btrfs: do not BUG_ON() on tree mod log failure at insert_new_root()
authorFilipe Manana <fdmanana@suse.com>
Thu, 8 Jun 2023 10:27:47 +0000 (11:27 +0100)
committerDavid Sterba <dsterba@suse.com>
Mon, 19 Jun 2023 11:59:39 +0000 (13:59 +0200)
At insert_new_root(), instead of doing a BUG_ON() in case we fail to
record the tree mod log operation, just return the error to the callers
after releasing the allocated tree block. At this point we haven't made
any changes to the b+tree, so we haven't left it in an inconsistent state
and therefore have no need to abort the transaction. All we need to do is
to unlock and free the extent buffer we just allocated with the purpose
of making it the new root.

Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
fs/btrfs/ctree.c

index dc60ece85664ca46d67b44dac124f821f4941d3f..58325b8f544f19423dcd2ff0082e7fafbaaaa89b 100644 (file)
@@ -2964,7 +2964,12 @@ static noinline int insert_new_root(struct btrfs_trans_handle *trans,
 
        old = root->node;
        ret = btrfs_tree_mod_log_insert_root(root->node, c, false);
-       BUG_ON(ret < 0);
+       if (ret < 0) {
+               btrfs_free_tree_block(trans, btrfs_root_id(root), c, 0, 1);
+               btrfs_tree_unlock(c);
+               free_extent_buffer(c);
+               return ret;
+       }
        rcu_assign_pointer(root->node, c);
 
        /* the super has an extra ref to root->node */