[Intel-gfx] [PATCH v2] drm/i915/dp: Update connector status for DP MST hotplugs

[Pandiyan, Dhinakaran]

On Thu, 2016-11-17 at 08:53 +0100, Daniel Vetter wrote:
> On Thu, Nov 17, 2016 at 01:53:31AM +0000, Pandiyan, Dhinakaran wrote:
> > On Sun, 2016-11-13 at 11:39 +0100, Daniel Vetter wrote:
> > > On Fri, Nov 11, 2016 at 10:21:39PM +0100, Daniel Vetter wrote:
> > > > On Mon, Nov 07, 2016 at 04:27:30PM -0800, Dhinakaran Pandiyan wrote:
> > > > > Hotplugging a monitor in DP MST configuration triggers short pulses.
> > > > > Although the short pulse handling path ends up in the MST hotplug function,
> > > > > we do not perform a detect before sending the hotplug uvent. This leads to
> > > > > the connector status not being updated when the userspace calls
> > > > > DRM_IOCTL_MODE_GETCONNECTOR.
> > > > > 
> > > > > So, let's call the connector function ->detect() to update the connector
> > > > > status before sending the uevent.
> > > > > 
> > > > > v2: Update connector status inside mode_config mutex (Ville)
> > > > > 
> > > > > Signed-off-by: Dhinakaran Pandiyan <dhinakaran.pandiyan at intel.com>
> > > > 
> > > > quick comments from me on irc:
> > > > 
> > > > <danvet> dhnkrn, really tricky I think
> > > > <danvet> dhnkrn, also feels wrong from an entire design pov
> > > > <danvet> I'd expect the topology manager could tell us from which exact downstream port the short pulse came
> > > > <danvet> which means we should be able to get at that connector without taking the global lock
> > > > <danvet> i.e. at least avoid the for_each_connector
> > > > <danvet> on the detect itself, I need to think
> > > > <danvet> but we might need to essentially postpone the ->detect work to the hotplug worker
> > > > <danvet> except that it's not a direct hpd from our chip
> > > > <danvet> but a downstream one
> > > > <danvet> this is going to be fun
> > > > <dhnkrn> Hmm .. Does updating the connector status need the lock or is it the detect() ?
> > > > <danvet> yup
> > > > <danvet> and we might need to do additional dpcd transactions (e.g. for load detect probing on downstream ports)
> > > > <danvet> or at least edid reads
> > > > <danvet> both recurse back into dp mst helper code like ville said
> > > > <danvet> dhnkrn, so instead of doing the detect work from the hotplug code
> > > > <danvet> put them on some list
> > > > <danvet> need to be careful in case they are already there
> > > > <danvet> list only protected with dedicated lock
> > > > <danvet> then launch the hpd worker
> > > > <danvet> and walk that list from there, with the mode_config.mutex to do the full-on probe
> > > > <danvet> we can't call ->detect from dig_port_work, and walking the entire connector list in there is also not a good idea imo
> > > > <danvet> dhnkrn, but that's just my preliminary analysis from like 10 minutes of thinking and checking what ville raised
> > > > <danvet> needs more thought probably
> > > > <danvet> dhnkrn, to clarify your question: both connector->status and calling ->detect need mode_config.mutex
> > > > <danvet> that mutex is what protects output probe state<Vtec234> ccr: i might have udev configured incorrectly then
> > > > 
> > > > Cheers, Daniel
> > > > > ---
> > > > >  drivers/gpu/drm/i915/intel_dp_mst.c | 22 +++++++++++++++++++++-
> > > > >  1 file changed, 21 insertions(+), 1 deletion(-)
> > > > > 
> > > > > diff --git a/drivers/gpu/drm/i915/intel_dp_mst.c b/drivers/gpu/drm/i915/intel_dp_mst.c
> > > > > index 3ffbd69..8e36a96 100644
> > > > > --- a/drivers/gpu/drm/i915/intel_dp_mst.c
> > > > > +++ b/drivers/gpu/drm/i915/intel_dp_mst.c
> > > > > @@ -492,8 +492,28 @@ static void intel_dp_mst_hotplug(struct drm_dp_mst_topology_mgr *mgr)
> > > > >  	struct intel_dp *intel_dp = container_of(mgr, struct intel_dp, mst_mgr);
> > > > >  	struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp);
> > > > >  	struct drm_device *dev = intel_dig_port->base.base.dev;
> > > > > +	struct intel_connector *intel_connector;
> > > > > +	bool changed = false;
> > > > > +	enum drm_connector_status old_status;
> > > > > +	struct drm_mode_config *mode_config = &dev->mode_config;
> > > > > +
> > > > > +	mutex_lock(&mode_config->mutex);
> > > > > +	for_each_intel_connector(dev, intel_connector) {
> > > > > +		struct drm_connector *connector = &(intel_connector->base);
> > > > > +
> > > > > +		if (intel_connector->mst_port != intel_dp)
> > > > > +			continue;
> > > > > +
> > > > > +		old_status = connector->status;
> > > > > +		connector->status = connector->funcs->detect(connector, false);
> > > > > +
> > > > > +		if (old_status != connector->status)
> > > > > +			changed = true;
> > > > > +	}
> > > > > +	mutex_unlock(&mode_config->mutex);
> > > > >  
> > > > > -	drm_kms_helper_hotplug_event(dev);
> > > 
> > > I just realized that this call here will call down into the fbdev helper,
> > > which already does the above ->detect calls. At least if there's no
> > > compositor running. So I wonder why the deadlock ville pointed out isn't
> > > blowing up already ...
> > > -Daniel
> > > 
> > 
> > You are right, that call sequence does end up taking the
> > mode_config.mutex lock recursively. Are you referring to CI results for
> > this patch?
> 
> Well I'm confused, since per our discussion here it should blow up, but it
> doesn't. What are we missing? CI can't be trusted on this since we can't
> (yet) do MST hotplug cycles, hence it doesn't exercise this codepath.
> -Daniel

The mutex is unlocked after the ->detect()'s that I added are completed.
The hotplug event is sent only after that. Or am I missing something
here?