Re: [PATCH v3 2/5] tests/intel/xe_exec_reset: Add timeslice preempt test

["Bernatowicz, Marcin" <marcin.bernatowicz@linux.intel.com>]

On 9/23/2025 11:13 PM, Matthew Brost wrote:
> Add tests in which submit multiple preemptable spinners to same hardware
> engine ensuring a cumulative longer timeout and validating the restart
> paths in the DRM scheduler TDR (Timeout Detection and Recovery).
> 
> v3:
>   - Spell out TDR (Kamal)
> 
> Signed-off-by: Matthew Brost <matthew.brost@intel.com>
> ---
>   lib/xe/xe_legacy.c          |  5 +++--
>   tests/intel/xe_exec_reset.c | 25 +++++++++++++++++++++++--
>   2 files changed, 26 insertions(+), 4 deletions(-)
> 
> diff --git a/lib/xe/xe_legacy.c b/lib/xe/xe_legacy.c
> index 669ecff7aa..9d2da7f413 100644
> --- a/lib/xe/xe_legacy.c
> +++ b/lib/xe/xe_legacy.c
> @@ -12,6 +12,7 @@
>   
>   /* Batch buffer element count, in number of dwords(u32) */
>   #define BATCH_DW_COUNT			16
> +#define CANCEL				(0x1 << 7)
>   #define PREEMPT				(0x1 << 6)
>   #define CAT_ERROR			(0x1 << 5)
>   #define CLOSE_EXEC_QUEUES		(0x1 << 2)
> @@ -100,7 +101,7 @@ xe_legacy_test_mode(int fd, struct drm_xe_engine_class_instance *eci,
>   		u64 exec_addr;
>   		int e = i % n_exec_queues;
>   
> -		if (!i) {
> +		if (!i || flags & CANCEL) {
>   			spin_opts.addr = base_addr + spin_offset;
>   			xe_spin_init(&data[i].spin, &spin_opts);
>   			exec_addr = spin_opts.addr;
> @@ -155,7 +156,7 @@ xe_legacy_test_mode(int fd, struct drm_xe_engine_class_instance *eci,
>   	xe_vm_unbind_async(fd, vm, 0, 0, addr, bo_size, sync, 1);
>   	igt_assert(syncobj_wait(fd, &sync[0].handle, 1, INT64_MAX, 0, NULL));
>   
> -	if (!use_capture_mode && !(flags & GT_RESET)) {
> +	if (!use_capture_mode && !(flags & (GT_RESET | CANCEL))) {
>   		for (i = 1; i < n_execs; i++)
>   			igt_assert_eq(data[i].data, 0xc0ffee);
>   	}
> diff --git a/tests/intel/xe_exec_reset.c b/tests/intel/xe_exec_reset.c
> index 72b85d3e7b..88e9df6fbb 100644
> --- a/tests/intel/xe_exec_reset.c
> +++ b/tests/intel/xe_exec_reset.c
> @@ -117,6 +117,7 @@ static void test_spin(int fd, struct drm_xe_engine_class_instance *eci,
>   #define PARALLEL			(0x1 << 4)
>   #define CAT_ERROR			(0x1 << 5)
>   #define PREEMPT				(0x1 << 6)
> +#define CANCEL				(0x1 << 7)
>   
>   /**
>    * SUBTEST: %s-cat-error
> @@ -302,6 +303,12 @@ test_balancer(int fd, int gt, int class, int n_exec_queues, int n_execs,
>    * SUBTEST: cancel-preempt
>    * Description: Test job cancel with a preemptable job
>    *
> + * SUBTEST: cancel-timeslice-preempt
> + * Description: Test job cancel with 2 preemptable jobs
> + *
> + * SUBTEST: cancel-timeslice-many-preempt
> + * Description: Test job cancel with many preemptable jobs
> + *
>    * SUBTEST: gt-reset
>    * Description: Test GT reset
>    *
> @@ -396,7 +403,7 @@ test_compute_mode(int fd, struct drm_xe_engine_class_instance *eci,
>   		uint64_t exec_addr;
>   		int e = i % n_exec_queues;
>   
> -		if (!i) {
> +		if (!i || flags & CANCEL) {
>   			spin_opts.addr = base_addr + spin_offset;
>   			xe_spin_init(&data[i].spin, &spin_opts);
>   			exec_addr = spin_opts.addr;
> @@ -451,7 +458,7 @@ test_compute_mode(int fd, struct drm_xe_engine_class_instance *eci,
>   	xe_vm_unbind_async(fd, vm, 0, 0, addr, bo_size, sync, 1);
>   	xe_wait_ufence(fd, &data[0].vm_sync, USER_FENCE_VALUE, 0, 3 * NSEC_PER_SEC);
>   
> -	if (!(flags & GT_RESET)) {
> +	if (!(flags & (GT_RESET | CANCEL))) {
>   		for (i = 1; i < n_execs; i++)
>   			igt_assert_eq(data[i].data, 0xc0ffee);
>   	}
> @@ -680,6 +687,20 @@ igt_main
>   			break;
>   		}
>   
> +	igt_subtest("cancel-timeslice-preempt")
> +		xe_for_each_engine(fd, hwe) {
> +			xe_legacy_test_mode(fd, hwe, 2, 2, CANCEL | PREEMPT,
> +					    LEGACY_MODE_ADDR, false);
> +			break;
> +		}
> +
> +	igt_subtest("cancel-timeslice-many-preempt")
> +		xe_for_each_engine(fd, hwe) {
> +			xe_legacy_test_mode(fd, hwe, 4, 4, CANCEL | PREEMPT,
> +					    LEGACY_MODE_ADDR, false);
> +			break;
> +		}
> +
LGTM,
Reviewed-by: Marcin Bernatowicz <marcin.bernatowicz@linux.intel.com>

>   	igt_subtest("gt-reset")
>   		xe_for_each_engine(fd, hwe)
>   			xe_legacy_test_mode(fd, hwe, 2, 2, GT_RESET,