From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751137AbVHXQay (ORCPT ); Wed, 24 Aug 2005 12:30:54 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1751138AbVHXQay (ORCPT ); Wed, 24 Aug 2005 12:30:54 -0400 Received: from e32.co.us.ibm.com ([32.97.110.130]:5082 "EHLO e32.co.us.ibm.com") by vger.kernel.org with ESMTP id S1751137AbVHXQax (ORCPT ); Wed, 24 Aug 2005 12:30:53 -0400 Date: Wed, 24 Aug 2005 11:29:59 -0500 To: John Rose Cc: paulus@samba.org, benh@kernel.crashing.org, akpm@osdl.org, Greg KH , linux-kernel@vger.kernel.org, linuxppc64-dev@ozlabs.org, linux-pci@atrey.karlin.mff.cuni.cz Subject: Re: [patch 8/8] PCI Error Recovery: PPC64 core recovery routines Message-ID: <20050824162959.GC25174@austin.ibm.com> References: <20050823231817.829359000@bilge> <20050823232143.003048000@bilge> <20050823234747.GI18113@austin.ibm.com> <1124898331.24668.33.camel@sinatra.austin.ibm.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1124898331.24668.33.camel@sinatra.austin.ibm.com> User-Agent: Mutt/1.5.9i From: Linas Vepstas Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Wed, Aug 24, 2005 at 10:45:31AM -0500, John Rose was heard to remark: > > +++ linux-2.6.13-rc6-git9/arch/ppc64/kernel/eeh_driver.c 2005-08-23 14:34:44.000000000 -0500 > > +/* > > + * PCI Hot Plug Controller Driver for RPA-compliant PPC64 platform. > > This probably isn't the right header description for this file :) Yes, this file is a little ball of ugliness that resulted from moving things out of the rpaphp directory; and, yes, it's rather un-reconstructed. I released it under the "release early" program. The meta-issue that I'd like to reach consensus on first is whether there should be any hot-plug recovery attempted at all. Removing hot-plug-recovery support will make many of the issues you raise to be moot. > > +++ linux-2.6.13-rc6-git9/include/asm-ppc64/prom.h 2005-08-23 13:31:52.000000000 -0500 > > int busno; /* for pci devices */ > > int bussubno; /* for pci devices */ > > int devfn; /* for pci devices */ > > How about a pointer to a struct of EEH fields? Folks are touchy about > adding anything PCI-specific to device nodes, especially since most DNs > aren't PCI at all. I attempted to remove all of the pci-related stuff from this struct, and got a crash in very very early boot (before the transition from real to virtual addressing). Not sure why, I was surprised. It seems related to the flattening of the device ndode tree. I'll try again soon. --linas