From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Michael Chan" Subject: Re: bnx2_poll panicking kernel Date: Thu, 10 Jul 2008 15:26:03 -0700 Message-ID: <1215728763.6901.23.camel@dell> References: <4863DA40.9050508@broadcom.com> <20080709164619.GA26098@orion.carnet.hr> <1215622625.5532.43.camel@dell> <20080709.164659.157912631.davem@davemloft.net> <1215723617.6901.3.camel@dell> <20080710212324.GA22521@orion.carnet.hr> <1215725921.6901.10.camel@dell> <20080710220030.GA29558@orion.carnet.hr> Mime-Version: 1.0 Content-Type: text/plain Content-Transfer-Encoding: 7bit Cc: "David Miller" , "billfink@mindspring.com" , "bhutchings@solarflare.com" , netdev , "mirrors@debian.org" To: "Josip Rodin" Return-path: Received: from mms3.broadcom.com ([216.31.210.19]:3284 "EHLO MMS3.broadcom.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751954AbYGJWXN (ORCPT ); Thu, 10 Jul 2008 18:23:13 -0400 In-Reply-To: <20080710220030.GA29558@orion.carnet.hr> Sender: netdev-owner@vger.kernel.org List-ID: On Thu, 2008-07-10 at 15:00 -0700, Josip Rodin wrote: > > This is the same crash that you originally reported. The driver was > > referencing a NULL SKB because the skb_shinfo(skb)->nr_frags was > > modified on the previous packet in the TX ring. > > > > If you had applied my debug patch #1 that I posted several weeks ago, it > > should have avoided this because we should have saved away the nr_frags > > for every packet on the TX ring. > > It was that same patched kernel that crashed nevertheless. > That's strange. You also saw lots of "Corrupted SKB..." dmesg as well, right? May be the debug patch is not bullet-proof although I cannot see how we can still get a NULL SKB in the TX completion path. Anyway, please try today's patch as we need to find the source of the problem. Thanks.