[XFS] make SYNC_DELWRI no longer use xfs_sync

Continue to de-multiplex xfs_sync be replacing all SYNC_DELWRI callers
with direct calls functions that do the work. Isolate the data quiesce
case to a function in xfs_sync.c. Isolate the FSDATA case with explicit
calls to xfs_sync_fsdata().

Version 2: o Push delwri related log forces into xfs_sync_inodes().

SGI-PV: 988140

SGI-Modid: xfs-linux-melb:xfs-kern:32309a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>

fs/xfs/linux-2.6/xfs_super.c

@@ -998,7 +998,6 @@ xfs_fs_put_super(
 	int			error;
 
 	xfs_syncd_stop(mp);
-	xfs_log_force(mp, 0, XFS_LOG_FORCE|XFS_LOG_SYNC);
 	xfs_sync_inodes(mp, SYNC_ATTR|SYNC_DELWRI);
 
 #ifdef HAVE_DMAPI
@@ -1057,7 +1056,7 @@ xfs_fs_write_super(
 	struct super_block	*sb)
 {
 	if (!(sb->s_flags & MS_RDONLY))
-		xfs_sync(XFS_M(sb), SYNC_FSDATA);
+		xfs_sync_fsdata(XFS_M(sb), 0);
 	sb->s_dirt = 0;
 }
 
@@ -1068,7 +1067,6 @@ xfs_fs_sync_super(
 {
 	struct xfs_mount	*mp = XFS_M(sb);
 	int			error;
-	int			flags;
 
 	/*
 	 * Treat a sync operation like a freeze.  This is to work
@@ -1082,20 +1080,10 @@ xfs_fs_sync_super(
 	 * dirty the Linux inode until after the transaction I/O
 	 * completes.
 	 */
-	if (wait || unlikely(sb->s_frozen == SB_FREEZE_WRITE)) {
+	if (wait || unlikely(sb->s_frozen == SB_FREEZE_WRITE))
-		/*
+		error = xfs_quiesce_data(mp);
-		 * First stage of freeze - no more writers will make progress
+	else
-		 * now we are here, so we flush delwri and delalloc buffers
+		error = xfs_sync_fsdata(mp, 0);
-		 * here, then wait for all I/O to complete.  Data is frozen at
-		 * that point. Metadata is not frozen, transactions can still
-		 * occur here so don't bother flushing the buftarg (i.e
-		 * SYNC_QUIESCE) because it'll just get dirty again.
-		 */
-		flags = SYNC_DATA_QUIESCE;
-	} else
-		flags = SYNC_FSDATA;
-
-	error = xfs_sync(mp, flags);
 	sb->s_dirt = 0;
 
 	if (unlikely(laptop_mode)) {
@@ -1233,8 +1221,7 @@ xfs_fs_remount(
 
 	/* rw -> ro */
 	if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
-		xfs_filestream_flush(mp);
+		xfs_quiesce_data(mp);
-		xfs_sync(mp, SYNC_DATA_QUIESCE);
 		xfs_attr_quiesce(mp);
 		mp->m_flags |= XFS_MOUNT_RDONLY;
 	}

fs/xfs/linux-2.6/xfs_sync.c

@@ -217,12 +217,16 @@ xfs_sync_inodes(
 	int		error;
 	int		last_error;
 	int		i;
+	int		lflags = XFS_LOG_FORCE;
 
 	if (mp->m_flags & XFS_MOUNT_RDONLY)
 		return 0;
 	error = 0;
 	last_error = 0;
 
+	if (flags & SYNC_WAIT)
+		lflags |= XFS_LOG_SYNC;
+
 	for (i = 0; i < mp->m_sb.sb_agcount; i++) {
 		if (!mp->m_perag[i].pag_ici_init)
 			continue;
@@ -232,6 +236,9 @@ xfs_sync_inodes(
 		if (error == EFSCORRUPTED)
 			break;
 	}
+	if (flags & SYNC_DELWRI)
+		xfs_log_force(mp, 0, lflags);
+
 	return XFS_ERROR(last_error);
 }
 
@@ -269,7 +276,7 @@ xfs_commit_dummy_trans(
 	return 0;
 }
 
-STATIC int
+int
 xfs_sync_fsdata(
 	struct xfs_mount	*mp,
 	int			flags)
@@ -322,6 +329,39 @@ xfs_sync_fsdata(
 	return error;
 }
 
+/*
+ * First stage of freeze - no more writers will make progress now we are here,
+ * so we flush delwri and delalloc buffers here, then wait for all I/O to
+ * complete.  Data is frozen at that point. Metadata is not frozen,
+ * transactions can still occur here so don't bother flushing the buftarg (i.e
+ * SYNC_QUIESCE) because it'll just get dirty again.
+ */
+int
+xfs_quiesce_data(
+	struct xfs_mount	*mp)
+{
+	int error;
+
+	/* push non-blocking */
+	xfs_sync_inodes(mp, SYNC_DELWRI|SYNC_BDFLUSH);
+	XFS_QM_DQSYNC(mp, SYNC_BDFLUSH);
+	xfs_filestream_flush(mp);
+
+	/* push and block */
+	xfs_sync_inodes(mp, SYNC_DELWRI|SYNC_WAIT|SYNC_IOWAIT);
+	XFS_QM_DQSYNC(mp, SYNC_WAIT);
+
+	/* write superblock and hoover shutdown errors */
+	error = xfs_sync_fsdata(mp, 0);
+
+	/* flush devices */
+	XFS_bflush(mp->m_ddev_targp);
+	if (mp->m_rtdev_targp)
+		XFS_bflush(mp->m_rtdev_targp);
+
+	return error;
+}
+
 /*
  * xfs_sync flushes any pending I/O to file system vfsp.
  *

fs/xfs/linux-2.6/xfs_sync.h

@@ -55,6 +55,9 @@ void xfs_syncd_stop(struct xfs_mount *mp);
 
 int xfs_sync(struct xfs_mount *mp, int flags);
 int xfs_sync_inodes(struct xfs_mount *mp, int flags);
+int xfs_sync_fsdata(struct xfs_mount *mp, int flags);
+
+int xfs_quiesce_data(struct xfs_mount *mp);
 
 void xfs_flush_inode(struct xfs_inode *ip);
 void xfs_flush_device(struct xfs_inode *ip);

fs/xfs/xfs_vfsops.c

@@ -75,7 +75,6 @@ xfs_quiesce_fs(
 	 * logged before we can write the unmount record.
 	 */
 	do {
-		xfs_log_force(mp, 0, XFS_LOG_FORCE|XFS_LOG_SYNC);
 		xfs_sync_inodes(mp, SYNC_ATTR|SYNC_WAIT);
 		pincount = xfs_flush_buftarg(mp->m_ddev_targp, 1);
 		if (!pincount) {