Merge branch 'master' into tests

Conflicts:
	ChangeLog

10 files changed, 253 insertions, 34 deletions

diff --git a/ChangeLog b/ChangeLog
index e74c7209..4540d6e6 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,7 @@
+- bugfix: potential abort with DA queue after high watermark is reached
+  There exists a race condition that can lead to a segfault. Thanks
+  go to vbernetr, who performed the analysis and provided patch, which
+  I only tweaked a very little bit.
 ---------------------------------------------------------------------------
 Version 4.1.6  [DEVEL] (rgerhards), 2009-03-??
 - implemented function support in RainerScript. That means the engine
@@ -19,6 +23,13 @@ Version 4.1.6  [DEVEL] (rgerhards), 2009-03-??
 - modified $ModLoad statement so that for modules whom's name starts with
   a dot, no path is prepended (this enables relative-pathes and should
   not break any valid current config)
+- fixed a bug that caused action retries not to work correctly
+  situation was only cleared by a restart
+- bugfix: closed dynafile was potentially never written until another
+  dynafile name was generated - potential loss of messages
+- improved omfile so that it properly suspends itself if there is an
+  i/o or file name generation error. This enables it to be used with
+  the full high availability features of rsyslog's engine
 ---------------------------------------------------------------------------
 Version 4.1.5  [DEVEL] (rgerhards), 2009-03-11
 - bugfix: parser did not correctly parse fields in UDP-received messages
diff --git a/action.c b/action.c
index cd4ba240..08755c13 100644
--- a/action.c
+++ b/action.c
@@ -353,7 +353,13 @@ static rsRetVal actionTryResume(action_t *pThis)
 
 	ASSERT(pThis != NULL);
 
-	ttNow = getActNow(pThis); /* cache "now" */
+	/* for resume handling, we must always obtain a fresh timestamp. We used
+	 * to use the action timestamp, but in this case we will never reach a
+	 * point where a resumption is actually tried, because the action timestamp
+	 * is always in the past. So we can not avoid doing a fresh time() call
+	 * here. -- rgerhards, 2009-03-18
+	 */
+	time(&ttNow); /* cache "now" */
 
 	/* first check if it is time for a re-try */
 	if(ttNow > pThis->ttResumeRtry) {
diff --git a/doc/rsyslog-vers.dot b/doc/rsyslog-vers.dot
new file mode 100644
index 00000000..a5563f94
--- /dev/null
+++ b/doc/rsyslog-vers.dot
@@ -0,0 +1,82 @@
+// This file is part of rsyslog.
+// 
+// rsyslog "family tree" compressed version
+//
+// see http://www.graphviz.org for how to obtain the graphviz processor
+// which is used to build the actual graph.
+//
+// generate the graph with
+// $ dot rsyslog-vers.dot -Tpng >rsyslog-vers.png
+
+digraph G {
+	label="\n\nrsyslog \"family tree\"\nhttp://www.rsyslog.com";
+	fontsize=20;
+
+	v1stable [label="v1-stable", shape=box, style=dotted];
+	v2stable [label="v2-stable", shape=box, style=filled];
+	v3stable [label="v3-stable", shape=box, style=filled];
+	beta     [label="beta", shape=box, style=filled];
+	devel    [label="devel", shape=box, style=filled];
+	"1.0.5"  [style=dotted];
+	perf     [style=dashed];
+	"imudp-select" [style=dashed label="imudp-\nselect"];
+	msgnolock [style=dashed];
+	"file-errHdlr" [style=dashed];
+	solaris [style=dashed];
+	tests   [style=dashed];
+	"0.x.y"   [group=master];
+	"1.10.0"  [group=master];
+	"1.21.2"  [group=master];
+	"3.10.0"  [group=master];
+	"3.15.1"  [group=master];
+	"3.17.0"  [group=master];
+	"3.19.x"  [group=master];
+	"3.21.x"  [group=master];
+	"4.1.0"   [group=master];
+	"4.1.4"   [group=master];
+	"4.1.5"   [group=master];
+	"4.1.6"   [group=master];
+
+	sysklogd -> "0.x.y" [color=red];
+	"0.x.y" -> "1.0.0";
+	"0.x.y" -> "1.10.0" [color=red];
+	"1.0.0" -> "1.0.5"  [style=dashed];
+	"1.10.0" -> "1.21.2" [color=red style=dashed];
+	"1.21.2" -> "2.0.0" [color=blue];
+	"2.0.0" -> "2.0.5" [style=dashed, color=blue];
+	"1.21.2" -> "3.10.0" [color=red];
+	"3.10.0" -> "3.15.1" [color=red style=dashed];
+	"3.15.1" -> "tests";
+	"3.15.1" -> "3.17.0" [color=red style=dashed];
+	"3.15.1" -> "3.16.x";
+	"3.16.x" -> "3.18.x" [color=blue, style=dashed];
+	"3.17.0" -> "3.18.x";
+	"3.17.0" -> "3.19.x" [color=red, style=dashed];
+	"3.19.x" -> "3.20.x";
+	"3.19.x" -> "3.21.x" [color=red];
+	"3.18.x" -> debian_lenny;
+	"3.18.x" -> "3.20.x" [color=blue, style=dashed];
+	"3.21.x" -> "3.21.11";
+	"3.21.x" -> "4.1.0" [color=red];
+	"3.21.x" -> "perf";
+	"perf" -> "4.1.0";
+	"4.1.0" -> "4.1.4" [color=red, style=dashed];
+	"4.1.4" -> "file-errHdlr";
+	"4.1.4" -> "4.1.5" [color=red];
+	"4.1.5" -> "4.1.6" [color=red];
+	"3.21.x" -> msgnolock
+	"3.21.x" -> "imudp-select";
+	"4.1.5" -> solaris;
+	"file-errHdlr" -> "4.1.6";
+	"tests" -> "4.1.6";
+
+	"1.0.5" -> v1stable [dir=none, style=dotted];
+	"2.0.5" -> v2stable [dir=none, style=dotted];
+	"3.20.x" -> v3stable [dir=none, style=dotted];
+	"3.21.11" -> beta    [dir=none, style=dotted];
+	"4.1.6" ->   devel   [dir=none, style=dotted];
+
+	{rank=same; "4.1.5" "solaris"}
+	{rank=same; "3.18.x" "debian_lenny"}
+	{rank=same; "1.0.5" "2.0.5" "3.20.x" "3.21.11" "4.1.6"}
+}
diff --git a/doc/rsyslog_conf_global.html b/doc/rsyslog_conf_global.html
index b0c1e400..d011bd2b 100644
--- a/doc/rsyslog_conf_global.html
+++ b/doc/rsyslog_conf_global.html
@@ -103,6 +103,7 @@ default 60000 (1 minute)]</li>
 <li>$DefaultNetstreamDriver &lt;drivername&gt;, the default <a href="netstream.html">network stream driver</a> to use. Defaults to&nbsp;ptcp.$DefaultNetstreamDriverCAFile &lt;/path/to/cafile.pem&gt;</li>
 <li>$DefaultNetstreamDriverCertFile &lt;/path/to/certfile.pem&gt;</li>
 <li>$DefaultNetstreamDriverKeyFile &lt;/path/to/keyfile.pem&gt;</li>
+<li><b>$CreateDirs</b> [<b>on</b>/off] - create directories on an as-needed basis</li>
 <li><a href="rsconf1_dircreatemode.html">$DirCreateMode</a></li>
 <li><a href="rsconf1_dirgroup.html">$DirGroup</a></li>
 <li><a href="rsconf1_dirowner.html">$DirOwner</a></li>
diff --git a/runtime/debug.c b/runtime/debug.c
index 53291fc2..96004e47 100644
--- a/runtime/debug.c
+++ b/runtime/debug.c
@@ -476,6 +476,55 @@ static inline void dbgMutexLockLog(pthread_mutex_t *pmut, dbgFuncDB_t *pFuncDB,
 		dbgprintf("%s:%d:%s: mutex %p aquired\n", pFuncDB->file, lockLn, pFuncDB->func, (void*)pmut); 
 }
 
+
+/* report trylock on a mutex and add it to the mutex log */
+static inline void dbgMutexPreTryLockLog(pthread_mutex_t *pmut, dbgFuncDB_t *pFuncDB, int ln)
+{
+   dbgMutLog_t *pHolder;
+   dbgMutLog_t *pLog;
+   char pszBuf[128];
+   char pszHolderThrdName[64];
+   char *pszHolder;
+
+   pthread_mutex_lock(&mutMutLog);
+   pHolder = dbgMutLogFindHolder(pmut);
+   pLog = dbgMutLogAddEntry(pmut, MUTOP_TRYLOCK, pFuncDB, ln);
+
+   if(pHolder == NULL)
+      pszHolder = "[NONE]";
+   else {
+      dbgGetThrdName(pszHolderThrdName, sizeof(pszHolderThrdName), pHolder->thrd, 1);
+      snprintf(pszBuf, sizeof(pszBuf)/sizeof(char), "%s:%d [%s]", pHolder->pFuncDB->file, pHolder->lockLn, pszHolderThrdName);
+      pszHolder = pszBuf;
+   }
+
+   if(bPrintMutexAction)
+      dbgprintf("%s:%d:%s: mutex %p trying to get lock, held by %s\n", pFuncDB->file, ln, pFuncDB->func, (void*)pmut, pszHolder);
+   pthread_mutex_unlock(&mutMutLog);
+}
+
+
+/* report attempted mutex lock */
+static inline void dbgMutexTryLockLog(pthread_mutex_t *pmut, dbgFuncDB_t *pFuncDB, int lockLn)
+{
+   dbgMutLog_t *pLog;
+
+   pthread_mutex_lock(&mutMutLog);
+
+   /* find and delete "trylock" entry */
+   pLog = dbgMutLogFindSpecific(pmut, MUTOP_TRYLOCK, pFuncDB);
+   assert(pLog != NULL);
+   dbgMutLogDelEntry(pLog);
+
+   /* add "lock" entry */
+   pLog = dbgMutLogAddEntry(pmut, MUTOP_LOCK, pFuncDB, lockLn);
+   dbgFuncDBAddMutexLock(pFuncDB, pmut, lockLn);
+   pthread_mutex_unlock(&mutMutLog);
+   if(bPrintMutexAction)
+      dbgprintf("%s:%d:%s: mutex %p aquired\n", pFuncDB->file, lockLn, pFuncDB->func, (void*)pmut);
+}
+
+
 /* if we unlock, we just remove the lock aquired entry from the log list */
 static inline void dbgMutexUnlockLog(pthread_mutex_t *pmut, dbgFuncDB_t *pFuncDB, int unlockLn)
 {
@@ -534,6 +583,26 @@ int dbgMutexLock(pthread_mutex_t *pmut, dbgFuncDB_t *pFuncDB, int ln, int iStack
 }
 
 
+/* wrapper for pthread_mutex_trylock() */
+int dbgMutexTryLock(pthread_mutex_t *pmut, dbgFuncDB_t *pFuncDB, int ln, int iStackPtr)
+{
+   int ret;
+
+   dbgRecordExecLocation(iStackPtr, ln);
+   dbgMutexPreLockLog(pmut, pFuncDB, ln); // TODO : update this
+   ret = pthread_mutex_trylock(pmut);
+   if(ret == 0 || ret == EBUSY) {
+      // TODO : update this
+      dbgMutexLockLog(pmut, pFuncDB, ln);
+   } else {
+      dbgprintf("%s:%d:%s: ERROR: pthread_mutex_trylock() for mutex %p failed with error %d\n",
+           pFuncDB->file, ln, pFuncDB->func, (void*)pmut, ret);
+   }
+
+   return ret;
+}
+
+
 /* wrapper for pthread_mutex_unlock() */
 int dbgMutexUnlock(pthread_mutex_t *pmut, dbgFuncDB_t *pFuncDB, int ln, int iStackPtr)
 {
diff --git a/runtime/debug.h b/runtime/debug.h
index 59b3e776..1375493d 100644
--- a/runtime/debug.h
+++ b/runtime/debug.h
@@ -89,6 +89,7 @@ void sigsegvHdlr(int signum);
 void dbgoprint(obj_t *pObj, char *fmt, ...) __attribute__((format(printf, 2, 3)));
 void dbgprintf(char *fmt, ...) __attribute__((format(printf, 1, 2)));
 int dbgMutexLock(pthread_mutex_t *pmut, dbgFuncDB_t *pFuncD, int ln, int iStackPtr);
+int dbgMutexTryLock(pthread_mutex_t *pmut, dbgFuncDB_t *pFuncD, int ln, int iStackPtr);
 int dbgMutexUnlock(pthread_mutex_t *pmut, dbgFuncDB_t *pFuncD, int ln, int iStackPtr);
 int dbgCondWait(pthread_cond_t *cond, pthread_mutex_t *pmut, dbgFuncDB_t *pFuncD, int ln, int iStackPtr);
 int dbgCondTimedWait(pthread_cond_t *cond, pthread_mutex_t *pmut, const struct timespec *abstime, dbgFuncDB_t *pFuncD, int ln, int iStackPtr);
@@ -129,18 +130,21 @@ void dbgPrintAllDebugInfo(void);
 #define MUTOP_LOCKWAIT		1
 #define MUTOP_LOCK		2
 #define MUTOP_UNLOCK		3
+#define MUTOP_TRYLOCK		4
 
 
 /* debug aides */
 //#ifdef RTINST
 #if 0 // temporarily removed for helgrind
 #define d_pthread_mutex_lock(x)      dbgMutexLock(x, pdbgFuncDB, __LINE__, dbgCALLStaCK_POP_POINT )
+#define d_pthread_mutex_trylock(x)   dbgMutexTryLock(x, pdbgFuncDB, __LINE__, dbgCALLStaCK_POP_POINT )
 #define d_pthread_mutex_unlock(x)    dbgMutexUnlock(x, pdbgFuncDB, __LINE__, dbgCALLStaCK_POP_POINT )
 #define d_pthread_cond_wait(cond, mut)   dbgCondWait(cond, mut, pdbgFuncDB, __LINE__, dbgCALLStaCK_POP_POINT )
 #define d_pthread_cond_timedwait(cond, mut, to)   dbgCondTimedWait(cond, mut, to, pdbgFuncDB, __LINE__, dbgCALLStaCK_POP_POINT )
 #define d_free(x)      dbgFree(x, pdbgFuncDB, __LINE__, dbgCALLStaCK_POP_POINT )
 #else
 #define d_pthread_mutex_lock(x)     pthread_mutex_lock(x)
+#define d_pthread_mutex_trylock(x)  pthread_mutex_trylock(x)
 #define d_pthread_mutex_unlock(x)   pthread_mutex_unlock(x)
 #define d_pthread_cond_wait(cond, mut)   pthread_cond_wait(cond, mut)
 #define d_pthread_cond_timedwait(cond, mut, to)   pthread_cond_timedwait(cond, mut, to)
diff --git a/runtime/wti.c b/runtime/wti.c
index 9b3450e6..544bffa7 100644
--- a/runtime/wti.c
+++ b/runtime/wti.c
@@ -245,10 +245,14 @@ wtiJoinThrd(wti_t *pThis)
 
 	ISOBJ_TYPE_assert(pThis, wti);
 	dbgprintf("waiting for worker %s termination, current state %d\n", wtiGetDbgHdr(pThis), pThis->tCurrCmd);
-	pthread_join(pThis->thrdID, NULL);
-	wtiSetState(pThis, eWRKTHRD_STOPPED, 0, MUTEX_ALREADY_LOCKED); /* back to virgin... */
-	pThis->thrdID = 0; /* invalidate the thread ID so that we do not accidently find reused ones */
+	if (pThis->thrdID == 0) {
+		dbgprintf("worker %s was already stopped\n", wtiGetDbgHdr(pThis));
+	} else {
+		pthread_join(pThis->thrdID, NULL);
+		wtiSetState(pThis, eWRKTHRD_STOPPED, 0, MUTEX_ALREADY_LOCKED); /* back to virgin... */
+		pThis->thrdID = 0; /* invalidate the thread ID so that we do not accidently find reused ones */
 	dbgprintf("worker %s has stopped\n", wtiGetDbgHdr(pThis));
+	}   
 
 	RETiRet;
 }
diff --git a/runtime/wtp.c b/runtime/wtp.c
index 41903576..9f54a9ab 100644
--- a/runtime/wtp.c
+++ b/runtime/wtp.c
@@ -84,6 +84,7 @@ static rsRetVal NotImplementedDummy() { return RS_RET_OK; }
 BEGINobjConstruct(wtp) /* be sure to specify the object type also in END macro! */
 	pThis->bOptimizeUniProc = glbl.GetOptimizeUniProc();
 	pthread_mutex_init(&pThis->mut, NULL);
+	pthread_mutex_init(&pThis->mutThrdShutdwn, NULL);
 	pthread_cond_init(&pThis->condThrdTrm, NULL);
 	/* set all function pointers to "not implemented" dummy so that we can safely call them */
 	pThis->pfChkStopWrkr = NotImplementedDummy;
@@ -148,6 +149,7 @@ CODESTARTobjDestruct(wtp)
 	/* actual destruction */
 	pthread_cond_destroy(&pThis->condThrdTrm);
 	pthread_mutex_destroy(&pThis->mut);
+	pthread_mutex_destroy(&pThis->mutThrdShutdwn);
 
 	if(pThis->pszDbgHdr != NULL)
 		free(pThis->pszDbgHdr);
@@ -199,11 +201,23 @@ wtpProcessThrdChanges(wtp_t *pThis)
 	if(pThis->bThrdStateChanged == 0)
 		FINALIZE;
 
-	/* go through all threads */
-	for(i = 0 ; i < pThis->iNumWorkerThreads ; ++i) {
-		wtiProcessThrdChanges(pThis->pWrkr[i], LOCK_MUTEX);
+	if(d_pthread_mutex_trylock(&(pThis->mutThrdShutdwn)) != 0) {
+		/* another thread is already in the loop */
+		FINALIZE;
 	}
 
+	do {
+		/* reset the change marker */
+		pThis->bThrdStateChanged = 0;
+		/* go through all threads */
+		for(i = 0 ; i < pThis->iNumWorkerThreads ; ++i) {
+			wtiProcessThrdChanges(pThis->pWrkr[i], LOCK_MUTEX);
+		}
+	/* restart if another change occured while we were processing the changes */
+	} while(pThis->bThrdStateChanged != 0);
+
+	d_pthread_mutex_unlock(&(pThis->mutThrdShutdwn));
+
 finalize_it:
 	RETiRet;
 }
diff --git a/runtime/wtp.h b/runtime/wtp.h
index 88d88afd..b9cb07c5 100644
--- a/runtime/wtp.h
+++ b/runtime/wtp.h
@@ -61,6 +61,7 @@ typedef struct wtp_s {
 	int	bInactivityGuard;/* prevents inactivity due to race condition */
 	rsRetVal (*pConsumer)(void *); /* user-supplied consumer function for dewtpd messages */
 	/* synchronization variables */
+	pthread_mutex_t mutThrdShutdwn; /* mutex to guard thread shutdown processing */
 	pthread_mutex_t mut; /* mutex for the wtp's thread management */
 	pthread_cond_t condThrdTrm;/* signalled when threads terminate */
 	int bThrdStateChanged;	/* at least one thread state has changed if 1 */
diff --git a/tools/omfile.c b/tools/omfile.c
index ea91d6ef..65306846 100644
--- a/tools/omfile.c
+++ b/tools/omfile.c
@@ -387,9 +387,12 @@ static void dynaFileFreeCache(instanceData *pData)
  * file access, which, among others, means the the file wil be opened
  * and any directories in between will be created (based on config, of
  * course). -- rgerhards, 2008-10-22
+ * changed to iRet interface - 2009-03-19
  */
-static void prepareFile(instanceData *pData, uchar *newFileName)
+static rsRetVal
+prepareFile(instanceData *pData, uchar *newFileName)
 {
+	DEFiRet;
 	if(pData->fileType == eTypePIPE) {
 		pData->fd = open((char*) pData->f_fname, O_RDWR|O_NONBLOCK);
 		FINALIZE; /* we are done in this case */
@@ -403,14 +406,14 @@ static void prepareFile(instanceData *pData, uchar *newFileName)
 		pData->fd = -1;
 		/* file does not exist, create it (and eventually parent directories */
 		if(pData->bCreateDirs) {
-			/* we fist need to create parent dirs if they are missing
+			/* We first need to create parent dirs if they are missing.
 			 * We do not report any errors here ourselfs but let the code
 			 * fall through to error handler below.
 			 */
 			if(makeFileParentDirs(newFileName, strlen((char*)newFileName),
 			     pData->fDirCreateMode, pData->dirUID,
 			     pData->dirGID, pData->bFailOnChown) != 0) {
-			     	return; /* we give up */
+			     	ABORT_FINALIZE(RS_RET_ERR); /* we give up */
 			}
 		}
 		/* no matter if we needed to create directories or not, we now try to create
@@ -422,8 +425,7 @@ static void prepareFile(instanceData *pData, uchar *newFileName)
 			/* check and set uid/gid */
 			if(pData->fileUID != (uid_t)-1 || pData->fileGID != (gid_t) -1) {
 				/* we need to set owner/group */
-				if(fchown(pData->fd, pData->fileUID,
-					  pData->fileGID) != 0) {
+				if(fchown(pData->fd, pData->fileUID, pData->fileGID) != 0) {
 					if(pData->bFailOnChown) {
 						int eSave = errno;
 						close(pData->fd);
@@ -438,11 +440,17 @@ static void prepareFile(instanceData *pData, uchar *newFileName)
 		}
 	}
 finalize_it:
-	if((pData->fd) != 0 && isatty(pData->fd)) {
+	/* this was "pData->fd != 0", which I think was a bug. I guess 0 was intended to mean
+	 * non-open file descriptor. Anyhow, I leave this comment for the time being to that if
+	 * problems surface, one at least knows what happened. -- rgerhards, 2009-03-19
+	 */
+	if(pData->fd != -1 && isatty(pData->fd)) {
 		DBGPRINTF("file %d is a tty file\n", pData->fd);
 		pData->fileType = eTypeTTY;
 		untty();
 	}
+
+	RETiRet;
 }
 
 
@@ -461,6 +469,8 @@ static int prepareDynFile(instanceData *pData, uchar *newFileName, unsigned iMsg
 	int i;
 	int iFirstFree;
 	dynaFileCacheEntry **pCache;
+	
+	BEGINfunc
 
 	ASSERT(pData != NULL);
 	ASSERT(newFileName != NULL);
@@ -522,7 +532,7 @@ static int prepareDynFile(instanceData *pData, uchar *newFileName, unsigned iMsg
 	}
 
 	/* Ok, we finally can open the file */
-	prepareFile(pData, newFileName);
+	prepareFile(pData, newFileName); /* ignore exact error, we check fd below */
 
 	/* file is either open now or an error state set */
 	if(pData->fd == -1) {
@@ -546,6 +556,8 @@ static int prepareDynFile(instanceData *pData, uchar *newFileName, unsigned iMsg
 	pData->iCurrElt = iFirstFree;
 	DBGPRINTF("Added new entry %d for file cache, file '%s'.\n", iFirstFree, newFileName);
 
+	ENDfunc
+
 	return 0;
 }
 
@@ -557,6 +569,7 @@ static int prepareDynFile(instanceData *pData, uchar *newFileName, unsigned iMsg
 static rsRetVal writeFile(uchar **ppString, unsigned iMsgOpts, instanceData *pData)
 {
 	off_t actualFileSize;
+	int iLenWritten;
 	DEFiRet;
 
 	ASSERT(pData != NULL);
@@ -566,9 +579,14 @@ static rsRetVal writeFile(uchar **ppString, unsigned iMsgOpts, instanceData *pDa
 	 */
 	if(pData->bDynamicName) {
 		if(prepareDynFile(pData, ppString[1], iMsgOpts) != 0)
-			ABORT_FINALIZE(RS_RET_ERR);
-	} else if(pData->fd == -1) {
-		prepareFile(pData, pData->f_fname);
+			ABORT_FINALIZE(RS_RET_SUSPENDED); /* whatever the failure was, we need to retry */
+	}
+	
+	if(pData->fd == -1) {
+		rsRetVal iRetLocal;
+		iRetLocal = prepareFile(pData, pData->f_fname);
+		if((iRetLocal != RS_RET_OK) || (pData->fd == -1))
+			ABORT_FINALIZE(RS_RET_SUSPENDED); /* whatever the failure was, we need to retry */
 	}
 
 	/* create the message based on format specified */
@@ -604,41 +622,47 @@ again:
 		}
 	}
 
-	if (write(pData->fd, ppString[0], strlen((char*)ppString[0])) < 0) {
+	iLenWritten = write(pData->fd, ppString[0], strlen((char*)ppString[0]));
+//dbgprintf("lenwritten: %d\n", iLenWritten);
+	if(iLenWritten < 0) {
 		int e = errno;
+		char errStr[1024];
+		rs_strerror_r(errno, errStr, sizeof(errStr));
+		DBGPRINTF("log file (%d) write error %d: %s\n", pData->fd, e, errStr);
 
-		/* If a named pipe is full, just ignore it for now
-		   - mrn 24 May 96 */
-		if (pData->fileType == eTypePIPE && e == EAGAIN)
-			ABORT_FINALIZE(RS_RET_OK);
-
-		/* If the filesystem is filled up, just ignore
-		 * it for now and continue writing when possible
-		 * based on patch for sysklogd by Martin Schulze on 2007-05-24
-		 */
-		if (pData->fileType == eTypeFILE && e == ENOSPC)
-			ABORT_FINALIZE(RS_RET_OK);
+		/* If a named pipe is full, we suspend this action for a while */
+		if(pData->fileType == eTypePIPE && e == EAGAIN)
+			ABORT_FINALIZE(RS_RET_SUSPENDED);
 
-		(void) close(pData->fd);
+		close(pData->fd);
+		pData->fd = -1; /* tell that fd is no longer open! */
+		if(pData->bDynamicName && pData->iCurrElt != -1) {
+			/* in this case, we need to invalidate the name in the cache, too
+			 * otherwise, an invalid fd may show up if we had a file name change.
+			 * rgerhards, 2009-03-19
+			 */
+			pData->dynCache[pData->iCurrElt]->fd = -1;
+		}
 		/* Check for EBADF on TTY's due to vhangup()
 		 * Linux uses EIO instead (mrn 12 May 96)
 		 */
 		if((pData->fileType == eTypeTTY || pData->fileType == eTypeCONSOLE)
 #ifdef linux
-			&& e == EIO) {
+			&& e == EIO
 #else
-			&& e == EBADF) {
+			&& e == EBADF
 #endif
+			) {
 			pData->fd = open((char*) pData->f_fname, O_WRONLY|O_APPEND|O_NOCTTY);
 			if (pData->fd < 0) {
-				iRet = RS_RET_DISABLE_ACTION;
+				iRet = RS_RET_SUSPENDED;
 				errmsg.LogError(0, NO_ERRCODE, "%s", pData->f_fname);
 			} else {
 				untty();
 				goto again;
 			}
 		} else {
-			iRet = RS_RET_DISABLE_ACTION;
+			iRet = RS_RET_SUSPENDED;
 			errno = e;
 			errmsg.LogError(0, NO_ERRCODE, "%s", pData->f_fname);
 		}
@@ -790,6 +814,9 @@ CODESTARTparseSelectorAct
 		pData->dirUID = dirUID;
 		pData->dirGID = dirGID;
 
+		/* at this stage, we ignore the return value of prepareFile, this is taken
+		 * care of in later steps. -- rgerhards, 2009-03-19
+		 */
 		prepareFile(pData, pData->f_fname);
 		        
 	  	if(pData->fd < 0 ) {