Date: Tue, 14 Apr 2009 15:23:19 +0200
From: Ingo Molnar <mingo@elte.hu>
To: Ed Tomlinson <edt@aei.ca>
Cc: Thomas Gleixner <tglx@linutronix.de>,
       Peter Zijlstra <a.p.zijlstra@chello.nl>, linux-kernel@vger.kernel.org
Subject: Re: 2.6.30-rc1 - nmi_watchdog broken?
Message-ID: <20090414132319.GB4403@elte.hu>
References: <200904140000.30155.edt@aei.ca> <20090414084232.GG27003@elte.hu> <200904140829.21493.edt@aei.ca>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <200904140829.21493.edt@aei.ca>
User-Agent: Mutt/1.5.18 (2008-05-17)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1747
Lines: 48


* Ed Tomlinson <edt@aei.ca> wrote:

> On Tuesday 14 April 2009 04:42:32 Ingo Molnar wrote:
> > 
> > * Ed Tomlinson <edt@aei.ca> wrote:
> > 
> > > Hi,
> > > 
> > > I've been having fun finding bugs in 30-rc1.  One of them is a 
> > > hard freeze.  I've not seen this type of problem on this hardware 
> > > before 30-rc1 - so I doubt if its hardware.  The best way I know 
> > > to debug a hard hang is with the nmi_watchdog.  I just cannot get 
> > > it to work.
> > 
> > [ Btw., have you tried CONFIG_PROVE_LOCKING=y - does it produce
> >   anything before or at the hard lockup point? ]
> > 
> > > The system is a 3 core amd cpu on a 790gx chipset.
> > > 
> > > If I boot the nmi_watchdog=1 it complains that lapci is not 
> > > available and the boot stops.  Same problem if I change the 
> > > clocksource to tsc, If I disable highres timers it panics.  If I 
> > > use nmi_watchdog=2 it panics. Am I doing something wrong or have I 
> > > hit a bug?
> > > 
> > > Logs of boots with and without highres timers inlined below.
> > 
> > hm, nmi_watchdog=1 acting funny is not unheard of. But 
> > nmi_watchdog=2 should really work. How does it panic, do
> > you have a capture of that?
> 
> I had not tried nmi_watchdog=2 highres=off.  This works.  Looks 
> like there is a conflict between highres timers and nmi_watchdog 
> here.

yes. Both use a limited resource of the lapic so we get one or the 
other.

( Might be fixable once we migrate the NMI watchdog code over to 
  perfcounters. )

	Ingo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/