From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S266614AbUGUUIa (ORCPT ); Wed, 21 Jul 2004 16:08:30 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S266616AbUGUUIa (ORCPT ); Wed, 21 Jul 2004 16:08:30 -0400 Received: from mtvcafw.sgi.com ([192.48.171.6]:16480 "EHLO omx3.sgi.com") by vger.kernel.org with ESMTP id S266614AbUGUUI2 (ORCPT ); Wed, 21 Jul 2004 16:08:28 -0400 From: Jesse Barnes To: Benjamin Herrenschmidt Subject: Re: reserve legacy io regions on powermac Date: Wed, 21 Jul 2004 16:07:11 -0400 User-Agent: KMail/1.6.2 Cc: Olaf Hering , Linux Kernel list , Paul Mackeras References: <20040721091249.GA1336@suse.de> <1090421466.2002.24.camel@gaston> In-Reply-To: <1090421466.2002.24.camel@gaston> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <200407211607.11915.jbarnes@engr.sgi.com> Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Wednesday, July 21, 2004 10:51 am, Benjamin Herrenschmidt wrote: > On Wed, 2004-07-21 at 05:12, Olaf Hering wrote: > > Anton pointed this out. > > > > ppc32 can boot one single binary on prep, chrp and pmac boards. > > pmac has no legacy io, probing for PC style legacy hardware leads to a > > hard crash. > > Several patches exist to prevent serial, floppy, ps2, parport and other > > drivers from probing these io ports. > > I think the simplest fix for 2.6 is a request_region of the problematic > > areas. > > PCMCIA is still missing. > > I found that partport_pc.c pokes at varios ports, without claiming the > > ports first. Should this be fixed? > > smsc_check(), winbond_check(), winbond_check2() > > Note that this is still all workarounds... Nothing prevents you (and some > people actually do that) to put a PCI card with legacy serial ports on it > inside a pmac.... Can you actually support this? Will it work? As for the patch, why not just reserve the first 64k of address space altogether instead of the individual ports? Jesse