[PATCH V5 RESEND] drm/rockchip: Reject AFBC for res >2560 on rk3399

[Andy Yan]

Hello Konstanin and  Daniel，

At 2025-06-04 17:00:53, "Konstantin Shabanov" <mail at etehtsea.me> wrote:
>As it isn't supported by rk3399. From the datasheet[1]
>("1.2.10 Video IN/OUT", "Display Interface", p. 17):
>
>  Support AFBC function co-operation with GPU
>    * support 2560x1600 UI
>
>Signed-off-by: Konstantin Shabanov <mail at etehtsea.me>
>Reported-by: Dan Callaghan <djc at djc.id.au>
>Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7968
>
>[1]: https://opensource.rock-chips.com/images/d/d7/Rockchip_RK3399_Datasheet_V2.1-20200323.pdf
>---
> V4 -> V5: Extract AFBC support check into drv
> V3 -> V4: Correct redundant header inclusion
> V2 -> V3: Run check only on rk3399
> V1 -> V2: Move the check to the fb_create callback
>
> AFBC check is implemented in a similar manner as in the malidp driver.
> Resend of the https://lore.kernel.org/all/20250521160556.14346-2-mail@etehtsea.me/.
> 
> drivers/gpu/drm/rockchip/rockchip_drm_drv.c | 16 ++++++++++++++++
> drivers/gpu/drm/rockchip/rockchip_drm_drv.h |  2 ++
> drivers/gpu/drm/rockchip/rockchip_drm_fb.c  |  3 +++
> 3 files changed, 21 insertions(+)
>
>diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_drv.c b/drivers/gpu/drm/rockchip/rockchip_drm_drv.c
>index 180fad5d49ad..9fb04022b2f3 100644
>--- a/drivers/gpu/drm/rockchip/rockchip_drm_drv.c
>+++ b/drivers/gpu/drm/rockchip/rockchip_drm_drv.c
>@@ -42,6 +42,8 @@
> #define DRIVER_MAJOR	1
> #define DRIVER_MINOR	0
> 
>+#define RK3399_AFBC_MAX_WIDTH		2560
>+
> static const struct drm_driver rockchip_drm_driver;
> 
> /*
>@@ -350,6 +352,20 @@ int rockchip_drm_endpoint_is_subdriver(struct device_node *ep)
> 	return false;
> }
> 
>+bool rockchip_verify_afbc_framebuffer_size(struct drm_device *dev,
>+					   const struct drm_mode_fb_cmd2 *mode_cmd)
>+{
>+	if (of_machine_is_compatible("rockchip,rk3399") &&
>+	    mode_cmd->width > RK3399_AFBC_MAX_WIDTH) {
>+		DRM_DEBUG_KMS("AFBC is not supported for the width %d (max %d)\n",
>+			      mode_cmd->width,
>+			      RK3399_AFBC_MAX_WIDTH);
>+		return false;
>+	}
>+
>+	return true;
>+}
>+
> static void rockchip_drm_match_remove(struct device *dev)
> {
> 	struct device_link *link;
>diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_drv.h b/drivers/gpu/drm/rockchip/rockchip_drm_drv.h
>index c183e82a42a5..5dabceaa4fd6 100644
>--- a/drivers/gpu/drm/rockchip/rockchip_drm_drv.h
>+++ b/drivers/gpu/drm/rockchip/rockchip_drm_drv.h
>@@ -86,6 +86,8 @@ int rockchip_drm_wait_vact_end(struct drm_crtc *crtc, unsigned int mstimeout);
> int rockchip_drm_encoder_set_crtc_endpoint_id(struct rockchip_encoder *rencoder,
> 					      struct device_node *np, int port, int reg);
> int rockchip_drm_endpoint_is_subdriver(struct device_node *ep);
>+bool rockchip_verify_afbc_framebuffer_size(struct drm_device *dev,
>+					   const struct drm_mode_fb_cmd2 *mode_cmd);
> extern struct platform_driver cdn_dp_driver;
> extern struct platform_driver dw_hdmi_rockchip_pltfm_driver;
> extern struct platform_driver dw_hdmi_qp_rockchip_pltfm_driver;
>diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_fb.c b/drivers/gpu/drm/rockchip/rockchip_drm_fb.c
>index 5829ee061c61..f0527f12f568 100644
>--- a/drivers/gpu/drm/rockchip/rockchip_drm_fb.c
>+++ b/drivers/gpu/drm/rockchip/rockchip_drm_fb.c
>@@ -52,6 +52,9 @@ rockchip_fb_create(struct drm_device *dev, struct drm_file *file,
> 	}
> 


I found that when calling drm_gem_fb_init_with_funcs earlier,  drm_any_plane_has_format would be called.

bool drm_plane_has_format(struct drm_plane *plane,
                          u32 format, u64 modifier)
{               
        unsigned int i;

        for (i = 0; i < plane->format_count; i++) {
                if (format == plane->format_types[i])
                        break; 
        }       
        if (i == plane->format_count)
                return false;

        if (plane->funcs->format_mod_supported) {
                if (!plane->funcs->format_mod_supported(plane, format, modifier))
                        return false;
        } else {
                if (!plane->modifier_count)

If we extend the format_mod_supported  by pass down fb， we still can do this
check by format_mod_supported at the fb create state ？



> 	if (drm_is_afbc(mode_cmd->modifier[0])) {
>+		if (!rockchip_verify_afbc_framebuffer_size(dev, mode_cmd))
>+			return ERR_PTR(-EINVAL);
>+
> 		ret = drm_gem_fb_afbc_init(dev, mode_cmd, afbc_fb);
> 		if (ret) {
> 			drm_framebuffer_put(&afbc_fb->base);
>
>base-commit: 685c407f168cb49a12cc703230d1e2d62767bfd2
>-- 
>2.49.0