btrfs: Remove unnecessary tree locking code in qgroup_rescan_leaf
authorNikolay Borisov <nborisov@suse.com>
Wed, 15 Aug 2018 15:26:56 +0000 (18:26 +0300)
committerDavid Sterba <dsterba@suse.com>
Mon, 17 Dec 2018 13:51:31 +0000 (14:51 +0100)
In qgroup_rescan_leaf a copy is made of the target leaf by calling
btrfs_clone_extent_buffer. The latter allocates a new buffer and
attaches a new set of pages and copies the content of the source buffer.
The new scratch buffer is only used to iterate it's items, it's not
published anywhere and cannot be accessed by a third party.

Hence, it's not necessary to perform any locking on it whatsoever.
Furthermore, remove the extra extent_buffer_get call since the new
buffer is always allocated with a reference count of 1 which is
sufficient here.  No functional changes.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
fs/btrfs/qgroup.c

index f70825af6438e9c48ebe782f6b8049d27ae807fc..b53d2e7b938d852cea62df7707366bf23c5d4db5 100644 (file)
@@ -3104,9 +3104,6 @@ static int qgroup_rescan_leaf(struct btrfs_trans_handle *trans,
                mutex_unlock(&fs_info->qgroup_rescan_lock);
                goto out;
        }
-       extent_buffer_get(scratch_leaf);
-       btrfs_tree_read_lock(scratch_leaf);
-       btrfs_set_lock_blocking_rw(scratch_leaf, BTRFS_READ_LOCK);
        slot = path->slots[0];
        btrfs_release_path(path);
        mutex_unlock(&fs_info->qgroup_rescan_lock);
@@ -3132,10 +3129,8 @@ static int qgroup_rescan_leaf(struct btrfs_trans_handle *trans,
                        goto out;
        }
 out:
-       if (scratch_leaf) {
-               btrfs_tree_read_unlock_blocking(scratch_leaf);
+       if (scratch_leaf)
                free_extent_buffer(scratch_leaf);
-       }
 
        if (done && !ret) {
                ret = 1;