Btrfs: do not call btrfs_update_inode in endio if nothing changed
authorJosef Bacik <josef@redhat.com>
Tue, 5 Apr 2011 23:25:36 +0000 (19:25 -0400)
committerJosef Bacik <josef@redhat.com>
Fri, 8 Apr 2011 17:00:36 +0000 (13:00 -0400)
In the DIO code we often don't update the i_disk_size because the i_size isn't
updated until after the DIO is completed, so basically we are allocating a path,
doing a search, and updating the inode item for no reason since nothing changed.
btrfs_ordered_update_i_size will return 1 if it didn't update i_disk_size, so
only run btrfs_update_inode if btrfs_ordered_update_i_size returns 0.  Thanks,

Signed-off-by: Josef Bacik <josef@redhat.com>
fs/btrfs/inode.c

index da26802..4a238d6 100644 (file)
@@ -1769,9 +1769,12 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end)
        add_pending_csums(trans, inode, ordered_extent->file_offset,
                          &ordered_extent->list);
 
-       btrfs_ordered_update_i_size(inode, 0, ordered_extent);
-       ret = btrfs_update_inode(trans, root, inode);
-       BUG_ON(ret);
+       ret = btrfs_ordered_update_i_size(inode, 0, ordered_extent);
+       if (!ret) {
+               ret = btrfs_update_inode(trans, root, inode);
+               BUG_ON(ret);
+       }
+       ret = 0;
 out:
        if (nolock) {
                if (trans)
@@ -5865,8 +5868,10 @@ again:
        }
 
        add_pending_csums(trans, inode, ordered->file_offset, &ordered->list);
-       btrfs_ordered_update_i_size(inode, 0, ordered);
-       btrfs_update_inode(trans, root, inode);
+       ret = btrfs_ordered_update_i_size(inode, 0, ordered);
+       if (!ret)
+               btrfs_update_inode(trans, root, inode);
+       ret = 0;
 out_unlock:
        unlock_extent_cached(&BTRFS_I(inode)->io_tree, ordered->file_offset,
                             ordered->file_offset + ordered->len - 1,