Re: [PATCH 1/3] drm/omap: Support for HDMI hot plug detection

[Laurent Pinchart]

Hi Tomi,

On Monday 22 May 2017 14:59:09 Tomi Valkeinen wrote:
> On 15/05/17 12:03, Peter Ujfalusi wrote:
> > The HPD signal can be used for detecting HDMI cable plug and unplug event
> > without the need for polling the status of the line.
> > This will speed up detecting such event because we do not need to wait for
> > the next poll event to notice the state change.
> > 
> > Signed-off-by: Peter Ujfalusi <peter.ujfalusi@xxxxxx>
> > ---
> > 
> >  drivers/gpu/drm/omapdrm/dss/omapdss.h    | 17 +++++++++++++++++
> >  drivers/gpu/drm/omapdrm/omap_connector.c | 32 ++++++++++++++++++++++++++-
> >  drivers/gpu/drm/omapdrm/omap_drv.c       | 29 ++++++++++++++++++++++++++
> >  3 files changed, 77 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/gpu/drm/omapdrm/dss/omapdss.h
> > b/drivers/gpu/drm/omapdrm/dss/omapdss.h index b19dae1fd6c5..1f01669eb610
> > 100644
> > --- a/drivers/gpu/drm/omapdrm/dss/omapdss.h
> > +++ b/drivers/gpu/drm/omapdrm/dss/omapdss.h
> > @@ -25,6 +25,7 @@
> > 
> >  #include <video/videomode.h>
> >  #include <linux/platform_data/omapdss.h>
> >  #include <uapi/drm/drm_mode.h>
> > 
> > +#include <drm/drm_crtc.h>
> > 
> >  #define DISPC_IRQ_FRAMEDONE		(1 << 0)
> >  #define DISPC_IRQ_VSYNC			(1 << 1)
> > 
> > @@ -555,6 +556,14 @@ struct omapdss_hdmi_ops {
> > 
> >  	int (*read_edid)(struct omap_dss_device *dssdev, u8 *buf, int len);
> >  	bool (*detect)(struct omap_dss_device *dssdev);
> > 
> > +	int (*register_hpd_cb)(struct omap_dss_device *dssdev,
> > +			       void (*cb)(void *cb_data,
> > +					  enum drm_connector_status status),
> > +			       void *cb_data);
> > +	void (*unregister_hpd_cb)(struct omap_dss_device *dssdev);
> > +	void (*enable_hpd)(struct omap_dss_device *dssdev);
> > +	void (*disable_hpd)(struct omap_dss_device *dssdev);
> > +
> > 
> >  	int (*set_hdmi_mode)(struct omap_dss_device *dssdev, bool hdmi_mode);
> >  	int (*set_infoframe)(struct omap_dss_device *dssdev,
> >  	
> >  		const struct hdmi_avi_infoframe *avi);
> > 
> > @@ -761,6 +770,14 @@ struct omap_dss_driver {
> > 
> >  	int (*read_edid)(struct omap_dss_device *dssdev, u8 *buf, int len);
> >  	bool (*detect)(struct omap_dss_device *dssdev);
> > 
> > +	int (*register_hpd_cb)(struct omap_dss_device *dssdev,
> > +			       void (*cb)(void *cb_data,
> > +					  enum drm_connector_status status),
> > +			       void *cb_data);
> > +	void (*unregister_hpd_cb)(struct omap_dss_device *dssdev);
> > +	void (*enable_hpd)(struct omap_dss_device *dssdev);
> > +	void (*disable_hpd)(struct omap_dss_device *dssdev);
> > +
> > 
> >  	int (*set_hdmi_mode)(struct omap_dss_device *dssdev, bool hdmi_mode);
> >  	int (*set_hdmi_infoframe)(struct omap_dss_device *dssdev,
> >  	
> >  		const struct hdmi_avi_infoframe *avi);
> > 
> > diff --git a/drivers/gpu/drm/omapdrm/omap_connector.c
> > b/drivers/gpu/drm/omapdrm/omap_connector.c index
> > c24b6b783e9a..5219e340ed9d 100644
> > --- a/drivers/gpu/drm/omapdrm/omap_connector.c
> > +++ b/drivers/gpu/drm/omapdrm/omap_connector.c
> > @@ -35,6 +35,18 @@ struct omap_connector {
> > 
> >  	bool hdmi_mode;
> >  
> >  };
> > 
> > +static void omap_connector_hpd_cb(void *cb_data,
> > +				  enum drm_connector_status status)
> > +{
> > +	struct omap_connector *omap_connector = cb_data;
> > +	struct drm_connector *connector = &omap_connector->base;
> > +
> > +	if (connector->status != status) {
> > +		connector->status = status;
> > +		drm_kms_helper_hotplug_event(omap_connector->base.dev);
> > +	}
> > +}
> 
> I'm not sure if this is racy or not... drm_kms_helper_hotplug_event()
> should be called without locks held, but is it safe to touch
> connector->status without any locks?

We should at least protect against internal race conditions if the hpd 
callback can be called concurrently (I haven't checked the callers yet). I 
think it would be safer to use the mode_config mutex the same way 
drm_helper_hpd_irq_event() does. Something like the following should do.

	struct omap_connector *omap_connector = cb_data;
	struct drm_connector *connector = &omap_connector->base;
	struct drm_device *dev = connector->dev;
	enum drm_connector_status old_status;

	mutex_lock(&dev->mode_config.mutex);
	old_status = connector->status;
	connector->status = status;
	mutex_unlock(&dev->mode_config.mutex);

	if (old_status != status)
		drm_kms_helper_hotplug_event(dev);

> > +	if (connector->status != status) {
> > +		connector->status = status;
> > +		drm_kms_helper_hotplug_event(omap_connector->base.dev);
> > +	}
> > +
> >  bool omap_connector_get_hdmi_mode(struct drm_connector *connector)
> >  {
> >  	struct omap_connector *omap_connector = to_omap_connector(connector);
> > @@ -75,6 +87,10 @@ static void omap_connector_destroy(struct drm_connector
> > *connector)
> >  	struct omap_dss_device *dssdev = omap_connector->dssdev;
> >  	
> >  	DBG("%s", omap_connector->dssdev->name);
> > +	if (connector->polled == DRM_CONNECTOR_POLL_HPD &&
> > +	    dssdev->driver->unregister_hpd_cb) {
> > +		dssdev->driver->unregister_hpd_cb(dssdev);
> > +	}
> >  	drm_connector_unregister(connector);
> >  	drm_connector_cleanup(connector);
> >  	kfree(omap_connector);
> > @@ -216,6 +232,7 @@ struct drm_connector *omap_connector_init(struct
> > drm_device *dev,
> >  {
> >  	struct drm_connector *connector = NULL;
> >  	struct omap_connector *omap_connector;
> > +	bool hpd_supported = false;
> > 
> >  	DBG("%s", dssdev->name);
> > 
> > @@ -233,7 +250,20 @@ struct drm_connector *omap_connector_init(struct
> > drm_device *dev,
> >  				connector_type);
> >  	
> >  	drm_connector_helper_add(connector, &omap_connector_helper_funcs);
> > 
> > -	if (dssdev->driver->detect)
> > +	if (dssdev->driver->register_hpd_cb) {
> > +		int ret = dssdev->driver->register_hpd_cb(dssdev,
> > +							  
omap_connector_hpd_cb,
> > +							  omap_connector);
> > +		if (!ret)
> > +			hpd_supported = true;
> > +		else if (ret != -ENOTSUPP)
> > +			DBG("%s: Failed to register HPD callback (%d).",
> > +			    dssdev->name, ret);
> 
> This should be an error print, shouldn't it?

Can it actually happen ? If it can't, I wouldn't bother with a message at all.

-- 
Regards,

Laurent Pinchart