Re: [PATCH net] tcp: allow again tcp_disconnect() when threads are waiting

[Paolo Abeni <pabeni@redhat.com>]

Hi,

Thank you for your review!

On Tue, 2023-10-10 at 18:21 +0200, Eric Dumazet wrote:
[...]
> > @@ -1585,7 +1596,11 @@ static int peekmsg(struct sock *sk, struct msghdr *msg,
> >                         release_sock(sk);
> >                         lock_sock(sk);
> >                 } else {
> > -                       sk_wait_data(sk, &timeo, NULL);
> > +                       ret = sk_wait_data(sk, &timeo, NULL);
> > +                       if (ret < 0) {
> > +                               copied = ret;
> 
>    if (!copied)
>       copied = ret;

I think we can infer 'copied' is zero here, as a few lines before we
have:

	if (copied)
		break;

> 
> > @@ -3092,6 +3092,7 @@ int tcp_disconnect(struct sock *sk, int flags)
> >                 sk->sk_frag.offset = 0;
> >         }
> >         sk_error_report(sk);
> > +       sk->sk_disconnects++;
> 
> Should we perform this generically, from the caller ?

Ok, I'll do that in v2.

Thanks!

Paolo