Re: Yet more softlockups.

[Dave Hansen <dave@sr71.net>]

On 07/12/2013 08:45 AM, Dave Jones wrote:
> On Fri, Jul 12, 2013 at 08:38:52AM -0700, Dave Hansen wrote:
>  > Dave, for your case, my suspicion would be that it got turned on
>  > inadvertently, or that we somehow have a bug which bumped up
>  > perf_event.c's 'active_events' and we're running some perf code that we
>  > don't have to.
>  
> What do you 'inadvertantly' ? I see this during bootup every time.
> Unless systemd or something has started playing with perf, (which afaik it isn't)

I mean that somebody turned 'active_events' on without actually wanting
perf to be on.  I'd be curious how it got set to something nonzero.
Could you stick a WARN_ONCE() or printk_ratelimit() on the three sites
that modify it?

>  > But, I'm suspicious.  I was having all kinds of issues with perf and
>  > NMIs taking hundreds of milliseconds.  I never isolated it to having a
>  > real, single, cause.  I attributed it to my large NUMA system just being
>  > slow.  Your description makes me wonder what I missed, though.
> 
> Here's a fun trick:
> 
> trinity -c perf_event_open -C4 -q -l off
> 
> Within about a minute, that brings any of my boxes to its knees.
> The softlockup detector starts going nuts, and then the box wedges solid.

On my box, the same happens with 'perf top'. ;)

*But* dropping the perf sample rate has been really effective at keeping
me from hitting it, and I've had to use _lots_ of CPUs (60-160) doing
those NMIs at once to trigger the lockups.

Being able to trigger it with so few CPUs is interesting though.  I'll
try on my hardware.