Re: RFC: futex_wait() can DoS the tick

[Thomas Gleixner <tglx@linutronix.de>]

On Wed, 10 Jun 2015, Steven Rostedt wrote:

> On Wed, 10 Jun 2015 20:59:49 +0200 (CEST)
> Thomas Gleixner <tglx@linutronix.de> wrote:
> 
> > On Wed, 10 Jun 2015, Mike Galbraith wrote:
> > 
> > > On Wed, 2015-06-10 at 17:12 +0200, Thomas Gleixner wrote:
> > > 
> > > > Some more information about your symptoms in form of configuration,
> > > > extra patches, kernel traces etc. would be appreciated.
> > > 
> > > homer:~/tmp # perf record -C 3 -eirq:*,irq_vectors:local* -- sleep 10
> 
> I would recommend using trace-cmd especially if you add function
> tracing, as it's made more for fast tracing.
> 
> -- Steve
> 
> > > [ perf record: Woken up 2 times to write data ]
> > > [ perf record: Captured and wrote 1380.507 MB perf.data (19530595 samples) ]
> > > homer:~/tmp # perf script | grep irq_vectors:local_timer
> > >      ksoftirqd/3    19 [003]  5586.564238: irq_vectors:local_timer_entry: vector=239
> > >      ksoftirqd/3    19 [003]  5586.564239: irq_vectors:local_timer_exit: vector=239
> > >      ksoftirqd/3    19 [003]  5586.568238: irq_vectors:local_timer_entry: vector=239
> > >      ksoftirqd/3    19 [003]  5586.568239: irq_vectors:local_timer_exit: vector=239
> > >      ksoftirqd/3    19 [003]  5586.632218: irq_vectors:local_timer_entry: vector=239
> > 
> > That gap is odd. Can you fire up the function tracer please?

And please enable hrtimer events as well. Output of that cpu is good enough.

Thanks,

	tglx