[PATCH i-g-t CI run 07/14] tests/intel/xe_oa: Simplify the buffer-fill test

Dixit, Ashutosh ashutosh.dixit at intel.com
Mon Feb 24 21:31:01 UTC 2025


On Tue, 18 Feb 2025 12:28:05 -0800, Umesh Nerlige Ramappa wrote:
>
> We only want to test that the BUFFER OVERFLOW status is set when we do
> not read the OA stream data in time. To do so, keeping reading zero
> bytes of data until you hit a buffer overflow.

The approach in this patch is great!

>
> Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa at intel.com>
> ---
>  tests/intel/xe_oa.c | 135 +++++++++-----------------------------------
>  1 file changed, 26 insertions(+), 109 deletions(-)
>
> diff --git a/tests/intel/xe_oa.c b/tests/intel/xe_oa.c
> index 5792ffec2..fc03b74af 100644
> --- a/tests/intel/xe_oa.c
> +++ b/tests/intel/xe_oa.c
> @@ -304,6 +304,7 @@ static size_t default_oa_buffer_size;
>  static struct intel_mmio_data mmio_data;
>  static igt_render_copyfunc_t render_copy;
>  static uint32_t rc_width, rc_height;
> +static uint32_t buffer_fill_size;

I am assuming these globals will be changed for slow platforms?

>
>  static struct intel_xe_perf_metric_set *metric_set(const struct drm_xe_engine_class_instance *hwe)
>  {
> @@ -1092,6 +1093,7 @@ init_sys_info(void)
>
>	rc_width = 1920;
>	rc_height = 1080;
> +	buffer_fill_size = SZ_16M;
>	oa_exponent_default = max_oa_exponent_for_period_lte(1000000);
>
>	default_oa_buffer_size = get_default_oa_buffer_size(drm_fd);
> @@ -2370,11 +2372,7 @@ test_oa_tlb_invalidate(const struct drm_xe_engine_class_instance *hwe)
>  static void
>  test_buffer_fill(const struct drm_xe_engine_class_instance *hwe)

There is this comment above this line:

Description: Test filling, wraparound and overflow of OA buffer

I think we should drop "wraparound", since this test never tested the
wraparound of the OA buffer, it just tested overflow.

To test wraparound I was using non-zero-reason.

>  {
> -	/* ~5 micro second period */
> -	int oa_exponent = max_oa_exponent_for_period_lte(5000);
> -	uint64_t oa_period = oa_exponent_to_ns(oa_exponent);
>	struct intel_xe_perf_metric_set *test_set = metric_set(hwe);
> -	uint64_t fmt = test_set->perf_oa_format;
>	uint64_t properties[] = {
>		DRM_XE_OA_PROPERTY_OA_UNIT_ID, 0,
>
> @@ -2383,126 +2381,45 @@ test_buffer_fill(const struct drm_xe_engine_class_instance *hwe)
>
>		/* OA unit configuration */
>		DRM_XE_OA_PROPERTY_OA_METRIC_SET, test_set->perf_oa_metrics_set,
> -		DRM_XE_OA_PROPERTY_OA_FORMAT, __ff(fmt),
> -		DRM_XE_OA_PROPERTY_OA_PERIOD_EXPONENT, oa_exponent,
> +		DRM_XE_OA_PROPERTY_OA_FORMAT, __ff(test_set->perf_oa_format),
> +		DRM_XE_OA_PROPERTY_OA_PERIOD_EXPONENT, oa_exponent_default,
>		DRM_XE_OA_PROPERTY_OA_ENGINE_INSTANCE, hwe->engine_instance,
> +		DRM_XE_OA_PROPERTY_OA_DISABLED, true,
> +		DRM_XE_OA_PROPERTY_OA_BUFFER_SIZE, buffer_fill_size,
>	};
>	struct intel_xe_oa_open_prop param = {
>		.num_properties = ARRAY_SIZE(properties) / 2,
>		.properties_ptr = to_user_pointer(properties),
>	};
> -	size_t report_size = get_oa_format(fmt).size;
> -	int buf_size = 65536 * report_size;
> -	uint8_t *buf = malloc(buf_size);
> -	int len;
> -	int n_full_oa_reports = default_oa_buffer_size / report_size;
> -	uint64_t fill_duration = n_full_oa_reports * oa_period;
> -	uint32_t *last_periodic_report = malloc(report_size);
> +	uint64_t oa_period = oa_exponent_to_ns(oa_exponent_default);
> +	char *buf = malloc(1024);
> +	bool overflow_seen;
>	u32 oa_status;
> +	int len;
>
> -	igt_assert(fill_duration < 1000000000);
> -
> +	igt_debug("oa_period %s\n", pretty_print_oa_period(oa_period));
>	stream_fd = __perf_open(drm_fd, &param, true /* prevent_pm */);
>          set_fd_flags(stream_fd, O_CLOEXEC);
>
> -	for (int i = 0; i < 5; i++) {
> -		bool overflow_seen;
> -		uint32_t n_periodic_reports;
> -		uint32_t first_timestamp = 0, last_timestamp = 0;
> -
> -		do_ioctl(stream_fd, DRM_XE_OBSERVATION_IOCTL_ENABLE, 0);
> -
> -		nanosleep(&(struct timespec){ .tv_sec = 0,
> -					      .tv_nsec = fill_duration * 1.25 },
> -			  NULL);
> -again:
> -		oa_status = 0;
> -		while ((len = read(stream_fd, buf, buf_size)) == -1 && errno == EINTR)
> -			;
> -
> -		if (errno == EIO) {
> -			oa_status = get_stream_status(stream_fd);
> -			igt_debug("oa_status %#x\n", oa_status);
> -			overflow_seen = oa_status & DRM_XE_OASTATUS_BUFFER_OVERFLOW;
> -			igt_assert_eq(overflow_seen, true);
> -			goto again;
> -		}
> -		igt_assert_neq(len, -1);
> -
> -		do_ioctl(stream_fd, DRM_XE_OBSERVATION_IOCTL_DISABLE, 0);
> -
> -		igt_debug("fill_duration = %"PRIu64"ns, oa_exponent = %u\n",
> -			  fill_duration, oa_exponent);
> -
> -		do_ioctl(stream_fd, DRM_XE_OBSERVATION_IOCTL_ENABLE, 0);
> -
> -		nanosleep(&(struct timespec){ .tv_sec = 0,
> -					.tv_nsec = fill_duration / 2 },
> -			NULL);
> -
> -		n_periodic_reports = 0;
> -
> -		/* Because of the race condition between notification of new
> -		 * reports and reports landing in memory, we need to rely on
> -		 * timestamps to figure whether we've read enough of them.
> -		 */
> -		while (((last_timestamp - first_timestamp) * oa_period) < (fill_duration / 2)) {
> -
> -			igt_debug("dts=%u elapsed=%"PRIu64" duration=%"PRIu64"\n",
> -				  last_timestamp - first_timestamp,
> -				  (last_timestamp - first_timestamp) * oa_period,
> -				  fill_duration / 2);
> -again_1:
> -			oa_status = 0;
> -			while ((len = read(stream_fd, buf, buf_size)) == -1 && errno == EINTR)
> -				;
> -			if (errno == EIO) {
> -				oa_status = get_stream_status(stream_fd);
> -				igt_debug("oa_status %#x\n", oa_status);
> -				igt_assert(!(oa_status & DRM_XE_OASTATUS_BUFFER_OVERFLOW));
> -				goto again_1;
> -			}
> -			igt_assert_neq(len, -1);
> -
> -			for (int offset = 0; offset < len; offset += report_size) {
> -				uint32_t *report = (void *) (buf + offset);
> -
> -				igt_debug(" > report ts=%"PRIu64""
> -					  " ts_delta_last_periodic=%"PRIu64" is_timer=%i ctx_id=%8x nb_periodic=%u\n",
> -					  oa_timestamp(report, fmt),
> -					  n_periodic_reports > 0 ?  oa_timestamp_delta(report, last_periodic_report, fmt) : 0,
> -					  oa_report_is_periodic(report),
> -					  oa_report_get_ctx_id(report),
> -					  n_periodic_reports);
> -
> -				if (first_timestamp == 0)
> -					first_timestamp = oa_timestamp(report, fmt);
> -				last_timestamp = oa_timestamp(report, fmt);
> -
> -				if (oa_report_is_periodic(report)) {
> -					memcpy(last_periodic_report, report, report_size);
> -					n_periodic_reports++;
> -				}
> -			}
> -		}
> -
> -		do_ioctl(stream_fd, DRM_XE_OBSERVATION_IOCTL_DISABLE, 0);
> -
> -		igt_debug("first ts = %u, last ts = %u\n", first_timestamp, last_timestamp);
> +	/* OA buffer is disable, we do not expect any error status */

disabled

> +	oa_status = get_stream_status(stream_fd);
> +	overflow_seen = !!(oa_status & DRM_XE_OASTATUS_BUFFER_OVERFLOW);

Don't need !!

> +	igt_assert_eq(overflow_seen, 0);
>
> -		igt_debug("%f < %zu < %f\n",
> -			  report_size * n_full_oa_reports * 0.45,
> -			  n_periodic_reports * report_size,
> -			  report_size * n_full_oa_reports * 0.55);
> +	do_ioctl(stream_fd, DRM_XE_OBSERVATION_IOCTL_ENABLE, 0);
>
> -		igt_assert(n_periodic_reports * report_size >
> -			   report_size * n_full_oa_reports * 0.45);
> -		igt_assert(n_periodic_reports * report_size <
> -			   report_size * n_full_oa_reports * 0.55);
> +	errno = 0;
> +	/* Read 0 bytes repeatedly until you see an EIO */
> +	while ((len = read(stream_fd, buf, 0)) == -1 && (errno == EINTR || errno == ENOSPC)) {
> +		usleep(100);
>	}
> +	igt_assert_eq(len, -1);
> +	igt_assert_eq(errno, EIO);
>
> -	free(last_periodic_report);
> -	free(buf);
> +	/* Ensure buffer overflowed */
> +	oa_status = get_stream_status(stream_fd);
> +	overflow_seen = !!(oa_status & DRM_XE_OASTATUS_BUFFER_OVERFLOW);

Don't need !!

> +	igt_assert(overflow_seen);

Also I am wondering if we disabled the stream now using
DRM_XE_OBSERVATION_IOCTL_DISABLE, would it clear the overflow status? But
anyway the test is good as is.

>
>	__perf_close(stream_fd);
>  }

Apart from the above nits, this is:

Reviewed-by: Ashutosh Dixit <ashutosh.dixit at intel.com>


More information about the igt-dev mailing list