Re: [PATCH v3 3/3] perf test: Add user space counter reading tests

From: Ian Rogers
Date: Wed Jul 20 2022 - 11:10:14 EST


On Wed, Jul 20, 2022 at 7:57 AM Vince Weaver <vincent.weaver@xxxxxxxxx> wrote:
>
> On Tue, 19 Jul 2022, Ian Rogers wrote:
>
> > These tests are based on test_stat_user_read in
> > tools/lib/perf/tests/test-evsel.c. The tests are modified to skip if
> > perf_event_open fails or rdpmc isn't supported.
> >
> > Signed-off-by: Ian Rogers <irogers@xxxxxxxxxx>
>
> > + .type = PERF_TYPE_HARDWARE,
> > + .config = event,
> > +#ifdef __aarch64__
> > + .config1 = 0x2, /* Request user access */
> > +#endif
>
> should this value have a name rather than being a "magic constant"?

Thanks! Firstly, this is just moving code around and so not a
regression introduced by this code:
https://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.git/tree/tools/lib/perf/tests/test-evsel.c?h=perf/core#n134
I don't believe there is an existing constant for this purpose. The
kernel isn't using one either:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/arm64/kernel/perf_event.c#n309

So I agree with you and hopefully this is something ARM will clean up.

Thanks,
Ian

> Vince