Re: [PATCH] net/rose: Fix Use-After-Free in rose_ioctl

[Eric Dumazet <edumazet@google.com>]

On Mon, Dec 4, 2023 at 7:57 AM Hyunwoo Kim <v4bel@theori.io> wrote:
>
> Because rose_ioctl() accesses sk->sk_receive_queue
> without holding a lock_sock, it can cause a race with
> rose_accept().
> A use-after-free for skb occurs with the following flow.
> ```
> rose_ioctl() -> skb_peek()
> rose_accept() -> skb_dequeue() -> kfree_skb()
> ```
> Add lock_sock to rose_ioctl() to fix this issue.
>
> Signed-off-by: Hyunwoo Kim <v4bel@theori.io>
> ---
>  net/rose/af_rose.c | 2 ++
>  1 file changed, 2 insertions(+)
>
> diff --git a/net/rose/af_rose.c b/net/rose/af_rose.c
> index 0cc5a4e19900..5fe9db64b6df 100644
> --- a/net/rose/af_rose.c
> +++ b/net/rose/af_rose.c
> @@ -1316,8 +1316,10 @@ static int rose_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
>                 struct sk_buff *skb;
>                 long amount = 0L;
>                 /* These two are safe on a single CPU system as only user tasks fiddle here */
> +               lock_sock(sk);

This is not correct.

You will have to lock sk->sk_receive_queue.lock instead.

Look at rose_recvmsg() for the reason why locking the socket itself is
not helping.

>                 if ((skb = skb_peek(&sk->sk_receive_queue)) != NULL)
>                         amount = skb->len;
> +               release_sock(sk);
>                 return put_user(amount, (unsigned int __user *) argp);
>         }
>
> --
> 2.25.1
>