writeback-v1.patch

diff --git a/configure b/configure
index af4f9a3..3da0771 100755
--- a/configure
+++ b/configure
@@ -19263,7 +19263,8 @@ fi
 
 
 
-for ac_func in cbrt dlopen fcvt fdatasync getifaddrs getpeerucred getrlimit memmove poll pstat readlink setproctitle setsid sigprocmask symlink sysconf towlower utime utimes waitpid wcstombs wcstombs_l
+
+for ac_func in cbrt dlopen fcvt fdatasync sync_file_range getifaddrs getpeerucred getrlimit memmove poll pstat readlink setproctitle setsid sigprocmask symlink sysconf towlower utime utimes waitpid wcstombs wcstombs_l
 do
 as_ac_var=`$as_echo "ac_cv_func_$ac_func" | $as_tr_sh`
 { $as_echo "$as_me:$LINENO: checking for $ac_func" >&5
diff --git a/configure.in b/configure.in
index 9cad436..2d1608d 100644
--- a/configure.in
+++ b/configure.in
@@ -1216,7 +1216,7 @@ PGAC_VAR_INT_TIMEZONE
 AC_FUNC_ACCEPT_ARGTYPES
 PGAC_FUNC_GETTIMEOFDAY_1ARG
 
-AC_CHECK_FUNCS([cbrt dlopen fcvt fdatasync getifaddrs getpeerucred getrlimit memmove poll pstat readlink setproctitle setsid sigprocmask symlink sysconf towlower utime utimes waitpid wcstombs wcstombs_l])
+AC_CHECK_FUNCS([cbrt dlopen fcvt fdatasync sync_file_range getifaddrs getpeerucred getrlimit memmove poll pstat readlink setproctitle setsid sigprocmask symlink sysconf towlower utime utimes waitpid wcstombs wcstombs_l])
 
 AC_REPLACE_FUNCS(fseeko)
 case $host_os in
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 8e65962..f6e20b8 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -70,6 +70,11 @@
 
 /* User-settable parameters */
 int			CheckPointSegments = 3;
+#ifdef USE_WRITEBACK
+bool		checkpoint_writeback = true;
+#else
+bool		checkpoint_writeback = false;
+#endif
 int			wal_keep_segments = 0;
 int			XLOGbuffers = -1;
 int			XLogArchiveTimeout = 0;
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 8f68bcc..0402f2c 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -1766,6 +1766,7 @@ CheckPointBuffers(int flags)
 {
 	TRACE_POSTGRESQL_BUFFER_CHECKPOINT_START(flags);
 	CheckpointStats.ckpt_write_t = GetCurrentTimestamp();
+	smgrwriteback();
 	BufferSync(flags);
 	CheckpointStats.ckpt_sync_t = GetCurrentTimestamp();
 	TRACE_POSTGRESQL_BUFFER_CHECKPOINT_SYNC_START();
diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c
index 43bc43a..e0344ce 100644
--- a/src/backend/storage/file/fd.c
+++ b/src/backend/storage/file/fd.c
@@ -347,6 +347,21 @@ pg_flush_data(int fd, off_t offset, off_t amount)
 #endif
 }
 
+/*
+ * pg_writeback --- advise OS that the data described won't be needed soon
+ *
+ * Treat as noop if no OS support is available.
+ */
+int
+pg_writeback(int fd)
+{
+#if defined(HAVE_SYNC_FILE_RANGE)
+	return sync_file_range(fd, 0, 0, SYNC_FILE_RANGE_WRITE);
+#else
+	return 0;
+#endif
+}
+
 
 /*
  * InitFileAccess --- initialize this module during backend startup
@@ -1336,6 +1351,23 @@ retry:
 }
 
 int
+FileWriteback(File file)
+{
+	int			returnCode;
+
+	Assert(FileIsValid(file));
+
+	DO_DB(elog(LOG, "FileWriteback: %d (%s)",
+			   file, VfdCache[file].fileName));
+
+	returnCode = FileAccess(file);
+	if (returnCode < 0)
+		return returnCode;
+
+	return pg_writeback(VfdCache[file].fd);
+}
+
+int
 FileSync(File file)
 {
 	int			returnCode;
diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c
index bfc9f06..d85e9bf 100644
--- a/src/backend/storage/smgr/md.c
+++ b/src/backend/storage/smgr/md.c
@@ -931,6 +931,95 @@ mdimmedsync(SMgrRelation reln, ForkNumber forknum)
 }
 
 /*
+ *	mdwriteback() -- Initiate writeback of data to stable storage.
+ */
+void
+mdwriteback(void)
+{
+#ifdef USE_WRITEBACK
+	HASH_SEQ_STATUS hstat;
+	PendingOperationEntry *entry;
+	int			absorb_counter;
+
+	/*
+	 * This is only called during checkpoints, and checkpoints should only
+	 * occur in processes that have created a pendingOpsTable.
+	 */
+	if (!pendingOpsTable)
+		elog(ERROR, "cannot sync without a pendingOpsTable");
+
+	/* Scan the hashtable for fsync requests. */
+	absorb_counter = FSYNCS_PER_ABSORB;
+	hash_seq_init(&hstat, pendingOpsTable);
+	while ((entry = (PendingOperationEntry *) hash_seq_search(&hstat)) != NULL)
+	{
+		SMgrRelation reln;
+		MdfdVec    *seg;
+
+		/*
+		 * If writeback is off then we don't have to bother opening the file at
+		 * all.  (We delay checking until this point so that changing this on
+		 * the fly behaves sensibly.)
+		 */
+		if (!checkpoint_writeback)
+			break;
+
+		/* Skip canceled entries. */
+		if (entry->canceled)
+			continue;
+
+		/* Absorb fsync requests so that the queue doesn't overflow. */
+		if (--absorb_counter <= 0)
+		{
+			AbsorbFsyncRequests();
+			absorb_counter = FSYNCS_PER_ABSORB;
+		}
+
+		/*
+		 * Find or create an smgr hash entry for this relation.  See
+		 * mdsync() a full explanation of why we go back through the smgr
+		 * layer here.
+		 */
+		reln = smgropen(entry->tag.rnode.node, entry->tag.rnode.backend);
+
+		/*
+		 * It is possible that the relation has been dropped or truncated
+		 * since the fsync request was entered.  Since writeback is just a
+		 * performance optimization, there's no harm in just skipping the
+		 * segment if it turns out not to exist any more.
+		 */
+		seg = _mdfd_getseg(reln, entry->tag.forknum,
+						   entry->tag.segno * ((BlockNumber) RELSEG_SIZE),
+						   false, EXTENSION_RETURN_NULL);
+		if (seg == NULL)
+			continue;
+
+		/*
+		 * Try to write it back.
+		 */
+		errno = FileWriteback(seg->mdfd_vfd);
+
+		/*
+		 * Since this is just a hint to the OS to get the file on disk,
+		 * there's no great harm if it fails.  Of course, failure here may be
+		 * a sign that the eventual fsync will also fail, but that's mdsync's
+		 * problem, not ours.
+		 */
+		if (errno != 0 && !FILE_POSSIBLY_DELETED(errno))
+		{
+			char	   *path;
+
+			path = _mdfd_segpath(reln, entry->tag.forknum,
+								 entry->tag.segno);
+			ereport(LOG,
+					(errcode_for_file_access(),
+				   errmsg("could not write back file \"%s\": %m", path)));
+		}
+	}
+#endif
+}
+
+/*
  *	mdsync() -- Sync previous writes to stable storage.
  */
 void
diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c
index 5f87543..ccd952a 100644
--- a/src/backend/storage/smgr/smgr.c
+++ b/src/backend/storage/smgr/smgr.c
@@ -58,6 +58,7 @@ typedef struct f_smgr
 											  BlockNumber nblocks);
 	void		(*smgr_immedsync) (SMgrRelation reln, ForkNumber forknum);
 	void		(*smgr_pre_ckpt) (void);		/* may be NULL */
+	void		(*smgr_writeback) (void);	/* may be NULL */
 	void		(*smgr_sync) (void);	/* may be NULL */
 	void		(*smgr_post_ckpt) (void);		/* may be NULL */
 } f_smgr;
@@ -67,7 +68,7 @@ static const f_smgr smgrsw[] = {
 	/* magnetic disk */
 	{mdinit, NULL, mdclose, mdcreate, mdexists, mdunlink, mdextend,
 		mdprefetch, mdread, mdwrite, mdnblocks, mdtruncate, mdimmedsync,
-		mdpreckpt, mdsync, mdpostckpt
+		mdpreckpt, mdwriteback, mdsync, mdpostckpt
 	}
 };
 
@@ -533,6 +534,21 @@ smgrpreckpt(void)
 }
 
 /*
+ *	smgrwriteback() -- Initial writeback during checkpoint.
+ */
+void
+smgrwriteback(void)
+{
+	int			i;
+
+	for (i = 0; i < NSmgr; i++)
+	{
+		if (smgrsw[i].smgr_writeback)
+			(*(smgrsw[i].smgr_writeback)) ();
+	}
+}
+
+/*
  *	smgrsync() -- Sync files to disk during checkpoint.
  */
 void
diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c
index 5c910dd..c1eec9c 100644
--- a/src/backend/utils/misc/guc.c
+++ b/src/backend/utils/misc/guc.c
@@ -182,6 +182,7 @@ static bool check_phony_autocommit(bool *newval, void **extra, GucSource source)
 static bool check_debug_assertions(bool *newval, void **extra, GucSource source);
 static bool check_bonjour(bool *newval, void **extra, GucSource source);
 static bool check_ssl(bool *newval, void **extra, GucSource source);
+static bool check_checkpoint_writeback(bool *newval, void **extra, GucSource source);
 static bool check_stage_log_stats(bool *newval, void **extra, GucSource source);
 static bool check_log_stats(bool *newval, void **extra, GucSource source);
 static bool check_canonical_path(char **newval, void **extra, GucSource source);
@@ -816,6 +817,25 @@ static struct config_bool ConfigureNamesBool[] =
 		NULL, NULL, NULL
 	},
 	{
+		{"checkpoint_writeback",
+#ifdef USE_WRITEBACK
+			PGC_SIGHUP,
+#else
+			PGC_INTERNAL,
+#endif
+			WAL_CHECKPOINTS,
+			gettext_noop("Initiates OS writeback of dirty data at checkpoint start."),
+			gettext_noop("For RAID arrays, this should be approximately the number of drive spindles in the array.")
+		},
+		&checkpoint_writeback,
+#ifdef USE_WRITEBACK
+		true,
+#else
+		false,
+#endif
+		check_checkpoint_writeback, NULL, NULL
+	},
+	{
 		{"zero_damaged_pages", PGC_SUSET, DEVELOPER_OPTIONS,
 			gettext_noop("Continues processing past damaged page headers."),
 			gettext_noop("Detection of a damaged page header normally causes PostgreSQL to "
@@ -8374,6 +8394,19 @@ check_bonjour(bool *newval, void **extra, GucSource source)
 }
 
 static bool
+check_checkpoint_writeback(bool *newval, void **extra, GucSource source)
+{
+#ifndef USE_WRITEBACK
+	if (*newval)
+	{
+		GUC_check_errmsg("writeback is not supported by this build");
+		return false;
+	}
+#endif
+	return true;
+}
+
+static bool
 check_ssl(bool *newval, void **extra, GucSource source)
 {
 #ifndef USE_SSL
diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample
index 315db46..bfaf0bc 100644
--- a/src/backend/utils/misc/postgresql.conf.sample
+++ b/src/backend/utils/misc/postgresql.conf.sample
@@ -180,6 +180,7 @@
 #checkpoint_timeout = 5min		# range 30s-1h
 #checkpoint_completion_target = 0.5	# checkpoint target duration, 0.0 - 1.0
 #checkpoint_warning = 30s		# 0 disables
+#checkpoint_writeback = true	# false if async writeback not supported
 
 # - Archiving -
 
diff --git a/src/include/access/xlog.h b/src/include/access/xlog.h
index 93622c4..a3fa124 100644
--- a/src/include/access/xlog.h
+++ b/src/include/access/xlog.h
@@ -186,6 +186,7 @@ extern bool reachedConsistency;
 
 /* these variables are GUC parameters related to XLOG */
 extern int	CheckPointSegments;
+extern bool	checkpoint_writeback;
 extern int	wal_keep_segments;
 extern int	XLOGbuffers;
 extern int	XLogArchiveTimeout;
diff --git a/src/include/pg_config.h.in b/src/include/pg_config.h.in
index db84f49..adee3e4 100644
--- a/src/include/pg_config.h.in
+++ b/src/include/pg_config.h.in
@@ -544,6 +544,9 @@
 /* Define to 1 if you have the `symlink' function. */
 #undef HAVE_SYMLINK
 
+/* Define to 1 if you have the `sync_file_range' function. */
+#undef HAVE_SYNC_FILE_RANGE
+
 /* Define to 1 if you have the `sysconf' function. */
 #undef HAVE_SYSCONF
 
diff --git a/src/include/pg_config_manual.h b/src/include/pg_config_manual.h
index ac45ee6..05cf2f2 100644
--- a/src/include/pg_config_manual.h
+++ b/src/include/pg_config_manual.h
@@ -129,14 +129,23 @@
 
 /*
  * USE_PREFETCH code should be compiled only if we have a way to implement
- * prefetching.  (This is decoupled from USE_POSIX_FADVISE because there
- * might in future be support for alternative low-level prefetch APIs.)
+ * prefetching.  (This is decoupled from HAVE_FILE_SYNC_RANGE because there
+ * might in future be support for alternative low-level writeback APIs.)
  */
 #ifdef USE_POSIX_FADVISE
 #define USE_PREFETCH
 #endif
 
 /*
+ * USE_WRITEBACK code should be compiled only if we have a way to implement
+ * writeback.  (This is decoupled from HAVE_SYNC_FILE_RANGE because there
+ * might in future be support for alternative low-level writeback APIs.)
+ */
+#ifdef HAVE_SYNC_FILE_RANGE
+#define USE_WRITEBACK
+#endif
+
+/*
  * This is the default directory in which AF_UNIX socket files are
  * placed.	Caution: changing this risks breaking your existing client
  * applications, which are likely to continue to look in the old
diff --git a/src/include/storage/fd.h b/src/include/storage/fd.h
index 22e7fe8..dbf74c0 100644
--- a/src/include/storage/fd.h
+++ b/src/include/storage/fd.h
@@ -66,6 +66,7 @@ extern void FileClose(File file);
 extern int	FilePrefetch(File file, off_t offset, int amount);
 extern int	FileRead(File file, char *buffer, int amount);
 extern int	FileWrite(File file, char *buffer, int amount);
+extern int	FileWriteback(File file);
 extern int	FileSync(File file);
 extern off_t FileSeek(File file, off_t offset, int whence);
 extern int	FileTruncate(File file, off_t offset);
@@ -100,6 +101,7 @@ extern int	pg_fsync_no_writethrough(int fd);
 extern int	pg_fsync_writethrough(int fd);
 extern int	pg_fdatasync(int fd);
 extern int	pg_flush_data(int fd, off_t offset, off_t amount);
+extern int	pg_writeback(int fd);
 
 /* Filename components for OpenTemporaryFile */
 #define PG_TEMP_FILES_DIR "pgsql_tmp"
diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h
index 46c8402..37c76fb 100644
--- a/src/include/storage/smgr.h
+++ b/src/include/storage/smgr.h
@@ -95,6 +95,7 @@ extern void smgrtruncate(SMgrRelation reln, ForkNumber forknum,
 			 BlockNumber nblocks);
 extern void smgrimmedsync(SMgrRelation reln, ForkNumber forknum);
 extern void smgrpreckpt(void);
+extern void smgrwriteback(void);
 extern void smgrsync(void);
 extern void smgrpostckpt(void);
 
@@ -120,6 +121,7 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum,
 		   BlockNumber nblocks);
 extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum);
 extern void mdpreckpt(void);
+extern void mdwriteback(void);
 extern void mdsync(void);
 extern void mdpostckpt(void);