Btrfs: fix protection between walking backrefs and root deletion

There is a race condition between resolving indirect ref and root deletion,
and we should gurantee that root can not be destroyed to avoid accessing
broken tree here.

Here we fix it by holding @subvol_srcu, and we will release it as soon
as we have held root node lock.

Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: Chris Mason <clm@fb.com>

diff --git a/fs/btrfs/backref.c b/fs/btrfs/backref.c
index 1538496..10ae570 100644 (file)
--- a/fs/btrfs/backref.c
+++ b/fs/btrfs/backref.c
@@ -301,23 +301,34 @@ static int __resolve_indirect_ref(struct btrfs_fs_info *fs_info,
        int ret = 0;
        int root_level;
        int level = ref->level;
+       int index;
 
        root_key.objectid = ref->root_id;
        root_key.type = BTRFS_ROOT_ITEM_KEY;
        root_key.offset = (u64)-1;
+
+       index = srcu_read_lock(&fs_info->subvol_srcu);
+
        root = btrfs_read_fs_root_no_name(fs_info, &root_key);
        if (IS_ERR(root)) {
+               srcu_read_unlock(&fs_info->subvol_srcu, index);
                ret = PTR_ERR(root);
                goto out;
        }
 
        root_level = btrfs_old_root_level(root, time_seq);
 
-       if (root_level + 1 == level)
+       if (root_level + 1 == level) {
+               srcu_read_unlock(&fs_info->subvol_srcu, index);
                goto out;
+       }
 
        path->lowest_level = level;
        ret = btrfs_search_old_slot(root, &ref->key_for_search, path, time_seq);
+
+       /* root node has been locked, we can release @subvol_srcu safely here */
+       srcu_read_unlock(&fs_info->subvol_srcu, index);
+
        pr_debug("search slot in root %llu (level %d, ref count %d) returned "
                 "%d for key (%llu %u %llu)\n",
                 ref->root_id, level, ref->count, ret,