Re: [PATCH 3/5] drm/msm/dp: suppress bridge hotplug events during MST operation

[Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>]

On Mon, Aug 17, 2026 at 04:01:36PM +0800, Yongxing Mou wrote:
> 
> 
> On 7/12/2026 6:44 PM, Dmitry Baryshkov wrote:
> > On Mon, Jun 29, 2026 at 10:48:05PM +0800, Yongxing Mou wrote:
> > > The DP MST framework already generates the required hotplug events for
> > > MST topology changes.
> > > 
> > > Suppress connector hotplug event generation from the bridge connector
> > > path while MST is active, and continue propagating HPD notifications to
> > > the DP driver.
> > > 
> > > Signed-off-by: Yongxing Mou <yongxing.mou@oss.qualcomm.com>
> > > ---
> > >   drivers/gpu/drm/display/drm_bridge_connector.c | 6 ++++--
> > >   drivers/gpu/drm/msm/dp/dp_display.c            | 9 ++++++++-
> > 
> > No, it can't go as this. The drm_bridge_connector part should have been
> > a part of the first patch.
> > 
> Got it.
> > >   2 files changed, 12 insertions(+), 3 deletions(-)
> > > 
> > > diff --git a/drivers/gpu/drm/display/drm_bridge_connector.c b/drivers/gpu/drm/display/drm_bridge_connector.c
> > > index 7334d6677604..82ed0dc450ab 100644
> > > --- a/drivers/gpu/drm/display/drm_bridge_connector.c
> > > +++ b/drivers/gpu/drm/display/drm_bridge_connector.c
> > > @@ -162,6 +162,7 @@ static void drm_bridge_connector_handle_hpd(struct drm_bridge_connector *drm_bri
> > >   {
> > >   	struct drm_connector *connector = &drm_bridge_connector->base;
> > >   	struct drm_device *dev = connector->dev;
> > > +	bool send_hotplug = true;
> > >   	/*
> > >   	 * IRQ-only notification: extra_status carries the event but
> > > @@ -179,9 +180,10 @@ static void drm_bridge_connector_handle_hpd(struct drm_bridge_connector *drm_bri
> > >   	connector->status = status;
> > >   	mutex_unlock(&dev->mode_config.mutex);
> > > -	drm_bridge_connector_hpd_notify(connector, status, extra_status, NULL);
> > > +	drm_bridge_connector_hpd_notify(connector, status, extra_status, &send_hotplug);
> > > -	drm_kms_helper_connector_hotplug_event(connector);
> > > +	if (send_hotplug)
> > > +		drm_kms_helper_connector_hotplug_event(connector);
> > 
> > But now I can also see that the idea seems to be incorrect. You are
> > preventing the kernel from sending the events, but it doesn't really
> > matter. If connection status oscillates, then other components might
> > notice it even without the event being sent.
> > 
> Here our original intent behind suppressing the hotplug was
> that, under MST, hotplugs should be sent independently by the MST
> framework, rather than being emitted along the IRQ handling path.
> 
> Following the extra_status approach you introduced, would it make sense
> to let the bridge_connector layer distinguish between long pulses (HPD)
> and short pulses (IRQ_HPD) -- a short pulse would not update
> connector->status and would not emit a hotplug, only forwarding the
> event to the bridge driver. Does this direction sound reasonable to
> you?

IRQ_HPD is not limited to the MST usecase.

> > 
> > >   }
> > >   static void drm_bridge_connector_hpd_cb(void *cb_data,
> > > diff --git a/drivers/gpu/drm/msm/dp/dp_display.c b/drivers/gpu/drm/msm/dp/dp_display.c
> > > index 6835c68fe510..bc93b566fbca 100644
> > > --- a/drivers/gpu/drm/msm/dp/dp_display.c
> > > +++ b/drivers/gpu/drm/msm/dp/dp_display.c
> > > @@ -1790,10 +1790,17 @@ void msm_dp_bridge_hpd_notify(struct drm_bridge *bridge,
> > >   			msm_dp_hpd_plug_handle(dp);
> > >   		} else {
> > >   			msm_dp_hpd_plug_handle(dp);
> > > +			/* mst_active is set in plug_handle; suppress SST hotplug */
> > > +			if (send_hotplug && msm_dp_display->mst_active)
> > > +				*send_hotplug = false;
> > >   		}
> > >   	} else {
> > > -		if (hpd_link_status == ISR_DISCONNECTED)
> > > +		if (!msm_dp_display->mst_active) {
> > >   			msm_dp_hpd_unplug_handle(dp);
> > > +		} else if (send_hotplug) {
> > > +			msm_dp_hpd_unplug_handle(dp);
> > > +			*send_hotplug = false;
> > 
> > Why? Disconnected events definitely should be reported further.
> > 
> In MST case, MST DRM framework will send the discconnected hotplug by
> msm_dp_mst_display_set_mgr_state(&dp->msm_dp_display, false);.

msm_dp_mst_display_set_mgr_state() is not a part of MST DRM framework.
But, while we are at it. The MST framwork needs to be updated to send
connector-specific HPD events.

> > > +		}
> > >   	}
> > >   	pm_runtime_put_sync(&msm_dp_display->pdev->dev);
> > > 
> > > -- 
> > > 2.43.0
> > > 
> > 
> 

-- 
With best wishes
Dmitry

_______________________________________________
linux-amlogic mailing list
linux-amlogic@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-amlogic