From mboxrd@z Thu Jan 1 00:00:00 1970 From: Rob Herring Subject: Re: [RFC PATCH] microblaze/irq: Change NO_IRQ to 0 Date: Wed, 07 Dec 2011 10:39:46 -0600 Message-ID: <4EDF96D2.9070808@gmail.com> References: <4EDEE2B2.7080606@gmail.com> <1323268911-6104-1-git-send-email-robherring2@gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: Sender: linux-kernel-owner@vger.kernel.org To: Linus Torvalds Cc: Michal Simek , John Linn , Grant Likely , Rob Herring , Ingo Molnar , Jeff Garzik , Stephen Rothwell , devicetree-discuss@lists.ozlabs.org, LKML , Randy Dunlap , Andrew Morton , Alan Cox , Anton Vorontsov List-Id: devicetree@vger.kernel.org Linus, On 12/07/2011 10:05 AM, Linus Torvalds wrote: > On Wed, Dec 7, 2011 at 6:41 AM, Rob Herring wrote: >> >> This patch changes the Microblaze NO_IRQ setting from -1 to 0 to bring >> it in line with most of the rest of the kernel. It also prepares for >> Microblaze eventually supporting multiple interrupt controllers by >> breaking the assumption that hwirq# == Linux IRQ#. The Linux IRQ >> number is just a cookie with no guarantee of a direct relationship >> with the hardware irq arrangement. > > This looks really nice and would probably be a great example for other > architectures with the same issue. > > My only (small) nit is that just from a debuggability standpoint I'd > actually suggest keeping the "translated" and "hardware" interrupt > numbers totally disjoint, which would seem to be trivial - make the > offset be 32 instead of 1. Sure, some NR_IRQ arrays would end up being > a bit bigger that way, but it would allow for trivially seeing whether > anybody reports the raw or translated irq numbers by just looking at > the number. That could reduce lots of confusion if somebody ends up > having the old non-translated number hardcoded etc. > > It would also potentially allow for actual debug checks - having code > like "WARN_ON_ONCE(irq < 32)" in some arch-specific irq controller > registration path etc, which is much harder to do with the overlapping > numbers. Agreed. I debated making it 16 just to skip the legacy interrupts as we're doing on ARM. So there's 2 benefits to skipping to 32. Anyway, I'll leave it to someone that can actually build and test this. > But it's nice to see how small the patch *can* be. If only ARM was this easy... Rob