All of lore.kernel.org
 help / color / mirror / Atom feed
From: Jani Nikula <jani.nikula@linux.intel.com>
To: Manasi Navare <manasi.d.navare@intel.com>,
	intel-gfx@lists.freedesktop.org
Cc: Daniel Vetter <daniel.vetter@intel.com>
Subject: Re: [PATCH 5/5] drm/i915: Link Rate fallback on Link training failure
Date: Tue, 25 Oct 2016 15:17:47 +0300	[thread overview]
Message-ID: <87r374mxlg.fsf@intel.com> (raw)
In-Reply-To: <1477093543-11622-6-git-send-email-manasi.d.navare@intel.com>

On Sat, 22 Oct 2016, Manasi Navare <manasi.d.navare@intel.com> 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;

I think eventually you'll need to store a list (array) of failing link
rate, lane count pairs, not just the last that failed. Now you restrict
the link config computation to only reducing the link rate. But
currently (for whatever reason, it's flip-flopped too many times) we
start with wide & slow, meaning that in many cases we've already
exhausted the option to go slower. If optimal fails, maybe we need to
try narrow & fast instead.

BR,
Jani.


> +		/* 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) {
> +		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;
> +		}
> +	} 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;
> +		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);

-- 
Jani Nikula, Intel Open Source Technology Center
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

  parent reply	other threads:[~2016-10-25 12:17 UTC|newest]

Thread overview: 33+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-10-21 23:45 [PATCH 0/5] Handle link training failure for DDI platforms Manasi Navare
2016-10-21 23:45 ` [PATCH 1/5] drm: Add atomic helper to redo a modeset on current mode Manasi Navare
2016-10-22  8:47   ` Daniel Vetter
2016-10-22 14:01     ` [Intel-gfx] " Daniel Vetter
2016-10-22 14:46       ` Ville Syrjälä
2016-10-24  6:00         ` Daniel Vetter
2016-10-24  6:12           ` Manasi Navare
2016-10-24  6:33             ` Daniel Vetter
2016-10-24  7:00               ` Manasi Navare
2016-10-24  7:12                 ` Daniel Vetter
2016-10-24 18:38                   ` [Intel-gfx] " Sean Paul
2016-10-25  6:35                     ` Daniel Vetter
2016-10-24 22:08                   ` Manasi Navare
2016-10-25  6:40                     ` [Intel-gfx] " Daniel Vetter
2016-10-25 12:09   ` Jani Nikula
2016-10-25 22:28     ` Manasi Navare
2016-10-25 22:38     ` Rodrigo Vivi
2016-10-21 23:45 ` [PATCH 2/5] drm: Define a work struct for scheduling a uevent for modeset retry Manasi Navare
2016-10-22  8:48   ` Daniel Vetter
2016-10-25  6:28     ` Manasi Navare
2016-10-25  6:30       ` Pandiyan, Dhinakaran
2016-10-25  6:45         ` [Intel-gfx] " Daniel Vetter
2016-10-21 23:45 ` [PATCH 3/5] drm/i915: Change the placement of some static functions in intel_dp.c Manasi Navare
2016-10-21 23:45 ` [PATCH 4/5] drm/i915; Add a function to return index of link rate Manasi Navare
2016-10-25  6:33   ` Pandiyan, Dhinakaran
2016-10-21 23:45 ` [PATCH 5/5] drm/i915: Link Rate fallback on Link training failure Manasi Navare
2016-10-24 17:53   ` Jim Bride
2016-10-25  6:23   ` Pandiyan, Dhinakaran
2016-10-25 18:32     ` Manasi Navare
2016-10-25 12:17   ` Jani Nikula [this message]
2016-10-25 18:00     ` Jim Bride
2016-10-25 18:37     ` Manasi Navare
2016-10-22  0:16 ` ✗ Fi.CI.BAT: warning for Handle link training failure for DDI platforms Patchwork

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=87r374mxlg.fsf@intel.com \
    --to=jani.nikula@linux.intel.com \
    --cc=daniel.vetter@intel.com \
    --cc=intel-gfx@lists.freedesktop.org \
    --cc=manasi.d.navare@intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.