Re: stpd kernel event filtering

["bibo,mao" <bibo dot mao at intel dot com>]

Frank Ch. Eigler wrote:
> "Mao, Bibo" <bibo.mao@intel.com> writes:
>
> > [... Maybe] there should be one additional option in stap. If this
> > option is on, all probepoint related with stpd will be discarded;
> > else user will get all real raw data including stpd process.
>
> That's a reasonable option.  The translator could emit the explicit
> conditional within the probe prologue, that is as long as the
> conditional can apply generally to all kinds of probes.  Actually,
> "begin" and "end" probes often run from the context of stpd, so they
> would have to skip the check.  And what about timers/profiling ...
> or interrupt handlers?
>
> - FChE
That actually is one problem, there is no explicit definition whether 
some kernel probepoint in interrupt context should be discarded, because 
  these events have no relation with stpd process.

bibo,mao