Re: [Intel-gfx] [PATCH] drm/i915/slpc: Fix PCODE IA Freq requests when using SLPC

["Dixit, Ashutosh" <ashutosh.dixit@intel.com>]

On Fri, 26 Aug 2022 10:44:34 -0700, Rodrigo Vivi wrote:
>
> Fixes: 7ba79a671568 ("drm/i915/guc/slpc: Gate Host RPS when SLPC is enabled")
> Cc: <stable@vger.kernel.org> # v5.15+
> Cc: Ashutosh Dixit <ashutosh.dixit@intel.com>
> Tested-by: Sushma Venkatesh Reddy <sushma.venkatesh.reddy@intel.com>
> Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
> ---
>  drivers/gpu/drm/i915/gt/intel_llc.c | 24 ++++++++++++++++--------
>  1 file changed, 16 insertions(+), 8 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/gt/intel_llc.c b/drivers/gpu/drm/i915/gt/intel_llc.c
> index 14fe65812e42..2677d62573d9 100644
> --- a/drivers/gpu/drm/i915/gt/intel_llc.c
> +++ b/drivers/gpu/drm/i915/gt/intel_llc.c
> @@ -49,13 +49,28 @@ static unsigned int cpu_max_MHz(void)
>  static bool get_ia_constants(struct intel_llc *llc,
>			     struct ia_constants *consts)
>  {
> +	struct intel_guc_slpc *slpc = &llc_to_gt(llc)->uc.guc.slpc;
>	struct drm_i915_private *i915 = llc_to_gt(llc)->i915;
>	struct intel_rps *rps = &llc_to_gt(llc)->rps;
>
>	if (!HAS_LLC(i915) || IS_DGFX(i915))
>		return false;
>
> -	if (rps->max_freq <= rps->min_freq)
> +	if (intel_uc_uses_guc_slpc(&llc_to_gt(llc)->uc)) {
> +		consts->min_gpu_freq = slpc->min_freq;
> +		consts->max_gpu_freq = slpc->rp0_freq;
> +	} else {
> +		consts->min_gpu_freq = rps->min_freq;
> +		consts->max_gpu_freq = rps->max_freq;
> +	}
> +
> +	if (GRAPHICS_VER(i915) >= 9) {
> +		/* Convert GT frequency to 50 HZ units */
> +		consts->min_gpu_freq /= GEN9_FREQ_SCALER;
> +		consts->max_gpu_freq /= GEN9_FREQ_SCALER;
> +	}
> +
> +	if (consts->max_gpu_freq <= consts->min_gpu_freq)
>		return false;

Hi Rodrigo, sorry, I missed this check previously too and the code is now
equivalent to the previous code.

But now, looking at the code in gen6_update_ring_freq, I am wondering if we
should return true in this case (i.e. remove the check) and we had a bug in
the previous code? Because if we return false, gen6_update_ring_freq will
skip the PCODE programming if 'max_gpu_freq == min_gpu_freq', but why
should we skip the PCODE programming if 'max_gpu_freq == min_gpu_freq'? The
case of 'max_gpu_freq < min_gpu_freq' is fine since the loop in
gen6_update_ring_freq is not entered in that case.

Thanks.
--
Ashutosh