Re: [PATCH 5/5] drm/i915: Link Rate fallback on Link training failure

[Manasi Navare <manasi.d.navare@intel.com>]

On Mon, Oct 24, 2016 at 11:23:39PM -0700, Pandiyan, Dhinakaran wrote:
> On Fri, 2016-10-21 at 16:45 -0700, Manasi Navare wrote:
> > If link training at a link rate optimal for a particular
> > mode fails during modeset's atomic commit phase, then we
> > let the modeset complete and then retry. We save the link rate
> > value at which link training failed and use a lower link rate
> > to prune the modes. It will redo the modeset on the current mode
> > at lower link rate or if the current mode gets pruned due to lower
> > link constraints then, it will send a hotplug uevent for userspace
> > to handle it.
> > 
> > This is also required to pass DP CTS tests 4.3.1.3, 4.3.1.4,
> > 4.3.1.6.
> > 
> > Cc: Jani Nikula <jani.nikula@linux.intel.com>
> > Cc: Daniel Vetter <daniel.vetter@intel.com>
> > Cc: Ville Syrjala <ville.syrjala@linux.intel.com>
> > Signed-off-by: Manasi Navare <manasi.d.navare@intel.com>
> > ---
> >  drivers/gpu/drm/i915/intel_ddi.c              | 15 +++++-
> >  drivers/gpu/drm/i915/intel_dp.c               | 69 ++++++++++++++++++++++++++-
> >  drivers/gpu/drm/i915/intel_dp_link_training.c | 12 +++--
> >  drivers/gpu/drm/i915/intel_drv.h              |  6 ++-
> >  4 files changed, 95 insertions(+), 7 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c
> > index fb18d69..451433b 100644
> > --- a/drivers/gpu/drm/i915/intel_ddi.c
> > +++ b/drivers/gpu/drm/i915/intel_ddi.c
> > @@ -1712,6 +1712,8 @@ static void intel_ddi_pre_enable_dp(struct intel_encoder *encoder,
> >  	struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base);
> >  	struct drm_i915_private *dev_priv = to_i915(encoder->base.dev);
> >  	enum port port = intel_ddi_get_encoder_port(encoder);
> > +	struct intel_connector *intel_connector = intel_dp->attached_connector;
> > +	struct drm_connector *connector = &intel_connector->base;
> >  
> >  	intel_dp_set_link_params(intel_dp, link_rate, lane_count,
> >  				 link_mst);
> > @@ -1722,7 +1724,18 @@ static void intel_ddi_pre_enable_dp(struct intel_encoder *encoder,
> >  	intel_prepare_dp_ddi_buffers(encoder);
> >  	intel_ddi_init_dp_buf_reg(encoder);
> >  	intel_dp_sink_dpms(intel_dp, DRM_MODE_DPMS_ON);
> > -	intel_dp_start_link_train(intel_dp);
> > +	if (!intel_dp_start_link_train(intel_dp)) {
> > +		DRM_ERROR("Link Training failed at link rate = %d, lane count = %d\n",
> > +			  link_rate, lane_count);
> > +		intel_dp->link_train_failed = true;
> > +		intel_dp->link_train_failed_link_rate = link_rate;
> > +		intel_dp->link_train_failed_lane_count = lane_count;
> > +		/* Schedule a Hotplug Uevent to userspace to start modeset */
> > +		schedule_work(&connector->i915_modeset_retry_work);
> > +	} else {
> > +		intel_dp->link_train_failed = false;
> > +	}
> > +
> >  	if (port != PORT_A || INTEL_GEN(dev_priv) >= 9)
> >  		intel_dp_stop_link_train(intel_dp);
> >  }
> > diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c
> > index c192e18..5d5f4a7 100644
> > --- a/drivers/gpu/drm/i915/intel_dp.c
> > +++ b/drivers/gpu/drm/i915/intel_dp.c
> > @@ -313,6 +313,7 @@ static int intel_dp_link_rate_index(struct intel_dp *intel_dp,
> >  	int target_clock = mode->clock;
> >  	int max_rate, mode_rate, max_lanes, max_link_clock;
> >  	int max_dotclk;
> > +	int common_rates[DP_MAX_SUPPORTED_RATES] = {};
> >  
> >  	max_dotclk = intel_dp_downstream_max_dotclock(intel_dp);
> >  
> > @@ -326,8 +327,27 @@ static int intel_dp_link_rate_index(struct intel_dp *intel_dp,
> >  		target_clock = fixed_mode->clock;
> >  	}
> >  
> > -	max_link_clock = intel_dp_max_link_rate(intel_dp);
> > -	max_lanes = intel_dp_max_lane_count(intel_dp);
> > +	/* Prune the modes based on the link rate that failed */
> > +	if (intel_dp->link_train_failed_link_rate) {
> 
> Shouldn't the bool link_train_failed that you are adding be used here?
> The naming indicates a check like this should use that.

Either that or the link_train_failed_link_rate can be used, both get
initialized and reset the same time. But using link_train_failed would be more
appropriate I guess, I will look at changing that.

Manasi
> 
> 
> > +		intel_dp->link_rate_index = intel_dp_link_rate_index(intel_dp,
> > +								     common_rates,
> > +								     intel_dp->link_train_failed_link_rate);
> > +		if (intel_dp->link_rate_index > 0) {
> > +			max_link_clock = common_rates[intel_dp->link_rate_index - 1];
> > +			max_lanes = intel_dp_max_lane_count(intel_dp);
> > +		} else {
> > +			/* Here we can lower the lane count, but that will be
> > +			 * added for DP Spec 1.3
> > +			 */
> > +			DRM_ERROR("No Valid Mode Supported for this Link\n");
> > +			intel_dp->link_train_failed_link_rate = 0;
> > +			intel_dp->link_rate_index = -1;
> > +			intel_dp->link_train_failed = false;
> 
> 1. You are not setting max_link_clock and max_lanes here. Won't this end
> up computing the max_rate based on uninitialized locals?
> 
> 2. Also, what is the correct return value in this case?
> 
> 3. You are resetting all values when a link rate cannot be found, what
> link rate is the subsequent modeset expected to start from? 
> 
> 
> -DK

The idea is to just fail the link training in this case since we cannot
further reduce the link rate/lane count so it shoudl just fail
the modeset and send a uevent to userspace. I think this logic will
need further tuning.


> > +		}
> > +	} else {
> > +		max_link_clock = intel_dp_max_link_rate(intel_dp);
> > +		max_lanes = intel_dp_max_lane_count(intel_dp);
> > +	}
> >  
> >  	max_rate = intel_dp_max_data_rate(max_link_clock, max_lanes);
> >  	mode_rate = intel_dp_link_required(target_clock, 18);
> > @@ -1619,6 +1639,14 @@ static int intel_dp_compute_bpp(struct intel_dp *intel_dp,
> >  	if (adjusted_mode->flags & DRM_MODE_FLAG_DBLCLK)
> >  		return false;
> >  
> > +	/* Fall back to lower link rate in case of failure in previous modeset */
> > +	if (intel_dp->link_train_failed_link_rate) {
> > +		min_lane_count = max_lane_count;
> > +		min_clock = max_clock = intel_dp->link_rate_index - 1;
> 
> 1. Why is min_clock same as max_clock? This forces the clock to a value
> instead of iterating it based on the mode rate.
> 
> 2. Why is min_lane_count set to max_lane_count? Is this done to be Spec
> compliant? 
> 
> 3. What about bpp? Looks like the we will still be iterating over bpp to
> find the link rate. 
> 
> -DK
> > +		intel_dp->link_train_failed_link_rate = 0;
> > +		intel_dp->link_rate_index = -1;
> > +	}
> > +
> >  	DRM_DEBUG_KMS("DP link computation with max lane count %i "
> >  		      "max bw %d pixel clock %iKHz\n",
> >  		      max_lane_count, common_rates[max_clock],
> > @@ -5689,6 +5717,39 @@ static bool intel_edp_init_connector(struct intel_dp *intel_dp,
> >  	return false;
> >  }
> >  
> > +static void intel_dp_modeset_retry_work_fn(struct work_struct *work)
> > +{
> > +	struct drm_connector *connector;
> > +	struct intel_dp *intel_dp;
> > +	struct drm_display_mode *mode;
> > +	bool verbose_prune = true;
> > +	bool reprobe = false;
> > +
> > +	connector = container_of(work, typeof(*connector),
> > +				 i915_modeset_retry_work);
> > +	intel_dp = intel_attached_dp(connector);
> > +
> > +	/* Grab the locks before changing connector property*/
> > +	mutex_lock(&connector->dev->mode_config.mutex);
> > +	DRM_DEBUG_KMS("[CONNECTOR:%d:%s]\n", connector->base.id,
> > +		      connector->name);
> > +	list_for_each_entry(mode, &connector->modes, head) {
> > +		mode->status = intel_dp_mode_valid(connector,
> > +						   mode);
> > +		if (mode->status != MODE_OK)
> > +			reprobe = true;
> > +	}
> > +	drm_mode_prune_invalid(connector->dev, &connector->modes,
> > +			       verbose_prune);
> > +	mutex_unlock(&connector->dev->mode_config.mutex);
> > +	if (reprobe) {
> > +		/* Send Hotplug uevent so userspace can reprobe */
> > +		drm_kms_helper_hotplug_event(connector->dev);
> > +	}
> > +	if (intel_dp->link_train_failed)
> > +		drm_atomic_helper_connector_modeset(connector);
> > +}
> > +
> >  bool
> >  intel_dp_init_connector(struct intel_digital_port *intel_dig_port,
> >  			struct intel_connector *intel_connector)
> > @@ -5701,6 +5762,10 @@ static bool intel_edp_init_connector(struct intel_dp *intel_dp,
> >  	enum port port = intel_dig_port->port;
> >  	int type;
> >  
> > +	/* Initialize the work for modeset in case of link train failure */
> > +	INIT_WORK(&connector->i915_modeset_retry_work,
> > +		  intel_dp_modeset_retry_work_fn);
> > +
> >  	if (WARN(intel_dig_port->max_lanes < 1,
> >  		 "Not enough lanes (%d) for DP on port %c\n",
> >  		 intel_dig_port->max_lanes, port_name(port)))
> > diff --git a/drivers/gpu/drm/i915/intel_dp_link_training.c b/drivers/gpu/drm/i915/intel_dp_link_training.c
> > index 0048b52..10f81ab 100644
> > --- a/drivers/gpu/drm/i915/intel_dp_link_training.c
> > +++ b/drivers/gpu/drm/i915/intel_dp_link_training.c
> > @@ -310,9 +310,15 @@ void intel_dp_stop_link_train(struct intel_dp *intel_dp)
> >  				DP_TRAINING_PATTERN_DISABLE);
> >  }
> >  
> > -void
> > +bool
> >  intel_dp_start_link_train(struct intel_dp *intel_dp)
> >  {
> > -	intel_dp_link_training_clock_recovery(intel_dp);
> > -	intel_dp_link_training_channel_equalization(intel_dp);
> > +	bool ret;
> > +
> > +	if (intel_dp_link_training_clock_recovery(intel_dp)) {
> > +		ret = intel_dp_link_training_channel_equalization(intel_dp);
> > +		if (ret)
> > +			return true;
> > +	}
> > +	return false;
> >  }
> > diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h
> > index 4e90b07..d3fcffc 100644
> > --- a/drivers/gpu/drm/i915/intel_drv.h
> > +++ b/drivers/gpu/drm/i915/intel_drv.h
> > @@ -890,6 +890,10 @@ struct intel_dp {
> >  	uint32_t DP;
> >  	int link_rate;
> >  	uint8_t lane_count;
> > +	int link_train_failed_link_rate;
> > +	uint8_t link_train_failed_lane_count;
> > +	int link_rate_index;
> > +	bool link_train_failed;
> >  	uint8_t sink_count;
> >  	bool link_mst;
> >  	bool has_audio;
> > @@ -1394,7 +1398,7 @@ bool intel_dp_init_connector(struct intel_digital_port *intel_dig_port,
> >  void intel_dp_set_link_params(struct intel_dp *intel_dp,
> >  			      int link_rate, uint8_t lane_count,
> >  			      bool link_mst);
> > -void intel_dp_start_link_train(struct intel_dp *intel_dp);
> > +bool intel_dp_start_link_train(struct intel_dp *intel_dp);
> >  void intel_dp_stop_link_train(struct intel_dp *intel_dp);
> >  void intel_dp_sink_dpms(struct intel_dp *intel_dp, int mode);
> >  void intel_dp_encoder_reset(struct drm_encoder *encoder);
> 
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx