[igt-dev] [PATCH i-g-t] tests/intel/gem_eio: Skip stress test on BCS0 for MTL

[Kamil Konieczny]

Hi Nirmoy,
On 2023-10-06 at 10:56:41 +0200, Nirmoy Das wrote:
> Hi Kamil,
> 
> On 10/5/2023 12:54 PM, Kamil Konieczny wrote:
> > Hi Nirmoy,
> > On 2023-10-05 at 12:39:07 +0200, Nirmoy Das wrote:
> > > ping
> > > 
> > > On 10/2/2023 12:48 PM, Nirmoy Das wrote:
> > > > We do GGTT update on MTL using bcs engine, blocking that would
> > > > fail the test so skip such subtests on bcs engine for MTL.
> > > > 
> > > > Cc: Kamil Konieczny <kamil.konieczny at linux.intel.com>
> > > > Signed-off-by: Nirmoy Das <nirmoy.das at intel.com>
> > Please ask Sai from bug filing team for respin
> > (look at BAT fail).
> 
> repin didn't help :/  unrelated failures. Let me check with Sai.
> 
> 
> > 
> > > > ---
> > > >    tests/intel/gem_eio.c | 10 +++++++++-
> > > >    1 file changed, 9 insertions(+), 1 deletion(-)
> > > > 
> > > > diff --git a/tests/intel/gem_eio.c b/tests/intel/gem_eio.c
> > > > index 6d4b8f7df..99891ad15 100644
> > > > --- a/tests/intel/gem_eio.c
> > > > +++ b/tests/intel/gem_eio.c
> > > > @@ -1111,8 +1111,16 @@ static void test_reset_stress(int fd, unsigned int flags)
> > > >    	const intel_ctx_t *ctx0 = context_create_safe(fd);
> > > >    	uint64_t ahnd = get_reloc_ahnd(fd, ctx0->id);
> > > > -	for_each_ring(e, fd)
> > > > +	for_each_ring(e, fd) {
> > > > +		struct intel_execution_engine2 engine;
> > > > +
> > > > +		engine = gem_eb_flags_to_engine(eb_ring(e));
> > > > +
> > > > +		if (gem_engine_can_block_ggtt_binder(fd, &engine))
> > > > +			continue;
> > Is it possible that this loop will have only bcs egines?
> > Or due to (future?) bug in gem_engine_can_block_ggtt_binder this
> > loop will never execute any test? In that case:
> 
> for_each_ring() iterate over the device so it is not possible to just have copy engine.
> Don't think we need to check that.

Reviewed-by: Kamil Konieczny <kamil.konieczny at linux.intel.com>

> 
> Regards,
> Nirmoy
> 
> > 
> >              ++counter;
> > 
> > > > +
> > > >    		reset_stress(fd, ahnd, ctx0, e->name, eb_ring(e), flags);
> > > > +	}
> > igt_require(counter);

Hmm, this should be after ctx_destroy and put_ahnd but as
it is not needed you may drop my suggestion.

Regards,
Kamil

> > 
> > Regards,
> > Kamil
> > 
> > > >    	intel_ctx_destroy(fd, ctx0);
> > > >    	put_ahnd(ahnd);