Some tests are on right on the limit of targetting a one second runtime, and some of those have recently been correctly marked as slow. Lets instead make them a bit quicker, by either reducing the amount of submitted jobs or waiting a bit less. For the latter a comment is added to explain why that should be fine, and for the former, they are tests which aim to hit races via stress testing, so reducing the runtime is also fine since the main thing is the number of test runs over time and machines.
Signed-off-by: Tvrtko Ursulin <[email protected]> Cc: Maxime Ripard <[email protected]> Cc: Danilo Krummrich <[email protected]> Cc: Philipp Stanner <[email protected]> Reviewed-by: Danilo Krummrich <[email protected]> --- v2: * Downgraded Reported-by: to Cc: * Dropped one test comment. --- drivers/gpu/drm/scheduler/tests/tests_basic.c | 24 ++++++++++++------- 1 file changed, 16 insertions(+), 8 deletions(-) diff --git a/drivers/gpu/drm/scheduler/tests/tests_basic.c b/drivers/gpu/drm/scheduler/tests/tests_basic.c index a5a5a35a87b0..76b44f5df3b1 100644 --- a/drivers/gpu/drm/scheduler/tests/tests_basic.c +++ b/drivers/gpu/drm/scheduler/tests/tests_basic.c @@ -382,7 +382,7 @@ static void drm_sched_change_priority(struct kunit *test) struct drm_mock_sched_entity *entity[DRM_SCHED_PRIORITY_COUNT]; struct drm_mock_scheduler *sched = test->priv; struct drm_mock_sched_job *job; - const unsigned int qd = 1000; + const unsigned int qd = 500; unsigned int i, cur_ent = 0; enum drm_sched_priority p; @@ -421,7 +421,7 @@ static void drm_sched_change_priority(struct kunit *test) static struct kunit_case drm_sched_priority_tests[] = { KUNIT_CASE(drm_sched_priorities), - KUNIT_CASE_SLOW(drm_sched_change_priority), + KUNIT_CASE(drm_sched_change_priority), {} }; @@ -438,7 +438,7 @@ static void drm_sched_test_modify_sched(struct kunit *test) struct drm_mock_sched_entity *entity[13]; struct drm_mock_scheduler *sched[3]; struct drm_mock_sched_job *job; - const unsigned int qd = 1000; + const unsigned int qd = 500; /* * Submit a bunch of jobs against entities configured with different @@ -500,6 +500,7 @@ static struct kunit_suite drm_sched_modify_sched = { static void drm_sched_test_credits(struct kunit *test) { + const long timeout = msecs_to_jiffies(200); struct drm_mock_sched_entity *entity; struct drm_mock_scheduler *sched; struct drm_mock_sched_job *job[2]; @@ -523,22 +524,29 @@ static void drm_sched_test_credits(struct kunit *test) drm_mock_sched_job_submit(job[0]); drm_mock_sched_job_submit(job[1]); - done = drm_mock_sched_job_wait_scheduled(job[0], HZ); + done = drm_mock_sched_job_wait_scheduled(job[0], timeout); KUNIT_ASSERT_TRUE(test, done); - done = drm_mock_sched_job_wait_scheduled(job[1], HZ); + /* + * Verify that the scheduler has not consumed more than the configured + * single credit. This can false negative if the system fails to + * execute the scheduler's workqueue two times between submitting two + * jobs and this wait expires, but this is so unlikely that we opt for a + * timeout which does not make the test excessively slow. + */ + done = drm_mock_sched_job_wait_scheduled(job[1], timeout); KUNIT_ASSERT_FALSE(test, done); i = drm_mock_sched_advance(sched, 1); KUNIT_ASSERT_EQ(test, i, 1); - done = drm_mock_sched_job_wait_scheduled(job[1], HZ); + done = drm_mock_sched_job_wait_scheduled(job[1], timeout); KUNIT_ASSERT_TRUE(test, done); i = drm_mock_sched_advance(sched, 1); KUNIT_ASSERT_EQ(test, i, 1); - done = drm_mock_sched_job_wait_finished(job[1], HZ); + done = drm_mock_sched_job_wait_finished(job[1], timeout); KUNIT_ASSERT_TRUE(test, done); drm_mock_sched_entity_free(entity); @@ -546,7 +554,7 @@ static void drm_sched_test_credits(struct kunit *test) } static struct kunit_case drm_sched_credits_tests[] = { - KUNIT_CASE_SLOW(drm_sched_test_credits), + KUNIT_CASE(drm_sched_test_credits), {} }; -- 2.52.0
