From mboxrd@z Thu Jan 1 00:00:00 1970 From: Thomas Petazzoni Subject: Re: [PATCH v2 0/7] of: setup dma parameters using dma-ranges and dma-coherent Date: Mon, 21 Apr 2014 20:19:11 +0200 Message-ID: <20140421201911.10783191@skate> References: <1397917972-6293-1-git-send-email-santosh.shilimkar@ti.com> <20140419182528.414b9b2e@skate> <53551E9D.9080107@ti.com> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <53551E9D.9080107@ti.com> Sender: linux-kernel-owner@vger.kernel.org To: Santosh Shilimkar Cc: "linux-kernel@vger.kernel.org" , "devicetree@vger.kernel.org" , "Strashko, Grygorii" , Russell King , Arnd Bergmann , Greg Kroah-Hartman , Linus Walleij , Grant Likely , Rob Herring , Catalin Marinas , Olof Johansson , "linux-arm-kernel@lists.infradead.org" , Gregory =?UTF-8?B?Q2zDqW1lbnQ=?= , Ezequiel Garcia , Lior Amsalem , Tawfik Bayouk List-Id: devicetree@vger.kernel.org Dear Santosh Shilimkar, On Mon, 21 Apr 2014 09:35:25 -0400, Santosh Shilimkar wrote: > > In mach-mvebu, what we do is that we register a bus notifier on the > > platform bus, so that we can set our custom DMA operations for all > > platform devices in the system. Should this be done in a different way > > after your series? > > > Nope. Since you have a very custom SOC specific case, you can continue > what you are doing. True, but as you said, the goal is to remove machine code. So instead of having just a 'dma-coherent' property, shouldn't we have a dma-method property, which could be dma-method = "coherent" or dma-method = "marvell,io-coherent" and therefore allow the DT binding to cover more use cases than just the default non-coherent and coherent DMA operations? Best regards, Thomas -- Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com