From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S264061AbUGOKSl (ORCPT ); Thu, 15 Jul 2004 06:18:41 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S264571AbUGOKSl (ORCPT ); Thu, 15 Jul 2004 06:18:41 -0400 Received: from mion.elka.pw.edu.pl ([194.29.160.35]:1175 "EHLO mion.elka.pw.edu.pl") by vger.kernel.org with ESMTP id S264061AbUGOKSk (ORCPT ); Thu, 15 Jul 2004 06:18:40 -0400 From: Bartlomiej Zolnierkiewicz To: "Dave Woods" , Subject: Re: Problems with DMA on IDE/ServerWorks/Seagate. Date: Thu, 15 Jul 2004 12:24:45 +0200 User-Agent: KMail/1.5.3 Cc: "Stephanie Marasciullo" References: <7632915A8F000C4FAEFCF272A880344105095A@Ehost067.exch005intermedia.net> In-Reply-To: <7632915A8F000C4FAEFCF272A880344105095A@Ehost067.exch005intermedia.net> MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200407151224.45341.bzolnier@elka.pw.edu.pl> Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Thursday 15 of July 2004 02:16, Dave Woods wrote: > Please CC me on all replies, as I am not subscribed to the list. > > We have diagnosed a problem (file corruption) using Ultra DMA & IDE with > > ServerWorks OSB4 Chipset and Seagate drives under heavy disk I/O. > We were able to find a thread that discusses this problem in detail, > but dates back to 2001. We have nearly 100 boxes using this > configuration > with kernal 2.4.20, and we can reproduce the problem easily with a perl > script using md5 checksums. > > The applicable thread is given below: > http://www.uwsg.iu.edu/hypermail/linux/kernel/0109.3/1006.html > > We've disabled DMA on several machines, and that appears to fix > the problem. However, despite seeing this workaround mentioned > several times, we have not had any luck downshifting to mdma2. > Every time I attempt it, all disk accesses freeze up, and the box > has to be rebooted. > > Our question: short of disabling DMA on all of our machines, what > other alternatives are there? I've played with using the hdparm -p > parameter prior to issuing hdparm -X34 and hdparm -d1, but it > doesn't like it. Is there a simple fix to this problem out there? Upgrade to 2.4.26 and see if it fixes your problem.