On Tue, 2015-12-01 at 16:04 +0900, Lorenzo Colitti wrote:
> + /* Don't race with BH socket closes such as inet_csk_listen_stop. */
> + local_bh_disable();
> + bh_lock_sock(sk);
> +
> + if (!sock_flag(sk, SOCK_DEAD)) {
> + smp_wmb(); /* Be consistent with tcp_reset */
This barrier is absolutely useless here, it must be moved _after_
sk->sk_err = ECONNABORTED.
Since you copied tcp_reset() code, please do not reorder barriers.
Also, comment in tcp_reset() is better IMO
/* This barrier is coupled with smp_rmb() in tcp_poll() */
> + sk->sk_err = ECONNABORTED;
> + sk->sk_error_report(sk);
Why not using tcp_need_reset() here, otherwise we might send spurious
RST for some tcp states.
Ideally, packetdrill tests would be nice.
> + tcp_send_active_reset(sk, GFP_ATOMIC);
> + tcp_done(sk);
> + }
Please do not send a v5 before letting other people comment on v3 & v4,
no need to flood netdev with many versions sent the same day.
Thanks.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
