Re: [PATCH RESEND] drm/sun4i: report coverage pixel blending for alpha-capable planes

From: Jernej Škrabec

Date: Tue Nov 17 2026 - 11:35:47 EST


Dne torek, 17. november 2026 ob 10:19:32 Srednjeevropski standardni čas je Yang Siyuan napisal(a):
> Linux 7.3 added validate_blend_mode_for_alpha_formats(), which warns once
> per plane that exposes a pixel format with an alpha channel but has no
> blend mode property:
>
> sun4i-drm display-engine: [drm] [PLANE:35:plane-0] pixel format with
> alpha exposed but blend mode not setup. Please fix.
>
> The mixer blends a layer using its own alpha channel: ALPHA_MODE is set
> to PIXEL when plane alpha is opaque and to COMBINED otherwise, while
> BLEND_PREMULTIPLY keeps its reset value. That is straight alpha, i.e. the
> DRM "coverage" blending equation.
>
> Advertise exactly DRM_MODE_BLEND_COVERAGE on every layer that can expose
> an alpha format (UI always, VI on DE3 and later). This matches the
> existing register programming, keeps userspace behaviour unchanged and
> satisfies the new validation. Premultiplied blending would require
> driving BLEND_PREMULTIPLY and is left as a follow-up.

The subject says drm/sun4i, but the DE1 backend planes (sun4i_layer.c)
also expose ARGB8888/ARGB4444/ARGB1555 without a blend mode property, so
A10/A20/A31/A33 still trigger the warning. Either handle them in a separate
patch or say in the commit message that this one only covers the DE2+
mixer.

>
> Fixes: 860e748bddcc ("drm: ensure blend mode supported if pixel format with alpha exposed")

This is incorrect. That patch only added general check. Drop the tag.

> Signed-off-by: Yang Siyuan <hsyysy@xxxxxxx>
> ---
> drivers/gpu/drm/sun4i/sun8i_ui_layer.c | 13 +++++++++++++
> drivers/gpu/drm/sun4i/sun8i_vi_layer.c | 14 ++++++++++++++
> 2 files changed, 27 insertions(+)
>
> diff --git a/drivers/gpu/drm/sun4i/sun8i_ui_layer.c b/drivers/gpu/drm/sun4i/sun8i_ui_layer.c
> index 7d894e1eb..ef059ac47 100644
> --- a/drivers/gpu/drm/sun4i/sun8i_ui_layer.c
> +++ b/drivers/gpu/drm/sun4i/sun8i_ui_layer.c
> @@ -283,6 +283,19 @@ struct sun8i_layer *sun8i_ui_layer_init_one(struct drm_device *drm,
> return ERR_PTR(ret);
> }
>
> + /*
> + * The mixer blends the layer using its own alpha channel with the
> + * pixels not premultiplied (BLEND_PREMULTIPLY keeps its reset value),
> + * which is the DRM "coverage" equation. Report it so that the blend
> + * mode validation in drm_mode_config_validate() is satisfied.
> + */

Above comment doesn't add much value.

> + ret = drm_plane_create_blend_mode_property(&layer->plane,
> + BIT(DRM_MODE_BLEND_COVERAGE));
> + if (ret) {
> + dev_err(drm->dev, "Couldn't add blend mode property\n");
> + return ERR_PTR(ret);
> + }
> +
> ret = drm_plane_create_zpos_property(&layer->plane, index,
> 0, plane_cnt - 1);
> if (ret) {
> diff --git a/drivers/gpu/drm/sun4i/sun8i_vi_layer.c b/drivers/gpu/drm/sun4i/sun8i_vi_layer.c
> index 92dd5a97f..b6771576b 100644
> --- a/drivers/gpu/drm/sun4i/sun8i_vi_layer.c
> +++ b/drivers/gpu/drm/sun4i/sun8i_vi_layer.c
> @@ -470,6 +470,20 @@ struct sun8i_layer *sun8i_vi_layer_init_one(struct drm_device *drm,
> }
> }
>
> + /*
> + * Only DE3 and later expose pixel formats with an alpha channel, and
> + * the mixer blends them with straight alpha (coverage), same as the UI
> + * layer.
> + */

And neither this.

Best regards,
Jernej

> + if (layer->cfg->de_type >= SUN8I_MIXER_DE3) {
> + ret = drm_plane_create_blend_mode_property(&layer->plane,
> + BIT(DRM_MODE_BLEND_COVERAGE));
> + if (ret) {
> + dev_err(drm->dev, "Couldn't add blend mode property\n");
> + return ERR_PTR(ret);
> + }
> + }
> +
> ret = drm_plane_create_zpos_property(&layer->plane, index,
> 0, plane_cnt - 1);
> if (ret) {
> base-commit: 091b06388b117fefdcb63336ac09956a4914b961
>