
diff --git a/ctdb/lib/tevent/tevent_timed.c b/ctdb/lib/tevent/tevent_timed.c
new file mode 100644
index 0000000000..920d39fe06
--- /dev/null
+++ b/ctdb/lib/tevent/tevent_timed.c
@@ -0,0 +1,355 @@
+/* 
+   Unix SMB/CIFS implementation.
+
+   common events code for timed events
+
+   Copyright (C) Andrew Tridgell	2003-2006
+   Copyright (C) Stefan Metzmacher	2005-2009
+
+     ** NOTE! The following LGPL license applies to the tevent
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "replace.h"
+#include "system/time.h"
+#include "tevent.h"
+#include "tevent_internal.h"
+#include "tevent_util.h"
+
+/**
+  compare two timeval structures. 
+  Return -1 if tv1 < tv2
+  Return 0 if tv1 == tv2
+  Return 1 if tv1 > tv2
+*/
+int tevent_timeval_compare(const struct timeval *tv1, const struct timeval *tv2)
+{
+	if (tv1->tv_sec  > tv2->tv_sec)  return 1;
+	if (tv1->tv_sec  < tv2->tv_sec)  return -1;
+	if (tv1->tv_usec > tv2->tv_usec) return 1;
+	if (tv1->tv_usec < tv2->tv_usec) return -1;
+	return 0;
+}
+
+/**
+  return a zero timeval
+*/
+struct timeval tevent_timeval_zero(void)
+{
+	struct timeval tv;
+	tv.tv_sec = 0;
+	tv.tv_usec = 0;
+	return tv;
+}
+
+/**
+  return a timeval for the current time
+*/
+struct timeval tevent_timeval_current(void)
+{
+	struct timeval tv;
+	gettimeofday(&tv, NULL);
+	return tv;
+}
+
+/**
+  return a timeval struct with the given elements
+*/
+struct timeval tevent_timeval_set(uint32_t secs, uint32_t usecs)
+{
+	struct timeval tv;
+	tv.tv_sec = secs;
+	tv.tv_usec = usecs;
+	return tv;
+}
+
+/**
+  return the difference between two timevals as a timeval
+  if tv1 comes after tv2, then return a zero timeval
+  (this is *tv2 - *tv1)
+*/
+struct timeval tevent_timeval_until(const struct timeval *tv1,
+				    const struct timeval *tv2)
+{
+	struct timeval t;
+	if (tevent_timeval_compare(tv1, tv2) >= 0) {
+		return tevent_timeval_zero();
+	}
+	t.tv_sec = tv2->tv_sec - tv1->tv_sec;
+	if (tv1->tv_usec > tv2->tv_usec) {
+		t.tv_sec--;
+		t.tv_usec = 1000000 - (tv1->tv_usec - tv2->tv_usec);
+	} else {
+		t.tv_usec = tv2->tv_usec - tv1->tv_usec;
+	}
+	return t;
+}
+
+/**
+  return true if a timeval is zero
+*/
+bool tevent_timeval_is_zero(const struct timeval *tv)
+{
+	return tv->tv_sec == 0 && tv->tv_usec == 0;
+}
+
+struct timeval tevent_timeval_add(const struct timeval *tv, uint32_t secs,
+				  uint32_t usecs)
+{
+	struct timeval tv2 = *tv;
+	tv2.tv_sec += secs;
+	tv2.tv_usec += usecs;
+	tv2.tv_sec += tv2.tv_usec / 1000000;
+	tv2.tv_usec = tv2.tv_usec % 1000000;
+
+	return tv2;
+}
+
+/**
+  return a timeval in the future with a specified offset
+*/
+struct timeval tevent_timeval_current_ofs(uint32_t secs, uint32_t usecs)
+{
+	struct timeval tv = tevent_timeval_current();
+	return tevent_timeval_add(&tv, secs, usecs);
+}
+
+/*
+  destroy a timed event
+*/
+static int tevent_common_timed_destructor(struct tevent_timer *te)
+{
+	if (te->event_ctx == NULL) {
+		return 0;
+	}
+
+	tevent_debug(te->event_ctx, TEVENT_DEBUG_TRACE,
+		     "Destroying timer event %p \"%s\"\n",
+		     te, te->handler_name);
+
+	if (te->event_ctx->last_zero_timer == te) {
+		te->event_ctx->last_zero_timer = DLIST_PREV(te);
+	}
+	DLIST_REMOVE(te->event_ctx->timer_events, te);
+
+	return 0;
+}
+
+static int tevent_common_timed_deny_destructor(struct tevent_timer *te)
+{
+	return -1;
+}
+
+/*
+  add a timed event
+  return NULL on failure (memory allocation error)
+*/
+static struct tevent_timer *tevent_common_add_timer_internal(
+					struct tevent_context *ev,
+					TALLOC_CTX *mem_ctx,
+					struct timeval next_event,
+					tevent_timer_handler_t handler,
+					void *private_data,
+					const char *handler_name,
+					const char *location,
+					bool optimize_zero)
+{
+	struct tevent_timer *te, *prev_te, *cur_te;
+
+	te = talloc(mem_ctx?mem_ctx:ev, struct tevent_timer);
+	if (te == NULL) return NULL;
+
+	te->event_ctx		= ev;
+	te->next_event		= next_event;
+	te->handler		= handler;
+	te->private_data	= private_data;
+	te->handler_name	= handler_name;
+	te->location		= location;
+	te->additional_data	= NULL;
+
+	if (ev->timer_events == NULL) {
+		ev->last_zero_timer = NULL;
+	}
+
+	/* keep the list ordered */
+	prev_te = NULL;
+	if (optimize_zero && tevent_timeval_is_zero(&te->next_event)) {
+		/*
+		 * Some callers use zero tevent_timer
+		 * instead of tevent_immediate events.
+		 *
+		 * As these can happen very often,
+		 * we remember the last zero timer
+		 * in the list.
+		 */
+		prev_te = ev->last_zero_timer;
+		ev->last_zero_timer = te;
+	} else {
+		/*
+		 * we traverse the list from the tail
+		 * because it's much more likely that
+		 * timers are added at the end of the list
+		 */
+		for (cur_te = DLIST_TAIL(ev->timer_events);
+		     cur_te != NULL;
+		     cur_te = DLIST_PREV(cur_te))
+		{
+			int ret;
+
+			/*
+			 * if the new event comes before the current
+			 * we continue searching
+			 */
+			ret = tevent_timeval_compare(&te->next_event,
+						     &cur_te->next_event);
+			if (ret < 0) {
+				continue;
+			}
+
+			break;
+		}
+
+		prev_te = cur_te;
+	}
+
+	DLIST_ADD_AFTER(ev->timer_events, te, prev_te);
+
+	talloc_set_destructor(te, tevent_common_timed_destructor);
+
+	tevent_debug(ev, TEVENT_DEBUG_TRACE,
+		     "Added timed event \"%s\": %p\n",
+		     handler_name, te);
+	return te;
+}
+
+struct tevent_timer *tevent_common_add_timer(struct tevent_context *ev,
+					     TALLOC_CTX *mem_ctx,
+					     struct timeval next_event,
+					     tevent_timer_handler_t handler,
+					     void *private_data,
+					     const char *handler_name,
+					     const char *location)
+{
+	/*
+	 * do not use optimization, there are broken Samba
+	 * versions which use tevent_common_add_timer()
+	 * without using tevent_common_loop_timer_delay(),
+	 * it just uses DLIST_REMOVE(ev->timer_events, te)
+	 * and would leave ev->last_zero_timer behind.
+	 */
+	return tevent_common_add_timer_internal(ev, mem_ctx, next_event,
+						handler, private_data,
+						handler_name, location,
+						false);
+}
+
+struct tevent_timer *tevent_common_add_timer_v2(struct tevent_context *ev,
+						TALLOC_CTX *mem_ctx,
+					        struct timeval next_event,
+					        tevent_timer_handler_t handler,
+					        void *private_data,
+					        const char *handler_name,
+					        const char *location)
+{
+	/*
+	 * Here we turn on last_zero_timer optimization
+	 */
+	return tevent_common_add_timer_internal(ev, mem_ctx, next_event,
+						handler, private_data,
+						handler_name, location,
+						true);
+}
+
+/*
+  do a single event loop using the events defined in ev
+
+  return the delay until the next timed event,
+  or zero if a timed event was triggered
+*/
+struct timeval tevent_common_loop_timer_delay(struct tevent_context *ev)
+{
+	struct timeval current_time = tevent_timeval_zero();
+	struct tevent_timer *te = ev->timer_events;
+
+	if (!te) {
+		/* have a default tick time of 30 seconds. This guarantees
+		   that code that uses its own timeout checking will be
+		   able to proceed eventually */
+		return tevent_timeval_set(30, 0);
+	}
+
+	/*
+	 * work out the right timeout for the next timed event
+	 *
+	 * avoid the syscall to gettimeofday() if the timed event should
+	 * be triggered directly
+	 *
+	 * if there's a delay till the next timed event, we're done
+	 * with just returning the delay
+	 */
+	if (!tevent_timeval_is_zero(&te->next_event)) {
+		struct timeval delay;
+
+		current_time = tevent_timeval_current();
+
+		delay = tevent_timeval_until(&current_time, &te->next_event);
+		if (!tevent_timeval_is_zero(&delay)) {
+			return delay;
+		}
+	}
+
+	/*
+	 * ok, we have a timed event that we'll process ...
+	 */
+
+	/* deny the handler to free the event */
+	talloc_set_destructor(te, tevent_common_timed_deny_destructor);
+
+	/* We need to remove the timer from the list before calling the
+	 * handler because in a semi-async inner event loop called from the
+	 * handler we don't want to come across this event again -- vl */
+	if (ev->last_zero_timer == te) {
+		ev->last_zero_timer = DLIST_PREV(te);
+	}
+	DLIST_REMOVE(ev->timer_events, te);
+
+	tevent_debug(te->event_ctx, TEVENT_DEBUG_TRACE,
+		     "Running timer event %p \"%s\"\n",
+		     te, te->handler_name);
+
+	/*
+	 * If the timed event was registered for a zero current_time,
+	 * then we pass a zero timeval here too! To avoid the
+	 * overhead of gettimeofday() calls.
+	 *
+	 * otherwise we pass the current time
+	 */
+	te->handler(ev, te, current_time, te->private_data);
+
+	/* The destructor isn't necessary anymore, we've already removed the
+	 * event from the list. */
+	talloc_set_destructor(te, NULL);
+
+	tevent_debug(te->event_ctx, TEVENT_DEBUG_TRACE,
+		     "Ending timer event %p \"%s\"\n",
+		     te, te->handler_name);
+
+	talloc_free(te);
+
+	return tevent_timeval_zero();
+}
+
diff --git a/ctdb/lib/tevent/tevent_util.c b/ctdb/lib/tevent/tevent_util.c
new file mode 100644
index 0000000000..16af8f3b90
--- /dev/null
+++ b/ctdb/lib/tevent/tevent_util.c
@@ -0,0 +1,107 @@
+/*
+   Unix SMB/CIFS implementation.
+
+   Copyright (C) Andrew Tridgell 2005
+   Copyright (C) Jelmer Vernooij 2005
+
+     ** NOTE! The following LGPL license applies to the tevent
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "replace.h"
+#include "talloc.h"
+#include "tevent.h"
+#include "tevent_internal.h"
+#include "tevent_util.h"
+#include <fcntl.h>
+
+/**
+  return the number of elements in a string list
+*/
+size_t ev_str_list_length(const char **list)
+{
+	size_t ret;
+	for (ret=0;list && list[ret];ret++) /* noop */ ;
+	return ret;
+}
+
+/**
+  add an entry to a string list
+*/
+const char **ev_str_list_add(const char **list, const char *s)
+{
+	size_t len = ev_str_list_length(list);
+	const char **ret;
+
+	ret = talloc_realloc(NULL, list, const char *, len+2);
+	if (ret == NULL) return NULL;
+
+	ret[len] = talloc_strdup(ret, s);
+	if (ret[len] == NULL) return NULL;
+
+	ret[len+1] = NULL;
+
+	return ret;
+}
+
+
+/**
+ Set a fd into blocking/nonblocking mode. Uses POSIX O_NONBLOCK if available,
+ else
+  if SYSV use O_NDELAY
+  if BSD use FNDELAY
+**/
+
+int ev_set_blocking(int fd, bool set)
+{
+	int val;
+#ifdef O_NONBLOCK
+#define FLAG_TO_SET O_NONBLOCK
+#else
+#ifdef SYSV
+#define FLAG_TO_SET O_NDELAY
+#else /* BSD */
+#define FLAG_TO_SET FNDELAY
+#endif
+#endif
+
+	if((val = fcntl(fd, F_GETFL, 0)) == -1)
+		return -1;
+	if(set) /* Turn blocking on - ie. clear nonblock flag */
+		val &= ~FLAG_TO_SET;
+	else
+		val |= FLAG_TO_SET;
+	return fcntl( fd, F_SETFL, val);
+#undef FLAG_TO_SET
+}
+
+bool ev_set_close_on_exec(int fd)
+{
+#ifdef FD_CLOEXEC
+	int val;
+
+	val = fcntl(fd, F_GETFD, 0);
+	if (val >= 0) {
+		val |= FD_CLOEXEC;
+		val = fcntl(fd, F_SETFD, val);
+		if (val != -1) {
+			return true;
+		}
+	}
+#endif
+	return false;
+}
diff --git a/ctdb/lib/tevent/tevent_util.h b/ctdb/lib/tevent/tevent_util.h
new file mode 100644
index 0000000000..311be604a0
--- /dev/null
+++ b/ctdb/lib/tevent/tevent_util.h
@@ -0,0 +1,192 @@
+/*
+   Unix SMB/CIFS implementation.
+
+   Copyright (C) Andrew Tridgell 1998-2010
+   Copyright (C) Jelmer Vernooij 2005
+
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.
+*/
+
+/* To use these macros you must have a structure containing a next and
+   prev pointer */
+
+#ifndef _DLINKLIST_H
+#define _DLINKLIST_H
+
+/*
+  February 2010 - changed list format to have a prev pointer from the
+  list head. This makes DLIST_ADD_END() O(1) even though we only have
+  one list pointer.
+
+  The scheme is as follows:
+
+     1) with no entries in the list:
+          list_head == NULL
+
+     2) with 1 entry in the list:
+          list_head->next == NULL
+          list_head->prev == list_head
+
+     3) with 2 entries in the list:
+          list_head->next == element2
+          list_head->prev == element2
+	  element2->prev == list_head
+	  element2->next == NULL
+
+     4) with N entries in the list:
+          list_head->next == element2
+          list_head->prev == elementN
+	  elementN->prev == element{N-1}
+	  elementN->next == NULL
+
+  This allows us to find the tail of the list by using
+  list_head->prev, which means we can add to the end of the list in
+  O(1) time
+
+
+  Note that the 'type' arguments below are no longer needed, but
+  are kept for now to prevent an incompatible argument change
+ */
+
+
+/*
+   add an element at the front of a list
+*/
+#define DLIST_ADD(list, p) \
+do { \
+        if (!(list)) { \
+		(p)->prev = (list) = (p);  \
+		(p)->next = NULL; \
+	} else { \
+		(p)->prev = (list)->prev; \
+		(list)->prev = (p); \
+		(p)->next = (list); \
+		(list) = (p); \
+	} \
+} while (0)
+
+/*
+   remove an element from a list
+   Note that the element doesn't have to be in the list. If it
+   isn't then this is a no-op
+*/
+#define DLIST_REMOVE(list, p) \
+do { \
+	if ((p) == (list)) { \
+		if ((p)->next) (p)->next->prev = (p)->prev; \
+		(list) = (p)->next; \
+	} else if ((list) && (p) == (list)->prev) {	\
+		(p)->prev->next = NULL; \
+		(list)->prev = (p)->prev; \
+	} else { \
+		if ((p)->prev) (p)->prev->next = (p)->next; \
+		if ((p)->next) (p)->next->prev = (p)->prev; \
+	} \
+	if ((p) != (list)) (p)->next = (p)->prev = NULL;	\
+} while (0)
+
+/*
+   find the head of the list given any element in it.
+   Note that this costs O(N), so you should avoid this macro
+   if at all possible!
+*/
+#define DLIST_HEAD(p, result_head) \
+do { \
+       (result_head) = (p); \
+       while (DLIST_PREV(result_head)) (result_head) = (result_head)->prev; \
+} while(0)
+
+/* return the last element in the list */
+#define DLIST_TAIL(list) ((list)?(list)->prev:NULL)
+
+/* return the previous element in the list. */
+#define DLIST_PREV(p) (((p)->prev && (p)->prev->next != NULL)?(p)->prev:NULL)
+
+/* insert 'p' after the given element 'el' in a list. If el is NULL then
+   this is the same as a DLIST_ADD() */
+#define DLIST_ADD_AFTER(list, p, el) \
+do { \
+        if (!(list) || !(el)) { \
+		DLIST_ADD(list, p); \
+	} else { \
+		(p)->prev = (el);   \
+		(p)->next = (el)->next;		\
+		(el)->next = (p);		\
+		if ((p)->next) (p)->next->prev = (p);	\
+		if ((list)->prev == (el)) (list)->prev = (p); \
+	}\
+} while (0)
+
+
+/*
+   add to the end of a list.
+   Note that 'type' is ignored
+*/
+#define DLIST_ADD_END(list, p, type)			\
+do { \
+	if (!(list)) { \
+		DLIST_ADD(list, p); \
+	} else { \
+		DLIST_ADD_AFTER(list, p, (list)->prev); \
+	} \
+} while (0)
+
+/* promote an element to the from of a list */
+#define DLIST_PROMOTE(list, p) \
+do { \
+          DLIST_REMOVE(list, p); \
+          DLIST_ADD(list, p); \
+} while (0)
+
+/*
+   demote an element to the end of a list.
+   Note that 'type' is ignored
+*/
+#define DLIST_DEMOTE(list, p, type)			\
+do { \
+	DLIST_REMOVE(list, p); \
+	DLIST_ADD_END(list, p, NULL);		\
+} while (0)
+
+/*
+   concatenate two lists - putting all elements of the 2nd list at the
+   end of the first list.
+   Note that 'type' is ignored
+*/
+#define DLIST_CONCATENATE(list1, list2, type)	\
+do { \
+	if (!(list1)) { \
+		(list1) = (list2); \
+	} else { \
+		(list1)->prev->next = (list2); \
+		if (list2) { \
+			void *_tmplist = (void *)(list1)->prev; \
+			(list1)->prev = (list2)->prev; \
+			(list2)->prev = _tmplist; \
+		} \
+	} \
+} while (0)
+
+#endif /* _DLINKLIST_H */
+
+const char **ev_str_list_add(const char **list, const char *s);
+int ev_set_blocking(int fd, bool set);
+size_t ev_str_list_length(const char **list);
+bool ev_set_close_on_exec(int fd);
+
+/* Defined here so we can build against older talloc versions that don't
+ * have this define yet. */
+#ifndef TALLOC_FREE
+#define TALLOC_FREE(ctx) do { talloc_free(ctx); ctx=NULL; } while(0)
+#endif
diff --git a/ctdb/lib/tevent/tevent_wakeup.c b/ctdb/lib/tevent/tevent_wakeup.c
new file mode 100644
index 0000000000..82c3942ba8
--- /dev/null
+++ b/ctdb/lib/tevent/tevent_wakeup.c
@@ -0,0 +1,70 @@
+/*
+   Unix SMB/CIFS implementation.
+   Infrastructure for async requests
+   Copyright (C) Volker Lendecke 2008
+   Copyright (C) Stefan Metzmacher 2009
+
+     ** NOTE! The following LGPL license applies to the tevent
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 3 of the License, or (at your option) any later version.
+
+   This library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "replace.h"
+#include "tevent.h"
+#include "tevent_internal.h"
+#include "tevent_util.h"
+
+struct tevent_wakeup_state {
+	struct timeval wakeup_time;
+};
+
+struct tevent_req *tevent_wakeup_send(TALLOC_CTX *mem_ctx,
+				      struct tevent_context *ev,
+				      struct timeval wakeup_time)
+{
+	struct tevent_req *req;
+	struct tevent_wakeup_state *state;
+
+	req = tevent_req_create(mem_ctx, &state,
+				struct tevent_wakeup_state);
+	if (!req) {
+		return NULL;
+	}
+	state->wakeup_time = wakeup_time;
+
+	if (!tevent_req_set_endtime(req, ev, wakeup_time)) {
+		goto post;
+	}
+
+	return req;
+post:
+	return tevent_req_post(req, ev);
+}
+
+bool tevent_wakeup_recv(struct tevent_req *req)
+{
+	enum tevent_req_state state;
+	uint64_t error;
+
+	if (tevent_req_is_error(req, &state, &error)) {
+		if (state == TEVENT_REQ_TIMED_OUT) {
+			return true;
+		}
+	}
+
+	return false;
+}
+
diff --git a/ctdb/lib/tevent/wscript b/ctdb/lib/tevent/wscript
new file mode 100755
index 0000000000..02bddb80d8
--- /dev/null
+++ b/ctdb/lib/tevent/wscript
@@ -0,0 +1,138 @@
+#!/usr/bin/env python
+
+APPNAME = 'tevent'
+VERSION = '0.9.18'
+
+blddir = 'bin'
+
+import sys, os
+
+# find the buildtools directory
+srcdir = '.'
+while not os.path.exists(srcdir+'/buildtools') and len(srcdir.split('/')) < 5:
+    srcdir = '../' + srcdir
+sys.path.insert(0, srcdir + '/buildtools/wafsamba')
+
+import wafsamba, samba_dist, Options, Logs
+
+samba_dist.DIST_DIRS('lib/tevent:. lib/replace:lib/replace lib/talloc:lib/talloc buildtools:buildtools')
+
+def set_options(opt):
+    opt.BUILTIN_DEFAULT('replace')
+    opt.PRIVATE_EXTENSION_DEFAULT('tevent', noextension='tevent')
+    opt.RECURSE('lib/replace')
+    opt.RECURSE('lib/talloc')
+    if opt.IN_LAUNCH_DIR():
+        opt.add_option('--disable-python',
+                       help=("disable the pytevent module"),
+                       action="store_true", dest='disable_python', default=False)
+
+
+def configure(conf):
+    conf.RECURSE('lib/replace')
+    conf.RECURSE('lib/talloc')
+
+    conf.env.standalone_tevent = conf.IN_LAUNCH_DIR()
+
+    if not conf.env.standalone_tevent:
+        if conf.CHECK_BUNDLED_SYSTEM_PKG('tevent', minversion=VERSION,
+                                     onlyif='talloc', implied_deps='replace talloc'):
+            conf.define('USING_SYSTEM_TEVENT', 1)
+            if conf.CHECK_BUNDLED_SYSTEM_PYTHON('pytevent', 'tevent', minversion=VERSION):
+                conf.define('USING_SYSTEM_PYTEVENT', 1)
+
+    if conf.CHECK_FUNCS('epoll_create', headers='sys/epoll.h'):
+        conf.DEFINE('HAVE_EPOLL', 1)
+
+    tevent_num_signals = 64
+    v = conf.CHECK_VALUEOF('NSIG', headers='signal.h')
+    if v is not None:
+        tevent_num_signals = max(tevent_num_signals, v)
+    v = conf.CHECK_VALUEOF('_NSIG', headers='signal.h')
+    if v is not None:
+        tevent_num_signals = max(tevent_num_signals, v)
+    v = conf.CHECK_VALUEOF('SIGRTMAX', headers='signal.h')
+    if v is not None:
+        tevent_num_signals = max(tevent_num_signals, v)
+    v = conf.CHECK_VALUEOF('SIGRTMIN', headers='signal.h')
+    if v is not None:
+        tevent_num_signals = max(tevent_num_signals, v*2)
+
+    if not conf.CONFIG_SET('USING_SYSTEM_TEVENT'):
+        conf.DEFINE('TEVENT_NUM_SIGNALS', tevent_num_signals)
+
+    conf.env.disable_python = getattr(Options.options, 'disable_python', False)
+
+    if not conf.env.disable_python:
+        # also disable if we don't have the python libs installed
+        conf.find_program('python', var='PYTHON')
+        conf.check_tool('python')
+        conf.check_python_version((2,4,2))
+        conf.SAMBA_CHECK_PYTHON_HEADERS(mandatory=False)
+        if not conf.env.HAVE_PYTHON_H:
+            Logs.warn('Disabling pytevent as python devel libs not found')
+            conf.env.disable_python = True
+
+    conf.SAMBA_CONFIG_H()
+
+    conf.SAMBA_CHECK_UNDEFINED_SYMBOL_FLAGS()
+
+def build(bld):
+    bld.RECURSE('lib/replace')
+    bld.RECURSE('lib/talloc')
+
+    SRC = '''tevent.c tevent_debug.c tevent_fd.c tevent_immediate.c
+             tevent_queue.c tevent_req.c tevent_select.c
+             tevent_poll.c
+             tevent_signal.c tevent_standard.c tevent_timed.c tevent_util.c tevent_wakeup.c'''
+
+    if bld.CONFIG_SET('HAVE_EPOLL'):
+        SRC += ' tevent_epoll.c'
+
+    if bld.env.standalone_tevent:
+        bld.env.PKGCONFIGDIR = '${LIBDIR}/pkgconfig'
+        private_library = False
+    else:
+        private_library = True
+
+    if not bld.CONFIG_SET('USING_SYSTEM_TEVENT'):
+        bld.SAMBA_LIBRARY('tevent',
+                          SRC,
+                          deps='replace talloc',
+                          enabled= not bld.CONFIG_SET('USING_SYSTEM_TEVENT'),
+                          includes='.',
+                          abi_directory='ABI',
+                          abi_match='tevent_* _tevent_*',
+                          vnum=VERSION,
+                          public_headers='tevent.h',
+                          public_headers_install=not private_library,
+                          pc_files='tevent.pc',
+                          private_library=private_library)
+
+    if not bld.CONFIG_SET('USING_SYSTEM_PYTEVENT') and not bld.env.disable_python:
+        bld.SAMBA_PYTHON('pytevent',
+                         'pytevent.c',
+                         deps='tevent',
+                         realname='_tevent.so',
+                         cflags='-DPACKAGE_VERSION=\"%s\"' % VERSION)
+        # install out various python scripts for use by make test
+        bld.SAMBA_SCRIPT('tevent_python',
+                         pattern='tevent.py',
+                         installdir='python')
+
+        bld.INSTALL_WILDCARD('${PYTHONARCHDIR}', 'tevent.py', flat=False)
+
+
+def test(ctx):
+    '''test tevent'''
+    print("The tevent testsuite is part of smbtorture in samba4")
+
+
+def dist():
+    '''makes a tarball for distribution'''
+    samba_dist.dist()
+
+def reconfigure(ctx):
+    '''reconfigure if config scripts have changed'''
+    import samba_utils
+    samba_utils.reconfigure(ctx)
diff --git a/ctdb/lib/util/db_wrap.c b/ctdb/lib/util/db_wrap.c
new file mode 100644
index 0000000000..1b2bf7e228
--- /dev/null
+++ b/ctdb/lib/util/db_wrap.c
@@ -0,0 +1,103 @@
+/* 
+   Unix SMB/CIFS implementation.
+
+   database wrap functions
+
+   Copyright (C) Andrew Tridgell 2004
+   
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+/*
+  the stupidity of the unix fcntl locking design forces us to never
+  allow a database file to be opened twice in the same process. These
+  wrappers provide convenient access to a tdb or ldb, taking advantage
+  of talloc destructors to ensure that only a single open is done
+*/
+
+#include "includes.h"
+#include "lib/util/dlinklist.h"
+#include "tdb.h"
+#include "db_wrap.h"
+
+static struct tdb_wrap *tdb_list;
+
+
+
+/* destroy the last connection to a tdb */
+static int tdb_wrap_destructor(struct tdb_wrap *w)
+{
+	tdb_close(w->tdb);
+	DLIST_REMOVE(tdb_list, w);
+	return 0;
+}				 
+
+static void log_fn(struct tdb_context *tdb, enum tdb_debug_level level, const char *fmt, ...)
+{
+	if (level <= TDB_DEBUG_ERROR) {
+		va_list ap;
+		this_log_level = level;
+		char newfmt[strlen(tdb_name(tdb)) + 1 + strlen(fmt) + 1];
+		sprintf(newfmt, "%s:%s", tdb_name(tdb), fmt);
+		va_start(ap, fmt);
+		do_debug_v(newfmt, ap);
+		va_end(ap);
+	}
+}
+
+
+/*
+  wrapped connection to a tdb database
+  to close just talloc_free() the tdb_wrap pointer
+ */
+struct tdb_wrap *tdb_wrap_open(TALLOC_CTX *mem_ctx,
+			       const char *name, int hash_size, int tdb_flags,
+			       int open_flags, mode_t mode)
+{
+	struct tdb_wrap *w;
+	struct tdb_logging_context log_ctx;
+
+	log_ctx.log_fn = log_fn;
+	log_ctx.log_private = NULL;
+
+	for (w=tdb_list;w;w=w->next) {
+		if (strcmp(name, w->name) == 0) {
+			return talloc_reference(mem_ctx, w);
+		}
+	}
+
+	w = talloc(mem_ctx, struct tdb_wrap);
+	if (w == NULL) {
+		return NULL;
+	}
+
+	w->name = talloc_strdup(w, name);
+	if (w->name == NULL) {
+		talloc_free(w);
+		return NULL;
+	}
+
+	w->tdb = tdb_open_ex(name, hash_size, tdb_flags, 
+			     open_flags, mode, &log_ctx, NULL);
+	if (w->tdb == NULL) {
+		talloc_free(w);
+		return NULL;
+	}
+
+	talloc_set_destructor(w, tdb_wrap_destructor);
+
+	DLIST_ADD(tdb_list, w);
+
+	return w;
+}
diff --git a/ctdb/lib/util/db_wrap.h b/ctdb/lib/util/db_wrap.h
new file mode 100644
index 0000000000..5ae5fd10ae
--- /dev/null
+++ b/ctdb/lib/util/db_wrap.h
@@ -0,0 +1,36 @@
+/* 
+   Unix SMB/CIFS implementation.
+
+   database wrap headers
+
+   Copyright (C) Andrew Tridgell 2004
+   
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#ifndef _DB_WRAP_H
+#define _DB_WRAP_H
+
+struct tdb_wrap {
+	struct tdb_context *tdb;
+
+	const char *name;
+	struct tdb_wrap *next, *prev;
+};
+
+struct tdb_wrap *tdb_wrap_open(TALLOC_CTX *mem_ctx,
+			       const char *name, int hash_size, int tdb_flags,
+			       int open_flags, mode_t mode);
+
+#endif /* _DB_WRAP_H */
diff --git a/ctdb/lib/util/debug.c b/ctdb/lib/util/debug.c
new file mode 100644
index 0000000000..e9365d88f6
--- /dev/null
+++ b/ctdb/lib/util/debug.c
@@ -0,0 +1,136 @@
+/*
+   Unix SMB/CIFS implementation.
+   ctdb debug functions
+   Copyright (C) Volker Lendecke 2007
+
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "includes.h"
+#include "system/time.h"
+#include <unistd.h>
+#include <ctype.h>
+
+static void _do_debug_v(const char *format, va_list ap)
+{
+	struct timeval t;
+	char *s = NULL;
+	struct tm *tm;
+	char tbuf[100];
+	int ret;
+
+	ret = vasprintf(&s, format, ap);
+	if (ret == -1) {
+		fprintf(stderr, "vasprintf failed in _do_debug_v, cannot print debug message.\n");
+		fflush(stderr);
+		return;
+	}
+
+	t = timeval_current();
+	tm = localtime(&t.tv_sec);
+
+	strftime(tbuf,sizeof(tbuf)-1,"%Y/%m/%d %H:%M:%S", tm);
+
+	fprintf(stderr, "%s.%06u [%s%5u]: %s", tbuf, (unsigned)t.tv_usec,
+		debug_extra, (unsigned)getpid(), s);
+	fflush(stderr);
+	free(s);
+}
+
+/* default logging function */
+void (*do_debug_v)(const char *, va_list ap) = _do_debug_v;
+const char *debug_extra = "";
+
+void do_debug(const char *format, ...)
+{
+	va_list ap;
+
+	va_start(ap, format);
+	do_debug_v(format, ap);
+	va_end(ap);
+}
+
+
+static void _do_debug_add_v(const char *format, va_list ap)
+{
+	char *s = NULL;
+	int ret;
+
+	ret = vasprintf(&s, format, ap);
+	if (ret == -1) {
+		fprintf(stderr, "vasprintf failed in _do_debug_add_v, cannot print debug message.\n");
+		fflush(stderr);
+		return;
+	}
+
+	fprintf(stderr, "%s", s);
+	fflush(stderr);
+	free(s);
+}
+
+/* default logging function */
+void (*do_debug_add_v)(const char *, va_list ap) = _do_debug_add_v;
+
+void do_debug_add(const char *format, ...)
+{
+	va_list ap;
+
+	va_start(ap, format);
+	do_debug_add_v(format, ap);
+	va_end(ap);
+}
+
+static void print_asc(int level, const uint8_t *buf, size_t len)
+{
+	int i;
+	for (i=0;i<len;i++) {
+		DEBUGADD(level,("%c", isprint(buf[i])?buf[i]:'.'));
+	}
+}
+
+void dump_data(int level, const uint8_t *buf, size_t len)
+{
+	int i=0;
+
+	if (len<=0) return;
+
+	if (!DEBUGLVL(level)) return;
+
+	DEBUG(level, (__location__ " dump data of size %i:\n", (int)len));
+	DEBUGADD(level,("[%03X] ",i));
+	for (i=0;i<len;) {
+		DEBUGADD(level,("%02X ",(int)buf[i]));
+		i++;
+		if (i%8 == 0) DEBUGADD(level,(" "));
+		if (i%16 == 0) {
+			print_asc(level,&buf[i-16],8); DEBUGADD(level,(" "));
+			print_asc(level,&buf[i-8],8); DEBUGADD(level,("\n"));
+			if (i<len) DEBUGADD(level,("[%03X] ",i));
+		}
+	}
+	if (i%16) {
+		int n;
+		n = 16 - (i%16);
+		DEBUGADD(level,(" "));
+		if (n>8) DEBUGADD(level,(" "));
+		while (n--) DEBUGADD(level,("   "));
+		n = MIN(8,i%16);
+		print_asc(level,&buf[i-(i%16)],n); DEBUGADD(level,( " " ));
+		n = (i%16) - n;
+		if (n>0) print_asc(level,&buf[i-n],n);
+		DEBUGADD(level,("\n"));
+	}
+	DEBUG(level, (__location__ " dump data of size %i finished\n", (int)len));
+}
+
diff --git a/ctdb/lib/util/debug.h b/ctdb/lib/util/debug.h
new file mode 100644
index 0000000000..27490a3dbf
--- /dev/null
+++ b/ctdb/lib/util/debug.h
@@ -0,0 +1,27 @@
+/*
+   Unix SMB/CIFS implementation.
+   ctdb debug functions
+   Copyright (C) Volker Lendecke 2007
+
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+void (*do_debug_v)(const char *, va_list ap);
+const char *debug_extra;
+void (*do_debug_add_v)(const char *, va_list ap);
+void log_ringbuffer(const char *format, ...);
+void do_debug(const char *format, ...) PRINTF_ATTRIBUTE(1, 2);
+void do_debug_add(const char *format, ...) PRINTF_ATTRIBUTE(1, 2);
+void dump_data(int level, const uint8_t *buf1, size_t len);
+
diff --git a/ctdb/lib/util/dlinklist.h b/ctdb/lib/util/dlinklist.h
new file mode 100644
index 0000000000..6d525f9036
--- /dev/null
+++ b/ctdb/lib/util/dlinklist.h
@@ -0,0 +1,181 @@
+/* 
+   Unix SMB/CIFS implementation.
+   some simple double linked list macros
+
+   Copyright (C) Andrew Tridgell 1998-2010
+   
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.
+*/
+
+/* To use these macros you must have a structure containing a next and
+   prev pointer */
+
+#ifndef _DLINKLIST_H
+#define _DLINKLIST_H
+
+/*
+  February 2010 - changed list format to have a prev pointer from the
+  list head. This makes DLIST_ADD_END() O(1) even though we only have
+  one list pointer.
+
+  The scheme is as follows:
+
+     1) with no entries in the list:
+          list_head == NULL
+
+     2) with 1 entry in the list:
+          list_head->next == NULL
+          list_head->prev == list_head
+
+     3) with 2 entries in the list:
+          list_head->next == element2
+          list_head->prev == element2
+	  element2->prev == list_head
+	  element2->next == NULL
+
+     4) with N entries in the list:
+          list_head->next == element2
+          list_head->prev == elementN
+	  elementN->prev == element{N-1}
+	  elementN->next == NULL
+
+  This allows us to find the tail of the list by using
+  list_head->prev, which means we can add to the end of the list in
+  O(1) time
+
+
+  Note that the 'type' arguments below are no longer needed, but
+  are kept for now to prevent an incompatible argument change
+ */
+
+
+/*
+   add an element at the front of a list
+*/
+#define DLIST_ADD(list, p) \
+do { \
+        if (!(list)) { \
+		(p)->prev = (list) = (p);  \
+		(p)->next = NULL; \
+	} else { \
+		(p)->prev = (list)->prev; \
+		(list)->prev = (p); \
+		(p)->next = (list); \
+		(list) = (p); \
+	} \
+} while (0)
+
+/*
+   remove an element from a list
+   Note that the element doesn't have to be in the list. If it
+   isn't then this is a no-op
+*/
+#define DLIST_REMOVE(list, p) \
+do { \
+	if ((p) == (list)) { \
+		if ((p)->next) (p)->next->prev = (p)->prev; \
+		(list) = (p)->next; \
+	} else if ((list) && (p) == (list)->prev) {	\
+		(p)->prev->next = NULL; \
+		(list)->prev = (p)->prev; \
+	} else { \
+		if ((p)->prev) (p)->prev->next = (p)->next; \
+		if ((p)->next) (p)->next->prev = (p)->prev; \
+	} \
+	if ((p) != (list)) (p)->next = (p)->prev = NULL;	\
+} while (0)
+
+/*
+   find the head of the list given any element in it.
+   Note that this costs O(N), so you should avoid this macro
+   if at all possible!
+*/
+#define DLIST_HEAD(p, result_head) \
+do { \
+       (result_head) = (p); \
+       while (DLIST_PREV(result_head)) (result_head) = (result_head)->prev; \
+} while(0)
+
+/* return the last element in the list */
+#define DLIST_TAIL(list) ((list)?(list)->prev:NULL)
+
+/* return the previous element in the list. */
+#define DLIST_PREV(p) (((p)->prev && (p)->prev->next != NULL)?(p)->prev:NULL)
+
+/* insert 'p' after the given element 'el' in a list. If el is NULL then
+   this is the same as a DLIST_ADD() */
+#define DLIST_ADD_AFTER(list, p, el) \
+do { \
+        if (!(list) || !(el)) { \
+		DLIST_ADD(list, p); \
+	} else { \
+		(p)->prev = (el);   \
+		(p)->next = (el)->next;		\
+		(el)->next = (p);		\
+		if ((p)->next) (p)->next->prev = (p);	\
+		if ((list)->prev == (el)) (list)->prev = (p); \
+	}\
+} while (0)
+
+
+/*
+   add to the end of a list.
+   Note that 'type' is ignored
+*/
+#define DLIST_ADD_END(list, p, type)			\
+do { \
+	if (!(list)) { \
+		DLIST_ADD(list, p); \
+	} else { \
+		DLIST_ADD_AFTER(list, p, (list)->prev); \
+	} \
+} while (0)
+
+/* promote an element to the from of a list */
+#define DLIST_PROMOTE(list, p) \
+do { \
+          DLIST_REMOVE(list, p); \
+          DLIST_ADD(list, p); \
+} while (0)
+
+/*
+   demote an element to the end of a list.
+   Note that 'type' is ignored
+*/
+#define DLIST_DEMOTE(list, p, type)			\
+do { \
+	DLIST_REMOVE(list, p); \
+	DLIST_ADD_END(list, p, NULL);		\
+} while (0)
+
+/*
+   concatenate two lists - putting all elements of the 2nd list at the
+   end of the first list.
+   Note that 'type' is ignored
+*/
+#define DLIST_CONCATENATE(list1, list2, type)	\
+do { \
+	if (!(list1)) { \
+		(list1) = (list2); \
+	} else { \
+		(list1)->prev->next = (list2); \
+		if (list2) { \
+			void *_tmplist = (void *)(list1)->prev; \
+			(list1)->prev = (list2)->prev; \
+			(list2)->prev = _tmplist; \
+		} \
+	} \
+} while (0)
+
+#endif /* _DLINKLIST_H */
diff --git a/ctdb/lib/util/fault.c b/ctdb/lib/util/fault.c
new file mode 100644
index 0000000000..3dddd0ecb2
--- /dev/null
+++ b/ctdb/lib/util/fault.c
@@ -0,0 +1,238 @@
+/* 
+   Unix SMB/CIFS implementation.
+   Critical Fault handling
+   Copyright (C) Andrew Tridgell 1992-1998
+   
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "includes.h"
+#include "system/wait.h"
+#include "system/filesys.h"
+
+/**
+ * @file
+ * @brief Fault handling
+ */
+
+/* the registered fault handler */
+static struct {
+	const char *name;
+	void (*fault_handler)(int sig);
+} fault_handlers;
+
+static const char *progname;
+
+#ifdef HAVE_BACKTRACE
+#include <execinfo.h>
+#elif HAVE_LIBEXC_H
+#include <libexc.h>
+#endif
+
+/**
+ * Write backtrace to debug log
+ */
+_PUBLIC_ void call_backtrace(void)
+{
+#ifdef HAVE_BACKTRACE
+#ifndef BACKTRACE_STACK_SIZE
+#define BACKTRACE_STACK_SIZE 64
+#endif
+	void *backtrace_stack[BACKTRACE_STACK_SIZE];
+	size_t backtrace_size;
+	char **backtrace_strings;
+
+	/* get the backtrace (stack frames) */
+	backtrace_size = backtrace(backtrace_stack,BACKTRACE_STACK_SIZE);
+	backtrace_strings = backtrace_symbols(backtrace_stack, backtrace_size);
+
+	DEBUG(0, ("BACKTRACE: %lu stack frames:\n", 
+		  (unsigned long)backtrace_size));
+	
+	if (backtrace_strings) {
+		int i;
+
+		for (i = 0; i < backtrace_size; i++)
+			DEBUGADD(0, (" #%u %s\n", i, backtrace_strings[i]));
+
+		/* Leak the backtrace_strings, rather than risk what free() might do */
+	}
+
+#elif HAVE_LIBEXC
+
+#define NAMESIZE 32 /* Arbitrary */
+#ifndef BACKTRACE_STACK_SIZE
+#define BACKTRACE_STACK_SIZE 64
+#endif
+
+	/* The IRIX libexc library provides an API for unwinding the stack. See
+	 * libexc(3) for details. Apparantly trace_back_stack leaks memory, but
+	 * since we are about to abort anyway, it hardly matters.
+	 *
+	 * Note that if we paniced due to a SIGSEGV or SIGBUS (or similar) this
+	 * will fail with a nasty message upon failing to open the /proc entry.
+	 */
+	{
+		uint64_t	addrs[BACKTRACE_STACK_SIZE];
+		char *      	names[BACKTRACE_STACK_SIZE];
+		char		namebuf[BACKTRACE_STACK_SIZE * NAMESIZE];
+
+		int		i;
+		int		levels;
+
+		ZERO_ARRAY(addrs);
+		ZERO_ARRAY(names);
+		ZERO_ARRAY(namebuf);
+
+		for (i = 0; i < BACKTRACE_STACK_SIZE; i++) {
+			names[i] = namebuf + (i * NAMESIZE);
+		}
+
+		levels = trace_back_stack(0, addrs, names,
+				BACKTRACE_STACK_SIZE, NAMESIZE);
+
+		DEBUG(0, ("BACKTRACE: %d stack frames:\n", levels));
+		for (i = 0; i < levels; i++) {
+			DEBUGADD(0, (" #%d 0x%llx %s\n", i, addrs[i], names[i]));
+		}
+     }
+#undef NAMESIZE
+#else
+	DEBUG(0, ("call_backtrace: not implemented\n"));
+#endif
+}
+
+_PUBLIC_ const char *panic_action = NULL;
+_PUBLIC_ void (*pre_panic_action_hook)(void) = NULL;
+_PUBLIC_ void (*post_panic_action_hook)(void) = NULL;
+
+/**
+ Something really nasty happened - panic !
+**/
+_PUBLIC_ void smb_panic(const char *why)
+{
+	int result;
+
+	if (panic_action && *panic_action) {
+		char pidstr[20];
+		char cmdstring[200];
+		strlcpy(cmdstring, panic_action, sizeof(cmdstring));
+		snprintf(pidstr, sizeof(pidstr), "%u", getpid());
+		all_string_sub(cmdstring, "%PID%", pidstr, sizeof(cmdstring));
+		if (progname) {
+			all_string_sub(cmdstring, "%PROG%", progname, sizeof(cmdstring));
+		}
+		DEBUG(0, ("smb_panic(): calling panic action [%s]\n", cmdstring));
+
+		if (pre_panic_action_hook) {
+			pre_panic_action_hook();
+		}
+
+		result = system(cmdstring);
+
+		if (post_panic_action_hook) {
+			post_panic_action_hook();
+		}
+
+		if (result == -1)
+			DEBUG(0, ("smb_panic(): fork failed in panic action: %s\n",
+				  strerror(errno)));
+		else
+			DEBUG(0, ("smb_panic(): action returned status %d\n",
+				  WEXITSTATUS(result)));
+	}
+	DEBUG(0,("PANIC: %s\n", why));
+
+	call_backtrace();
+
+#ifdef SIGABRT
+	CatchSignal(SIGABRT, SIG_DFL);
+#endif
+	abort();
+}
+
+/**
+report a fault
+**/
+_NORETURN_ static void fault_report(int sig)
+{
+	static int counter;
+	
+	if (counter) _exit(1);
+
+	DEBUG(0,("=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=\n"));
+	DEBUG(0,("INTERNAL ERROR: Signal %d in %s pid %d",sig, progname, (int)getpid()));
+	DEBUG(0,("\nPlease read the file BUGS.txt in the distribution\n"));
+	DEBUG(0,("=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=\n"));
+
+	smb_panic("internal error");
+
+	exit(1);
+}
+
+/**
+catch serious errors
+**/
+_NORETURN_ static void sig_fault(int sig)
+{
+	if (fault_handlers.fault_handler) {
+		/* we have a fault handler, call it. It may not return. */
+		fault_handlers.fault_handler(sig);
+	}
+	/* If it returns or doesn't exist, use regular reporter */
+	fault_report(sig);
+}
+
+/**
+setup our fault handlers
+**/
+_PUBLIC_ void fault_setup(const char *pname)
+{
+	if (progname == NULL) {
+		progname = pname;
+	}
+#ifdef SIGSEGV
+	CatchSignal(SIGSEGV, sig_fault);
+#endif
+#ifdef SIGBUS
+	CatchSignal(SIGBUS, sig_fault);
+#endif
+#ifdef SIGABRT
+	CatchSignal(SIGABRT, sig_fault);
+#endif
+#ifdef SIGFPE
+	CatchSignal(SIGFPE, sig_fault);
+#endif
+}
+
+/**
+  register a fault handler. 
+  Should only be called once in the execution of smbd.
+*/
+_PUBLIC_ bool register_fault_handler(const char *name, 
+				     void (*fault_handler)(int sig))
+{
+	if (fault_handlers.name != NULL) {
+		/* it's already registered! */
+		DEBUG(2,("fault handler '%s' already registered - failed '%s'\n", 
+			 fault_handlers.name, name));
+		return false;
+	}
+
+	fault_handlers.name = name;
+	fault_handlers.fault_handler = fault_handler;
+
+	DEBUG(2,("fault handler '%s' registered\n", name));
+	return true;
+}
diff --git a/ctdb/lib/util/fault.m4 b/ctdb/lib/util/fault.m4
new file mode 100644
index 0000000000..da077af31d
--- /dev/null
+++ b/ctdb/lib/util/fault.m4
@@ -0,0 +1,15 @@
+AC_CHECK_HEADERS(execinfo.h)
+AC_SEARCH_LIBS_EXT(backtrace, [execinfo], EXECINFO_LIBS)
+AC_CHECK_FUNC_EXT(backtrace, $EXECINFO_LIBS)
+
+
+if test x"$ac_cv_header_execinfo_h" = x"yes" -a x"$ac_cv_func_ext_backtrace" = x"yes";then
+	SMB_ENABLE(EXECINFO, YES)
+	EXECINFO_CFLAGS="$CFLAGS"
+	EXECINFO_CPPFLAGS="$CPPFLAGS"
+	EXECINFO_LDFLAGS="$LDFLAGS"
+else
+	SMB_ENABLE(EXECINFO,NO)
+fi
+
+SMB_EXT_LIB(EXECINFO, [${EXECINFO_LIBS}], [${EXECINFO_CFLAGS}], [${EXECINFO_CPPFLAGS}], [${EXECINFO_LDFLAGS}])
diff --git a/ctdb/lib/util/idtree.c b/ctdb/lib/util/idtree.c
new file mode 100644
index 0000000000..09dc237f83
--- /dev/null
+++ b/ctdb/lib/util/idtree.c
@@ -0,0 +1,387 @@
+/* 
+   Unix SMB/CIFS implementation.
+
+   very efficient functions to manage mapping a id (such as a fnum) to
+   a pointer. This is used for fnum and search id allocation.
+
+   Copyright (C) Andrew Tridgell 2004
+
+   This code is derived from lib/idr.c in the 2.6 Linux kernel, which was 
+   written by Jim Houston jim.houston@ccur.com, and is
+   Copyright (C) 2002 by Concurrent Computer Corporation
+    
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 2 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+/*
+  see the section marked "public interface" below for documentation
+*/
+
+/**
+ * @file
+ */
+
+#include "includes.h"
+
+#define IDR_BITS 5
+#define IDR_FULL 0xfffffffful
+#if 0 /* unused */
+#define TOP_LEVEL_FULL (IDR_FULL >> 30)
+#endif
+#define IDR_SIZE (1 << IDR_BITS)
+#define IDR_MASK ((1 << IDR_BITS)-1)
+#define MAX_ID_SHIFT (sizeof(int)*8 - 1)
+#define MAX_ID_BIT (1U << MAX_ID_SHIFT)
+#define MAX_ID_MASK (MAX_ID_BIT - 1)
+#define MAX_LEVEL (MAX_ID_SHIFT + IDR_BITS - 1) / IDR_BITS
+#define IDR_FREE_MAX MAX_LEVEL + MAX_LEVEL
+
+#define set_bit(bit, v) (v) |= (1<<(bit))
+#define clear_bit(bit, v) (v) &= ~(1<<(bit))
+#define test_bit(bit, v) ((v) & (1<<(bit)))
+				   
+struct idr_layer {
+	uint32_t		 bitmap;
+	struct idr_layer	*ary[IDR_SIZE];
+	int			 count;
+};
+
+struct idr_context {
+	struct idr_layer *top;
+	struct idr_layer *id_free;
+	int		  layers;
+	int		  id_free_cnt;
+};
+
+static struct idr_layer *alloc_layer(struct idr_context *idp)
+{
+	struct idr_layer *p;
+
+	if (!(p = idp->id_free))
+		return NULL;
+	idp->id_free = p->ary[0];
+	idp->id_free_cnt--;
+	p->ary[0] = NULL;
+	return p;
+}
+
+static int find_next_bit(uint32_t bm, int maxid, int n)
+{
+	while (n<maxid && !test_bit(n, bm)) n++;
+	return n;
+}
+
+static void free_layer(struct idr_context *idp, struct idr_layer *p)
+{
+	p->ary[0] = idp->id_free;
+	idp->id_free = p;
+	idp->id_free_cnt++;
+}
+
+static int idr_pre_get(struct idr_context *idp)
+{
+	while (idp->id_free_cnt < IDR_FREE_MAX) {
+		struct idr_layer *new = talloc_zero(idp, struct idr_layer);
+		if(new == NULL)
+			return (0);
+		free_layer(idp, new);
+	}
+	return 1;
+}
+
+static int sub_alloc(struct idr_context *idp, void *ptr, int *starting_id)
+{
+	int n, m, sh;
+	struct idr_layer *p, *new;
+	struct idr_layer *pa[MAX_LEVEL+1];
+	unsigned int l, id, oid;
+	uint32_t bm;
+
+	memset(pa, 0, sizeof(pa));
+
+	id = *starting_id;
+restart:
+	p = idp->top;
+	l = idp->layers;
+	pa[l--] = NULL;
+	while (1) {
+		/*
+		 * We run around this while until we reach the leaf node...
+		 */
+		n = (id >> (IDR_BITS*l)) & IDR_MASK;
+		bm = ~p->bitmap;
+		m = find_next_bit(bm, IDR_SIZE, n);
+		if (m == IDR_SIZE) {
+			/* no space available go back to previous layer. */
+			l++;
+			oid = id;
+			id = (id | ((1 << (IDR_BITS*l))-1)) + 1;
+
+			/* if already at the top layer, we need to grow */
+			if (!(p = pa[l])) {
+				*starting_id = id;
+				return -2;
+			}
+
+			/* If we need to go up one layer, continue the
+			 * loop; otherwise, restart from the top.
+			 */
+			sh = IDR_BITS * (l + 1);
+			if (oid >> sh == id >> sh)
+				continue;
+			else
+				goto restart;
+		}
+		if (m != n) {
+			sh = IDR_BITS*l;
+			id = ((id >> sh) ^ n ^ m) << sh;
+		}
+		if ((id >= MAX_ID_BIT) || (id < 0))
+			return -1;
+		if (l == 0)
+			break;
+		/*
+		 * Create the layer below if it is missing.
+		 */
+		if (!p->ary[m]) {
+			if (!(new = alloc_layer(idp)))
+				return -1;
+			p->ary[m] = new;
+			p->count++;
+		}
+		pa[l--] = p;
+		p = p->ary[m];
+	}
+	/*
+	 * We have reached the leaf node, plant the
+	 * users pointer and return the raw id.
+	 */
+	p->ary[m] = (struct idr_layer *)ptr;
+	set_bit(m, p->bitmap);
+	p->count++;
+	/*
+	 * If this layer is full mark the bit in the layer above
+	 * to show that this part of the radix tree is full.
+	 * This may complete the layer above and require walking
+	 * up the radix tree.
+	 */
+	n = id;
+	while (p->bitmap == IDR_FULL) {
+		if (!(p = pa[++l]))
+			break;
+		n = n >> IDR_BITS;
+		set_bit((n & IDR_MASK), p->bitmap);
+	}
+	return(id);
+}
+
+static int idr_get_new_above_int(struct idr_context *idp, void *ptr, int starting_id)
+{
+	struct idr_layer *p, *new;
+	int layers, v, id;
+
+	idr_pre_get(idp);
+	
+	id = starting_id;
+build_up:
+	p = idp->top;
+	layers = idp->layers;
+	if (!p) {
+		if (!(p = alloc_layer(idp)))
+			return -1;
+		layers = 1;
+	}
+	/*
+	 * Add a new layer to the top of the tree if the requested
+	 * id is larger than the currently allocated space.
+	 */
+	while ((layers < MAX_LEVEL) && (id >= (1 << (layers*IDR_BITS)))) {
+		layers++;
+		if (!p->count)
+			continue;
+		if (!(new = alloc_layer(idp))) {
+			/*
+			 * The allocation failed.  If we built part of
+			 * the structure tear it down.
+			 */
+			for (new = p; p && p != idp->top; new = p) {
+				p = p->ary[0];
+				new->ary[0] = NULL;
+				new->bitmap = new->count = 0;
+				free_layer(idp, new);
+			}
+			return -1;
+		}
+		new->ary[0] = p;
+		new->count = 1;
+		if (p->bitmap == IDR_FULL)
+			set_bit(0, new->bitmap);
+		p = new;
+	}
+	idp->top = p;
+	idp->layers = layers;
+	v = sub_alloc(idp, ptr, &id);
+	if (v == -2)
+		goto build_up;
+	return(v);
+}
+
+static int sub_remove(struct idr_context *idp, int shift, int id)
+{
+	struct idr_layer *p = idp->top;
+	struct idr_layer **pa[1+MAX_LEVEL];
+	struct idr_layer ***paa = &pa[0];
+	int n;
+
+	*paa = NULL;
+	*++paa = &idp->top;
+
+	while ((shift > 0) && p) {
+		n = (id >> shift) & IDR_MASK;
+		clear_bit(n, p->bitmap);
+		*++paa = &p->ary[n];
+		p = p->ary[n];
+		shift -= IDR_BITS;
+	}
+	n = id & IDR_MASK;
+	if (p != NULL && test_bit(n, p->bitmap)) {
+		clear_bit(n, p->bitmap);
+		p->ary[n] = NULL;
+		while(*paa && ! --((**paa)->count)){
+			free_layer(idp, **paa);
+			**paa-- = NULL;
+		}
+		if ( ! *paa )
+			idp->layers = 0;
+		return 0;
+	}
+	return -1;
+}
+
+static void *_idr_find(struct idr_context *idp, int id)
+{
+	int n;
+	struct idr_layer *p;
+
+	n = idp->layers * IDR_BITS;
+	p = idp->top;
+	/*
+	 * This tests to see if bits outside the current tree are
+	 * present.  If so, tain't one of ours!
+	 */
+	if (n + IDR_BITS < 31 &&
+	    ((id & ~(~0 << MAX_ID_SHIFT)) >> (n + IDR_BITS))) {
+	     return NULL;
+	}
+
+	/* Mask off upper bits we don't use for the search. */
+	id &= MAX_ID_MASK;
+
+	while (n >= IDR_BITS && p) {
+		n -= IDR_BITS;
+		p = p->ary[(id >> n) & IDR_MASK];
+	}
+	return((void *)p);
+}
+
+static int _idr_remove(struct idr_context *idp, int id)
+{
+	struct idr_layer *p;
+
+	/* Mask off upper bits we don't use for the search. */
+	id &= MAX_ID_MASK;
+
+	if (sub_remove(idp, (idp->layers - 1) * IDR_BITS, id) == -1) {
+		return -1;
+	}
+
+	if ( idp->top && idp->top->count == 1 && 
+	     (idp->layers > 1) &&
+	     idp->top->ary[0]) {
+		/* We can drop a layer */
+		p = idp->top->ary[0];
+		idp->top->bitmap = idp->top->count = 0;
+		free_layer(idp, idp->top);
+		idp->top = p;
+		--idp->layers;
+	}
+	while (idp->id_free_cnt >= IDR_FREE_MAX) {
+		p = alloc_layer(idp);
+		talloc_free(p);
+	}
+	return 0;
+}
+
+/************************************************************************
+  this is the public interface
+**************************************************************************/
+
+/**
+  initialise a idr tree. The context return value must be passed to
+  all subsequent idr calls. To destroy the idr tree use talloc_free()
+  on this context
+ */
+_PUBLIC_ struct idr_context *idr_init(TALLOC_CTX *mem_ctx)
+{
+	return talloc_zero(mem_ctx, struct idr_context);
+}
+
+/**
+  allocate the next available id, and assign 'ptr' into its slot.
+  you can retrieve later this pointer using idr_find()
+*/
+_PUBLIC_ int idr_get_new(struct idr_context *idp, void *ptr, int limit)
+{
+	int ret = idr_get_new_above_int(idp, ptr, 0);
+	if (ret > limit) {
+		idr_remove(idp, ret);
+		return -1;
+	}
+	return ret;
+}
+
+/**
+   allocate a new id, giving the first available value greater than or
+   equal to the given starting id
+*/
+_PUBLIC_ int idr_get_new_above(struct idr_context *idp, void *ptr, int starting_id, int limit)
+{
+	int ret = idr_get_new_above_int(idp, ptr, starting_id);
+	if (ret > limit) {
+		idr_remove(idp, ret);
+		return -1;
+	}
+	return ret;
+}
+
+/**
+  find a pointer value previously set with idr_get_new given an id
+*/
+_PUBLIC_ void *idr_find(struct idr_context *idp, int id)
+{
+	return _idr_find(idp, id);
+}
+
+/**
+  remove an id from the idr tree
+*/
+_PUBLIC_ int idr_remove(struct idr_context *idp, int id)
+{
+	int ret;
+	ret = _idr_remove((struct idr_context *)idp, id);
+	if (ret != 0) {
+		DEBUG(0,("WARNING: attempt to remove unset id %d in idtree\n", id));
+	}
+	return ret;
+}
diff --git a/ctdb/lib/util/signal.c b/ctdb/lib/util/signal.c
new file mode 100644
index 0000000000..ead947eb5e
--- /dev/null
+++ b/ctdb/lib/util/signal.c
@@ -0,0 +1,144 @@
+/* 
+   Unix SMB/CIFS implementation.
+   signal handling functions
+
+   Copyright (C) Andrew Tridgell 1998
+   
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "includes.h"
+#include "system/wait.h"
+
+/**
+ * @file
+ * @brief Signal handling
+ */
+
+/****************************************************************************
+ Catch child exits and reap the child zombie status.
+****************************************************************************/
+
+static void sig_cld(int signum)
+{
+	while (waitpid((pid_t)-1,(int *)NULL, WNOHANG) > 0)
+		;
+
+	/*
+	 * Turns out it's *really* important not to
+	 * restore the signal handler here if we have real POSIX
+	 * signal handling. If we do, then we get the signal re-delivered
+	 * immediately - hey presto - instant loop ! JRA.
+	 */
+
+#if !defined(HAVE_SIGACTION)
+	CatchSignal(SIGCLD, sig_cld);
+#endif
+}
+
+/****************************************************************************
+catch child exits - leave status;
+****************************************************************************/
+
+static void sig_cld_leave_status(int signum)
+{
+	/*
+	 * Turns out it's *really* important not to
+	 * restore the signal handler here if we have real POSIX
+	 * signal handling. If we do, then we get the signal re-delivered
+	 * immediately - hey presto - instant loop ! JRA.
+	 */
+
+#if !defined(HAVE_SIGACTION)
+	CatchSignal(SIGCLD, sig_cld_leave_status);
+#endif
+}
+
+/**
+ Block sigs.
+**/
+
+void BlockSignals(bool block, int signum)
+{
+#ifdef HAVE_SIGPROCMASK
+	sigset_t set;
+	sigemptyset(&set);
+	sigaddset(&set,signum);
+	sigprocmask(block?SIG_BLOCK:SIG_UNBLOCK,&set,NULL);
+#elif defined(HAVE_SIGBLOCK)
+	if (block) {
+		sigblock(sigmask(signum));
+	} else {
+		sigsetmask(siggetmask() & ~sigmask(signum));
+	}
+#else
+	/* yikes! This platform can't block signals? */
+	static int done;
+	if (!done) {
+		DEBUG(0,("WARNING: No signal blocking available\n"));
+		done=1;
+	}
+#endif
+}
+
+/**
+ Catch a signal. This should implement the following semantics:
+
+ 1) The handler remains installed after being called.
+ 2) The signal should be blocked during handler execution.
+**/
+
+void (*CatchSignal(int signum,void (*handler)(int )))(int)
+{
+#ifdef HAVE_SIGACTION
+	struct sigaction act;
+	struct sigaction oldact;
+
+	ZERO_STRUCT(act);
+
+	act.sa_handler = handler;
+#ifdef SA_RESTART
+	/*
+	 * We *want* SIGALRM to interrupt a system call.
+	 */
+	if(signum != SIGALRM)
+		act.sa_flags = SA_RESTART;
+#endif
+	sigemptyset(&act.sa_mask);
+	sigaddset(&act.sa_mask,signum);
+	sigaction(signum,&act,&oldact);
+	return oldact.sa_handler;
+#else /* !HAVE_SIGACTION */
+	/* FIXME: need to handle sigvec and systems with broken signal() */
+	return signal(signum, handler);
+#endif
+}
+
+/**
+ Ignore SIGCLD via whatever means is necessary for this OS.
+**/
+
+void CatchChild(void)
+{
+	CatchSignal(SIGCLD, sig_cld);
+}
+
+/**
+ Catch SIGCLD but leave the child around so it's status can be reaped.
+**/
+
+void CatchChildLeaveStatus(void)
+{
+	CatchSignal(SIGCLD, sig_cld_leave_status);
+}
diff --git a/ctdb/lib/util/signal.m4 b/ctdb/lib/util/signal.m4
new file mode 100644
index 0000000000..c6d7f72f68
--- /dev/null
+++ b/ctdb/lib/util/signal.m4
@@ -0,0 +1 @@
+AC_CHECK_FUNCS(sigprocmask sigblock sigaction)
diff --git a/ctdb/lib/util/strlist.c b/ctdb/lib/util/strlist.c
new file mode 100644
index 0000000000..48a6e45ace
--- /dev/null
+++ b/ctdb/lib/util/strlist.c
@@ -0,0 +1,52 @@
+/* 
+   Unix SMB/CIFS implementation.
+   
+   Copyright (C) Andrew Tridgell 2005
+   Copyright (C) Jelmer Vernooij 2005
+   
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "includes.h"
+#include "system/locale.h"
+
+/**
+  return the number of elements in a string list
+*/
+_PUBLIC_ size_t str_list_length(const char **list)
+{
+	size_t ret;
+	for (ret=0;list && list[ret];ret++) /* noop */ ;
+	return ret;
+}
+
+
+/**
+  add an entry to a string list
+*/
+_PUBLIC_ const char **str_list_add(const char **list, const char *s)
+{
+	size_t len = str_list_length(list);
+	const char **ret;
+
+	ret = talloc_realloc(NULL, list, const char *, len+2);
+	if (ret == NULL) return NULL;
+
+	ret[len] = talloc_strdup(ret, s);
+	if (ret[len] == NULL) return NULL;
+
+	ret[len+1] = NULL;
+
+	return ret;
+}
diff --git a/ctdb/lib/util/substitute.c b/ctdb/lib/util/substitute.c
new file mode 100644
index 0000000000..32945a7213
--- /dev/null
+++ b/ctdb/lib/util/substitute.c
@@ -0,0 +1,167 @@
+/* 
+   Unix SMB/CIFS implementation.
+   Samba utility functions
+   
+   Copyright (C) Andrew Tridgell 1992-2001
+   Copyright (C) Simo Sorce      2001-2002
+   Copyright (C) Martin Pool     2003
+   Copyright (C) James Peach	 2005
+   
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "includes.h"
+
+/**
+ * @file
+ * @brief Substitute utilities.
+ **/
+
+/**
+ Substitute a string for a pattern in another string. Make sure there is 
+ enough room!
+
+ This routine looks for pattern in s and replaces it with 
+ insert. It may do multiple replacements.
+
+ Any of " ; ' $ or ` in the insert string are replaced with _
+ if len==0 then the string cannot be extended. This is different from the old
+ use of len==0 which was for no length checks to be done.
+**/
+
+_PUBLIC_ void string_sub(char *s, const char *pattern, const char *insert, size_t len)
+{
+	char *p;
+	ssize_t ls, lp, li, i;
+
+	if (!insert || !pattern || !*pattern || !s)
+		return;
+
+	ls = (ssize_t)strlen(s);
+	lp = (ssize_t)strlen(pattern);
+	li = (ssize_t)strlen(insert);
+
+	if (len == 0)
+		len = ls + 1; /* len is number of *bytes* */
+
+	while (lp <= ls && (p = strstr(s, pattern))) {
+		if (ls + (li-lp) >= len) {
+			DEBUG(0,("ERROR: string overflow by %d in string_sub(%.50s, %d)\n", 
+				 (int)(ls + (li-lp) - len),
+				 pattern, (int)len));
+			break;
+		}
+		if (li != lp) {
+			memmove(p+li,p+lp,strlen(p+lp)+1);
+		}
+		for (i=0;i<li;i++) {
+			switch (insert[i]) {
+			case '`':
+			case '"':
+			case '\'':
+			case ';':
+			case '$':
+			case '%':
+			case '\r':
+			case '\n':
+				p[i] = '_';
+				break;
+			default:
+				p[i] = insert[i];
+			}
+		}
+		s = p + li;
+		ls += (li-lp);
+	}
+}
+
+/**
+ * Talloc'ed version of string_sub
+ */
+_PUBLIC_ char *string_sub_talloc(TALLOC_CTX *mem_ctx, const char *s, 
+				const char *pattern, const char *insert)
+{
+	const char *p;
+	char *ret;
+	size_t len, alloc_len;
+
+	if (insert == NULL || pattern == NULL || !*pattern || s == NULL)
+		return NULL;
+
+	/* determine length needed */
+	len = strlen(s);
+	
+	for (p = strstr(s, pattern); p != NULL; 
+	     p = strstr(p+strlen(pattern), pattern)) {
+		len += strlen(insert) - strlen(pattern);
+	}
+
+	alloc_len = MAX(len, strlen(s))+1;
+	ret = talloc_array(mem_ctx, char, alloc_len);
+	if (ret == NULL)
+		return NULL;
+	strncpy(ret, s, alloc_len);
+	string_sub(ret, pattern, insert, alloc_len);
+
+	ret = talloc_realloc(mem_ctx, ret, char, len+1);
+	if (ret == NULL)
+		return NULL;
+
+	SMB_ASSERT(ret[len] == '\0');
+
+	talloc_set_name_const(ret, ret);
+
+	return ret;
+}
+
+/**
+ Similar to string_sub() but allows for any character to be substituted. 
+ Use with caution!
+ if len==0 then the string cannot be extended. This is different from the old
+ use of len==0 which was for no length checks to be done.
+**/
+
+_PUBLIC_ void all_string_sub(char *s,const char *pattern,const char *insert, size_t len)
+{
+	char *p;
+	ssize_t ls,lp,li;
+
+	if (!insert || !pattern || !s)
+		return;
+
+	ls = (ssize_t)strlen(s);
+	lp = (ssize_t)strlen(pattern);
+	li = (ssize_t)strlen(insert);
+
+	if (!*pattern)
+		return;
+	
+	if (len == 0)
+		len = ls + 1; /* len is number of *bytes* */
+	
+	while (lp <= ls && (p = strstr(s,pattern))) {
+		if (ls + (li-lp) >= len) {
+			DEBUG(0,("ERROR: string overflow by %d in all_string_sub(%.50s, %d)\n", 
+				 (int)(ls + (li-lp) - len),
+				 pattern, (int)len));
+			break;
+		}
+		if (li != lp) {
+			memmove(p+li,p+lp,strlen(p+lp)+1);
+		}
+		memcpy(p, insert, li);
+		s = p + li;
+		ls += (li-lp);
+	}
+}
diff --git a/ctdb/lib/util/util.c b/ctdb/lib/util/util.c
new file mode 100644
index 0000000000..af52805995
--- /dev/null
+++ b/ctdb/lib/util/util.c
@@ -0,0 +1,52 @@
+/* 
+   Unix SMB/CIFS implementation.
+   
+   Copyright (C) Andrew Tridgell 2005
+   
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "includes.h"
+#include "system/filesys.h"
+
+
+/**
+ Set a fd into blocking/nonblocking mode. Uses POSIX O_NONBLOCK if available,
+ else
+  if SYSV use O_NDELAY
+  if BSD use FNDELAY
+**/
+
+_PUBLIC_ int set_blocking(int fd, bool set)
+{
+	int val;
+#ifdef O_NONBLOCK
+#define FLAG_TO_SET O_NONBLOCK
+#else
+#ifdef SYSV
+#define FLAG_TO_SET O_NDELAY
+#else /* BSD */
+#define FLAG_TO_SET FNDELAY
+#endif
+#endif
+
+	if((val = fcntl(fd, F_GETFL, 0)) == -1)
+		return -1;
+	if(set) /* Turn blocking on - ie. clear nonblock flag */
+		val &= ~FLAG_TO_SET;
+	else
+		val |= FLAG_TO_SET;
+	return fcntl( fd, F_SETFL, val);
+#undef FLAG_TO_SET
+}
diff --git a/ctdb/lib/util/util.h b/ctdb/lib/util/util.h
new file mode 100644
index 0000000000..467fba88ad
--- /dev/null
+++ b/ctdb/lib/util/util.h
@@ -0,0 +1,651 @@
+/* 
+   Unix SMB/CIFS implementation.
+   Utility functions for Samba
+   Copyright (C) Andrew Tridgell 1992-1999
+   Copyright (C) Jelmer Vernooij 2005
+    
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.
+*/
+
+#ifndef _SAMBA_UTIL_H_
+#define _SAMBA_UTIL_H_
+
+/**
+ * @file
+ * @brief Helpful macros
+ */
+
+struct smbsrv_tcon;
+
+#ifdef _SAMBA_BUILD_
+extern const char *logfile;
+#endif
+extern const char *panic_action;
+extern void (*pre_panic_action_hook)(void);
+extern void (*post_panic_action_hook)(void);
+
+/**
+ * assert macros 
+ */
+#ifdef DEVELOPER
+#define SMB_ASSERT(b) do { if (!(b)) { \
+        DEBUG(0,("PANIC: assert failed at %s(%d): %s\n", \
+		 __FILE__, __LINE__, #b)); smb_panic("assert failed: " #b); }} while(0)
+#else
+/* redefine the assert macro for non-developer builds */
+#define SMB_ASSERT(b) do { if (!(b)) { \
+        DEBUG(0,("PANIC: assert failed at %s(%d): %s\n", \
+	    __FILE__, __LINE__, #b)); }} while (0)
+#endif
+
+#if _SAMBA_BUILD_ == 4
+#ifdef VALGRIND
+#define strlen(x) valgrind_strlen(x)
+size_t valgrind_strlen(const char *s);
+#endif
+#endif
+
+#ifndef ABS
+#define ABS(a) ((a)>0?(a):(-(a)))
+#endif
+
+/**
+ * Write backtrace to debug log
+ */
+_PUBLIC_ void call_backtrace(void);
+
+/**
+ Something really nasty happened - panic !
+**/
+_PUBLIC_ _NORETURN_ void smb_panic(const char *why);
+
+/**
+setup our fault handlers
+**/
+_PUBLIC_ void fault_setup(const char *pname);
+
+/**
+  register a fault handler. 
+  Should only be called once in the execution of smbd.
+*/
+_PUBLIC_ bool register_fault_handler(const char *name, void (*fault_handler)(int sig));
+
+/* The following definitions come from lib/util/signal.c  */
+
+
+/**
+ Block sigs.
+**/
+void BlockSignals(bool block, int signum);
+
+/**
+ Catch a signal. This should implement the following semantics:
+
+ 1) The handler remains installed after being called.
+ 2) The signal should be blocked during handler execution.
+**/
+void (*CatchSignal(int signum,void (*handler)(int )))(int);
+
+/**
+ Ignore SIGCLD via whatever means is necessary for this OS.
+**/
+void CatchChild(void);
+
+/**
+ Catch SIGCLD but leave the child around so it's status can be reaped.
+**/
+void CatchChildLeaveStatus(void);
+
+
+/* The following definitions come from lib/util/util_str.c  */
+
+
+/**
+ Trim the specified elements off the front and back of a string.
+**/
+_PUBLIC_ bool trim_string(char *s, const char *front, const char *back);
+
+/**
+ Find the number of 'c' chars in a string
+**/
+_PUBLIC_ _PURE_ size_t count_chars(const char *s, char c);
+
+/**
+ Safe string copy into a known length string. maxlength does not
+ include the terminating zero.
+**/
+_PUBLIC_ char *safe_strcpy(char *dest,const char *src, size_t maxlength);
+
+/**
+ Safe string cat into a string. maxlength does not
+ include the terminating zero.
+**/
+_PUBLIC_ char *safe_strcat(char *dest, const char *src, size_t maxlength);
+
+/**
+ Routine to get hex characters and turn them into a 16 byte array.
+ the array can be variable length, and any non-hex-numeric
+ characters are skipped.  "0xnn" or "0Xnn" is specially catered
+ for.
+
+ valid examples: "0A5D15"; "0x15, 0x49, 0xa2"; "59\ta9\te3\n"
+
+
+**/
+_PUBLIC_ size_t strhex_to_str(char *p, size_t p_len, const char *strhex, size_t strhex_len);
+
+#ifdef _SAMBA_BUILD_
+/** 
+ * Parse a hex string and return a data blob. 
+ */
+_PUBLIC_ _PURE_ DATA_BLOB strhex_to_data_blob(TALLOC_CTX *mem_ctx, const char *strhex) ;
+#endif
+
+/**
+ * Routine to print a buffer as HEX digits, into an allocated string.
+ */
+_PUBLIC_ void hex_encode(const unsigned char *buff_in, size_t len, char **out_hex_buffer);
+
+/**
+ * talloc version of hex_encode()
+ */
+_PUBLIC_ char *hex_encode_talloc(TALLOC_CTX *mem_ctx, const unsigned char *buff_in, size_t len);
+
+/**
+ Substitute a string for a pattern in another string. Make sure there is 
+ enough room!
+
+ This routine looks for pattern in s and replaces it with 
+ insert. It may do multiple replacements.
+
+ Any of " ; ' $ or ` in the insert string are replaced with _
+ if len==0 then the string cannot be extended. This is different from the old
+ use of len==0 which was for no length checks to be done.
+**/
+_PUBLIC_ void string_sub(char *s,const char *pattern, const char *insert, size_t len);
+
+
+_PUBLIC_ char *string_sub_talloc(TALLOC_CTX *mem_ctx, const char *s, 
+				const char *pattern, const char *insert);
+
+/**
+ Similar to string_sub() but allows for any character to be substituted. 
+ Use with caution!
+ if len==0 then the string cannot be extended. This is different from the old
+ use of len==0 which was for no length checks to be done.
+**/
+_PUBLIC_ void all_string_sub(char *s,const char *pattern,const char *insert, size_t len);
+
+/**
+ Unescape a URL encoded string, in place.
+**/
+_PUBLIC_ void rfc1738_unescape(char *buf);
+
+/**
+  format a string into length-prefixed dotted domain format, as used in NBT
+  and in some ADS structures
+**/
+_PUBLIC_ const char *str_format_nbt_domain(TALLOC_CTX *mem_ctx, const char *s);
+
+/**
+ * Add a string to an array of strings.
+ *
+ * num should be a pointer to an integer that holds the current 
+ * number of elements in strings. It will be updated by this function.
+ */
+_PUBLIC_ bool add_string_to_array(TALLOC_CTX *mem_ctx,
+			 const char *str, const char ***strings, int *num);
+
+/**
+  varient of strcmp() that handles NULL ptrs
+**/
+_PUBLIC_ int strcmp_safe(const char *s1, const char *s2);
+
+/**
+return the number of bytes occupied by a buffer in ASCII format
+the result includes the null termination
+limited by 'n' bytes
+**/
+_PUBLIC_ size_t ascii_len_n(const char *src, size_t n);
+
+/**
+ Set a boolean variable from the text value stored in the passed string.
+ Returns true in success, false if the passed string does not correctly 
+ represent a boolean.
+**/
+_PUBLIC_ bool set_boolean(const char *boolean_string, bool *boolean);
+
+/**
+ * Parse a string containing a boolean value.
+ *
+ * val will be set to the read value.
+ *
+ * @retval true if a boolean value was parsed, false otherwise.
+ */
+_PUBLIC_ bool conv_str_bool(const char * str, bool * val);
+
+#if _SAMBA_BUILD_ == 4
+/**
+ * Convert a size specification like 16K into an integral number of bytes. 
+ **/
+_PUBLIC_ bool conv_str_size(const char * str, uint64_t * val);
+#endif
+
+/**
+ * Parse a uint64_t value from a string
+ *
+ * val will be set to the value read.
+ *
+ * @retval true if parsing was successful, false otherwise
+ */
+_PUBLIC_ bool conv_str_u64(const char * str, uint64_t * val);
+
+/**
+return the number of bytes occupied by a buffer in CH_UTF16 format
+the result includes the null termination
+**/
+_PUBLIC_ size_t utf16_len(const void *buf);
+
+/**
+return the number of bytes occupied by a buffer in CH_UTF16 format
+the result includes the null termination
+limited by 'n' bytes
+**/
+_PUBLIC_ size_t utf16_len_n(const void *src, size_t n);
+_PUBLIC_ size_t ucs2_align(const void *base_ptr, const void *p, int flags);
+
+/**
+Do a case-insensitive, whitespace-ignoring string compare.
+**/
+_PUBLIC_ int strwicmp(const char *psz1, const char *psz2);
+
+/**
+ String replace.
+**/
+_PUBLIC_ void string_replace(char *s, char oldc, char newc);
+
+/**
+ * Compare 2 strings.
+ *
+ * @note The comparison is case-insensitive.
+ **/
+_PUBLIC_ bool strequal(const char *s1, const char *s2);
+
+/* The following definitions come from lib/util/util_strlist.c  */
+#ifdef _SAMBA_BUILD_
+
+/* separators for lists */
+#ifndef LIST_SEP
+#define LIST_SEP " \t,\n\r"
+#endif
+
+/**
+  build a null terminated list of strings from a input string and a
+  separator list. The separator list must contain characters less than
+  or equal to 0x2f for this to work correctly on multi-byte strings
+*/
+_PUBLIC_ char **str_list_make(TALLOC_CTX *mem_ctx, const char *string, const char *sep);
+
+/**
+ * build a null terminated list of strings from an argv-like input string 
+ * Entries are seperated by spaces and can be enclosed by quotes. 
+ * Does NOT support escaping
+ */
+_PUBLIC_ const char **str_list_make_shell(TALLOC_CTX *mem_ctx, const char *string, const char *sep);
+
+/**
+ * join a list back to one string 
+ */
+_PUBLIC_ char *str_list_join(TALLOC_CTX *mem_ctx, const char **list, char seperator);
+
+/** join a list back to one (shell-like) string; entries 
+ * seperated by spaces, using quotes where necessary */
+_PUBLIC_ char *str_list_join_shell(TALLOC_CTX *mem_ctx, const char **list, char sep);
+
+/**
+  return the number of elements in a string list
+*/
+_PUBLIC_ size_t str_list_length(const char * const *list);
+
+/**
+  copy a string list
+*/
+_PUBLIC_ char **str_list_copy(TALLOC_CTX *mem_ctx, const char **list);
+
+/**
+   Return true if all the elements of the list match exactly.
+ */
+_PUBLIC_ bool str_list_equal(const char **list1, const char **list2);
+
+/**
+  add an entry to a string list
+*/
+_PUBLIC_ const char **str_list_add(const char **list, const char *s);
+
+/**
+  remove an entry from a string list
+*/
+_PUBLIC_ void str_list_remove(const char **list, const char *s);
+
+/**
+  return true if a string is in a list
+*/
+_PUBLIC_ bool str_list_check(const char **list, const char *s);
+
+/**
+  return true if a string is in a list, case insensitively
+*/
+_PUBLIC_ bool str_list_check_ci(const char **list, const char *s);
+#endif
+
+/* The following definitions come from lib/util/util_file.c  */
+
+
+#ifdef _SAMBA_BUILD_
+/**
+read a line from a file with possible \ continuation chars. 
+Blanks at the start or end of a line are stripped.
+The string will be allocated if s2 is NULL
+**/
+_PUBLIC_ char *fgets_slash(char *s2,int maxlen,XFILE *f);
+#endif
+
+/**
+ * Read one line (data until next newline or eof) and allocate it 
+ */
+_PUBLIC_ char *afdgets(int fd, TALLOC_CTX *mem_ctx, size_t hint);
+
+#ifdef _SAMBA_BUILD_
+/**
+load a file into memory from a fd.
+**/
+_PUBLIC_ char *fd_load(int fd, size_t *size, size_t maxsize, TALLOC_CTX *mem_ctx);
+
+
+char **file_lines_parse(char *p, size_t size, int *numlines, TALLOC_CTX *mem_ctx);
+
+/**
+load a file into memory
+**/
+_PUBLIC_ char *file_load(const char *fname, size_t *size, size_t maxsize, TALLOC_CTX *mem_ctx);
+#endif
+
+/**
+mmap (if possible) or read a file
+**/
+_PUBLIC_ void *map_file(const char *fname, size_t size);
+
+#ifdef _SAMBA_BUILD_
+/**
+load a file into memory and return an array of pointers to lines in the file
+must be freed with talloc_free(). 
+**/
+_PUBLIC_ char **file_lines_load(const char *fname, int *numlines, size_t maxsize, TALLOC_CTX *mem_ctx);
+#endif
+
+/**
+load a fd into memory and return an array of pointers to lines in the file
+must be freed with talloc_free(). If convert is true calls unix_to_dos on
+the list.
+**/
+_PUBLIC_ char **fd_lines_load(int fd, int *numlines, size_t maxsize, TALLOC_CTX *mem_ctx);
+
+/**
+take a list of lines and modify them to produce a list where \ continues
+a line
+**/
+_PUBLIC_ void file_lines_slashcont(char **lines);
+
+/**
+  save a lump of data into a file. Mostly used for debugging 
+*/
+_PUBLIC_ bool file_save(const char *fname, const void *packet, size_t length);
+_PUBLIC_ int vfdprintf(int fd, const char *format, va_list ap) PRINTF_ATTRIBUTE(2,0);
+_PUBLIC_ int fdprintf(int fd, const char *format, ...) PRINTF_ATTRIBUTE(2,3);
+_PUBLIC_ bool large_file_support(const char *path);
+
+/* The following definitions come from lib/util/util.c  */
+
+
+/**
+ Find a suitable temporary directory. The result should be copied immediately
+ as it may be overwritten by a subsequent call.
+**/
+_PUBLIC_ const char *tmpdir(void);
+
+/**
+ Check if a file exists - call vfs_file_exist for samba files.
+**/
+_PUBLIC_ bool file_exist(const char *fname);
+
+/**
+ Check a files mod time.
+**/
+_PUBLIC_ time_t file_modtime(const char *fname);
+
+/**
+ Check if a directory exists.
+**/
+_PUBLIC_ bool directory_exist(const char *dname);
+
+/**
+ * Try to create the specified directory if it didn't exist.
+ *
+ * @retval true if the directory already existed and has the right permissions 
+ * or was successfully created.
+ */
+_PUBLIC_ bool directory_create_or_exist(const char *dname, uid_t uid, 
+			       mode_t dir_perms);
+
+/**
+ Set a fd into blocking/nonblocking mode. Uses POSIX O_NONBLOCK if available,
+ else
+  if SYSV use O_NDELAY
+  if BSD use FNDELAY
+**/
+_PUBLIC_ int set_blocking(int fd, bool set);
+
+/**
+ Sleep for a specified number of milliseconds.
+**/
+_PUBLIC_ void msleep(unsigned int t);
+
+/**
+ Get my own name, return in malloc'ed storage.
+**/
+_PUBLIC_ char* get_myname(void);
+
+/**
+ Return true if a string could be a pure IP address.
+**/
+_PUBLIC_ bool is_ipaddress(const char *str);
+
+/**
+ Interpret an internet address or name into an IP address in 4 byte form.
+**/
+_PUBLIC_ uint32_t interpret_addr(const char *str);
+
+/**
+ A convenient addition to interpret_addr().
+**/
+_PUBLIC_ struct in_addr interpret_addr2(const char *str);
+
+/**
+ Check if an IP is the 0.0.0.0.
+**/
+_PUBLIC_ bool is_zero_ip_v4(struct in_addr ip);
+
+/**
+ Are two IPs on the same subnet?
+**/
+_PUBLIC_ bool same_net_v4(struct in_addr ip1,struct in_addr ip2,struct in_addr mask);
+
+_PUBLIC_ bool is_ipaddress_v4(const char *str);
+
+/**
+ Check if a process exists. Does this work on all unixes?
+**/
+_PUBLIC_ bool process_exists_by_pid(pid_t pid);
+
+/**
+ Simple routine to do POSIX file locking. Cruft in NFS and 64->32 bit mapping
+ is dealt with in posix.c
+**/
+_PUBLIC_ bool fcntl_lock(int fd, int op, off_t offset, off_t count, int type);
+
+/**
+ malloc that aborts with smb_panic on fail or zero size.
+**/
+_PUBLIC_ void *smb_xmalloc(size_t size);
+
+/**
+ Memdup with smb_panic on fail.
+**/
+_PUBLIC_ void *smb_xmemdup(const void *p, size_t size);
+
+/**
+ strdup that aborts on malloc fail.
+**/
+_PUBLIC_ char *smb_xstrdup(const char *s);
+
+char *smb_xstrndup(const char *s, size_t n);
+
+/**
+ Like strdup but for memory.
+**/
+_PUBLIC_ void *memdup(const void *p, size_t size);
+
+/**
+ * see if a range of memory is all zero. A NULL pointer is considered
+ * to be all zero 
+ */
+_PUBLIC_ bool all_zero(const uint8_t *ptr, size_t size);
+
+/**
+  realloc an array, checking for integer overflow in the array size
+*/
+_PUBLIC_ void *realloc_array(void *ptr, size_t el_size, unsigned count, bool free_on_fail);
+
+void *malloc_array(size_t el_size, unsigned int count);
+
+/* The following definitions come from lib/util/fsusage.c  */
+
+
+/**
+ * Retrieve amount of free disk space.
+ * this does all of the system specific guff to get the free disk space.
+ * It is derived from code in the GNU fileutils package, but has been
+ * considerably mangled for use here 
+ *
+ * results are returned in *dfree and *dsize, in 512 byte units
+*/
+_PUBLIC_ int sys_fsusage(const char *path, uint64_t *dfree, uint64_t *dsize);
+
+/* The following definitions come from lib/util/ms_fnmatch.c  */
+
+
+/**
+ * @file
+ * @brief MS-style Filename matching
+ */
+
+#if _SAMBA_BUILD_ == 4
+/* protocol types. It assumes that higher protocols include lower protocols
+   as subsets. FIXME: Move to one of the smb-specific headers */
+enum protocol_types {
+	PROTOCOL_NONE,
+	PROTOCOL_CORE,
+	PROTOCOL_COREPLUS,
+	PROTOCOL_LANMAN1,
+	PROTOCOL_LANMAN2,
+	PROTOCOL_NT1,
+	PROTOCOL_SMB2
+};
+
+int ms_fnmatch(const char *pattern, const char *string, enum protocol_types protocol);
+
+/** a generic fnmatch function - uses for non-CIFS pattern matching */
+int gen_fnmatch(const char *pattern, const char *string);
+#endif
+
+/* The following definitions come from lib/util/mutex.c  */
+
+
+#ifdef _SAMBA_BUILD_
+/**
+  register a set of mutex/rwlock handlers. 
+  Should only be called once in the execution of smbd.
+*/
+_PUBLIC_ bool register_mutex_handlers(const char *name, struct mutex_ops *ops);
+#endif
+
+/* The following definitions come from lib/util/idtree.c  */
+
+
+/**
+  initialise a idr tree. The context return value must be passed to
+  all subsequent idr calls. To destroy the idr tree use talloc_free()
+  on this context
+ */
+_PUBLIC_ struct idr_context *idr_init(TALLOC_CTX *mem_ctx);
+
+/**
+  allocate the next available id, and assign 'ptr' into its slot.
+  you can retrieve later this pointer using idr_find()
+*/
+_PUBLIC_ int idr_get_new(struct idr_context *idp, void *ptr, int limit);
+
+/**
+   allocate a new id, giving the first available value greater than or
+   equal to the given starting id
+*/
+_PUBLIC_ int idr_get_new_above(struct idr_context *idp, void *ptr, int starting_id, int limit);
+
+/**
+  allocate a new id randomly in the given range
+*/
+_PUBLIC_ int idr_get_new_random(struct idr_context *idp, void *ptr, int limit);
+
+/**
+  find a pointer value previously set with idr_get_new given an id
+*/
+_PUBLIC_ void *idr_find(struct idr_context *idp, int id);
+
+/**
+  remove an id from the idr tree
+*/
+_PUBLIC_ int idr_remove(struct idr_context *idp, int id);
+
+/* The following definitions come from lib/util/become_daemon.c  */
+
+#if _SAMBA_BUILD_ == 4
+/**
+ Become a daemon, discarding the controlling terminal.
+**/
+_PUBLIC_ void become_daemon(bool fork);
+#endif
+
+/**
+ * Load a ini-style file.
+ */
+bool pm_process( const char *fileName,
+                 bool (*sfunc)(const char *, void *),
+                 bool (*pfunc)(const char *, const char *, void *),
+				 void *userdata);
+
+bool unmap_file(void *start, size_t size);
+
+#define CONST_DISCARD(type, ptr)      ((type) ((void *) (ptr)))
+
+#endif /* _SAMBA_UTIL_H_ */
diff --git a/ctdb/lib/util/util_file.c b/ctdb/lib/util/util_file.c
new file mode 100644
index 0000000000..3a90201cce
--- /dev/null
+++ b/ctdb/lib/util/util_file.c
@@ -0,0 +1,132 @@
+/*
+  functions taken from samba4 for quick prototyping of ctdb. These are
+  not intended to remain part of ctdb
+*/
+
+#include "includes.h"
+#include "system/filesys.h"
+
+
+static char *fd_load(int fd, size_t *size, TALLOC_CTX *mem_ctx)
+{
+	struct stat sbuf;
+	char *p;
+
+	if (fstat(fd, &sbuf) != 0) return NULL;
+
+	p = (char *)talloc_size(mem_ctx, sbuf.st_size+1);
+	if (!p) return NULL;
+
+	if (read(fd, p, sbuf.st_size) != sbuf.st_size) {
+		talloc_free(p);
+		return NULL;
+	}
+	p[sbuf.st_size] = 0;
+
+	if (size) *size = sbuf.st_size;
+
+	return p;
+}
+
+
+static char *file_load(const char *fname, size_t *size, TALLOC_CTX *mem_ctx)
+{
+	int fd;
+	char *p;
+
+	if (!fname || !*fname) return NULL;
+	
+	fd = open(fname,O_RDONLY);
+	if (fd == -1) return NULL;
+
+	p = fd_load(fd, size, mem_ctx);
+
+	close(fd);
+
+	return p;
+}
+
+
+/**
+parse a buffer into lines
+'p' will be freed on error, and otherwise will be made a child of the returned array
+**/
+static char **file_lines_parse(char *p, size_t size, int *numlines, TALLOC_CTX *mem_ctx)
+{
+	int i;
+	char *s, **ret;
+
+	if (!p) return NULL;
+
+	for (s = p, i=0; s < p+size; s++) {
+		if (s[0] == '\n') i++;
+	}
+
+	ret = talloc_array(mem_ctx, char *, i+2);
+	if (!ret) {
+		talloc_free(p);
+		return NULL;
+	}	
+	
+	talloc_steal(ret, p);
+	
+	memset(ret, 0, sizeof(ret[0])*(i+2));
+	if (numlines) *numlines = i;
+
+	ret[0] = p;
+	for (s = p, i=0; s < p+size; s++) {
+		if (s[0] == '\n') {
+			s[0] = 0;
+			i++;
+			ret[i] = s+1;
+		}
+		if (s[0] == '\r') s[0] = 0;
+	}
+
+	return ret;
+}
+
+
+/**
+load a file into memory and return an array of pointers to lines in the file
+must be freed with talloc_free(). 
+**/
+_PUBLIC_ char **file_lines_load(const char *fname, int *numlines, TALLOC_CTX *mem_ctx)
+{
+	char *p;
+	size_t size;
+
+	p = file_load(fname, &size, mem_ctx);
+	if (!p) return NULL;
+
+	return file_lines_parse(p, size, numlines, mem_ctx);
+}
+
+char *hex_encode_talloc(TALLOC_CTX *mem_ctx, const unsigned char *buff_in, size_t len)
+{
+	int i;
+	char *hex_buffer;
+
+	hex_buffer = talloc_array(mem_ctx, char, (len*2)+1);
+
+	for (i = 0; i < len; i++)
+		slprintf(&hex_buffer[i*2], 3, "%02X", buff_in[i]);
+
+	return hex_buffer;
+}
+
+uint8_t *hex_decode_talloc(TALLOC_CTX *mem_ctx, const char *hex_in, size_t *len)
+{
+	int i, num;
+	uint8_t *buffer;
+
+	*len = strlen(hex_in) / 2;
+	buffer = talloc_array(mem_ctx, unsigned char, *len);
+
+	for (i=0; i<*len; i++) {
+		sscanf(&hex_in[i*2], "%02X", &num);
+		buffer[i] = (uint8_t)num;
+	}
+
+	return buffer;
+}
diff --git a/ctdb/lib/util/util_time.c b/ctdb/lib/util/util_time.c
new file mode 100644
index 0000000000..be14f26458
--- /dev/null
+++ b/ctdb/lib/util/util_time.c
@@ -0,0 +1,108 @@
+/*
+  functions taken from samba4 for quick prototyping of ctdb. These are
+  not intended to remain part of ctdb
+*/
+
+#include "includes.h"
+#include "system/time.h"
+#include "system/filesys.h"
+
+
+/**
+  return a zero timeval
+*/
+struct timeval timeval_zero(void)
+{
+	struct timeval tv;
+	tv.tv_sec = 0;
+	tv.tv_usec = 0;
+	return tv;
+}
+
+/**
+  return True if a timeval is zero
+*/
+bool timeval_is_zero(const struct timeval *tv)
+{
+	return tv->tv_sec == 0 && tv->tv_usec == 0;
+}
+
+/**
+  return a timeval for the current time
+*/
+struct timeval timeval_current(void)
+{
+	struct timeval tv;
+	gettimeofday(&tv, NULL);
+	return tv;
+}
+
+double timeval_elapsed(struct timeval *tv)
+{
+	struct timeval tv2 = timeval_current();
+	return (tv2.tv_sec - tv->tv_sec) + 
+	       (tv2.tv_usec - tv->tv_usec)*1.0e-6;
+}
+
+double timeval_delta(struct timeval *tv2, struct timeval *tv)
+{
+	return (tv2->tv_sec - tv->tv_sec) + 
+	       (tv2->tv_usec - tv->tv_usec)*1.0e-6;
+}
+
+/**
+  return a timeval struct with the given elements
+*/
+_PUBLIC_ struct timeval timeval_set(uint32_t secs, uint32_t usecs)
+{
+	struct timeval tv;
+	tv.tv_sec = secs;
+	tv.tv_usec = usecs;
+	return tv;
+}
+
+_PUBLIC_ int timeval_compare(const struct timeval *tv1, const struct timeval *tv2)
+{
+	if (tv1->tv_sec  > tv2->tv_sec)  return 1;
+	if (tv1->tv_sec  < tv2->tv_sec)  return -1;
+	if (tv1->tv_usec > tv2->tv_usec) return 1;
+	if (tv1->tv_usec < tv2->tv_usec) return -1;
+	return 0;
+}
+
+_PUBLIC_ struct timeval timeval_until(const struct timeval *tv1,
+				      const struct timeval *tv2)
+{
+	struct timeval t;
+	if (timeval_compare(tv1, tv2) >= 0) {
+		return timeval_zero();
+	}
+	t.tv_sec = tv2->tv_sec - tv1->tv_sec;
+	if (tv1->tv_usec > tv2->tv_usec) {
+		t.tv_sec--;
+		t.tv_usec = 1000000 - (tv1->tv_usec - tv2->tv_usec);
+	} else {
+		t.tv_usec = tv2->tv_usec - tv1->tv_usec;
+	}
+	return t;
+}
+
+static struct timeval timeval_add(const struct timeval *tv,
+			   uint32_t secs, uint32_t usecs)
+{
+	struct timeval tv2 = *tv;
+	const unsigned int million = 1000000;
+	tv2.tv_sec += secs;
+	tv2.tv_usec += usecs;
+	tv2.tv_sec += tv2.tv_usec / million;
+	tv2.tv_usec = tv2.tv_usec % million;
+	return tv2;
+}
+
+
+_PUBLIC_ struct timeval timeval_current_ofs(uint32_t secs, uint32_t usecs)
+{
+	struct timeval tv = timeval_current();
+	return timeval_add(&tv, secs, usecs);
+}
+