On 07/12/2022 11:47, Andrzej Hajda wrote:
With introduction of long running compute context preemption timeouts
in some engines (rcs and ccs) can be quite big (by default 7.5s).
Keeping spinner timeout to at least twice bigger than engines preemption
timeout should allow to finish the test without -ETIME error.
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/2410
Signed-off-by: Andrzej Hajda <andrzej.hajda@xxxxxxxxx>
---
tests/i915/gem_ctx_persistence.c | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/tests/i915/gem_ctx_persistence.c b/tests/i915/gem_ctx_persistence.c
index a844439de19..9fb52818616 100644
--- a/tests/i915/gem_ctx_persistence.c
+++ b/tests/i915/gem_ctx_persistence.c
@@ -1173,6 +1173,14 @@ static void many_contexts(int i915, const intel_ctx_cfg_t *cfg)
gem_sync(i915, spin->handle);
igt_spin_reset(spin);
+ for_each_ctx_cfg_engine(i915, cfg, e) {
+ int t = 0;
+
+ gem_engine_property_scanf(i915, e->name,
+ "preempt_timeout_ms", "%d", &t);
+ timeout = max_t(int64_t, timeout, 2000000ll * t);
+ }
+
igt_until_timeout(30) {
for_each_ctx_cfg_engine(i915, cfg, e) {
const intel_ctx_t *ctx;
LGTM. And over time we will see if there will be remaining unexplained
failures.
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@xxxxxxxxx>
Regards,
Tvrtko
