xfs: Fix wrong return value of xfs_file_aio_write
authorMarkus Trippelsdorf <markus@trippelsdorf.de>
Sun, 24 Jul 2011 12:03:30 +0000 (14:03 +0200)
committerAl Viro <viro@zeniv.linux.org.uk>
Mon, 25 Jul 2011 18:23:21 +0000 (14:23 -0400)
The fsync prototype change commit 02c24a82187d accidentally overwrote
the ssize_t return value of xfs_file_aio_write with 0 for SYNC type
writes. Fix this by checking if an error occured when calling
xfs_file_fsync and only change the return value in this case.
In addition xfs_file_fsync actually returns a normal negative error, so
fix this, too.

Signed-off-by: Markus Trippelsdorf <markus@trippelsdorf.de>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Tested-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
fs/xfs/linux-2.6/xfs_file.c

index cca00f49e092a75f2726a79f4c1076469f0b2f36..825390e1c13850985c634fcbaec19c88ce2d6e46 100644 (file)
@@ -881,11 +881,14 @@ xfs_file_aio_write(
        /* Handle various SYNC-type writes */
        if ((file->f_flags & O_DSYNC) || IS_SYNC(inode)) {
                loff_t end = pos + ret - 1;
+               int error;
 
                xfs_rw_iunlock(ip, iolock);
-               ret = -xfs_file_fsync(file, pos, end,
+               error = xfs_file_fsync(file, pos, end,
                                      (file->f_flags & __O_SYNC) ? 0 : 1);
                xfs_rw_ilock(ip, iolock);
+               if (error)
+                       ret = error;
        }
 
 out_unlock: