From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753941AbaCJOAA (ORCPT ); Mon, 10 Mar 2014 10:00:00 -0400 Received: from devils.ext.ti.com ([198.47.26.153]:53522 "EHLO devils.ext.ti.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752651AbaCJN76 (ORCPT ); Mon, 10 Mar 2014 09:59:58 -0400 Message-ID: <531DC51F.50309@ti.com> Date: Mon, 10 Mar 2014 19:28:55 +0530 From: Sricharan R User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:11.0) Gecko/20120410 Thunderbird/11.0.1 MIME-Version: 1.0 To: Rob Herring CC: "devicetree@vger.kernel.org" , "linux-kernel@vger.kernel.org" , "linux-arm-kernel@lists.infradead.org" , linux-omap , , Nishanth Menon , Russell King - ARM Linux , Tony Lindgren , , Sekhar Nori , Vinod Koul , Santosh Shilimkar , Dan Williams Subject: Re: [RFC PATCH 0/9] drivers: dma: Add support for dma-channel router References: <1394194576-23741-1-git-send-email-r.sricharan@ti.com> In-Reply-To: Content-Type: text/plain; charset="ISO-8859-1" Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Hi Rob, On Sunday 09 March 2014 01:23 AM, Rob Herring wrote: > On Fri, Mar 7, 2014 at 6:16 AM, Sricharan R wrote: >> In some SoCs the dma request lines from the peripherals are >> routed to the dma-controller through a crossbar. With this the >> dma controller's available request lines are shared between the >> peripherals. >> >> This adds support to register the crossbar router associated with >> a dma-channel and let the dma-controller driver map/unmap >> the peripheral dma crossbar line to dma-controller's request >> line. > This is not a unique concept and should follow some existing pattern. > Perhaps something like interrupt-map. For the dmas property, the > parent should be the crossbar and then the crossbar block has to > translate that into the DMA controller request. All the DMA ctrlr > request connections should be described in the crossbar node. In > theory, you could have chained crossbars. This should be documented as > part of the generic DMA binding. The idea here was to do the map/unmap crossbar<->dma-request from the dma-controller at runtime just like any other resource, unlike interrupt-map fixed in DTS. I did not think about the chained crossbars case. I will add this here. And as you have suggested will make it work like irq_parse_and_map. Regards, Sricharan > There are also other similar IP like CoreSight CTI which are just > signal routers. So there is probably some possibility of common code > here. > > Rob