From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S933937AbXCFEWl (ORCPT ); Mon, 5 Mar 2007 23:22:41 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S933938AbXCFEWl (ORCPT ); Mon, 5 Mar 2007 23:22:41 -0500 Received: from smtp.cce.hp.com ([161.114.21.24]:46090 "EHLO ccerelrim03.cce.hp.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S933937AbXCFEWk (ORCPT ); Mon, 5 Mar 2007 23:22:40 -0500 Subject: Re: [patch] Add insmod option to force the use of the backup timer. From: Alex Williamson To: Andrew Morton Cc: Gerd Hoffmann , Dave Jones , linux-kernel@vger.kernel.org, Kevin Stansell In-Reply-To: <20070305190333.c01a7ffc.akpm@linux-foundation.org> References: <20070228102346.115707000@suse.de> <20070228185505.GA17193@redhat.com> <45E6D309.5050409@suse.de> <20070305190333.c01a7ffc.akpm@linux-foundation.org> Content-Type: text/plain Organization: HP OSLO R&D Date: Mon, 05 Mar 2007 21:22:32 -0700 Message-Id: <1173154952.5941.142.camel@bling> Mime-Version: 1.0 X-Mailer: Evolution 2.8.1 Content-Transfer-Encoding: 7bit X-PMX-Version: 5.2.1.279297, Antispam-Engine: 2.4.0.264935, Antispam-Data: 2007.3.5.200935 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Mon, 2007-03-05 at 19:03 -0800, Andrew Morton wrote: > > Perhaps Alex can suggest some debugging steps we can take to work out > why the test isn't triggering? I was lucky and had a hardware description of the bug I was trying to work around with the 8250 backup timer patch. If the UART on the system in question is an integrated component, it might be worthwhile to start with the errata documented in the hardware manual. The detection test is simply looking for UARTs that don't re-assert the transmit holding register empty interrupt when it becomes re-enabled. Since the backup timer is successfully kicking the UART back into action, it would be interesting to know whether this is because the "Diva test" is detecting work to do or if it's just a matter of reading the IIR bits (the interrupt is there, but not getting delivered). The state of the UART at that point may be a clue how to detect the problem. UARTs often seem to be the most troubling part of a system, so I'm not opposed to a boot/module option, but auto-detection provides a much better user experience. Thanks, Alex -- Alex Williamson HP Open Source & Linux Org.