From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754638Ab2G3SD4 (ORCPT ); Mon, 30 Jul 2012 14:03:56 -0400 Received: from opensource.wolfsonmicro.com ([80.75.67.52]:44801 "EHLO opensource.wolfsonmicro.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754563Ab2G3SDz (ORCPT ); Mon, 30 Jul 2012 14:03:55 -0400 Date: Mon, 30 Jul 2012 19:03:52 +0100 From: Mark Brown To: Stephen Warren Cc: Samuel Ortiz , linux-kernel@vger.kernel.org, Stephen Warren Subject: Re: [PATCH] mfd: arizona: convert to regmap_add_irq_chips Message-ID: <20120730180352.GR4468@opensource.wolfsonmicro.com> References: <1343415776-27176-1-git-send-email-swarren@wwwdotorg.org> <20120729203854.GF4384@opensource.wolfsonmicro.com> <5016BEC4.5000600@wwwdotorg.org> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <5016BEC4.5000600@wwwdotorg.org> X-Cookie: Give him an evasive answer. User-Agent: Mutt/1.5.21 (2010-09-15) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Mon, Jul 30, 2012 at 11:05:08AM -0600, Stephen Warren wrote: > On 07/29/2012 02:38 PM, Mark Brown wrote: > > On Fri, Jul 27, 2012 at 01:02:56PM -0600, Stephen Warren wrote: > >> 1) regmap_add_irq_chips() calls regmap_add_irq_chip() with irq==0 rather > >> than -1, so in turn irq_domain_add_linear() is called rather than > >> irq_domain_add_legacy(). This change could be avoided by providing an > >> irq_bases array to regmap_add_irq_chips(). > > This is a problem. > OK, can you explain why? Is the problem the difference between the two > types of IRQ domain? I would have assumed this was an internal detail of > the driver hence not an issue. I assume there's no issue with > known/static IRQ numbers, since both 0 and -1 end up dynamically > allocating the IRQ bases IIRC. We have GPIOs we might want to do interrupts on, if the API doesn't support providing a base we've got an issue. > >> 2) regmap_add_irq_chips() requests the top-level interrupt itself, so this > >> happens before the Arizona driver hooks the child BOOT_DONE and > >> CTRLIF_ERR interrupts. In the original, all the IRQ chips were created > >> first, and then the top-level IRQ was requested. This may cause a > >> functional difference if those interrupts are pending at probe() time. > > Boot done is very likely to be asserted. > Hmmm. Perhaps that means regmap_add_irq_chips() should be split into two > parts; one to create all the IRQ chips and hook everything together, and > the second to actually enable the interrupt. That's what the existing arizona code does, it only requests the primary IRQ line once it's completed setup.