From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id ; Fri, 9 Aug 2002 05:31:28 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id ; Fri, 9 Aug 2002 05:31:26 -0400 Received: from wsip68-15-8-100.sd.sd.cox.net ([68.15.8.100]:12928 "EHLO gnuppy.monkey.org") by vger.kernel.org with ESMTP id ; Fri, 9 Aug 2002 05:31:23 -0400 Date: Fri, 9 Aug 2002 02:34:55 -0700 To: adilger@clusterfs.com Cc: "Bill Huey (Hui)" , mingo@elte.hu, linux-kernel@vger.kernel.org Subject: Re: ext3 journal/IDE problems ? (softirq + sched assert ?) Message-ID: <20020809093455.GA1000@gnuppy.monkey.org> References: <20020809040456.GA786@gnuppy.monkey.org> <20020809063725.GB12482@clusterfs.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20020809063725.GB12482@clusterfs.com> User-Agent: Mutt/1.4i From: Bill Huey (Hui) Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Fri, Aug 09, 2002 at 12:37:25AM -0600, Andreas Dilger wrote: > On Aug 08, 2002 21:04 -0700, Bill Huey wrote: > > What's going on with this ? > > > > I get: > > EXT3-fs error (device ide0(3,5)): ext3_free_blocks: bit already cleared for block 753210 > > : > > : > > EXT3-fs error (device ide0(3,5)): ext3_free_blocks: bit already cleared for block 753273 > > ext3_free_blocks: aborting transaction: Journal has aborted in __ext3_journal_get_undo_access<2>EXT3-fs error (device ide0(3,5)) in ext3_free_blocks: Journal has aborted > > Looks like you got a block of zeros from disk when it should have been a > block bitmap, or your filesystem is otherwise corrupted. You need to do > a full fsck on this filesystem. > > As for cause, I have no idea. IDE DMA, IDE cables, memory, kernel bug... It's certainly a kernel bug. This machine has been reliable for years with stable kernels and I started to run the 2.5 series over the last couple of days just to test out the softirq stuff (+ preempt) and latency. I applied Mingo's patch for a recent scheduler assertion bug and it looks like some kind of race is possibly happening in the IDE layer. And Mingo, yes, the patch you sent me did help, but it looks like there's other problems. Thanks bill