Re: [PATCH -v3] perf, x86: try to handle unknown nmis with running perfctrs

[Cyrill Gorcunov <gorcunov@gmail.com>]

On Mon, Aug 23, 2010 at 10:53:39AM +0200, Ingo Molnar wrote:
> 
> * Ingo Molnar <mingo@elte.hu> wrote:
> 
> > 
> > * Don Zickus <dzickus@redhat.com> wrote:
> > 
> > > I'll test tip later today to see if I can reproduce it.
> > > 
> > > Cheers,
> > > Don
> > > 
> > > Ingo Molnar <mingo@elte.hu> wrote:
> > > 
> > > >
> > > >it's not working so well, i'm getting:
> > > >
> > > > Uhhuh. NMI received for unknown reason 00 on CPU 9.
> > > > Do you have a strange power saving mode enabled?
> > > > Dazed and confused, but trying to continue
> > > >
> > > >on a nehalem box, after a perf top and perf stat run.
> > 
> > FYI, it does not trigger on an AMD box.
> 
> Ok, to not hold up the perf/urgent flow i zapped these two commits for 
> the time being:
> 
>  4a31beb: perf, x86: Fix handle_irq return values
>  8e3e42b: perf, x86: Try to handle unknown nmis with an enabled PMU
> 
> We can apply them if they take a form that dont introduce a different 
> kind of (and more visible) regression.
> 
> Thanks,
> 
> 	Ingo
> 

Btw, guys, I fail to see how new nmi_watchdog work, we have
default_do_nmi
  if (!(reason & 0xc0)) {
    if (notify_die(DIE_NMI, "nmi", regs, reason, 2, SIGINT) == == NOTIFY_STOP)
      return
    if (nmi_watchdog_tick(regs, reason))
      return

but perf_event_nmi_handler returns NOTIFY_STOP when watchdog is perf event
and nmi_watchdog_tick _never_ called, or (most probably) I miss something?

	-- Cyrill