Re: [RFC -tip] perf, x86: Add PERF_COUNT_HW_NMI_WATCHDOG event v2

[Cyrill Gorcunov <gorcunov@gmail.com>]

On Tue, Jun 21, 2011 at 08:06:54PM +0200, Stephane Eranian wrote:
...
> 
> If in arch/x86/kernel/cpu/perf_event.c I add a callback which
> checks:
> 
>    if (netburst_host) {
>        attr.type = PERF_TYPE_RAW
>        attr.config = 0x3c
>    } else {
>       attr.type = PERF_TYPE_HARDWARE
>       attr.config = PERF_COUNT_HW__CPU_CYCLES
>   }
> 
> I simply don't think PERF_COUNT_HW_NMI_WATCHDOG should
> ever be exposed to users.

yes, and peterz pointed it out as well, i agree of course.

> 

So, you mean to put it into x86_pmu_event_init, right? How would
we know if this event comes from watchdog.c or not? And what if
user wanted to pass _exactly_ this crazy values to RAW event but
I will threat it as a sign for nmi-watchdog event and will map
it internally to P4_EVENT_EXECUTION_EVENT?

Don't get me wrong, Stephane, this approach will work and it's
not a problem (actually I think I would put -1ULL into config
since it's kinda "icorrect" and can be treated as special one)
but I think it introduce more complexity, it should not be like
this.

Gimme some time, probably I invent something more convenient.

	Cyrill