[PATCH 0/3] drm/i915: MST link training locking and cleanups

[PATCH 0/3] drm/i915: MST link training locking and cleanups

[ville.syrjala]

From: Ville Syrjälä <ville.syrjala@linux.intel.com>

While glancing at the MST hpd code I noticed we have no locking in there
to protect the link retraining against a concurrent modeset. So I added some,
and then ocd struct and had to clean up the code a bit.

These patches are available at 
git://github.com/vsyrjala/linux.git mst_fixes_cleanups

In the repo I've also included an earlier patch which attempts to enable
the DDI PLL for MST on SKL [1]

[1] http://lists.freedesktop.org/archives/intel-gfx/2015-August/073965.html

Ville Syrjälä (3):
  drm/i915: Protect MST retraining with connection_mutex
  drm/i915: Flatten the mst suspend/resume functions a bit
  drm/i915: Flatten intel_dp_check_mst_status()

 drivers/gpu/drm/i915/intel_dp.c | 127 ++++++++++++++++++++--------------------
 1 file changed, 65 insertions(+), 62 deletions(-)

-- 
2.4.6

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH 1/3] drm/i915: Protect MST retraining with connection_mutex

[ville.syrjala]

From: Ville Syrjälä <ville.syrjala@linux.intel.com>

Grab the connection_mutex around MSR link retraining to protect it
against a concurrent modeset. We already do the same for SST.

DP hpd_pulse can still otherwise race against modeset and ->detect(), so
it's not clear what will happen when both want to scribble into eg.
intel_dp->dpcd[] at the same time. But sorting it all out requires way
more thought than I'm willing to expend now.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
---
 drivers/gpu/drm/i915/intel_dp.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c
index 9e90a2b..1259c9b 100644
--- a/drivers/gpu/drm/i915/intel_dp.c
+++ b/drivers/gpu/drm/i915/intel_dp.c
@@ -4391,6 +4391,7 @@ update_status:
 static int
 intel_dp_check_mst_status(struct intel_dp *intel_dp)
 {
+	struct drm_device *dev = dp_to_dig_port(intel_dp)->base.base.dev;
 	bool bret;
 
 	if (intel_dp->is_mst) {
@@ -4402,6 +4403,8 @@ intel_dp_check_mst_status(struct intel_dp *intel_dp)
 go_again:
 		if (bret == true) {
 
+			drm_modeset_lock(&dev->mode_config.connection_mutex, NULL);
+
 			/* check link status - esi[10] = 0x200c */
 			if (intel_dp->active_mst_links &&
 			    !drm_dp_channel_eq_ok(&esi[10], intel_dp->lane_count)) {
@@ -4411,6 +4414,8 @@ go_again:
 				intel_dp_stop_link_train(intel_dp);
 			}
 
+			drm_modeset_unlock(&dev->mode_config.connection_mutex);
+
 			DRM_DEBUG_KMS("got esi %3ph\n", esi);
 			ret = drm_dp_mst_hpd_irq(&intel_dp->mst_mgr, esi, &handled);
 
-- 
2.4.6

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH 1/3] drm/i915: Protect MST retraining with connection_mutex

[Ville Syrjälä]

On Thu, Aug 27, 2015 at 06:36:48PM +0300, ville.syrjala@linux.intel.com wrote:
> From: Ville Syrjälä <ville.syrjala@linux.intel.com>
> 
> Grab the connection_mutex around MSR link retraining to protect it
> against a concurrent modeset. We already do the same for SST.
> 
> DP hpd_pulse can still otherwise race against modeset and ->detect(), so
> it's not clear what will happen when both want to scribble into eg.
> intel_dp->dpcd[] at the same time. But sorting it all out requires way
> more thought than I'm willing to expend now.

Actually I suppose this might not work out so well after all. I suppose
MST depends on short HPDs during modeset due to the sideband stuff.

So if we want to grab modeset locks for retraining, I suppose we'd
need to move the retraining to happen from .hot_plug() which gets run
from the other hotplug work, and so wouldn't interfere with sideband.

> 
> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
> ---
>  drivers/gpu/drm/i915/intel_dp.c | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c
> index 9e90a2b..1259c9b 100644
> --- a/drivers/gpu/drm/i915/intel_dp.c
> +++ b/drivers/gpu/drm/i915/intel_dp.c
> @@ -4391,6 +4391,7 @@ update_status:
>  static int
>  intel_dp_check_mst_status(struct intel_dp *intel_dp)
>  {
> +	struct drm_device *dev = dp_to_dig_port(intel_dp)->base.base.dev;
>  	bool bret;
>  
>  	if (intel_dp->is_mst) {
> @@ -4402,6 +4403,8 @@ intel_dp_check_mst_status(struct intel_dp *intel_dp)
>  go_again:
>  		if (bret == true) {
>  
> +			drm_modeset_lock(&dev->mode_config.connection_mutex, NULL);
> +
>  			/* check link status - esi[10] = 0x200c */
>  			if (intel_dp->active_mst_links &&
>  			    !drm_dp_channel_eq_ok(&esi[10], intel_dp->lane_count)) {
> @@ -4411,6 +4414,8 @@ go_again:
>  				intel_dp_stop_link_train(intel_dp);
>  			}
>  
> +			drm_modeset_unlock(&dev->mode_config.connection_mutex);
> +
>  			DRM_DEBUG_KMS("got esi %3ph\n", esi);
>  			ret = drm_dp_mst_hpd_irq(&intel_dp->mst_mgr, esi, &handled);
>  
> -- 
> 2.4.6

-- 
Ville Syrjälä
Intel OTC
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH 1/3] drm/i915: Protect MST retraining with connection_mutex

[Maarten Lankhorst]

Hey,

Op 03-09-15 om 14:11 schreef Ville Syrjälä:
> On Thu, Aug 27, 2015 at 06:36:48PM +0300, ville.syrjala@linux.intel.com wrote:
>> From: Ville Syrjälä <ville.syrjala@linux.intel.com>
>>
>> Grab the connection_mutex around MSR link retraining to protect it
>> against a concurrent modeset. We already do the same for SST.
>>
>> DP hpd_pulse can still otherwise race against modeset and ->detect(), so
>> it's not clear what will happen when both want to scribble into eg.
>> intel_dp->dpcd[] at the same time. But sorting it all out requires way
>> more thought than I'm willing to expend now.
> Actually I suppose this might not work out so well after all. I suppose
> MST depends on short HPDs during modeset due to the sideband stuff.
>
> So if we want to grab modeset locks for retraining, I suppose we'd
> need to move the retraining to happen from .hot_plug() which gets run
> from the other hotplug work, and so wouldn't interfere with sideband.
>
I think it would be better to have a per encoder mutex. In the future we may want to run modeset
disable/enable async, in which case it may not hold the connection_mutex.

If we do decide on a separate mutex then it would also be useful to also think about how to
protect intel_mst_*(dis,en)able_dp. :)

~Maarten
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH 1/3] drm/i915: Protect MST retraining with connection_mutex

[Daniel Vetter]

On Wed, Sep 16, 2015 at 01:48:49PM +0200, Maarten Lankhorst wrote:
> Hey,
> 
> Op 03-09-15 om 14:11 schreef Ville Syrjälä:
> > On Thu, Aug 27, 2015 at 06:36:48PM +0300, ville.syrjala@linux.intel.com wrote:
> >> From: Ville Syrjälä <ville.syrjala@linux.intel.com>
> >>
> >> Grab the connection_mutex around MSR link retraining to protect it
> >> against a concurrent modeset. We already do the same for SST.
> >>
> >> DP hpd_pulse can still otherwise race against modeset and ->detect(), so
> >> it's not clear what will happen when both want to scribble into eg.
> >> intel_dp->dpcd[] at the same time. But sorting it all out requires way
> >> more thought than I'm willing to expend now.
> > Actually I suppose this might not work out so well after all. I suppose
> > MST depends on short HPDs during modeset due to the sideband stuff.
> >
> > So if we want to grab modeset locks for retraining, I suppose we'd
> > need to move the retraining to happen from .hot_plug() which gets run
> > from the other hotplug work, and so wouldn't interfere with sideband.
> >
> I think it would be better to have a per encoder mutex. In the future we may want to run modeset
> disable/enable async, in which case it may not hold the connection_mutex.
> 
> If we do decide on a separate mutex then it would also be useful to also think about how to
> protect intel_mst_*(dis,en)able_dp. :)

We need to sync re-training with any other modeset work. Which means
another thing to get right for async modesets.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH 2/3] drm/i915: Flatten the mst suspend/resume functions a bit

[ville.syrjala]

From: Ville Syrjälä <ville.syrjala@linux.intel.com>

I'm not a fan of deeply nested ifs. Just pull most of the conditions
into a single place to flatten things a bit.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
---
 drivers/gpu/drm/i915/intel_dp.c | 32 ++++++++++++++------------------
 1 file changed, 14 insertions(+), 18 deletions(-)

diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c
index 1259c9b..6c34784 100644
--- a/drivers/gpu/drm/i915/intel_dp.c
+++ b/drivers/gpu/drm/i915/intel_dp.c
@@ -6252,15 +6252,14 @@ void intel_dp_mst_suspend(struct drm_device *dev)
 	/* disable MST */
 	for (i = 0; i < I915_MAX_PORTS; i++) {
 		struct intel_digital_port *intel_dig_port = dev_priv->hotplug.irq_port[i];
-		if (!intel_dig_port)
+
+		if (!intel_dig_port ||
+		    intel_dig_port->base.type != INTEL_OUTPUT_DISPLAYPORT ||
+		    !intel_dig_port->dp.can_mst)
 			continue;
 
-		if (intel_dig_port->base.type == INTEL_OUTPUT_DISPLAYPORT) {
-			if (!intel_dig_port->dp.can_mst)
-				continue;
-			if (intel_dig_port->dp.is_mst)
-				drm_dp_mst_topology_mgr_suspend(&intel_dig_port->dp.mst_mgr);
-		}
+		if (intel_dig_port->dp.is_mst)
+			drm_dp_mst_topology_mgr_suspend(&intel_dig_port->dp.mst_mgr);
 	}
 }
 
@@ -6271,18 +6270,15 @@ void intel_dp_mst_resume(struct drm_device *dev)
 
 	for (i = 0; i < I915_MAX_PORTS; i++) {
 		struct intel_digital_port *intel_dig_port = dev_priv->hotplug.irq_port[i];
-		if (!intel_dig_port)
-			continue;
-		if (intel_dig_port->base.type == INTEL_OUTPUT_DISPLAYPORT) {
-			int ret;
+		int ret;
 
-			if (!intel_dig_port->dp.can_mst)
-				continue;
+		if (!intel_dig_port ||
+		    intel_dig_port->base.type != INTEL_OUTPUT_DISPLAYPORT ||
+		    !intel_dig_port->dp.can_mst)
+			continue;
 
-			ret = drm_dp_mst_topology_mgr_resume(&intel_dig_port->dp.mst_mgr);
-			if (ret != 0) {
-				intel_dp_check_mst_status(&intel_dig_port->dp);
-			}
-		}
+		ret = drm_dp_mst_topology_mgr_resume(&intel_dig_port->dp.mst_mgr);
+		if (ret)
+			intel_dp_check_mst_status(&intel_dig_port->dp);
 	}
 }
-- 
2.4.6

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH 2/3] drm/i915: Flatten the mst suspend/resume functions a bit

[Maarten Lankhorst]

Hey,

Op 27-08-15 om 17:36 schreef ville.syrjala@linux.intel.com:
> From: Ville Syrjälä <ville.syrjala@linux.intel.com>
>
> I'm not a fan of deeply nested ifs. Just pull most of the conditions
> into a single place to flatten things a bit.
I like flat..

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH 3/3] drm/i915: Flatten intel_dp_check_mst_status()

[ville.syrjala]

From: Ville Syrjälä <ville.syrjala@linux.intel.com>

Restructure intel_dp_check_mst_status() to be more straightforward to
read.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
---
 drivers/gpu/drm/i915/intel_dp.c | 92 +++++++++++++++++++++--------------------
 1 file changed, 47 insertions(+), 45 deletions(-)

diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c
index 6c34784..033ee20 100644
--- a/drivers/gpu/drm/i915/intel_dp.c
+++ b/drivers/gpu/drm/i915/intel_dp.c
@@ -4392,62 +4392,64 @@ static int
 intel_dp_check_mst_status(struct intel_dp *intel_dp)
 {
 	struct drm_device *dev = dp_to_dig_port(intel_dp)->base.base.dev;
+	u8 esi[16];
 	bool bret;
 
-	if (intel_dp->is_mst) {
-		u8 esi[16] = { 0 };
-		int ret = 0;
-		int retry;
+	if (!intel_dp->is_mst)
+		return -EINVAL;
+
+	bret = intel_dp_get_sink_irq_esi(intel_dp, esi);
+
+	if (!bret) {
+		DRM_DEBUG_KMS("failed to get ESI - device may have failed\n");
+		intel_dp->is_mst = false;
+		drm_dp_mst_topology_mgr_set_mst(&intel_dp->mst_mgr, intel_dp->is_mst);
+		/* send a hotplug event */
+		drm_kms_helper_hotplug_event(dev);
+
+		return -EINVAL;
+	}
+
+	for (;;) {
 		bool handled;
-		bret = intel_dp_get_sink_irq_esi(intel_dp, esi);
-go_again:
-		if (bret == true) {
+		int retry;
+		int ret;
 
-			drm_modeset_lock(&dev->mode_config.connection_mutex, NULL);
+		drm_modeset_lock(&dev->mode_config.connection_mutex, NULL);
 
-			/* check link status - esi[10] = 0x200c */
-			if (intel_dp->active_mst_links &&
-			    !drm_dp_channel_eq_ok(&esi[10], intel_dp->lane_count)) {
-				DRM_DEBUG_KMS("channel EQ not ok, retraining\n");
-				intel_dp_start_link_train(intel_dp);
-				intel_dp_complete_link_train(intel_dp);
-				intel_dp_stop_link_train(intel_dp);
-			}
+		/* check link status - esi[10] = 0x200c */
+		if (intel_dp->active_mst_links &&
+		    !drm_dp_channel_eq_ok(&esi[10], intel_dp->lane_count)) {
+			DRM_DEBUG_KMS("channel EQ not ok, retraining\n");
+			intel_dp_start_link_train(intel_dp);
+			intel_dp_complete_link_train(intel_dp);
+			intel_dp_stop_link_train(intel_dp);
+		}
 
-			drm_modeset_unlock(&dev->mode_config.connection_mutex);
 
-			DRM_DEBUG_KMS("got esi %3ph\n", esi);
-			ret = drm_dp_mst_hpd_irq(&intel_dp->mst_mgr, esi, &handled);
-
-			if (handled) {
-				for (retry = 0; retry < 3; retry++) {
-					int wret;
-					wret = drm_dp_dpcd_write(&intel_dp->aux,
-								 DP_SINK_COUNT_ESI+1,
-								 &esi[1], 3);
-					if (wret == 3) {
-						break;
-					}
-				}
+		drm_modeset_unlock(&dev->mode_config.connection_mutex);
 
-				bret = intel_dp_get_sink_irq_esi(intel_dp, esi);
-				if (bret == true) {
-					DRM_DEBUG_KMS("got esi2 %3ph\n", esi);
-					goto go_again;
-				}
-			} else
-				ret = 0;
+		DRM_DEBUG_KMS("got esi %3ph\n", esi);
+		ret = drm_dp_mst_hpd_irq(&intel_dp->mst_mgr, esi, &handled);
 
-			return ret;
-		} else {
-			struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp);
-			DRM_DEBUG_KMS("failed to get ESI - device may have failed\n");
-			intel_dp->is_mst = false;
-			drm_dp_mst_topology_mgr_set_mst(&intel_dp->mst_mgr, intel_dp->is_mst);
-			/* send a hotplug event */
-			drm_kms_helper_hotplug_event(intel_dig_port->base.base.dev);
+		if (!handled)
+			return 0;
+
+		for (retry = 0; retry < 3; retry++) {
+			int wret = drm_dp_dpcd_write(&intel_dp->aux,
+						     DP_SINK_COUNT_ESI+1,
+						     &esi[1], 3);
+			if (wret == 3)
+				break;
 		}
+
+		bret = intel_dp_get_sink_irq_esi(intel_dp, esi);
+		if (!bret)
+			return ret;
+
+		DRM_DEBUG_KMS("got esi2 %3ph\n", esi);
 	}
+
 	return -EINVAL;
 }
 
-- 
2.4.6

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH 3/3] drm/i915: Flatten intel_dp_check_mst_status()

[Maarten Lankhorst]

Op 27-08-15 om 17:36 schreef ville.syrjala@linux.intel.com:
> From: Ville Syrjälä <ville.syrjala@linux.intel.com>
>
> Restructure intel_dp_check_mst_status() to be more straightforward to
> read.
>
> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
> ---
>  drivers/gpu/drm/i915/intel_dp.c | 92 +++++++++++++++++++++--------------------
>  1 file changed, 47 insertions(+), 45 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c
> index 6c34784..033ee20 100644
> --- a/drivers/gpu/drm/i915/intel_dp.c
> +++ b/drivers/gpu/drm/i915/intel_dp.c
> <snip>

> +	for (;;) {
>  		bool handled;
> -		bret = intel_dp_get_sink_irq_esi(intel_dp, esi);
> -go_again:
> -		if (bret == true) {
> +		int retry;
> +		int ret;
>  
> -			drm_modeset_lock(&dev->mode_config.connection_mutex, NULL);
> +		drm_modeset_lock(&dev->mode_config.connection_mutex, NULL);
>  
> -			/* check link status - esi[10] = 0x200c */
> -			if (intel_dp->active_mst_links &&
> -			    !drm_dp_channel_eq_ok(&esi[10], intel_dp->lane_count)) {
> -				DRM_DEBUG_KMS("channel EQ not ok, retraining\n");
> -				intel_dp_start_link_train(intel_dp);
> -				intel_dp_complete_link_train(intel_dp);
> -				intel_dp_stop_link_train(intel_dp);
> -			}
> +		/* check link status - esi[10] = 0x200c */
> +		if (intel_dp->active_mst_links &&
> +		    !drm_dp_channel_eq_ok(&esi[10], intel_dp->lane_count)) {
> +			DRM_DEBUG_KMS("channel EQ not ok, retraining\n");
> +			intel_dp_start_link_train(intel_dp);
> +			intel_dp_complete_link_train(intel_dp);
> +			intel_dp_stop_link_train(intel_dp);
> +		}
>  
> -			drm_modeset_unlock(&dev->mode_config.connection_mutex);
>  
> -			DRM_DEBUG_KMS("got esi %3ph\n", esi);
> -			ret = drm_dp_mst_hpd_irq(&intel_dp->mst_mgr, esi, &handled);
> -
> -			if (handled) {
> -				for (retry = 0; retry < 3; retry++) {
> -					int wret;
> -					wret = drm_dp_dpcd_write(&intel_dp->aux,
> -								 DP_SINK_COUNT_ESI+1,
> -								 &esi[1], 3);
> -					if (wret == 3) {
> -						break;
> -					}
> -				}
> +		drm_modeset_unlock(&dev->mode_config.connection_mutex);
>  
> -				bret = intel_dp_get_sink_irq_esi(intel_dp, esi);
> -				if (bret == true) {
> -					DRM_DEBUG_KMS("got esi2 %3ph\n", esi);
> -					goto go_again;
> -				}
> -			} else
> -				ret = 0;
> +		DRM_DEBUG_KMS("got esi %3ph\n", esi);
> +		ret = drm_dp_mst_hpd_irq(&intel_dp->mst_mgr, esi, &handled);
>  
> -			return ret;
> -		} else {
> -			struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp);
> -			DRM_DEBUG_KMS("failed to get ESI - device may have failed\n");
> -			intel_dp->is_mst = false;
> -			drm_dp_mst_topology_mgr_set_mst(&intel_dp->mst_mgr, intel_dp->is_mst);
> -			/* send a hotplug event */
> -			drm_kms_helper_hotplug_event(intel_dig_port->base.base.dev);
> +		if (!handled)
> +			return 0;
> +
> +		for (retry = 0; retry < 3; retry++) {
> +			int wret = drm_dp_dpcd_write(&intel_dp->aux,
> +						     DP_SINK_COUNT_ESI+1,
> +						     &esi[1], 3);
> +			if (wret == 3)
> +				break;
>  		}
> +
> +		bret = intel_dp_get_sink_irq_esi(intel_dp, esi);
> +		if (!bret)
> +			return ret;
^This seemed like a bug the first time I looked at it with the indent changes.

Original indent with if (handled) {.. } seems better here, but with a continue instead of a goto.
I think a single return ret; would make it more clear when the loop finishes.

The original code sets ret = 0 when handled = false, but looking a  drm_dp_mst_hpd_irq
this is unneeded.


> +
> +		DRM_DEBUG_KMS("got esi2 %3ph\n", esi);
>  	}
> +
>  	return -EINVAL;
Can this -EINVAL be removed? It cannot be reached any more.
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx