From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1030235AbWGYXF2 (ORCPT ); Tue, 25 Jul 2006 19:05:28 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1030236AbWGYXF2 (ORCPT ); Tue, 25 Jul 2006 19:05:28 -0400 Received: from sccrmhc14.comcast.net ([204.127.200.84]:38339 "EHLO sccrmhc14.comcast.net") by vger.kernel.org with ESMTP id S1030235AbWGYXF1 (ORCPT ); Tue, 25 Jul 2006 19:05:27 -0400 Subject: Re: [discuss] Re: VIA x86-64 bootlogs needed From: Nicholas Miell To: Andi Kleen Cc: discuss@x86-64.org, linux-kernel@vger.kernel.org In-Reply-To: <200607260018.04851.ak@suse.de> References: <200607251824.30504.ak@suse.de> <1153848759.2661.5.camel@entropy> <200607260018.04851.ak@suse.de> Content-Type: text/plain Date: Tue, 25 Jul 2006 16:05:14 -0700 Message-Id: <1153868714.2661.18.camel@entropy> Mime-Version: 1.0 X-Mailer: Evolution 2.6.2 (2.6.2-1.fc5.5.0.njm.1) Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Wed, 2006-07-26 at 00:18 +0200, Andi Kleen wrote: > > MSI Master2-FAR with VIA K8T800, kernel-2.6.17-1.2157_FC5 > > Thanks. > > > With any luck, this APIC rework will fix this board's habit of > > spontaneously turning off interrupts at the IOAPIC level without the > > kernel's knowledge. > > Unlikely unfortunately. How does it look like when they get turned off? > I stop getting any interrupts from the device in question (either the on-board UHCI controllers or the on-board SATA controller) -- i.e. mouse movement gets jerky because uhci-hcd is operating purely by polling or I start getting libata timeouts in my logs. I have an abomination of a systemtap script which calls unmask_IO_APIC_irq directly which instantly "fixes" it when things go wrong. I also have another stap script which records stack traces when mask_IO_APIC_irq and unmask_IO_APIC_irq called in order to check if the IRQs are actually getting turned off without the kernel's knowledge, but when I run that in the background, the problem doesn't ever manifest itself. -- Nicholas Miell