From mboxrd@z Thu Jan 1 00:00:00 1970 From: Linus Torvalds Subject: Re: linux-next: origin tree build failure Date: Tue, 14 Oct 2008 17:34:57 -0700 (PDT) Message-ID: References: <20081015095916.f30c0979.sfr@canb.auug.org.au> <20081015000541.GA22570@opensource.wolfsonmicro.com> Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Return-path: Received: from smtp1.linux-foundation.org ([140.211.169.13]:47965 "EHLO smtp1.linux-foundation.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751434AbYJOAfL (ORCPT ); Tue, 14 Oct 2008 20:35:11 -0400 In-Reply-To: <20081015000541.GA22570@opensource.wolfsonmicro.com> Sender: linux-next-owner@vger.kernel.org List-ID: To: Mark Brown Cc: Stephen Rothwell , linux-next@vger.kernel.org, LKML , Liam Girdwood On Wed, 15 Oct 2008, Mark Brown wrote: > > That would be people on the ARM list. Unless I misremember I first > heard about using NO_IRQ from the ARM list in connection with another > (much less ARM-specific) driver only within the past six months. ARM should just do #define NO_IRQ (0) and be done with it. It's what powerpc did long ago. That allows old broken drivers to work, and doesn't require new drivers to introduce more NO_IRQ insanity. If there is a "hardware irq 0", either it can be special (eg some system-only irq not used by any drivers) or the irq numbers can/must be remapped. ARM already converted to the generic irq layer, and there are actually parts of that that know that irq 0 has special meaning. Like the autoprobing and spurious irq handling. Of course, ARM may not care, and the generic code is generally trying to avoid any assumptions, but it's still true that all generic _drivers_ do the whole "if (irq)" thing. Yeah, yeah, we've let NO_IRQ poison some subsystems (like IDE and apparently parts of i2c), but that was a mistake. Linus