From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756166AbYEOJoo (ORCPT ); Thu, 15 May 2008 05:44:44 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1752658AbYEOJoe (ORCPT ); Thu, 15 May 2008 05:44:34 -0400 Received: from mo11.iij4u.or.jp ([210.138.174.79]:51328 "EHLO mo11.iij4u.or.jp" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751699AbYEOJoe (ORCPT ); Thu, 15 May 2008 05:44:34 -0400 Date: Thu, 15 May 2008 18:41:26 +0900 To: andi@firstfloor.org Cc: fujita.tomonori@lab.ntt.co.jp, linux-kernel@vger.kernel.org, akpm@linux-foundation.org, muli@il.ibm.com, alexisb@us.ibm.com Subject: Re: [PATCH 0/2] x86: per-device dma_mapping_ops From: FUJITA Tomonori In-Reply-To: <482C02A4.80101@firstfloor.org> References: <87y76cym96.fsf@basil.nowhere.org> <20080515181619R.tomof@acm.org> <482C02A4.80101@firstfloor.org> Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Message-Id: <20080515184126O.tomof@acm.org> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Thu, 15 May 2008 11:30:12 +0200 Andi Kleen wrote: > FUJITA Tomonori wrote: > > On Thu, 15 May 2008 11:01:09 +0200 > > Andi Kleen wrote: > > > >> FUJITA Tomonori writes: > >> > >>> This patchset adds per-device dma_mapping_ops support for > >>> CONFIG_X86_64 like POWER architecture does. This change enables us to > >>> cleanly fix the Calgary IOMMU issue that some devices are not behind > >>> the IOMMU [1]. It also would be helpful to handle KVM PCI passthrough. > >> This makes it basically impossible to do stack ops, which some > >> people have been doing. > > > > Seems that I misunderstand what those people want. > > > > What those people want to do and how the stack ops achieve it? Or can > > you tell me where their patches are? > > I've seen it in two cases: first was for KVM IO bypass and the other was a > (unfinished) patch to support the NoDMA bitmaps on some systems. > > In this case you really want to do a wrapper around the existing ops > and extend the mapping. > > That worked fine by just replacing the global pointer, but will > be quite hard in your set up. Thanks, I thought that KVM people want to do it per device (in the first case). So with my patchse, they can replace the dma_ops pointer in dev_archdata with what they want.