Re: [PATCH net] l2tp: Don't sleep and disable BH under writer-side sk_callback_lock

[Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>]

On 2022/11/21 18:00, Jakub Sitnicki wrote:
>> Is it safe to temporarily set a dummy pointer like below?
>> If it is not safe, what makes assignments done by
>> setup_udp_tunnel_sock() safe?
> 
> Yes, I think so. Great idea. I've used it in v2.

So, you are sure that e.g.

	udp_sk(sk)->gro_receive = cfg->gro_receive;

in setup_udp_tunnel_sock() (where the caller is passing cfg->gro_receive == NULL)
never races with e.g. below check (because the socket might be sending/receiving
in progress since the socket is retrieved via user-specified file descriptor) ?

Then, v2 patch would be OK for fixing this regression. (But I think we still should
avoid retrieving a socket from user-specified file descriptor in order to avoid
lockdep race window.)


struct sk_buff *udp_gro_receive(struct list_head *head, struct sk_buff *skb,
                                struct udphdr *uh, struct sock *sk)
{
	(...snipped...)
        if (!sk || !udp_sk(sk)->gro_receive) {
		(...snipped...)
                /* no GRO, be sure flush the current packet */
                goto out;
        }
	(...snipped...)
        pp = call_gro_receive_sk(udp_sk(sk)->gro_receive, sk, head, skb);
out:
        skb_gro_flush_final(skb, pp, flush);
        return pp;
}

> 
> We can check & assign sk_user_data under sk_callback_lock, and then just
> let setup_udp_tunnel_sock overwrite it with the same value, without
> holding the lock.

Given that sk_user_data is RCU-protected on reader-side, don't we need to
wait for RCU grace period after resetting to NULL?

> 
> I still think that it's best to keep the critical section as short as
> possible, though.