From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Dumazet Subject: Re: [PATCH] [PATCH] Fix deadlock in af_packet while stressing raw ethernet socket interface Date: Tue, 12 Jul 2011 10:23:06 +0200 Message-ID: <1310458986.4763.4.camel@edumazet-HP-Compaq-6005-Pro-SFF-PC> References: <20110711.202717.1764669415135298036.davem@davemloft.net> <1310453928.2860.32.camel@edumazet-laptop> <20110712.001953.709404117315345178.davem@davemloft.net> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: ronny.meeus@gmail.com, netdev@vger.kernel.org To: David Miller Return-path: Received: from mail-ww0-f44.google.com ([74.125.82.44]:37829 "EHLO mail-ww0-f44.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751259Ab1GLIXS (ORCPT ); Tue, 12 Jul 2011 04:23:18 -0400 Received: by wwe5 with SMTP id 5so4685985wwe.1 for ; Tue, 12 Jul 2011 01:23:17 -0700 (PDT) In-Reply-To: <20110712.001953.709404117315345178.davem@davemloft.net> Sender: netdev-owner@vger.kernel.org List-ID: Le mardi 12 juillet 2011 =C3=A0 00:19 -0700, David Miller a =C3=A9crit = : > From: Eric Dumazet > Date: Tue, 12 Jul 2011 08:58:48 +0200 >=20 > > This seems a bug, but in softirq handling in your arch >=20 > That's not the problem, it's the dev_queue_xmit_nit() code > path that interrupt is interrupting. Isnt it what I said ? Unless I am mistaken, dev_queue_xmit_nit() is called from dev_hard_start_xmit(), and from dev_queue_xmit(), therefore BH are/should_be masked. Calling do_softirq() while BH are masked is the bug. Something must be messing the !in_interrupt() test done from irq_exit()