From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1761558AbXGLJRS (ORCPT ); Thu, 12 Jul 2007 05:17:18 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1754489AbXGLJRG (ORCPT ); Thu, 12 Jul 2007 05:17:06 -0400 Received: from mtagate7.uk.ibm.com ([195.212.29.140]:64678 "EHLO mtagate7.uk.ibm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752513AbXGLJRD (ORCPT ); Thu, 12 Jul 2007 05:17:03 -0400 Date: Thu, 12 Jul 2007 12:17:00 +0300 From: Muli Ben-Yehuda To: Arjan van de Ven Cc: Andi Kleen , Jeff Garzik , Yinghai Lu , Andrew Morton , Linux-Kernel , Jon Mason Subject: Re: [PATCH 1/1] x86-64: introduce struct pci_sysdata to facilitate sharing of ->sysdata Message-ID: <20070712091700.GL4189@rhun.haifa.ibm.com> References: <20070711134514.GA4789@rhun.haifa.ibm.com> <1184203051.2693.3.camel@laptopd505.fenrus.org> <20070712091029.GH4189@rhun.haifa.ibm.com> <1184231585.2720.8.camel@laptopd505.fenrus.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1184231585.2720.8.camel@laptopd505.fenrus.org> User-Agent: Mutt/1.5.11 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Thu, Jul 12, 2007 at 05:13:05PM +0800, Arjan van de Ven wrote: > > > I very much start to dislike the untyped "sysdata"... I much > > > rather have separate fields for the different uses (like a IOMMU > > > field) that aren't going to share ever. Possibly even typed, but > > > for IOMMU that may be tricky.... > > > > Could you elaborate on what you mean here? > > > > If you mean instead of hanging a pci_sysdata off of struct > > pci_bus, hang a 'void *iommu' and an 'int node' directly off of > > the pci_bus, > > yes that is what I mean. > > > that looks like a lot of churn (most architecture use sysdata as > > an opaque pointer to an arch specific structure) and bloat (you > > would > > it's only a few bytes per pci device/bus. not all that bad for > having something which involves splitting a shared pointer into > logical users. > > (and heck, if you really really are worried we could make it a union > for some of them, although that defeats the "split is safer" > benefit) It's not just bloat and code churn, I also don't like teaching every architecture about every other architecture's use of sysdata, which is inherently architecture specific. I think the benefits of modularity here outweigh the benefits of a little type safety. Cheers, Muli