The 'clear' subtest exercises correctness of object memory clearing on passing a batch with the object to GPU for processing. The exercise is executed in several parallel threads, one per CPU. Each thread repeats the exercise in a time only limited loop, with no delay between consecutive iterations. In case of system memory objects, that happens to exhaust all available physical memory, which is neither the goal nor requirement of the exercise. Make sure sufficient amount of physical memory is available before calling another execbuf. v2: Limit the scope of the fix to SMEM exercise. Closes: https://gitlab.freedesktop.org/drm/i915/kernel/-/issues/11738 Signed-off-by: Janusz Krzysztofik <janusz.krzysztofik@xxxxxxxxxxxxxxx> --- tests/intel/gem_mmap_offset.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/tests/intel/gem_mmap_offset.c b/tests/intel/gem_mmap_offset.c index 20dca17378..8ce10d6144 100644 --- a/tests/intel/gem_mmap_offset.c +++ b/tests/intel/gem_mmap_offset.c @@ -745,6 +745,9 @@ static void *thread_clear(void *data) npages = get_npages(&arg->max, npages); size = npages << 12; + /* Execbuf requires sufficient amount of free physical memory */ + if (arg->flags & CLEAR_IN_EXECBUF && arg->region.memory_class == I915_MEMORY_CLASS_SYSTEM) + igt_require_memory(1, size, CHECK_RAM); igt_assert_eq(__gem_create_in_memory_region_list(i915, &handle, &size, 0, &arg->region, 1), 0); /* Zero-init bo in execbuf or pagefault handler path as requested */ if (arg->flags & CLEAR_IN_EXECBUF) -- 2.47.0
