From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S263775AbTJ1ACL (ORCPT ); Mon, 27 Oct 2003 19:02:11 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S263776AbTJ1ACL (ORCPT ); Mon, 27 Oct 2003 19:02:11 -0500 Received: from e5.ny.us.ibm.com ([32.97.182.105]:45954 "EHLO e5.ny.us.ibm.com") by vger.kernel.org with ESMTP id S263775AbTJ1ACH (ORCPT ); Mon, 27 Oct 2003 19:02:07 -0500 Date: Mon, 27 Oct 2003 16:01:13 -0800 From: Patrick Mansfield To: Alan Stern Cc: Pedro Larroy , Matthew Dharm , USB development list , linux-kernel@vger.kernel.org Subject: Re: aborts in usb-storage in branch 2.6 Message-ID: <20031027160113.A2582@beaverton.ibm.com> References: <20031023165953.GA1410@81.38.200.176> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.2.5.1i In-Reply-To: ; from stern@rowland.harvard.edu on Fri, Oct 24, 2003 at 05:15:13PM -0400 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Fri, Oct 24, 2003 at 05:15:13PM -0400, Alan Stern wrote: > On Thu, 23 Oct 2003, Pedro Larroy wrote: > > abort.log -> 2.6.0test8 > > devices -> 2.6.0test8 > > kernel.log -> 2.6.0testx where x < 8 > > > > In abort.log, is the paste of the log with kernel 2.6.0test8 in kern.log > > there is a log from 2.6.0test4 or 2.6.0test4-mm > > > > On previous 2.6.0testx kernels the bug appeared to trigger when ACPI debug > > statements were printed, notice there are aborts after ACPI printks. > > > > On 2.6.0test8 it seems to be triggered when doing concurrent access to the > > filesystem. > > > > I would like to help in anyway I can, but ACPI and USB is far beyond my > > kernel knowledge, so I'm not currently able to fix the bugs myself. > I looked at your logs, but I can't add much to what you already know. > This certainly does appear to be some sort of interrupt sharing/routing > problem -- the logs don't indicate anything specific to USB. However > there's no easy way to tell what the source of that problem is. The logs look like a SCSI timeout - a 30 second gap, followed by abort, test unit ready, and a resend of the timed out write. That does not help much. It could certainly be a lost interrupt. I did not see any timeout/aborts happening (immediately) after the acpi messages. > Maybe someone who knows more about ACPI and interrupt handling can help. I don't know ACPI, but I thought interrupt problems it might cause happened more consistently. -- Patrick Mansfield