drm/i915/selftest: Bump up sample period for busy stats selftest
authorUmesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
Thu, 10 Nov 2022 17:19:13 +0000 (17:19 +0000)
committerJohn Harrison <John.C.Harrison@Intel.com>
Mon, 21 Nov 2022 20:53:35 +0000 (12:53 -0800)
commit529d95a6067b74da9d4d5d9ab3009b35c98c5fce
treed95e196f68c3feaf4d0a222d2163ee2ba068e2ae
parente746f84b8e813816951b63485134927ed6763a1b
drm/i915/selftest: Bump up sample period for busy stats selftest

Engine busyness samples around a 10ms period is failing with busyness
ranging approx. from 87% to 115% as shown below. The expected range is
+/- 5% of the sample period. Fail 10% of the time.

rcs0: reported 11716042ns [91%] busyness while spinning [for 12805719ns]

When determining busyness of active engine, the GuC based engine
busyness implementation relies on a 64 bit timestamp register read. The
latency incurred by this register read causes the failure.

On DG1, when the test fails, the observed latencies range from 900us -
1.5ms.

Optimizing the 2x32 read by acquiring the lock and forcewake prior to
all reg reads reduces the rate of failure to around 2%, but does not
eliminate it.

In order to make the selftest more robust and always account for such
latencies, increase the sample period to 100 ms. This eliminates the
issue as seen in a 1000 runs.

v2: (Ashutosh)
- Add error to commit msg
- Include gitlab bug
- Update commit for inclusion of 2x32 optimized read

Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/4418
Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
Acked-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Ashutosh Dixit <ashutosh.dixit@intel.com>
Signed-off-by: John Harrison <John.C.Harrison@Intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20221110171913.670286-3-umesh.nerlige.ramappa@intel.com
drivers/gpu/drm/i915/gt/selftest_engine_pm.c