Re: [PATCH] btrfs: Fix btrfs_block_group refcnt leak

From: Filipe Manana
Date: Mon Apr 20 2020 - 13:31:28 EST


On Mon, Apr 20, 2020 at 6:48 AM Xiyu Yang <xiyuyang19@xxxxxxxxxxxx> wrote:
>
> btrfs_remove_block_group() invokes btrfs_lookup_block_group(), which
> returns a local reference of the blcok group that contains the given
> bytenr to "block_group" with increased refcount.
>
> When btrfs_remove_block_group() returns, "block_group" becomes invalid,
> so the refcount should be decreased to keep refcount balanced.
>
> The reference counting issue happens in several exception handling paths
> of btrfs_remove_block_group(). When those error scenarios occur such as
> btrfs_alloc_path() returns NULL, the function forgets to decrease its
> refcnt increased by btrfs_lookup_block_group() and will cause a refcnt
> leak.
>
> Fix this issue by jumping to "out_put_group" label and calling
> btrfs_put_block_group() when those error scenarios occur.
>
> Signed-off-by: Xiyu Yang <xiyuyang19@xxxxxxxxxxxx>
> Signed-off-by: Xin Tan <tanxin.ctf@xxxxxxxxx>

Seems correct to me.
I would change the subject to something more clear like: "btrfs: fix
block group leak after failure to remove it"

One more suggestion below.

> ---
> fs/btrfs/block-group.c | 13 ++++++++-----
> 1 file changed, 8 insertions(+), 5 deletions(-)
>
> diff --git a/fs/btrfs/block-group.c b/fs/btrfs/block-group.c
> index 404e050ce8ee..d9f432bd3329 100644
> --- a/fs/btrfs/block-group.c
> +++ b/fs/btrfs/block-group.c
> @@ -916,7 +916,7 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
> path = btrfs_alloc_path();
> if (!path) {
> ret = -ENOMEM;
> - goto out;
> + goto out_put_group;
> }
>
> /*
> @@ -954,7 +954,7 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
> ret = btrfs_orphan_add(trans, BTRFS_I(inode));
> if (ret) {
> btrfs_add_delayed_iput(inode);
> - goto out;
> + goto out_put_group;
> }
> clear_nlink(inode);
> /* One for the block groups ref */
> @@ -977,13 +977,13 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
>
> ret = btrfs_search_slot(trans, tree_root, &key, path, -1, 1);
> if (ret < 0)
> - goto out;
> + goto out_put_group;
> if (ret > 0)
> btrfs_release_path(path);
> if (ret == 0) {
> ret = btrfs_del_item(trans, tree_root, path);
> if (ret)
> - goto out;
> + goto out_put_group;
> btrfs_release_path(path);
> }
>
> @@ -1102,7 +1102,7 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
>
> ret = remove_block_group_free_space(trans, block_group);
> if (ret)
> - goto out;
> + goto out_put_group;
>
> btrfs_put_block_group(block_group);
> btrfs_put_block_group(block_group);
> @@ -1132,6 +1132,9 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
> btrfs_delayed_refs_rsv_release(fs_info, 1);
> btrfs_free_path(path);
> return ret;
> +out_put_group:
> + btrfs_put_block_group(block_group);
> + goto out;

Instead of this double goto, which tends to be error prone and harder to follow,
I suggest placing a call to btrfs_put_block_group() right after the
'out' label, with a comment
above it saying something like "once for the lookup reference" and
removing one of the btrfs_put_block_group()
calls right after calling remove_block_group_free_space(), and leaving
a comment above the other one saying "once for the block groups
rbtree".

Thanks.

> }
>
> struct btrfs_trans_handle *btrfs_start_trans_remove_block_group(
> --
> 2.7.4
>


--
Filipe David Manana,

âWhether you think you can, or you think you can't â you're right.â