Re: [PATCH v2 04/14] drm/i915/cdclk: Notify DG2 pcode about pipe power wells regardless of CDCLK

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [High] The function `dg2_power_well_count` incorrectly returns a `bool` instead of an integer type, causing the active pipe count to be silently truncated to 1 for any multi-pipe configuration. This breaks the new equality checks introduced in this patch, preventing PCODE notifications during multi-pipe display transitions.
--

commit fcc366768f01016ec317b427777b27612fa4fe66
Author: Ville Syrjälä <ville.syrjala@linux.intel.com>

drm/i915/cdclk: Notify DG2 pcode about pipe power wells regardless of CDCLK

This patch updates the early return logic in CDCLK pre/post plane update
functions. It ensures DG2 pcode notifications are sent when pipe power well
counts change, even if the actual CDCLK remains unchanged.

> diff --git a/drivers/gpu/drm/i915/display/intel_cdclk.c b/drivers/gpu/drm/i915/display/intel_cdclk.c
> index d96067d7346b7..570c257626066 100644
> --- a/drivers/gpu/drm/i915/display/intel_cdclk.c
> +++ b/drivers/gpu/drm/i915/display/intel_cdclk.c
> @@ -2841,7 +2841,9 @@ intel_set_cdclk_pre_plane_update(struct intel_atomic_state *state)
>  		return;
>  
>  	if (!intel_cdclk_changed(&old_cdclk_state->actual,
> -				 &new_cdclk_state->actual))
> +				 &new_cdclk_state->actual) &&
> +	    dg2_power_well_count(display, old_cdclk_state) ==
> +	    dg2_power_well_count(display, new_cdclk_state))

[Severity: High]
This isn't a bug introduced by this patch, but does this new equality check
work correctly given the return type of dg2_power_well_count()?

Looking at dg2_power_well_count() in intel_cdclk.c:

static bool dg2_power_well_count(struct intel_display *display,
				 const struct intel_cdclk_state *cdclk_state)
{
	return display->platform.dg2 ? hweight8(cdclk_state->active_pipes) : 0;
}

Because it returns a bool instead of an integer type, any non-zero pipe count
returned by hweight8() will be truncated to 1 (true).

If a user transitions between non-zero active pipe counts (for example, going
from 1 to 2 pipes during a hotplug), the check will evaluate to true == true.

Could this falsely trigger the early return and skip the required pcode
notification on DG2?

[ ... ]

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260826134033.13976-1-ville.syrjala@linux.intel.com?part=4