From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S261513AbULBATy (ORCPT ); Wed, 1 Dec 2004 19:19:54 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S261521AbULBATo (ORCPT ); Wed, 1 Dec 2004 19:19:44 -0500 Received: from atlrel7.hp.com ([156.153.255.213]:29371 "EHLO atlrel7.hp.com") by vger.kernel.org with ESMTP id S261513AbULBASv (ORCPT ); Wed, 1 Dec 2004 19:18:51 -0500 From: Bjorn Helgaas To: Adrian Bunk Subject: Re: 2.6.10-rc2-mm4 Date: Wed, 1 Dec 2004 17:18:42 -0700 User-Agent: KMail/1.7.1 Cc: Andrew Morton , Linus Torvalds , linux-kernel@vger.kernel.org, Len Brown References: <20041130095045.090de5ea.akpm@osdl.org> <20041201211036.GQ2650@stusta.de> In-Reply-To: <20041201211036.GQ2650@stusta.de> MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200412011718.42740.bjorn.helgaas@hp.com> Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Wednesday 01 December 2004 2:10 pm, Adrian Bunk wrote: > > add-acpi-based-floppy-controller-enumeration.patch > > Add ACPI-based floppy controller enumeration. > > As far as I understood the discussion, this patch should be dropped. As I understand it, Len & Linus have figured out how to fiddle with ELCR[1] in such a way that when ACPI disables a PCI link device that happens to be on IRQ6, the ELCR polarity doesn't get screwed up. So the floppy driver can still blindly probe for its device without getting an interrupt storm. But the BIOS is still telling the OS that there's no floppy controller, and Linux still isn't listening. In the case of floppy, maybe that's OK, because all arches that support floppy seem to make it safe to do blind probing. But in the case of i8042, IDE, and IPMI, I think we definitely *should* do either ACPI or PNP-ACPI enumeration. These devices are all optional on ia64, and at least on HP hardware, the only reason we configure the box to allow blind probing is so these deaf drivers continue to work. [1] http://linux.bkbits.net:8080/linux-2.5/cset%4041a2c479tEbbKs1AxXHrR-LgHzPXzA