From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1759324Ab2EINNX (ORCPT ); Wed, 9 May 2012 09:13:23 -0400 Received: from cantor2.suse.de ([195.135.220.15]:41785 "EHLO mx2.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1759275Ab2EINNW (ORCPT ); Wed, 9 May 2012 09:13:22 -0400 Date: Wed, 9 May 2012 12:44:46 +0200 From: Jan Kara To: =?iso-8859-1?Q?Llu=EDs?= Batlle i Rossell Cc: Jan Kara , linux-kernel@vger.kernel.org Subject: Re: BUG on fs/inode.c:1442 (linux 3.3.1 and 3.3.2) Message-ID: <20120509104446.GB5092@quack.suse.cz> References: <20120415215601.GV1927@vicerveza.homeunix.net> <20120418114844.GA22295@quack.suse.cz> <20120506123100.GI1927@vicerveza.homeunix.net> MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Disposition: inline Content-Transfer-Encoding: 8bit In-Reply-To: <20120506123100.GI1927@vicerveza.homeunix.net> User-Agent: Mutt/1.5.20 (2009-06-14) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Sun 06-05-12 14:31:00, Lluís Batlle i Rossell wrote: > On Wed, Apr 18, 2012 at 01:48:44PM +0200, Jan Kara wrote: > > Hello, > > > > On Sun 15-04-12 23:56:01, Lluís Batlle i Rossell wrote: > > > destroying my openvpn client connection (SIGINT to openvp), in linux 3.3.1 and > > > now also in 3.3.2, I noticed this BUG in dmesg (attached). > > > > > > It's a vanilla 3.3.2, at this shot. > > > > > > I know it never happened to me in any 3.2, but I did not try 3.3.0. > > > > > > I attach the .config. And I have the debug info for this kernel too, if this > > > helps someone find a fix. But I imagine it's easy to reproduce. > > From the first look it would seem as use after free bug but can you > > please post disassembly of iput() function from your kernel? I.e. you load > > vmlinux in gdb and run 'disass iput'. Thanks. > > Sorry for the delay. Here it is, for 3.3.2: Thanks for the disassembly. I was hoping that the contents of inode->i_state would be loaded in some register so we can see what value it has. Sadly that's not the case so I cannot say anything more. Did the bug happen to you again after the original report? Honza -- Jan Kara SUSE Labs, CR