Re: [PATCH] fuse: Fix a potential double free in virtio_fs_get_tree

[Connor Kuehl]

On 3/23/21 12:18 AM, Lv Yunlong wrote:
> In virtio_fs_get_tree, fm is allocated by kzalloc() and
> assigned to fsc->s_fs_info by fsc->s_fs_info=fm statement.
> If the kzalloc() failed, it will goto err directly, so that

Right, I follow this so far.

> fsc->s_fs_info must be non-NULL and fm will be freed.

But this I don't follow in the context of the stuff that happens in out_err.

> But later fm is freed again when virtio_fs_fill_super() fialed.
> I think the statement if (fsc->s_fs_info) {kfree(fm);} is
> misplaced.

I'm not sure this can double free, because:

* If fm = kzalloc[..] fails, the function bails early.

* If sget_fc() fails, the function cleans up fm and fc and bails early.

* If sget_fc() succeeds and allocated a new superblock, fc->s_fs_info 
pointer is moved to sb->s_fs_info and fc->s_fs_info is set to NULL, so 
the first free hasn't happened yet.

* If sget_fc() succeeds and somehow returns an existing superblock 
(which I think is tested by checking if fc->s_fs_info is not NULL, since 
otherwise it'd have been moved to the superblock and set to NULL in 
sget_fc), I think sb->s_root would not be NULL, therefore the flow of 
control wouldn't enter the if-block where virtio_fs_fill_super could 
fail which means the code won't reach the double free.

That's just my reading of it though, and I'm wondering if that makes 
sense to others :-)

One last comment inline:

> My patch puts this statement in the correct palce to avoid
> double free.
>
> Signed-off-by: Lv Yunlong <lyl2019@xxxxxxxxxxxxxxxx>
> ---
>   fs/fuse/virtio_fs.c | 10 ++++++----
>   1 file changed, 6 insertions(+), 4 deletions(-)
>
> diff --git a/fs/fuse/virtio_fs.c b/fs/fuse/virtio_fs.c
> index 8868ac31a3c0..727cf436828f 100644
> --- a/fs/fuse/virtio_fs.c
> +++ b/fs/fuse/virtio_fs.c
> @@ -1437,10 +1437,7 @@ static int virtio_fs_get_tree(struct fs_context *fsc)
>   
>   	fsc->s_fs_info = fm;
>   	sb = sget_fc(fsc, virtio_fs_test_super, set_anon_super_fc);
> -	if (fsc->s_fs_info) {
> -		fuse_conn_put(fc);
> -		kfree(fm);
> -	}
> +
>   	if (IS_ERR(sb))
>   		return PTR_ERR(sb);

By removing the check from here, it now looks like if sget_fc() fails, 
then this early return will leak fm's memory and fc's reference.

Connor

>   
> @@ -1457,6 +1454,11 @@ static int virtio_fs_get_tree(struct fs_context *fsc)
>   		sb->s_flags |= SB_ACTIVE;
>   	}
>   
> +	if (fsc->s_fs_info) {
> +		fuse_conn_put(fc);
> +		kfree(fm);
> +	}
> +
>   	WARN_ON(fsc->root);
>   	fsc->root = dget(sb->s_root);
>   	return 0;