From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Arve Knudsen" Subject: Re: System freeze when transferring large files (libata) Date: Sat, 10 Jul 2004 18:52:45 +0200 Sender: linux-ide-owner@vger.kernel.org Message-ID: References: <40EFEAE4.3060201@pobox.com> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII; format=flowed delsp=yes Content-Transfer-Encoding: 7BIT Return-path: Received: from mail.broadpark.no ([217.13.4.2]:52950 "EHLO mail.broadpark.no") by vger.kernel.org with ESMTP id S266296AbUGJQvR (ORCPT ); Sat, 10 Jul 2004 12:51:17 -0400 In-Reply-To: <40EFEAE4.3060201@pobox.com> List-Id: linux-ide@vger.kernel.org To: Jeff Garzik Cc: "linux-ide@vger.kernel.org" On Sat, 10 Jul 2004 09:11:00 -0400, Jeff Garzik wrote: > Arve Knudsen wrote: >> I thought I'd add that this is on an Asus A7N8X-X motherboard, and the >> filesystem is ext3. >> On Fri, 09 Jul 2004 18:03:18 +0200, Arve Knudsen >> wrote: >> >>> I've been having this problem lately when transferring large amounts >>> of data to my Maxtor DiamondMax Plus 9 SATA 120GB drive, which is >>> connected to a Promise SATA150 TX2 Plus controller. It happened once >>> with an -mm kernel, but I figured it was probably down to instability >>> in the -mm patchset. However it just happened to me again with a >>> vanilla 2.6.6 kernel, when untarring a backup. The system froze >>> completely, I couldn't even use alt+sysrq+b to reboot. Are there any >>> known issues with the Promise driver that can lead to such lockups? >>> I've recently run a disc check which found no problems (it wasn't >>> Maxtor's own though, but Seagate's as the Maxtor tools won't work >>> with the controller). > > > No known bugs at least. Promise seems to be more stable than most, in > fact. > > Maybe you have an intermittent problem with your SATA cables. I've reproduced the error after untarring approximately 1.7GB to the Maxtor disk (after adding noapic and nolapic to my kernel parameters, as suggested by Patrick Dreker), but I was able to untar 12GB to my Seagate disk (the main drive) successfully on the same controller. I will try to swap cables/ports later, and see if writing to the Maxtor still hangs. Question: Is it possible to have the kernel abort an i/o operation which is obviously stuck, so that a possible disk problem doesn't cause the whole system to hang? Thanks Arve Knudsen