xfs: remove invalid barrier optimization from xfs_fsync
authorChristoph Hellwig <hch@lst.de>
Mon, 1 Feb 2010 23:16:26 +0000 (10:16 +1100)
committerDave Chinner <david@fromorbit.com>
Mon, 1 Feb 2010 23:16:26 +0000 (10:16 +1100)
We always need to flush the disk write cache and can't skip it just because
the no inode attributes have changed.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <david@fromorbit.com>
fs/xfs/xfs_vnodeops.c

index fd108b7385594cac9310518caf74b6e592ea7152..43241e2898006ce5ee3ab2a83e2da9f7f4385305 100644 (file)
@@ -597,7 +597,7 @@ xfs_fsync(
 {
        xfs_trans_t     *tp;
        int             error = 0;
-       int             log_flushed = 0, changed = 1;
+       int             log_flushed = 0;
 
        xfs_itrace_entry(ip);
 
@@ -627,18 +627,10 @@ xfs_fsync(
                 * disk yet, the inode will be still be pinned.  If it is,
                 * force the log.
                 */
-
                xfs_iunlock(ip, XFS_ILOCK_SHARED);
-
                if (xfs_ipincount(ip)) {
                        error = _xfs_log_force(ip->i_mount, XFS_LOG_SYNC,
                                               &log_flushed);
-               } else {
-                       /*
-                        * If the inode is not pinned and nothing has changed
-                        * we don't need to flush the cache.
-                        */
-                       changed = 0;
                }
        } else  {
                /*
@@ -673,7 +665,7 @@ xfs_fsync(
                xfs_iunlock(ip, XFS_ILOCK_EXCL);
        }
 
-       if ((ip->i_mount->m_flags & XFS_MOUNT_BARRIER) && changed) {
+       if (ip->i_mount->m_flags & XFS_MOUNT_BARRIER) {
                /*
                 * If the log write didn't issue an ordered tag we need
                 * to flush the disk cache for the data device now.