Re: [patch/rfc] perf on raspberry-pi without overflow interrupt

[Peter Zijlstra <peterz@infradead.org>]

On Thu, Jan 16, 2014 at 12:13:10PM -0500, Vince Weaver wrote:
> On Tue, 14 Jan 2014, Peter Zijlstra wrote:
> 
> > On Mon, Jan 13, 2014 at 11:55:17PM -0500, Vince Weaver wrote:
> 
> > > I don't see an existing pmu capabilities struct... or do you mean
> > > coming up with one?
> > 
> > Yeah, adding one.
> 
> So would it be a struct, or just an integer with flags?

I hadn't really considered that; per the below proposed names I suppose
I was thinking of 'unsigned long flags' with #define PMU_HAS_flags. But
a named struct would work too I suppose.

> > > Would it only hold an "overflow_interrupt_available" flag, or are
> > > there other generic capabilities it would be handy to know about?
> > 
> > Possible (other) flags could be:
> > 
> >   PMU_HAS_INT      -- would allow sampling events
> >   PMU_HAS_PRECISE  -- would allow any ::precise value
> >   PMU_HAS_FILTER   -- would allow all os/user/etc. flags
> 
> should we export these to userspace somehow?
> 
> It would be handy to be able to tell you're getting EOPNOTSUP because
> PMU_HAS_INT is not set for your pmu, rather than trying to guess why 
> things are failing.

Yeah I suppose we could do something like that. Maybe something like:

# cat /sys/bus/event_source/devices/cpu/flags
int precise filter

?