Re: [PATCH 2/3] perf/x86/pebs: add workaround for broken OVFL status on HSW

[Peter Zijlstra]

On Thu, Mar 03, 2016 at 03:40:49PM -0800, Stephane Eranian wrote:
> On Thu, Mar 3, 2016 at 1:43 PM, Andi Kleen <ak@xxxxxxxxxxxxxxx> wrote:
> >
> > > +             /*
> > > +              * There are cases where, even though, the PEBS ovfl bit is set in
> > > +              * GLOBAL_OVF_STATUS, the PEBS events may also have their overflow bits
> > > +              * set for their counters. We must clear them here because they have
> > > +              * been processed as exact samples in the drain_pebs() routine. They
> > > +              * must not be processed again in the for_each_bit_set() loop for
> > > +              * regular samples below.
> > > +              */
> > > +             status &= ~cpuc->pebs_enabled;
> > > +             status &= x86_pmu.intel_ctrl | GLOBAL_STATUS_TRACE_TOPAPMI;
> >
> > If you just clear the bits here they will not be acked and stay around
> > forever in GLOBAL_STATUS, which causes other problems.
> >
> > You need a separate ack_status variable that contains all bits and is always
> > acked.
> 
> I understand that. You mean I need to that has all the bits that were set to
> call intel_pmu_ack_status().
> 
> But if you look at the code, and where I made the change, there is no more
> intel_pmu_ack_status() BEFORE you read the status again via
> intel_pmu_get_status().
> 
> So why would I need to keep another variable around?

I suspect Andi is having something along:

 lkml.kernel.org/r/1445458568-16956-1-git-send-email-andi@xxxxxxxxxxxxxx

applied to his tree.