[igt-dev] [PATCH i-g-t v2 1/2] tests/gem_tiled_blits: fix sigsegv on TGL when allocator is in use

Zbigniew Kempczyński zbigniew.kempczynski at intel.com
Wed Apr 14 09:04:44 UTC 2021


TGL is rare case when intel-bb assumes it should use allocator
infrastructure (even if it supports relocations). The main reason
of such decision is building aux ccs tables for rendercopy.
These tables requires filling address a-priori for surfaces so
softpin and allocator allow to build them without calling warming
execbuf.

We need to ensure intel-bb here can be called in multiprocess
environment so multiprocess start()/stop() around normal and
interruptible tests are necessary.

Signed-off-by: Zbigniew Kempczyński <zbigniew.kempczynski at intel.com>
Cc: Petri Latvala <petri.latvala at intel.com>
---
 tests/i915/gem_tiled_blits.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/tests/i915/gem_tiled_blits.c b/tests/i915/gem_tiled_blits.c
index 91537ac48..c93d04d67 100644
--- a/tests/i915/gem_tiled_blits.c
+++ b/tests/i915/gem_tiled_blits.c
@@ -196,17 +196,21 @@ igt_main
 		run_test(fd, 2);
 
 	igt_subtest("normal") {
+		intel_allocator_multiprocess_start();
 		igt_fork(child, ncpus)
 			run_test(fd, count);
 		igt_waitchildren();
+		intel_allocator_multiprocess_stop();
 	}
 
 	igt_subtest("interruptible") {
+		intel_allocator_multiprocess_start();
 		igt_fork_signal_helper();
 		igt_fork(child, ncpus)
 			run_test(fd, count);
 		igt_waitchildren();
 		igt_stop_signal_helper();
+		intel_allocator_multiprocess_stop();
 	}
 
 	igt_fixture {
-- 
2.26.0



More information about the igt-dev mailing list