Btrfs: flush data on snapshot creation
authorSage Weil <sage@newdream.net>
Fri, 19 Feb 2010 22:13:50 +0000 (14:13 -0800)
committerChris Mason <chris.mason@oracle.com>
Mon, 15 Mar 2010 15:00:11 +0000 (11:00 -0400)
Flush any delalloc extents when we create a snapshot, so that recently
written file data is always included in the snapshot.

A later commit will add the ability to snapshot without the flush, but
most people expect flushing.

Signed-off-by: Sage Weil <sage@newdream.net>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
fs/btrfs/transaction.c

index 2a36e236a4924c978d7b74c416e062f7d92dc921..2d654c1c794d2f7690409a53bfcbc02ae95890a9 100644 (file)
@@ -997,13 +997,10 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 
                mutex_unlock(&root->fs_info->trans_mutex);
 
-               if (flush_on_commit) {
+               if (flush_on_commit || snap_pending) {
                        btrfs_start_delalloc_inodes(root, 1);
                        ret = btrfs_wait_ordered_extents(root, 0, 1);
                        BUG_ON(ret);
-               } else if (snap_pending) {
-                       ret = btrfs_wait_ordered_extents(root, 0, 1);
-                       BUG_ON(ret);
                }
 
                /*