From mboxrd@z Thu Jan 1 00:00:00 1970 From: Pawel Moll Subject: Re: [PATCH v4 5/6] ARM: vexpress: DT-based support for Cortex-A5 and Cortex-A9 based tiles Date: Wed, 07 Dec 2011 19:06:28 +0000 Message-ID: <1323284788.32116.11.camel@hornet.cambridge.arm.com> References: <1323186229-22054-1-git-send-email-pawel.moll@arm.com> <1323186229-22054-6-git-send-email-pawel.moll@arm.com> <3527082.WMy5vX2qZB@wuerfel> <27584177.XzoGq1stOI@wuerfel> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <27584177.XzoGq1stOI@wuerfel> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: linux-arm-kernel-bounces@lists.infradead.org Errors-To: linux-arm-kernel-bounces+linux-arm-kernel=m.gmane.org@lists.infradead.org To: Arnd Bergmann Cc: "devicetree-discuss@lists.ozlabs.org" , "linux-arm-kernel@lists.infradead.org" List-Id: devicetree@vger.kernel.org On Tue, 2011-12-06 at 23:13 +0000, Arnd Bergmann wrote: > Actually, I have to take that back. Looking at both patch 5 and 6, > the dt-ca*.c files are almost identical and all the differences are about stuff > that you can find in the device tree: You are obviously right - I was rushing with that. Less is better then more ;-) Will respin tomorrow. > * The iotable gets initialized from "mrc p15, 4, %0, c15, c0, 0", which would > be fine if that worked on all machines, but in order to unify the two > files, I would recommend searching the flat device tree for the respective > node and only map it if present. > > * You have two ways of finding out the number of cores, but looking in > the device tree would just work either way. Those two are actually related, as it's all about memory mapped SCU in case of A5/A9 and CP15-controlled on A7/15 (the *smp_enable() is different as well). I'll probably just define a "scu" node compatible with "arm,cortex-a9-scu" and use the A5/A9 SMP callbacks if it's present (and create static mapping for it) or the A7/15 if it's missing. > * You set the twd_base unconditionally on a5/a9 but never on a7/a15. > This looks correct, but you could just as well see if a twd node exists > and use its base address. I'll reuse Rob's Highbank solution and binding. > * You only initialize the l2x0 on a5/a9. If a7/a15 don't have a matching > l2x0 device, then calling the same function unconditionally should be harmless. Yep, that's not a problem. Thanks for all the other ack-bys and review-bys! Pawel