Re: [PATCH v2] tests/intel/xe_oa: Wa_14026633728

[Gustavo Sousa <gustavo.sousa@intel.com>]

Shekhar Chauhan <shekhar.chauhan@intel.com> writes:

> For MERTOA in CRI, oa buffer can be in device memory. Because of slower
> device mem access, OA exponent values lower than 8 can result in buffer
> overflows. Bump the OA exponent value.
>
> v2: Fix commit title, description and code description.
>
> Signed-off-by: Shekhar Chauhan <shekhar.chauhan@intel.com>
> ---
>  tests/intel/xe_oa.c | 10 ++++++++++
>  1 file changed, 10 insertions(+)
>
> diff --git a/tests/intel/xe_oa.c b/tests/intel/xe_oa.c
> index 988c46df6..cbc1e0d3e 100644
> --- a/tests/intel/xe_oa.c
> +++ b/tests/intel/xe_oa.c
> @@ -2645,6 +2645,16 @@ test_non_zero_reason(const struct drm_xe_oa_unit *oau, size_t oa_buffer_size)
>  {
>  	/* ~20 micro second period */
>  	int oa_exponent = max_oa_exponent_for_period_lte(20000);
> +
> +	/*
> +	 * Wa_14026633728: For MERTOA in CRI, oa buffer can be in device memory.
> +	 * Because of slower device mem access, OA exponent values lower than 8 can
> +	 * result in buffer overflows.
> +	 */
> +
> +	if (oau->oa_unit_type == DRM_XE_OA_UNIT_TYPE_MERT && IS_CRESCENTISLAND(devid))
> +		oa_exponent = max(oa_exponent, 8);

We have a way of querying for workarounds via debugfs (intel_wa.h). It
is probably better to use that interface instead of replicating
workaround checks in IGT.

It currently only checks for GT workarounds, but can be easily extended
to also check for device workarounds.

--
Gustavo Sousa

> +
>  	struct intel_xe_perf_metric_set *test_set = oa_unit_metric_set(oau);
>  	uint64_t fmt = test_set->perf_oa_format;
>  	size_t report_size = get_oa_format(fmt).size;
> -- 
> 2.53.0