From mboxrd@z Thu Jan 1 00:00:00 1970 From: Dave Chinner Subject: Re: BUG: sleeping function called from invalid context at kernel/rwsem.c:131 XFS? (was: Re: linux-next: Tree for October 17) Date: Wed, 22 Oct 2008 18:58:38 +1100 Message-ID: <20081022075838.GK18495@disturbed> References: <20081017165738.GA20818@infradead.org> <20081017203710.GA27187@infradead.org> <20081017135510.7127c4e7@infradead.org> <20081020163327.GA15651@infradead.org> <20081020223549.GA21152@disturbed> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Received: from ipmail01.adl6.internode.on.net ([203.16.214.146]:37881 "EHLO ipmail01.adl6.internode.on.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752219AbYJVH6q (ORCPT ); Wed, 22 Oct 2008 03:58:46 -0400 Content-Disposition: inline In-Reply-To: Sender: linux-next-owner@vger.kernel.org List-ID: To: Alexander Beregalov Cc: lachlan@sgi.com, Christoph Hellwig , Arjan van de Ven , xfs@oss.sgi.com, linux-next@vger.kernel.org, LKML On Tue, Oct 21, 2008 at 03:42:16PM +0400, Alexander Beregalov wrote: > 2008/10/21 Dave Chinner : > > On Mon, Oct 20, 2008 at 09:13:05PM +0400, Alexander Beregalov wrote: > >> 2008/10/20 Christoph Hellwig : > >> > On Mon, Oct 20, 2008 at 06:58:09PM +0400, Alexander Beregalov wrote: > >> >> Hi Christoph > >> >> > >> >> I have the same result with next-1020 and today's xfs-2.6.git/master > >> >> ( > >> >> commit bfd2bd10da76378dc4afd87d7d204a1d3d70b347 > >> >> Author: David Chinner > >> >> Date: Fri Oct 17 15:36:23 2008 +1000 > >> >> Inode: Allow external list initialisation > >> >> ) > >> Ha, that kernel (xfs/master) has made my system dead locked. > >> SysRQ-d: > >> Show all locks held in the system > >> 1 lock held by pdflush > >> (&type->s_umount_key#18{----}, at writeback_inodes > >> 1 lock held by login > >> (&(&ip->i_iolock)->mr_lock){----} at xfs_ilock > >> and so on ( many locks at xfs_ilock) > > > > Curious. Can you post the full stack traces? > I can not reproduce it, yet. > > Bisected to: > dd509097cb0b76d3836385f80d6b2d6fd3b97757 is first bad commit > commit dd509097cb0b76d3836385f80d6b2d6fd3b97757 > Author: Lachlan McIlroy > Date: Mon Sep 29 14:56:40 2008 +1000 > > [XFS] Unlock inode before calling xfs_idestroy() > > Lock debugging reported the ilock was being destroyed without being > unlocked. We don't need to lock the inode until we are going to insert it > into the radix tree. Ah, OK, I see the problem, though I don't understand why I'm not seeing the might_sleep() triggering all the time given that I always build with: $ grep SLEEP .config CONFIG_DEBUG_SPINLOCK_SLEEP=y Basically the above commit moved xfs_ilock() inside radix_tree_preload()/radix_tree_preload_end(), which means we are taking a rwsem() while we have an elevated preempt count. I'll get a patch out to fix it. The question is why isn't this triggering for me during QA runs? Cheers, Dave. -- Dave Chinner david@fromorbit.com