Btrfs: unreserve space if our ordered extent fails to work
authorJosef Bacik <jbacik@fusionio.com>
Thu, 31 Jan 2013 19:58:00 +0000 (14:58 -0500)
committerJosef Bacik <jbacik@fusionio.com>
Wed, 20 Feb 2013 17:59:29 +0000 (12:59 -0500)
When a transaction aborts or there's an EIO on an ordered extent or any
error really we will not free up the space we reserved for this ordered
extent.  This results in warnings from the block group cache cleanup in the
case of a transaction abort, or leaking space in the case of EIO on an
ordered extent.  Fix this up by free'ing the reserved space if we have an
error at all trying to complete an ordered extent.  Thanks,

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

index 82c7c66f8523b2d42273e11bfdd2ab602ea07fb8..2b48d2d51d50b81862eca4e725d76e43049e1153 100644 (file)
@@ -2022,11 +2022,23 @@ out:
        if (trans)
                btrfs_end_transaction(trans, root);
 
-       if (ret)
+       if (ret) {
                clear_extent_uptodate(io_tree, ordered_extent->file_offset,
                                      ordered_extent->file_offset +
                                      ordered_extent->len - 1, NULL, GFP_NOFS);
 
+               /*
+                * If the ordered extent had an IOERR or something else went
+                * wrong we need to return the space for this ordered extent
+                * back to the allocator.
+                */
+               if (!test_bit(BTRFS_ORDERED_NOCOW, &ordered_extent->flags) &&
+                   !test_bit(BTRFS_ORDERED_PREALLOC, &ordered_extent->flags))
+                       btrfs_free_reserved_extent(root, ordered_extent->start,
+                                                  ordered_extent->disk_len);
+       }
+
+
        /*
         * This needs to be done to make sure anybody waiting knows we are done
         * updating everything for this ordered extent.