Re: [PATCH] btrfs: simplify error handling in btrfs_lookup_dentry

[Rohit Singh <rh0@fb.com>]

On Mon, Jul 11, 2022 at 06:16:18PM +0300, Nikolay Borisov wrote:
> In btrfs_lookup_dentry releasing the reference of the sub_root and the
> running orphan cleanup should only happen if the dentry found actually
> represents a subvolume. This can only be true in the 'else' branch as
> otherwise either fixup_tree_root_location returned an ENOENT error, in
> which case sub_root wouldn't have been changed or if we got a different
> errno this means btrfs_get_fs_root couldn't have executed successfully
> again meaning sub_root will equal to root. So simplify all the branches
> by moving the code into the 'else'.
> 
> Signed-off-by: Nikolay Borisov <nborisov@suse.com>
> ---
>  fs/btrfs/inode.c | 4 ----
>  1 file changed, 4 deletions(-)
> 
> diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
> index 0b17335555e0..1dd073e96696 100644
> --- a/fs/btrfs/inode.c
> +++ b/fs/btrfs/inode.c
> @@ -5821,11 +5821,7 @@ struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry)
>  			inode = new_simple_dir(dir->i_sb, &location, sub_root);
>  	} else {
>  		inode = btrfs_iget(dir->i_sb, location.objectid, sub_root);
> -	}
> -	if (root != sub_root)
>  		btrfs_put_root(sub_root);
> -
> -	if (!IS_ERR(inode) && root != sub_root) {

It looks like the root != sub_root stuff looks correct.

Can't the btrfs inode have an error state on it though?