Re: [patch 1/2] Btrfs: double unlock bug in error handling

[Jan Schmidt <list.btrfs@jan-o-sch.net>]

On 18.04.2012 08:59, Dan Carpenter wrote:
> The caller expects this function to return with the lock held and
> releases it immediately on error.
>
> Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
>
> diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
> index 2b35f8d..a0bb9dc 100644
> --- a/fs/btrfs/extent-tree.c
> +++ b/fs/btrfs/extent-tree.c
> @@ -2301,6 +2301,7 @@ static noinline int run_clustered_refs(struct
btrfs_trans_handle *trans,
>
>  				if (ret) {
>  					printk(KERN_DEBUG "btrfs: run_delayed_extent_op returned %d\n",
ret);
> +					spin_lock(&delayed_refs->lock);
>  					return ret;
>  				}
>
> @@ -2331,6 +2332,7 @@ static noinline int run_clustered_refs(struct
btrfs_trans_handle *trans,
>
>  		if (ret) {
>  			printk(KERN_DEBUG "btrfs: run_one_delayed_ref returned %d\n", ret);
> +			spin_lock(&delayed_refs->lock);
>  			return ret;
>  		}

I think the correct way to fix this is:

diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index a844204..9af261a 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -2450,7 +2450,6 @@ again:

                ret = run_clustered_refs(trans, root, &cluster);
                if (ret < 0) {
-                       spin_unlock(&delayed_refs->lock);
                        btrfs_abort_transaction(trans, root, ret);
                        return ret;
                }

-Jan