Re: [PATCH v9 17/61] drm/display: hdmi-state-helper: Sync SCDC state on hotplug

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] The HDMI disconnect path fails to implement SCDC scrambler synchronization, resulting in dead code and potential display corruption on hotplug.
--

--- Patch [17]: [PATCH v9 17/61] drm/display: hdmi-state-helper: Sync SCDC state on hotplug ---
commit c0cc8e9a78dd0bbc36a54e19e44a65a37553ae23
Author: Cristian Ciocaltea <[email protected]>

drm/display: hdmi-state-helper: Sync SCDC state on hotplug

This commit hooks drm_connector_hdmi_sync_scdc() into both the connect and
disconnect paths in drm_atomic_helper_connector_hdmi_update() to synchronize
the SCDC status on hotplug events.

> diff --git a/drivers/gpu/drm/display/drm_hdmi_state_helper.c b/drivers/gpu/drm/display/drm_hdmi_state_helper.c
> index 04d5196a2010f..ad59318a3e85f 100644
> --- a/drivers/gpu/drm/display/drm_hdmi_state_helper.c
> +++ b/drivers/gpu/drm/display/drm_hdmi_state_helper.c
> @@ -1205,13 +1205,16 @@ drm_atomic_helper_connector_hdmi_update(struct drm_connector *connector,
>  					enum drm_connector_status status)
>  {
>  	const struct drm_edid *drm_edid;
> +	int ret = 0;
>  
>  	if (status == connector_status_disconnected) {
> -		// TODO: also handle scramber, HDMI sink disconnected.
> -		drm_connector_hdmi_audio_plugged_notify(connector, false);
> -		drm_edid_connector_update(connector, NULL);
> -		drm_connector_cec_phys_addr_invalidate(connector);
> -		return 0;
> +		ret = drm_connector_hdmi_sync_scdc(connector, false, ctx);
> +		if (ret != -EDEADLK) {

[Severity: Medium]
Does this code actually synchronize the scrambler on disconnect?

Looking at drm_connector_hdmi_sync_scdc(), it has an early return when
plugged is false:

drivers/gpu/drm/display/drm_hdmi_helper.c:drm_connector_hdmi_sync_scdc() {
	...
	if (!connector || !plugged || !ctx)
		return 0;
	...
}

Because of this early return, calling drm_connector_hdmi_sync_scdc() with
false for the plugged parameter makes it a no-op, which bypasses the
scrambler synchronization logic.

Additionally, since it always returns 0 in this case, doesn't this make the
if (ret != -EDEADLK) check dead code?

> +			drm_connector_hdmi_audio_plugged_notify(connector, false);
> +			drm_edid_connector_update(connector, NULL);
> +			drm_connector_cec_phys_addr_invalidate(connector);
> +		}
> +		return ret;
>  	}
>  

[ ... ]

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=17