Re: [PATCH 4/6] watchdog: Boot-disable by default on full dynticks

[Frederic Weisbecker]

On Thu, Jun 13, 2013 at 10:45:15AM -0400, Don Zickus wrote:
> On Thu, Jun 13, 2013 at 04:22:12PM +0200, Frederic Weisbecker wrote:
> > On Thu, Jun 13, 2013 at 10:02:07AM -0400, Don Zickus wrote:
> > > On Thu, Jun 13, 2013 at 03:10:59PM +0200, Frederic Weisbecker wrote:
> > > > On Wed, Jun 12, 2013 at 01:03:16PM -0400, Don Zickus wrote:
> > > > > On Wed, Jun 12, 2013 at 04:02:36PM +0200, Frederic Weisbecker wrote:
> > > > > > When the watchdog runs, it prevents the full dynticks
> > > > > > CPUs from stopping their tick because the hard lockup
> > > > > > detector uses perf events internally, which in turn
> > > > > > rely on the periodic tick.
> > > > > > 
> > > > > > Since this is a rather confusing behaviour that is not
> > > > > > easy to track down and identify for those who want to
> > > > > > test CONFIG_NO_HZ_FULL, let's default disable the
> > > > > > watchdog on boot time when full dynticks is enabled.
> > > > > > 
> > > > > > The user can still enable it later on runtime using
> > > > > > proc or sysctl.
> > > > > 
> > > > > I thought we had a conversation awhile ago, where we agreed this was going
> > > > > to be fixed for 3.11?  Didn't Peter find the patch and apply it to his
> > > > > tree?  I am confused why this is still needed?
> > > > 
> > > > We agreed on the patch but it hasn't been applied yet. I'm trying to get
> > > > a sane series of nohz patches before sending to Ingo.
> > > 
> > > Peter,
> > > 
> > > Where is this patch?
> > 
> > Which patch? The old version of the current one? It was part of a previous series
> > that needed improvements so it hasn't been applied yet.
> 
> I guess I am confused.  I thought Peter said in an email awhile ago, that
> he found Stephan's patch (that converted perf to use hrtimers) and applied
> it to his tree.  Are you saying it was unapplied because it needed
> improvements?

I think it doesn't completely remove the perf tick. Besides it doesn't really
solve the problem since perf events will require periodic hrtimers to work,
which still defeats the purpose of full dynticks.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/