Re: [PATCH 2/3] perf: Move exclude_kernel tracepoint check to init event

[Arnaldo Carvalho de Melo <acme@kernel.org>]

Em Thu, Nov 26, 2015 at 08:49:10PM +0100, Jiri Olsa escreveu:
> On Thu, Nov 26, 2015 at 04:41:06PM -0300, Arnaldo Carvalho de Melo wrote:
> > Em Thu, Nov 26, 2015 at 08:34:48PM +0100, Jiri Olsa escreveu:
> > > We suppress events with attr::exclude_kernel set when
> > > the event is generated, so following capture will
> > > give no warning and produce no data:
> > > 
> > >   $ sudo perf record -e sched:sched_switch:u ls
> > >   $ sudo /perf script | wc -l
> > >   0
> > > 
> > > Checking the attr::exclude_kernel at the event init
> > > time and failing right away:
> > 
> > We can as well provide a better warning in the tooling side and don't
> > even ask this nonsensical combo to the kernel, right?
> 
> yep.. I'll update my todo ;-)

Thanks!
 
> > Arguably the kernel was doing what was asked for, its just that no
> > sched:sched_switch took place while in user space... ;-)
> 
> I think it's better to fail like that.. rather than spending
> time on figuring why my test did not give me any data ;-)

Up to peterz ;-)

- Arnaldo