Merge branch 'omfile-errHandler'

Conflicts:
	ChangeLog

4 files changed, 69 insertions, 28 deletions

diff --git a/ChangeLog b/ChangeLog
index 8c8442e2..d8be1f56 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -8,6 +8,13 @@ Version 4.1.6  [DEVEL] (rgerhards), 2009-03-??
   issue with the function call VM instruction set design.
 - implemented the strlen() RainerScript function
 - added a template output module
+- 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_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/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 ) {