From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1760746Ab3B0St7 (ORCPT ); Wed, 27 Feb 2013 13:49:59 -0500 Received: from mail-pa0-f52.google.com ([209.85.220.52]:58474 "EHLO mail-pa0-f52.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1759862Ab3B0St6 (ORCPT ); Wed, 27 Feb 2013 13:49:58 -0500 Date: Wed, 27 Feb 2013 10:49:53 -0800 From: Dmitry Torokhov To: Mathieu Poirier Cc: Linus Torvalds , Dave Airlie , Linux Kernel Mailing List , arve@android.com, kernel-team@android.com, John Stultz , Alan Cox Subject: Re: [PATCH v4] drivers/tty: Folding Android's keyreset driver in sysRQ Message-ID: <20130227184953.GA28638@core.coreip.homeip.net> References: <1352665488-13818-1-git-send-email-mathieu.poirier@linaro.org> <20121119072525.GA16730@core.coreip.homeip.net> <512E4954.6060601@linaro.org> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <512E4954.6060601@linaro.org> 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 Wed, Feb 27, 2013 at 10:58:44AM -0700, Mathieu Poirier wrote: > On 13-02-27 09:57 AM, Linus Torvalds wrote: > > On Tue, Feb 26, 2013 at 11:33 PM, Dave Airlie wrote: > >> > >> It looks to me like the weak bit isn't working so well > >> > >> if (platform_sysrq_reset_seq) { > >> for (i = 0; i < ARRAY_SIZE(sysrq_reset_seq); i++) { > >> key = platform_sysrq_reset_seq[i]; > >> 6d: 66 8b 8c 00 00 00 00 mov 0x0(%eax,%eax,1),%cx > >> 74: 00 > >> > >> is around where it craps out. > >> gcc version 4.7.2 20121109 (Red Hat 4.7.2-8) (GCC) > >> Fedora 18 machine. > > > > Hmm. I would love to blame gcc, but no, I think the code is crap. > > > > The whole 'platform_sysrq_reset_seq[]' thing is broken in current git, > > and it apparently only happens to work by mistake for most of us. > > > > Doing a "grep" for it shows all three uses: > > > > git grep platform_sysrq_reset_seq > > > > extern unsigned short platform_sysrq_reset_seq[] __weak; > > if (platform_sysrq_reset_seq) { > > key = platform_sysrq_reset_seq[i]; > > > > and the thing is, if it is declared as an array (not a pointer), then > > I think it is perfectly understandable that when then testing the > > *address* of that array, gcc just says "you're stupid, you're testing > > something that cannot possibly be NULL, so I'll throw your idiotic > > test away". > > > > And gcc would be completely correct. That test is moronic. You just > > said that platform_sysrq_reset_seq[] was an external array, there is > > no way in hell that is NULL. > > > > Now, if it was a _pointer_, that would be a different thing entirely. > > A pointer can have a NULL value. A named array, not so much. > > > > So I *think* the fix might be something like the attached. Totally > > untested. It may compile, or it may not. > > > > Linus > > > > Your fix is compiling, running and yielding the correct results - > apologies about that. > Actually defining platform_sysrq_reset_seq as __weak array was my doing so the fault is actually mine. Mathieu had a function there originally. Thanks. -- Dmitry