Re: [PATCH 2/8] drm/i915/cdclk: Add and use xe2lpd_mdclk_source_sel()

From: Matt Roper <matthew.d.roper@intel.com>
To: Gustavo Sousa <gustavo.sousa@intel.com>
Cc: <intel-gfx@lists.freedesktop.org>,
	<intel-xe@lists.freedesktop.org>,
	"Lucas De Marchi" <lucas.demarchi@intel.com>,
	<vinod.govindapillai@intel.com>, <stanislav.lisovskiy@intel.com>
Subject: Re: [PATCH 2/8] drm/i915/cdclk: Add and use xe2lpd_mdclk_source_sel()
Date: Mon, 4 Mar 2024 13:58:34 -0800	[thread overview]
Message-ID: <20240304215834.GW718896@mdroper-desk1.amr.corp.intel.com> (raw)
In-Reply-To: <20240304183028.195057-3-gustavo.sousa@intel.com>

On Mon, Mar 04, 2024 at 03:30:21PM -0300, Gustavo Sousa wrote:
> There will be future changes that rely on the source of the MDCLK. Let's
> have xe2lpd_mdclk_source_sel() as the function responsible for reporting
> that information.
> 
> Bspec: 69090
> Signed-off-by: Gustavo Sousa <gustavo.sousa@intel.com>
> ---
>  drivers/gpu/drm/i915/display/intel_cdclk.c | 17 ++++++++++++++++-
>  drivers/gpu/drm/i915/i915_reg.h            |  4 +++-
>  2 files changed, 19 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/display/intel_cdclk.c b/drivers/gpu/drm/i915/display/intel_cdclk.c
> index 407bd541eb46..bf84bf27213f 100644
> --- a/drivers/gpu/drm/i915/display/intel_cdclk.c
> +++ b/drivers/gpu/drm/i915/display/intel_cdclk.c
> @@ -1876,6 +1876,21 @@ static bool cdclk_pll_is_unknown(unsigned int vco)
>  	return vco == ~0;
>  }
>  
> +static u32 xe2lpd_mdclk_source_sel(struct drm_i915_private *i915)
> +{
> +	if (DISPLAY_VER(i915) >= 20)
> +		return MDCLK_SOURCE_SEL_CDCLK_PLL;
> +
> +	/*
> +	 * Earlier display IPs do not provide means of selecting the
> +	 * MDCLK source, but MDCLK_SOURCE_SEL_CD2XCLK is a nice default,
> +	 * since it reflects the source used for those and allows
> +	 * xe2lpd_mdclk_source_sel() to be used in logic that depends on
> +	 * it.
> +	 */
> +	return MDCLK_SOURCE_SEL_CD2XCLK;

At the moment this function only gets called on Xe2 and beyond where the
register field exists; if that's going to change soon, then wouldn't it
be more natural to just use an early exit to highlight that there's
nothing we need to OR into the CDCLK_CTL for earlier platforms?  

    /* Not configurable for older platforms; they always use CD2XCLK */
    if (DISPLAY_VER(i915) < 20)
            return 0;

Functionally it's the same, but it feels more intuitive to me.

If we aren't expecting to call this from common codepaths that aren't
already protected by a display version check, then we could make this a
drm_WARN_ON() to assert that we haven't deviated from expected behavior.

Matt

> +}
> +
>  static bool cdclk_compute_crawl_and_squash_midpoint(struct drm_i915_private *i915,
>  						    const struct intel_cdclk_config *old_cdclk_config,
>  						    const struct intel_cdclk_config *new_cdclk_config,
> @@ -1980,7 +1995,7 @@ static u32 bxt_cdclk_ctl(struct drm_i915_private *i915,
>  		val |= BXT_CDCLK_SSA_PRECHARGE_ENABLE;
>  
>  	if (DISPLAY_VER(i915) >= 20)
> -		val |= MDCLK_SOURCE_SEL_CDCLK_PLL;
> +		val |= xe2lpd_mdclk_source_sel(i915);
>  	else
>  		val |= skl_cdclk_decimal(cdclk);
>  
> diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
> index e00557e1a57f..eb953ed1f113 100644
> --- a/drivers/gpu/drm/i915/i915_reg.h
> +++ b/drivers/gpu/drm/i915/i915_reg.h
> @@ -5900,7 +5900,9 @@ enum skl_power_gate {
>  #define  CDCLK_FREQ_540		REG_FIELD_PREP(CDCLK_FREQ_SEL_MASK, 1)
>  #define  CDCLK_FREQ_337_308		REG_FIELD_PREP(CDCLK_FREQ_SEL_MASK, 2)
>  #define  CDCLK_FREQ_675_617		REG_FIELD_PREP(CDCLK_FREQ_SEL_MASK, 3)
> -#define  MDCLK_SOURCE_SEL_CDCLK_PLL	REG_BIT(25)
> +#define  MDCLK_SOURCE_SEL_MASK		REG_GENMASK(25, 25)
> +#define  MDCLK_SOURCE_SEL_CD2XCLK	REG_FIELD_PREP(MDCLK_SOURCE_SEL_MASK, 0)
> +#define  MDCLK_SOURCE_SEL_CDCLK_PLL	REG_FIELD_PREP(MDCLK_SOURCE_SEL_MASK, 1)
>  #define  BXT_CDCLK_CD2X_DIV_SEL_MASK	REG_GENMASK(23, 22)
>  #define  BXT_CDCLK_CD2X_DIV_SEL_1	REG_FIELD_PREP(BXT_CDCLK_CD2X_DIV_SEL_MASK, 0)
>  #define  BXT_CDCLK_CD2X_DIV_SEL_1_5	REG_FIELD_PREP(BXT_CDCLK_CD2X_DIV_SEL_MASK, 1)
> -- 
> 2.44.0
> 

-- 
Matt Roper
Graphics Software Engineer
Linux GPU Platform Enablement
Intel Corporation