Re: [PATCH v6 17/22] drm: bridge: dw_hdmi: Drop call to drm_bridge_hpd_notify()

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] Removing drm_bridge_hpd_notify() violates the DRM bridge API contract and causes hotplug events to be permanently lost during system suspend.
--

commit 6b3491e39082015ab62431ea0c93c74c9344b995
Author: Jonas Karlman <jonas@kwiboo.se>

drm: bridge: dw_hdmi: Drop call to drm_bridge_hpd_notify()

This commit removes the drm_bridge_hpd_notify() call from the dw_hdmi IRQ
handler, relying instead only on drm_helper_hpd_irq_event() to avoid sending
duplicate hotplug uevents to userspace.

> diff --git a/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c b/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c
> index e9c4e24c090ca..6cc7b2a860bd8 100644
> --- a/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c
> +++ b/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c
> @@ -3101,10 +3101,8 @@ static irqreturn_t dw_hdmi_irq(int irq, void *dev_id)
>  			status == connector_status_connected ?
>  			"plugin" : "plugout");
>  
> -		if (hdmi->bridge.dev) {
> +		if (hdmi->bridge.dev)
>  			drm_helper_hpd_irq_event(hdmi->bridge.dev);
> -			drm_bridge_hpd_notify(&hdmi->bridge, status);
> -		}
>  	}

Will removing drm_bridge_hpd_notify() cause hotplug events to be permanently
lost if they occur during system suspend?

Since the driver advertises DRM_BRIDGE_OP_HPD, does the DRM bridge API
contract still require it to call drm_bridge_hpd_notify() to properly
propagate hotplug events to the rest of the bridge chain?

Looking at drm_helper_hpd_irq_event(), it aborts early if polling is
disabled:

drivers/gpu/drm/drm_probe_helper.c:drm_helper_hpd_irq_event() {
	struct drm_connector *connector, *first_changed_connector = NULL;
	struct drm_connector_list_iter conn_iter;
	int changed = 0;

	if (!dev->mode_config.poll_enabled)
		return false;
...
}

During system suspend, drm_kms_helper_poll_disable() sets poll_enabled
to false. If an HPD interrupt fires while the system is suspended, won't
drm_helper_hpd_irq_event() silently drop the event without updating the
connector status or scheduling any deferred work?

Furthermore, when the system resumes, output_poll_execute() explicitly skips
connectors that use the DRM_CONNECTOR_POLL_HPD flag, and the dw_hdmi resume
handler does not appear to manually trigger a hotplug event check.

Could this result in the system being completely unaware of any display
connections or disconnections that happened while suspended, at least until
a manual reprobe is forced by userspace?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260516183838.2024991-1-jonas@kwiboo.se?part=17