Re: [PATCH] tcp: possible race between tcp_done() and tcp_poll()

[Eric Dumazet]

On Thu, 2017-03-30 at 11:55 +0900, Seiichi Ikarashi wrote:

> I got a report that receiving a RST packet but poll() got only POLLERR, no 
> POLLIN|POLLRDHUP .
> It was an old x86_64 kernel which does not include sk_state_{load,store} 
> functions.
> I suspected some race might have occur above.

It looks that a one-liner patch would be enough in tcp_done()

No need adding extra barriers

Untested patch :

diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 
1e319a525d51b0b603a5ccc5143381c752b9f2c7..4107d3fc00a53f768cdb885971a1067c810f5c9f
 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -3285,12 +3285,12 @@ void tcp_done(struct sock *sk)
        if (sk->sk_state == TCP_SYN_SENT || sk->sk_state == TCP_SYN_RECV)
                TCP_INC_STATS(sock_net(sk), TCP_MIB_ATTEMPTFAILS);
 
+       sk->sk_shutdown = SHUTDOWN_MASK;
        tcp_set_state(sk, TCP_CLOSE);
        tcp_clear_xmit_timers(sk);
        if (req)
                reqsk_fastopen_remove(sk, req, false);
 
-       sk->sk_shutdown = SHUTDOWN_MASK;
 
        if (!sock_flag(sk, SOCK_DEAD))
                sk->sk_state_change(sk);