Btrfs: check for an extent_op on the locked ref
authorJosef Bacik <jbacik@fb.com>
Thu, 27 Mar 2014 23:41:34 +0000 (19:41 -0400)
committerChris Mason <clm@fb.com>
Mon, 7 Apr 2014 00:34:36 +0000 (17:34 -0700)
We could have possibly added an extent_op to the locked_ref while we dropped
locked_ref->lock, so check for this case as well and loop around.  Otherwise we
could lose flag updates which would lead to extent tree corruption.  Thanks,

cc: stable@vger.kernel.org
Signed-off-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: Chris Mason <clm@fb.com>
fs/btrfs/extent-tree.c

index c6b6a6e3e735ce73bf06a735a9ee85533d94e4bf..e09db2c2f3b48d578cefd5bc528b2b3e9bcba333 100644 (file)
@@ -2444,7 +2444,8 @@ static noinline int __btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
                        spin_unlock(&locked_ref->lock);
                        spin_lock(&delayed_refs->lock);
                        spin_lock(&locked_ref->lock);
-                       if (rb_first(&locked_ref->ref_root)) {
+                       if (rb_first(&locked_ref->ref_root) ||
+                           locked_ref->extent_op) {
                                spin_unlock(&locked_ref->lock);
                                spin_unlock(&delayed_refs->lock);
                                continue;