From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <linux-kernel-owner@vger.kernel.org>
Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand
	id S1755184Ab3C2PMe (ORCPT <rfc822;w@1wt.eu>);
	Fri, 29 Mar 2013 11:12:34 -0400
Received: from mx1.redhat.com ([209.132.183.28]:12995 "EHLO mx1.redhat.com"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S1752917Ab3C2PMd (ORCPT <rfc822;linux-kernel@vger.kernel.org>);
	Fri, 29 Mar 2013 11:12:33 -0400
Message-ID: <5155AF24.3090305@redhat.com>
Date: Fri, 29 Mar 2013 16:11:32 +0100
From: Ivan Vecera <ivecera@redhat.com>
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:17.0) Gecko/20130318 Thunderbird/17.0.4
MIME-Version: 1.0
To: Eric Dumazet <eric.dumazet@gmail.com>
CC: Jiri Pirko <jpirko@redhat.com>, Steven Rostedt <rostedt@goodmis.org>,
        Andy Gospodarek <andy@greyhouse.net>,
        "David S. Miller" <davem@davemloft.net>,
        LKML <linux-kernel@vger.kernel.org>, netdev <netdev@vger.kernel.org>,
        =?UTF-8?B?Tmljb2xhcyBkZSBQZXNsb8O8YW4=?= 
	<nicolas.2p.debian@gmail.com>,
        Thomas Gleixner <tglx@linutronix.de>,
        Guy Streeter <streeter@redhat.com>,
        "Paul E. McKenney" <paulmck@us.ibm.com>, stephen@networkplumber.org
Subject: Re: [PATCH] net: add a synchronize_net() in netdev_rx_handler_unregister()
References: <1364490997.6345.237.camel@gandalf.local.home>  <1364491792.15753.47.camel@edumazet-glaptop>  <20130329094856.GB1677@minipsycho.orion> <1364562082.5113.16.camel@edumazet-glaptop>
In-Reply-To: <1364562082.5113.16.camel@edumazet-glaptop>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org

On 03/29/2013 02:01 PM, Eric Dumazet wrote:
>> CPU0 will see rx_handler set and yet, rx_handler_data nulled. Write
>> >barrier in rcu_assign_pointer() might prevent this reorder from happening.
>> >Therefore I suggest:
>> >
>> >diff --git a/net/core/dev.c b/net/core/dev.c
>> >index 0caa38e..c16b829 100644
>> >--- a/net/core/dev.c
>> >+++ b/net/core/dev.c
>> >@@ -3332,8 +3332,8 @@ void netdev_rx_handler_unregister(struct net_device *dev)
>> >  {
>> >
>> >  	ASSERT_RTNL();
>> >-	RCU_INIT_POINTER(dev->rx_handler, NULL);
>> >-	RCU_INIT_POINTER(dev->rx_handler_data, NULL);
>> >+	rcu_assign_pointer(dev->rx_handler, NULL);
>> >+	rcu_assign_pointer(dev->rx_handler_data, NULL);
>> >  }
>> >  EXPORT_SYMBOL_GPL(netdev_rx_handler_unregister);
>> >
>> >
> Nope this changes nothing at all.
Erik, why doesn't help the write barrier between the assignments. It 
should guarantee their orders... or not?

Thanks,
Ivan