On Mon, Dec 09, 2019 at 11:41:14AM +0800, Dinghao Liu wrote:
> The return value of link_free_space is checked out-sync.
> One branch of an if statement uses an extra check after
> WARN_ON() but its peer branch does not. WARN_ON() does
> not change the control flow, thus only using this check
> might be insufficient.
>
> Fix this by simply adding a check on ret.
>
> Signed-off-by: Dinghao Liu <dinghao.liu@xxxxxxxxxx>
> --
> Changes in v2:
> - Add memory free for free space entry.
> ---
> fs/btrfs/free-space-cache.c | 4 ++++
> 1 file changed, 4 insertions(+)
>
> diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c
> index 3283da419200..ba2e6cea5233 100644
> --- a/fs/btrfs/free-space-cache.c
> +++ b/fs/btrfs/free-space-cache.c
> @@ -2437,6 +2437,10 @@ int btrfs_remove_free_space(struct btrfs_block_group *block_group,
> if (info->bytes) {
> ret = link_free_space(ctl, info);
> WARN_ON(ret);
> + if (ret) {
> + kmem_cache_free(btrfs_free_space_cachep, info);
> + goto out_lock;
> + }
> } else {
> kmem_cache_free(btrfs_free_space_cachep, info);
There are two link_free_space followed by WARN_ON instances in the
function, please remove both.
In the above case, the branches can be merged together so there's not
repeated kmem_cache_free, like
if (info->bytes)
ret = link_free_space(...);
kmem_cache_free(btrfs_free_space_cachep, info);
if (ret)
goto out_unlock;
