multiple_inprogress_backups1.patch

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 5b6a230..cc4d46e 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -338,7 +338,8 @@ typedef struct XLogCtlInsert
 	XLogPageHeader currpage;	/* points to header of block in cache */
 	char	   *currpos;		/* current insertion point in cache */
 	XLogRecPtr	RedoRecPtr;		/* current redo point for insertions */
-	bool		forcePageWrites;	/* forcing full-page writes for PITR? */
+	int			forcePageWrites;	/* forcing full-page writes for PITR? */
+	bool		exclusiveBackup;	/* a backup was started with pg_start_backup() */
 } XLogCtlInsert;
 
 /*
@@ -8313,7 +8314,7 @@ pg_start_backup(PG_FUNCTION_ARGS)
 
 	backupidstr = text_to_cstring(backupid);
 
-	startpoint = do_pg_start_backup(backupidstr, fast);
+	startpoint = do_pg_start_backup(backupidstr, fast, true, NULL);
 
 	snprintf(startxlogstr, sizeof(startxlogstr), "%X/%X",
 			 startpoint.xlogid, startpoint.xrecoff);
@@ -8321,7 +8322,7 @@ pg_start_backup(PG_FUNCTION_ARGS)
 }
 
 XLogRecPtr
-do_pg_start_backup(const char *backupidstr, bool fast)
+do_pg_start_backup(const char *backupidstr, bool fast, bool exclusive, char **labelfile)
 {
 	XLogRecPtr	checkpointloc;
 	XLogRecPtr	startpoint;
@@ -8332,6 +8333,7 @@ do_pg_start_backup(const char *backupidstr, bool fast)
 	uint32		_logSeg;
 	struct stat stat_buf;
 	FILE	   *fp;
+	StringInfoData labelfbuf;
 
 	if (!superuser() && !is_authenticated_user_replication_role())
 		ereport(ERROR,
@@ -8368,15 +8370,19 @@ do_pg_start_backup(const char *backupidstr, bool fast)
 	 * ensure adequate interlocking against XLogInsert().
 	 */
 	LWLockAcquire(WALInsertLock, LW_EXCLUSIVE);
-	if (XLogCtl->Insert.forcePageWrites)
+	if (exclusive)
 	{
-		LWLockRelease(WALInsertLock);
-		ereport(ERROR,
-				(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
-				 errmsg("a backup is already in progress"),
-				 errhint("Run pg_stop_backup() and try again.")));
+		if (XLogCtl->Insert.exclusiveBackup)
+		{
+			LWLockRelease(WALInsertLock);
+			ereport(ERROR,
+					(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
+					 errmsg("a backup is already in progress"),
+					 errhint("Run pg_stop_backup() and try again.")));
+		}
+		XLogCtl->Insert.exclusiveBackup = true;
 	}
-	XLogCtl->Insert.forcePageWrites = true;
+	XLogCtl->Insert.forcePageWrites++;
 	LWLockRelease(WALInsertLock);
 
 	/*
@@ -8393,7 +8399,7 @@ do_pg_start_backup(const char *backupidstr, bool fast)
 	RequestXLogSwitch();
 
 	/* Ensure we release forcePageWrites if fail below */
-	PG_ENSURE_ERROR_CLEANUP(pg_start_backup_callback, (Datum) 0);
+	PG_ENSURE_ERROR_CLEANUP(pg_start_backup_callback, (Datum) BoolGetDatum(exclusive));
 	{
 		/*
 		 * Force a CHECKPOINT.	Aside from being necessary to prevent torn
@@ -8420,54 +8426,67 @@ do_pg_start_backup(const char *backupidstr, bool fast)
 		XLByteToSeg(startpoint, _logId, _logSeg);
 		XLogFileName(xlogfilename, ThisTimeLineID, _logId, _logSeg);
 
+		/*
+		 * Construct backup label file 
+		 */
+		initStringInfo(&labelfbuf);
+
 		/* Use the log timezone here, not the session timezone */
 		stamp_time = (pg_time_t) time(NULL);
 		pg_strftime(strfbuf, sizeof(strfbuf),
 					"%Y-%m-%d %H:%M:%S %Z",
 					pg_localtime(&stamp_time, log_timezone));
+		appendStringInfo(&labelfbuf, "START WAL LOCATION: %X/%X (file %s)\n",
+						 startpoint.xlogid, startpoint.xrecoff, xlogfilename);
+		appendStringInfo(&labelfbuf, "CHECKPOINT LOCATION: %X/%X\n",
+						 checkpointloc.xlogid, checkpointloc.xrecoff);
+		appendStringInfo(&labelfbuf, "START TIME: %s\n", strfbuf);
+		appendStringInfo(&labelfbuf, "LABEL: %s\n", backupidstr);
 
 		/*
-		 * Check for existing backup label --- implies a backup is already
-		 * running.  (XXX given that we checked forcePageWrites above, maybe
-		 * it would be OK to just unlink any such label file?)
+		 * Okay, write the file, or return it to caller.
 		 */
-		if (stat(BACKUP_LABEL_FILE, &stat_buf) != 0)
+		if (exclusive)
 		{
-			if (errno != ENOENT)
+			/*
+			 * Check for existing backup label --- implies a backup is already
+			 * running.  (XXX given that we checked exclusiveBackup above, maybe
+			 * it would be OK to just unlink any such label file?)
+			 */
+			if (stat(BACKUP_LABEL_FILE, &stat_buf) != 0)
+			{
+				if (errno != ENOENT)
+					ereport(ERROR,
+							(errcode_for_file_access(),
+							 errmsg("could not stat file \"%s\": %m",
+									BACKUP_LABEL_FILE)));
+			}
+			else
+				ereport(ERROR,
+						(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
+						 errmsg("a backup is already in progress"),
+						 errhint("If you're sure there is no backup in progress, remove file \"%s\" and try again.",
+								 BACKUP_LABEL_FILE)));
+
+			fp = AllocateFile(BACKUP_LABEL_FILE, "w");
+
+			if (!fp)
 				ereport(ERROR,
 						(errcode_for_file_access(),
-						 errmsg("could not stat file \"%s\": %m",
+						 errmsg("could not create file \"%s\": %m",
 								BACKUP_LABEL_FILE)));
+			fwrite(labelfbuf.data, labelfbuf.len, 1, fp);
+			if (fflush(fp) || ferror(fp) || FreeFile(fp))
+				ereport(ERROR,
+						(errcode_for_file_access(),
+						 errmsg("could not write file \"%s\": %m",
+								BACKUP_LABEL_FILE)));
+			pfree(labelfbuf.data);
 		}
 		else
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
-					 errmsg("a backup is already in progress"),
-					 errhint("If you're sure there is no backup in progress, remove file \"%s\" and try again.",
-							 BACKUP_LABEL_FILE)));
-
-		/*
-		 * Okay, write the file
-		 */
-		fp = AllocateFile(BACKUP_LABEL_FILE, "w");
-		if (!fp)
-			ereport(ERROR,
-					(errcode_for_file_access(),
-					 errmsg("could not create file \"%s\": %m",
-							BACKUP_LABEL_FILE)));
-		fprintf(fp, "START WAL LOCATION: %X/%X (file %s)\n",
-				startpoint.xlogid, startpoint.xrecoff, xlogfilename);
-		fprintf(fp, "CHECKPOINT LOCATION: %X/%X\n",
-				checkpointloc.xlogid, checkpointloc.xrecoff);
-		fprintf(fp, "START TIME: %s\n", strfbuf);
-		fprintf(fp, "LABEL: %s\n", backupidstr);
-		if (fflush(fp) || ferror(fp) || FreeFile(fp))
-			ereport(ERROR,
-					(errcode_for_file_access(),
-					 errmsg("could not write file \"%s\": %m",
-							BACKUP_LABEL_FILE)));
+			*labelfile = labelfbuf.data;
 	}
-	PG_END_ENSURE_ERROR_CLEANUP(pg_start_backup_callback, (Datum) 0);
+	PG_END_ENSURE_ERROR_CLEANUP(pg_start_backup_callback, (Datum) BoolGetDatum(exclusive));
 
 	/*
 	 * We're done.  As a convenience, return the starting WAL location.
@@ -8479,8 +8498,15 @@ do_pg_start_backup(const char *backupidstr, bool fast)
 static void
 pg_start_backup_callback(int code, Datum arg)
 {
-	/* Turn off forcePageWrites on failure */
+	bool exclusive = DatumGetBool(arg);
+
+	/* Decrement forcePageWrites on failure */
 	LWLockAcquire(WALInsertLock, LW_EXCLUSIVE);
+	if (exclusive)
+	{
+		Assert(XLogCtl->Insert.exclusiveBackup);
+		XLogCtl->Insert.exclusiveBackup = false;
+	}
 	XLogCtl->Insert.forcePageWrites = false;
 	LWLockRelease(WALInsertLock);
 }
@@ -8504,16 +8530,20 @@ pg_stop_backup(PG_FUNCTION_ARGS)
 	XLogRecPtr	stoppoint;
 	char		stopxlogstr[MAXFNAMELEN];
 
-	stoppoint = do_pg_stop_backup();
+	stoppoint = do_pg_stop_backup(NULL);
 
 	snprintf(stopxlogstr, sizeof(stopxlogstr), "%X/%X",
 			 stoppoint.xlogid, stoppoint.xrecoff);
 	PG_RETURN_TEXT_P(cstring_to_text(stopxlogstr));
 }
 
+/*
+ * If labelfile is NULL, this stops an exclusive backup.
+ */
 XLogRecPtr
-do_pg_stop_backup(void)
+do_pg_stop_backup(char *labelfile)
 {
+	bool		exclusive = (labelfile == NULL);
 	XLogRecPtr	startpoint;
 	XLogRecPtr	stoppoint;
 	XLogRecData rdata;
@@ -8529,10 +8559,10 @@ do_pg_stop_backup(void)
 	FILE	   *lfp;
 	FILE	   *fp;
 	char		ch;
-	int			ich;
 	int			seconds_before_warning;
 	int			waits = 0;
 	bool		reported_waiting = false;
+	char	   *remaining;
 
 	if (!superuser() && !is_authenticated_user_replication_role())
 		ereport(ERROR,
@@ -8555,35 +8585,71 @@ do_pg_stop_backup(void)
 	 * OK to clear forcePageWrites
 	 */
 	LWLockAcquire(WALInsertLock, LW_EXCLUSIVE);
-	XLogCtl->Insert.forcePageWrites = false;
+	if (exclusive)
+	{
+		if (XLogCtl->Insert.exclusiveBackup)
+		{
+			XLogCtl->Insert.forcePageWrites--;
+			XLogCtl->Insert.exclusiveBackup = false;
+		}
+	}
+	XLogCtl->Insert.forcePageWrites--;
 	LWLockRelease(WALInsertLock);
 
-	/*
-	 * Open the existing label file
-	 */
-	lfp = AllocateFile(BACKUP_LABEL_FILE, "r");
-	if (!lfp)
+	if (!labelfile)
 	{
-		if (errno != ENOENT)
+		int len;
+
+		/*
+		 * Read the existing label file
+		 */
+		lfp = AllocateFile(BACKUP_LABEL_FILE, "r");
+		if (!lfp)
+		{
+			if (errno != ENOENT)
+				ereport(ERROR,
+						(errcode_for_file_access(),
+						 errmsg("could not read file \"%s\": %m",
+								labelfile)));
+			ereport(ERROR,
+					(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
+					 errmsg("a backup is not in progress")));
+		}
+
+		/* The label file should be small */
+		labelfile = palloc(1024);
+		len = fread(labelfile, 1, 1023, lfp);
+		if (len >= 1023)
+			ereport(ERROR,
+					(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
+					 errmsg("backup_label file \"%s\" is too long", BACKUP_LABEL_FILE)));
+
+		/*
+		 * Close and remove the backup label file
+		 */
+		if (ferror(lfp) || FreeFile(lfp))
 			ereport(ERROR,
 					(errcode_for_file_access(),
 					 errmsg("could not read file \"%s\": %m",
 							BACKUP_LABEL_FILE)));
-		ereport(ERROR,
-				(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
-				 errmsg("a backup is not in progress")));
+		if (unlink(BACKUP_LABEL_FILE) != 0)
+			ereport(ERROR,
+					(errcode_for_file_access(),
+					 errmsg("could not remove file \"%s\": %m",
+							BACKUP_LABEL_FILE)));
 	}
 
 	/*
 	 * Read and parse the START WAL LOCATION line (this code is pretty crude,
 	 * but we are not expecting any variability in the file format).
 	 */
-	if (fscanf(lfp, "START WAL LOCATION: %X/%X (file %24s)%c",
+	if (sscanf(labelfile, "START WAL LOCATION: %X/%X (file %24s)%c",
 			   &startpoint.xlogid, &startpoint.xrecoff, startxlogfilename,
 			   &ch) != 4 || ch != '\n')
 		ereport(ERROR,
 				(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
 				 errmsg("invalid data in file \"%s\"", BACKUP_LABEL_FILE)));
+	remaining = strchr(labelfile, '\n') + 1;
 
 	/*
 	 * Write the backup-end xlog record
@@ -8626,8 +8692,7 @@ do_pg_stop_backup(void)
 	fprintf(fp, "STOP WAL LOCATION: %X/%X (file %s)\n",
 			stoppoint.xlogid, stoppoint.xrecoff, stopxlogfilename);
 	/* transfer remaining lines from label to history file */
-	while ((ich = fgetc(lfp)) != EOF)
-		fputc(ich, fp);
+	fwrite(remaining, strlen(remaining), 1, fp);
 	fprintf(fp, "STOP TIME: %s\n", strfbuf);
 	if (fflush(fp) || ferror(fp) || FreeFile(fp))
 		ereport(ERROR,
@@ -8636,20 +8701,6 @@ do_pg_stop_backup(void)
 						histfilepath)));
 
 	/*
-	 * Close and remove the backup label file
-	 */
-	if (ferror(lfp) || FreeFile(lfp))
-		ereport(ERROR,
-				(errcode_for_file_access(),
-				 errmsg("could not read file \"%s\": %m",
-						BACKUP_LABEL_FILE)));
-	if (unlink(BACKUP_LABEL_FILE) != 0)
-		ereport(ERROR,
-				(errcode_for_file_access(),
-				 errmsg("could not remove file \"%s\": %m",
-						BACKUP_LABEL_FILE)));
-
-	/*
 	 * Clean out any no-longer-needed history files.  As a side effect, this
 	 * will post a .ready file for the newly created history file, notifying
 	 * the archiver that history file may be archived immediately.
@@ -8730,9 +8781,13 @@ do_pg_stop_backup(void)
 /*
  * do_pg_abort_backup: abort a running backup
  *
- * This does just the most basic steps of pg_stop_backup(), by taking the
+ * This does just the most basic steps of do_pg_stop_backup(), by taking the
  * system out of backup mode, thus making it a lot more safe to call from
  * an error handler.
+ *
+ * NB: This is only for aborting a non-exclusive backup that doesn't write
+ * backup_label. A backup started with pg_stop_backup() needs to be finished
+ * with pg_stop_backup().
  */
 void
 do_pg_abort_backup(void)
@@ -8741,17 +8796,14 @@ do_pg_abort_backup(void)
 	 * OK to clear forcePageWrites
 	 */
 	LWLockAcquire(WALInsertLock, LW_EXCLUSIVE);
-	XLogCtl->Insert.forcePageWrites = false;
-	LWLockRelease(WALInsertLock);
-
 	/*
-	 * Remove backup label file
+	 * The caller should take care to not call us only if a backup, but
+	 * be extra paranoid to avoid wrapping forcePageWrites around if the
+	 * caller screws up.
 	 */
-	if (unlink(BACKUP_LABEL_FILE) != 0)
-		ereport(ERROR,
-				(errcode_for_file_access(),
-				 errmsg("could not remove file \"%s\": %m",
-						BACKUP_LABEL_FILE)));
+	if (XLogCtl->Insert.forcePageWrites > 0)
+		XLogCtl->Insert.forcePageWrites--;
+	LWLockRelease(WALInsertLock);
 }
 
 /*
diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c
index de5fa22..c31b091 100644
--- a/src/backend/replication/basebackup.c
+++ b/src/backend/replication/basebackup.c
@@ -31,13 +31,17 @@
 #include "utils/memutils.h"
 #include "utils/ps_status.h"
 
+/* XXX: copied from xlog.c */
+#define BACKUP_LABEL_FILE		"backup_label"
+
 static int64 sendDir(char *path, int basepathlen, bool sizeonly);
-static void sendFile(char *path, int basepathlen, struct stat * statbuf);
+static void sendFile(char *readfilename, char *tarfilename,
+		 struct stat * statbuf);
 static void _tarWriteHeader(char *filename, char *linktarget,
 				struct stat * statbuf);
 static void send_int8_string(StringInfoData *buf, int64 intval);
 static void SendBackupHeader(List *tablespaces);
-static void SendBackupDirectory(char *location, char *spcoid);
+static void SendBackupDirectories(List *tablespaces, char *labelfilepath);
 static void base_backup_cleanup(int code, Datum arg);
 
 typedef struct
@@ -75,6 +79,7 @@ SendBaseBackup(const char *options)
 	tablespaceinfo *ti;
 	MemoryContext backup_context;
 	MemoryContext old_context;
+	char	   *labelfile;
 
 	backup_context = AllocSetContextCreate(CurrentMemoryContext,
 										   "Streaming base backup context",
@@ -145,26 +150,19 @@ SendBaseBackup(const char *options)
 	}
 	FreeDir(dir);
 
-	do_pg_start_backup(backup_label, true);
+	do_pg_start_backup(backup_label, true, false, &labelfile);
 
 	PG_ENSURE_ERROR_CLEANUP(base_backup_cleanup, (Datum) 0);
 	{
-		ListCell   *lc;
-
 		/* Send tablespace header */
 		SendBackupHeader(tablespaces);
 
-		/* Send off our tablespaces one by one */
-		foreach(lc, tablespaces)
-		{
-			ti = (tablespaceinfo *) lfirst(lc);
-
-			SendBackupDirectory(ti->path, ti->oid);
-		}
+		/* Send the tars */
+		SendBackupDirectories(tablespaces, labelfile);
 	}
 	PG_END_ENSURE_ERROR_CLEANUP(base_backup_cleanup, (Datum) 0);
 
-	do_pg_stop_backup();
+	do_pg_stop_backup(labelfile);
 
 	MemoryContextSwitchTo(old_context);
 	MemoryContextDelete(backup_context);
@@ -250,20 +248,65 @@ SendBackupHeader(List *tablespaces)
 }
 
 static void
-SendBackupDirectory(char *location, char *spcoid)
+SendBackupDirectories(List *tablespaces, char *labelfile)
 {
 	StringInfoData buf;
+	ListCell *lc;
 
-	/* Send CopyOutResponse message */
-	pq_beginmessage(&buf, 'H');
-	pq_sendbyte(&buf, 0);		/* overall format */
-	pq_sendint(&buf, 0, 2);		/* natts */
-	pq_endmessage(&buf);
+	/* Send off our tablespaces one by one */
+	foreach(lc, tablespaces)
+	{
+		tablespaceinfo *ti = (tablespaceinfo *) lfirst(lc);
+		char *location = ti->path;
+
+		/* Send CopyOutResponse message */
+		pq_beginmessage(&buf, 'H');
+		pq_sendbyte(&buf, 0);		/* overall format */
+		pq_sendint(&buf, 0, 2);		/* natts */
+		pq_endmessage(&buf);
+
+		/*
+		 * tar up the data directory if NULL, otherwise the tablespace
+		 * In the main tar, also include the backup_label.
+		 */
+		if (location == NULL)
+		{
+			struct stat statbuf;
+			int pad, len;
 
-	/* tar up the data directory if NULL, otherwise the tablespace */
-	sendDir(location == NULL ? "." : location,
-			location == NULL ? 1 : strlen(location),
-			false);
+			len = strlen(labelfile);
+
+			/*
+			 * We use PG_VERSION as a template for the uid and gid and mode
+			 * to use for the backup_label file. XXX: would "." be better?
+			 * Or something else?
+			 */
+			if (lstat("PG_VERSION", &statbuf) != 0)
+				ereport(ERROR,
+						(errcode(errcode_for_file_access()),
+						 errmsg("could not stat file \"%s\": %m",
+								"PG_VERSION")));
+			statbuf.st_mtime = time(NULL);
+			statbuf.st_size = len;
+
+			_tarWriteHeader(BACKUP_LABEL_FILE, NULL, &statbuf);
+			/* Send the contents as a CopyData message */
+			pq_putmessage('d', labelfile, len);
+
+			/* Pad to 512 byte boundary, per tar format requirements */
+			pad = ((len + 511) & ~511) - len;
+			if (pad > 0)
+			{
+				char buf[512];
+				MemSet(buf, 0, pad);
+				pq_putmessage('d', buf, pad);
+			}
+
+			location = ".";
+		}
+
+		sendDir(location, strlen(location),	false);
+	}
 
 	/* Send CopyDone message */
 	pq_putemptymessage('c');
@@ -360,7 +403,7 @@ sendDir(char *path, int basepathlen, bool sizeonly)
 			/* Add size, rounded up to 512byte block */
 			size += ((statbuf.st_size + 511) & ~511);
 			if (!sizeonly)
-				sendFile(pathbuf, basepathlen, &statbuf);
+				sendFile(pathbuf, pathbuf + basepathlen + 1, &statbuf);
 			size += 512;		/* Size of the header of the file */
 		}
 		else
@@ -418,7 +461,7 @@ _tarChecksum(char *header)
 
 /* Given the member, write the TAR header & send the file */
 static void
-sendFile(char *filename, int basepathlen, struct stat * statbuf)
+sendFile(char *readfilename, char *tarfilename, struct stat * statbuf)
 {
 	FILE	   *fp;
 	char		buf[32768];
@@ -426,11 +469,11 @@ sendFile(char *filename, int basepathlen, struct stat * statbuf)
 	pgoff_t		len = 0;
 	size_t		pad;
 
-	fp = AllocateFile(filename, "rb");
+	fp = AllocateFile(readfilename, "rb");
 	if (fp == NULL)
 		ereport(ERROR,
 				(errcode(errcode_for_file_access()),
-				 errmsg("could not open file \"%s\": %m", filename)));
+				 errmsg("could not open file \"%s\": %m", readfilename)));
 
 	/*
 	 * Some compilers will throw a warning knowing this test can never be true
@@ -439,9 +482,9 @@ sendFile(char *filename, int basepathlen, struct stat * statbuf)
 	if (statbuf->st_size > MAX_TAR_MEMBER_FILELEN)
 		ereport(ERROR,
 				(errmsg("archive member \"%s\" too large for tar format",
-						filename)));
+						tarfilename)));
 
-	_tarWriteHeader(filename + basepathlen + 1, NULL, statbuf);
+	_tarWriteHeader(tarfilename, NULL, statbuf);
 
 	while ((cnt = fread(buf, 1, Min(sizeof(buf), statbuf->st_size - len), fp)) > 0)
 	{
diff --git a/src/include/access/xlog.h b/src/include/access/xlog.h
index 74d3427..babaaa6 100644
--- a/src/include/access/xlog.h
+++ b/src/include/access/xlog.h
@@ -312,8 +312,8 @@ extern void HandleStartupProcInterrupts(void);
 extern void StartupProcessMain(void);
 extern void WakeupRecovery(void);
 
-extern XLogRecPtr do_pg_start_backup(const char *backupidstr, bool fast);
-extern XLogRecPtr do_pg_stop_backup(void);
+extern XLogRecPtr do_pg_start_backup(const char *backupidstr, bool fast, bool exclusive, char **labelfile);
+extern XLogRecPtr do_pg_stop_backup(char *labelfile);
 extern void do_pg_abort_backup(void);
 
 #endif   /* XLOG_H */