From mboxrd@z Thu Jan 1 00:00:00 1970 From: Bruno =?UTF-8?B?UHLDqW1vbnQ=?= Subject: Re: XFS shutting down due to IO timeout on SATA disk (pata_via for CX700) Date: Sun, 21 Sep 2008 21:51:57 +0200 Message-ID: <20080921215157.449ef5b7@neptune.home> References: <20080911193511.7960bc82@neptune.home> <48CE22E5.9090403@kernel.org> <20080915190242.58d21a8f@neptune.home> <48CEC8F2.4040904@kernel.org> <20080916102152.3f35439f@pluto.restena.lu> <48CFEE51.2070200@kernel.org> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Return-path: Received: from ppp-111-195.adsl.restena.lu ([158.64.111.195]:40240 "EHLO bonbons.gotdns.org" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1750877AbYIUUCy (ORCPT ); Sun, 21 Sep 2008 16:02:54 -0400 In-Reply-To: <48CFEE51.2070200@kernel.org> Sender: linux-ide-owner@vger.kernel.org List-Id: linux-ide@vger.kernel.org To: Tejun Heo Cc: Linux Kernel , linux-ide@vger.kernel.org, Jeff Garzik , Mark Lord Hello, On Tue, 16 September 2008 Tejun Heo wrote: > > Just for the record, on the VIA there were no errors yet with > > 2.6.27-rc6-git3 though I ran some compiling on it. As such it > > *could* be that one of the patches pushed by Jeff helped: > > [libata] LBA28/LBA48 off-by-one bug in ata.h > > sata_inic162x: enable LED blinking > > ata: duplicate variable sparse warning > > The only patch which could have affected the result is the off-by-one > one; however, FLUSH timeout is a bit unexpected failure mode for the > problem. Can you please keep the system running for a whlie and see > whether the problem occurs? I haven't seen the error happen yet with 2.6.27-rc6-git3, so that off by one bug might have had an incidence. No warranty on that as I did also slightly change power-saving settings for the disk around the time I updated kernel (-B254, -S241 parameters to hdparm) and have not seen the error yet on 2.6.27-rc5-git9 since then. Will do more testing as time permits over the next days and report if I find anything that looks useful (e.g. better way than time to trigger the error or some of the hdparm options influencing the probability) In case there is a patch that would eventually be able to get more info when the error occurs I would be happy to apply it. Bruno