Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932353AbVKRTq4 (ORCPT ); Fri, 18 Nov 2005 14:46:56 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S932362AbVKRTq4 (ORCPT ); Fri, 18 Nov 2005 14:46:56 -0500 Received: from mtaout2.012.net.il ([84.95.2.4]:45360 "EHLO mtaout2.012.net.il") by vger.kernel.org with ESMTP id S932353AbVKRTqz (ORCPT ); Fri, 18 Nov 2005 14:46:55 -0500 Date: Fri, 18 Nov 2005 21:45:48 +0200 From: Muli Ben-Yehuda Subject: Re: [PATCH] x86-64: dma_ops for DMA mapping -K4 In-reply-to: <20051117220348.GA9297@infradead.org> To: Christoph Hellwig , Andi Kleen , Linux-Kernel , Ravikiran G Thirumalai , niv@us.ibm.com, Jon Mason , Jimi Xenidis , Muli Ben-Yehuda , "Shai Fultheim (shai@scalex86.org)" Message-id: <20051118194548.GB3070@granada.merseine.nu> MIME-version: 1.0 Content-type: text/plain; charset=us-ascii Content-transfer-encoding: 7BIT Content-disposition: inline References: <20051117131622.GC11966@granada.merseine.nu> <20051117220348.GA9297@infradead.org> User-Agent: Mutt/1.5.11 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 1469 Lines: 34 On Thu, Nov 17, 2005 at 10:03:48PM +0000, Christoph Hellwig wrote: > On Thu, Nov 17, 2005 at 03:16:22PM +0200, Muli Ben-Yehuda wrote: > > Hi Andi, > > > > Here's the latest version of the dma_ops patch. The patch is against > > Linus's tree as of yesterday and applies cleanly to > > 2.6.15-rc1-git5. Tested on AMD64 and Intel EM64 (x366) with gart, > > swiotlb, nommu and iommu=off. Please apply... > > Any chance you could move struct dma_mapping_ops to generic code and > implement the dma_ operations ontop of them in linux/dma-mapping.h if > the arch sets a WANT_DMA_MAPPING_OPS symbol? This kind of switch is > duplicated in far too many architectures. I'm looking at this now. PPC, IA64 and parisc have something like this, as well as x86-64 with my patch applied. The others either have dma_ops with completely different semantics or choose the dma ops implementation at compile time. However, each arch needs a slightly different definition of dma_mapping_ops - I'm not sure having one definition catering to all archs is better than what we have now, considering the DMA mapping implementation is inherently arch specific. Cheers, Muli -- Muli Ben-Yehuda http://www.mulix.org | http://mulix.livejournal.com/ - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/