From mboxrd@z Thu Jan  1 00:00:00 1970
From: David Miller <davem@davemloft.net>
Subject: Re: 2.6.27-rc4: icmp_sk() uses smp_processor_id() in preemptible
 code
Date: Fri, 22 Aug 2008 04:54:02 -0700 (PDT)
Message-ID: <20080822.045402.32679253.davem@davemloft.net>
References: <20080821122040.GA2497@x200.localdomain>
	<E1KW9c0-0002Kp-00@gondolin.me.apana.org.au>
Mime-Version: 1.0
Content-Type: Text/Plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Cc: adobriyan@gmail.com, den@openvz.org, dlezcano@fr.ibm.com,
	netdev@vger.kernel.org
To: herbert@gondor.apana.org.au
Return-path: <netdev-owner@vger.kernel.org>
Received: from 74-93-104-97-Washington.hfc.comcastbusiness.net ([74.93.104.97]:47678
	"EHLO sunset.davemloft.net" rhost-flags-OK-FAIL-OK-OK)
	by vger.kernel.org with ESMTP id S1752896AbYHVLyF (ORCPT
	<rfc822;netdev@vger.kernel.org>); Fri, 22 Aug 2008 07:54:05 -0400
In-Reply-To: <E1KW9c0-0002Kp-00@gondolin.me.apana.org.au>
Sender: netdev-owner@vger.kernel.org
List-ID: <netdev.vger.kernel.org>

From: Herbert Xu <herbert@gondor.apana.org.au>
Date: Thu, 21 Aug 2008 22:50:00 +1000

> adobriyan@gmail.com wrote:
> > Steps to reproduce:
> > 
> >        CONFIG_PREEMPT=y
> >        CONFIG_DEBUG_PREEMPT=y
> >        tracepath <something>
> > 
> > icmp6_sk() is similar (I haven't checked at runtime, though).
> > 
> > BUG: using smp_processor_id() in preemptible [00000000] code: tracepath/3205
> > caller is icmp_sk+0x15/0x30
> > Pid: 3205, comm: tracepath Not tainted 2.6.27-rc4 #1
> > 
> > Call Trace:
> > [<ffffffff8031af14>] debug_smp_processor_id+0xe4/0xf0
> > [<ffffffff80409405>] icmp_sk+0x15/0x30
> > [<ffffffff8040a17b>] icmp_send+0x4b/0x3f0
> 
> Caused by:
> 
> commit 405666db84b984b68fc75794069f424c02e5796c
> Author: Denis V. Lunev <den@openvz.org>
> Date:   Fri Feb 29 11:16:46 2008 -0800
> 
>     [ICMP]: Pass proper ICMP socket into icmp(v6)_xmit_(un)lock.
> 
> We need to disable preemption (was provided by disabling BH) before
> getting the per-cpu ICMP socket.

Denis, please fix this.

The icmp_sk was always evaluated inside of icmp*_xmit_lock()
exactly because we needed the BH disabling in order for the
smp_processor_id() evaluation to be safe.

It's probably better to pass the namespace into the lock and
unlock routines.