[igt-dev] [PATCH i-g-t v3 2/2] tests/i915/gem_ctx_exec: Skip test for shared reset domain

[Dixit, Ashutosh]

On Mon, 17 Jan 2022 00:57:00 -0800, <priyanka.dandamudi at intel.com> wrote:
>
> @@ -292,11 +293,15 @@ static void nohangcheck_hostile(int i915)
>	ahnd = get_reloc_ahnd(i915, ctx->id);
>
>	igt_require(__enable_hangcheck(dir, false));
> +	has_sh_do = has_shared_reset_domain(i915, ctx);
>
>	for_each_ctx_engine(i915, ctx, e) {
>		igt_spin_t *spin;
>		int new;
>
> +		if (has_sh_do && (e->class == I915_ENGINE_CLASS_RENDER ||
> +			    e->class == I915_ENGINE_CLASS_COMPUTE))
> +			continue;

If we need to do this, let's not try to overly optimize things (which will
clutter up the code since has_shared_reset_domain() may get called from
multiple places) and just pass the engine into the function, something
like:

bool has_shared_reset_domain(int fd, struct intel_execution_engine2 *e2,
			     const intel_ctx_t *ctx)
{
	const struct intel_execution_engine2 *e;
	int count = 0;

	if (!(e2->class == I915_ENGINE_CLASS_RENDER ||
		e2->class == I915_ENGINE_CLASS_COMPUTE))
		return false;

	for_each_ctx_engine(fd, ctx, e)
		if (e->class == I915_ENGINE_CLASS_RENDER ||
			e->class == I915_ENGINE_CLASS_COMPUTE)
		count++;

	return count >= 2;
}

And then in the loop above just have:

	if (has_shared_reset_domain(fd, e, ctx))
		continue;

Also, I may be wrong, but it looks to me that this test should pass even if
we have shared resets, the test is not even using engine resets. What
failure are we seeing and on which platforms and with how many gt's? Does
the test fail only on DG2 or also on other products with RCS + CCS? Can you
paste the output after running the failing test?

Any idea anyone why this test would fail with a shared reset domain? Could
it be failing due to a different reason than a shared reset domain as
people pointed out on previous versions of the patch? In that case this
patch is incorrect.