From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755430AbcBBRap (ORCPT ); Tue, 2 Feb 2016 12:30:45 -0500 Received: from mx1.redhat.com ([209.132.183.28]:54256 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755067AbcBBRan (ORCPT ); Tue, 2 Feb 2016 12:30:43 -0500 Date: Tue, 2 Feb 2016 12:30:34 -0500 From: Don Zickus To: Jeffrey Merkey Cc: linux-kernel@vger.kernel.org, akpm@linux-foundation.org, atomlin@redhat.com, cmetcalf@ezchip.com, fweisbec@gmail.com, hidehiro.kawai.ez@hitachi.com, mhocko@suse.cz, tj@kernel.org, uobergfe@redhat.com Subject: Re: [PATCH v5 3/3] Add BUG_XX() debugging hard/soft lockup detection Message-ID: <20160202173034.GG26637@redhat.com> References: <1454380428-31474-1-git-send-email-jeffmerkey@gmail.com> <1454380428-31474-3-git-send-email-jeffmerkey@gmail.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1454380428-31474-3-git-send-email-jeffmerkey@gmail.com> User-Agent: Mutt/1.5.23.1 (2014-03-12) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Mon, Feb 01, 2016 at 07:33:48PM -0700, Jeffrey Merkey wrote: > This patch series adds config options which can be set during compile to > direct the compiler to output a breakpoint instruction anywhere a BUG() > macro has been placed in the kernel to trigger the system to > enter a debugger if a bug is detected by the system. Use of this > compile time option also allows conditional breakpoints to be set in the > kernel with these currently used macros. > > This addition is extremely useful for debugging hard and soft lockups > real time and quickly from a console debugger, and other areas of the > kernel. Please remember to add version history, so I can tell what changed. > > Signed-off-by: Jeffrey Merkey > --- > kernel/watchdog.c | 10 ++++++++++ > 1 file changed, 10 insertions(+) > > diff --git a/kernel/watchdog.c b/kernel/watchdog.c > index b3ace6e..c504f7c 100644 > --- a/kernel/watchdog.c > +++ b/kernel/watchdog.c > @@ -44,6 +44,12 @@ > #define NMI_WATCHDOG_ENABLED (1 << NMI_WATCHDOG_ENABLED_BIT) > #define SOFT_WATCHDOG_ENABLED (1 << SOFT_WATCHDOG_ENABLED_BIT) > > +#ifdef CONFIG_DEBUG_BUG > +#define WATCHDOG_DEBUG_LOCKUP 1 > +#else > +#define WATCHDOG_DEBUG_LOCKUP 0 > +#endif > + > static DEFINE_MUTEX(watchdog_proc_mutex); > > #ifdef CONFIG_HARDLOCKUP_DETECTOR > @@ -358,6 +364,8 @@ static void watchdog_overflow_callback(struct perf_event *event, > else > dump_stack(); > > + BUG_ON(WATCHDOG_DEBUG_LOCKUP); > + I am not sure I am a fan of this. You are taking a known macro BUG_ON with known expectations and perversely converting it into an 'asm'. So now when folks read the code they scratch their heads why we are dumping the stack twice when in fact we are not. It seems misleading. :-/ I still don't understand why we can't use Ingo's or tglx's approach? Your changelog doesn't point out the problems there. Cheers, Don > /* > * Perform all-CPU dump only once to avoid multiple hardlockups > * generating interleaving traces > @@ -478,6 +486,8 @@ static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer) > else > dump_stack(); > > + BUG_ON(WATCHDOG_DEBUG_LOCKUP); > + > if (softlockup_all_cpu_backtrace) { > /* Avoid generating two back traces for current > * given that one is already made above > -- > 1.8.3.1 >