From mboxrd@z Thu Jan 1 00:00:00 1970 From: Grant Grundler Subject: Re: bnx2 fails to compile on parisc because of missing get_dma_ops() Date: Tue, 22 Jun 2010 11:14:45 -0600 Message-ID: <20100622171445.GB8644@lackof.org> References: <20100617.073653.193708702.davem@davemloft.net> <20100617234520S.fujita.tomonori@lab.ntt.co.jp> <20100617153051.GB8964@linux-sh.org> <20100622152823R.fujita.tomonori@lab.ntt.co.jp> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: lethal@linux-sh.org, davem@davemloft.net, mchan@broadcom.com, vapier@gentoo.org, JBottomley@novell.com, netdev@vger.kernel.org, linux-parisc@vger.kernel.org, linux-kernel@vger.kernel.org To: FUJITA Tomonori Return-path: Content-Disposition: inline In-Reply-To: <20100622152823R.fujita.tomonori@lab.ntt.co.jp> Sender: linux-parisc-owner@vger.kernel.org List-Id: netdev.vger.kernel.org On Tue, Jun 22, 2010 at 03:30:08PM +0900, FUJITA Tomonori wrote: ... > > I don't see any harm in letting drivers know whether we can support > > consistent DMA allocs for a given struct device or not though, even if > > the micro-optimization is marginal at best. > > I'm happier with exporting less DMA APIs to drivers because looks like > new original ways to use the APIs wrongly can be always invented. Agree. ... > There are architectures that need to turn off the CPU cache for > coherent memory, I can't find none of them that see if an address is > coherent or not in dma_is_consistent(). parisc "knows" primarily based on chipset and then checks CPU model. We hook in the correct dma_ops early in boot before any device drivers are probed. hth, grant