From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932881AbbFIG66 (ORCPT ); Tue, 9 Jun 2015 02:58:58 -0400 Received: from verein.lst.de ([213.95.11.211]:46169 "EHLO newverein.lst.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932418AbbFIG64 (ORCPT ); Tue, 9 Jun 2015 02:58:56 -0400 Date: Tue, 9 Jun 2015 08:58:54 +0200 From: Christoph Hellwig To: Dan Williams Cc: linux-kernel@vger.kernel.org, axboe@kernel.dk, boaz@plexistor.com, david@fromorbit.com, linux-arch@vger.kernel.org, linux-fsdevel@vger.kernel.org, arnd@arndb.de, ross.zwisler@linux.intel.com, Konrad Rzeszutek Wilk , benh@kernel.crashing.org, linux-nvdimm@ml01.01.org, heiko.carstens@de.ibm.com, hch@lst.de, tj@kernel.org, paulus@samba.org, hpa@zytor.com, schwidefsky@de.ibm.com, willy@linux.intel.com, akpm@linux-foundation.org, torvalds@linux-foundation.org, mingo@kernel.org Subject: Re: [PATCH v4 9/9] x86: convert dma_map_ops to support mapping a __pfn_t. Message-ID: <20150609065854.GD10167@lst.de> References: <20150605205052.20751.77149.stgit@dwillia2-desk3.amr.corp.intel.com> <20150605211955.20751.15047.stgit@dwillia2-desk3.amr.corp.intel.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20150605211955.20751.15047.stgit@dwillia2-desk3.amr.corp.intel.com> User-Agent: Mutt/1.5.17 (2007-11-01) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Fri, Jun 05, 2015 at 05:19:55PM -0400, Dan Williams wrote: > As long as a dma_map_sg() implementation avoids sg_page() conversions it > can support scatterlists that carry "page-less" __pfn_t entries. > However, a couple implementations require that __pfn_t_has_page() is > always true. The Xen swiotlb implementation's entanglements with ARM and > the Calgary MMUs requirement to have a pre-existing virtual mapping make > them unable to support this conversion (i.e. these now have 'depends on > !HAVE_DMA_PFN'). That's why we really need a whole kernel conversion and not just a piecemail one. Given how trivial this patch is that doesn't look like a too big task ayway.