[PATCH i-g-t 17/28] tests/intel/xe_oa: Exclusive/concurrent access, rc6 and stress open close

[Dixit, Ashutosh]

On Thu, 20 Jun 2024 16:37:47 -0700, Umesh Nerlige Ramappa wrote:
>

Hi Umesh,

> > +/**
> > + * SUBTEST: oa-unit-concurrent-oa-buffer-read
> > + * Description: Test that we can read streams concurrently on all OA units
> > + */
> > +static void
> > +test_oa_unit_concurrent_oa_buffer_read(void)
> > +{
> > +	struct drm_xe_query_oa_units *qoa = xe_oa_units(drm_fd);
> > +
> > +	igt_fork(child, qoa->num_oa_units) {
> > +		struct drm_xe_engine_class_instance *hwe = oa_unit_engine(drm_fd, child);
> > +
> > +		/* No OAM support yet */
> > +		if (nth_oa_unit(drm_fd, child)->oa_unit_type != DRM_XE_OA_UNIT_TYPE_OAG)
> > +			exit(0);
> > +
> > +		test_blocking(40 * 1000 * 1000, false, 5 * 1000 * 1000, hwe);
>
> If this fails, there are no debug logs since this is being run from within
> a fork. Do you know if that can be fixed? It would be worthwhile to get the
> debug messages for this. Maybe the igt_asserts in the test can be changed
> to igt_assert_f with a message. If not now, we can create a enhancement
> issue and track it later.

For now I have just created this IGT issue for this:

https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/2178

So let's revisit this later since this is a pre-existing issue we see for
i915 too.

>
> Reviewed-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa at intel.com>

Thanks.
--
Ashutosh