From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752690AbbCWNtH (ORCPT ); Mon, 23 Mar 2015 09:49:07 -0400 Received: from mail-pa0-f50.google.com ([209.85.220.50]:35334 "EHLO mail-pa0-f50.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752570AbbCWNtF (ORCPT ); Mon, 23 Mar 2015 09:49:05 -0400 Message-ID: <1427118542.25985.62.camel@edumazet-glaptop2.roam.corp.google.com> Subject: Re: [PATCH net] tcp: prevent fetching dst twice in early demux code From: Eric Dumazet To: Michal Kubecek Cc: "David S. Miller" , netdev@vger.kernel.org, linux-kernel@vger.kernel.org, Eric Dumazet , Alexey Kuznetsov , James Morris , Hideaki YOSHIFUJI , Patrick McHardy Date: Mon, 23 Mar 2015 06:49:02 -0700 In-Reply-To: <20150323082718.265EDA0BE9@unicorn.suse.cz> References: <20150323082718.265EDA0BE9@unicorn.suse.cz> Content-Type: text/plain; charset="UTF-8" X-Mailer: Evolution 3.10.4-0ubuntu2 Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Mon, 2015-03-23 at 09:27 +0100, Michal Kubecek wrote: > On s390x, gcc 4.8 compiles this part of tcp_v6_early_demux() > > struct dst_entry *dst = sk->sk_rx_dst; > > if (dst) > dst = dst_check(dst, inet6_sk(sk)->rx_dst_cookie); > > to code reading sk->sk_rx_dst twice, once for the test and once for > the argument of ip6_dst_check() (dst_check() is inline). This allows > ip6_dst_check() to be called with null first argument, causing a crash. > > Protect sk->sk_rx_dst access by ACCESS_ONCE() both in IPv4 and IPv6 > TCP early demux code. > > Fixes: 41063e9dd119 ("ipv4: Early TCP socket demux.") > Fixes: c7109986db3c ("ipv6: Early TCP socket demux") > Please remove this empty line. "Fixes" tags should be in the final group of tags. > Signed-off-by: Michal Kubecek > --- > net/ipv4/tcp_ipv4.c | 2 +- > net/ipv6/tcp_ipv6.c | 2 +- > 2 files changed, 2 insertions(+), 2 deletions(-) > > diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c > index 5a2dfed..3d42b45 100644 > --- a/net/ipv4/tcp_ipv4.c > +++ b/net/ipv4/tcp_ipv4.c > @@ -1518,7 +1518,7 @@ void tcp_v4_early_demux(struct sk_buff *skb) > skb->sk = sk; > skb->destructor = sock_edemux; > if (sk->sk_state != TCP_TIME_WAIT) { > - struct dst_entry *dst = sk->sk_rx_dst; > + struct dst_entry *dst = ACCESS_ONCE(sk->sk_rx_dst); > > if (dst) > dst = dst_check(dst, 0); > diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c > index 5d46832..50eb5ff 100644 > --- a/net/ipv6/tcp_ipv6.c > +++ b/net/ipv6/tcp_ipv6.c > @@ -1585,7 +1585,7 @@ static void tcp_v6_early_demux(struct sk_buff *skb) > skb->sk = sk; > skb->destructor = sock_edemux; > if (sk->sk_state != TCP_TIME_WAIT) { > - struct dst_entry *dst = sk->sk_rx_dst; > + struct dst_entry *dst = ACCESS_ONCE(sk->sk_rx_dst); > > if (dst) > dst = dst_check(dst, inet6_sk(sk)->rx_dst_cookie); Good catch, (or would I say, strange compiler ? ;) ) Please use READ_ONCE(). Only stable ports will need ACCESS_ONCE() Also, this probably points a problem on this arch, because its looks different cpu are handling TCP traffic for the same TCP flow.