Skip to contentSkip to conversation

drm/probe_helper: Don't bother probing when connectors are forced off

PATCH In review Lyude Paul 3 participants 7 messages Sep 17, 2018 – Sep 21, 2018

Subsystems (2)
lore
Share & tools
lore

Lyude Paul

author
#0 lore raw

PATCH 1/1 drm/probe_helper: Don't bother probing when connectors are forced off

Userspace asked them to be forced off, so why would we care about what a probe tells us?

---
 drivers/gpu/drm/drm_probe_helper.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
drivers/gpu/drm/drm_probe_helper.c +6 -1
index a1bb157bfdfa..56d2b5dd1f58 100644
@@ -269,7 +269,9 @@ drm_helper_probe_detect_ctx(struct drm_connector *connector, bool force)
 retry:
 	ret = drm_modeset_lock(&connector->dev->mode_config.connection_mutex, &ctx);
 	if (!ret) {
-		if (funcs->detect_ctx)
+		if (connector->force == DRM_FORCE_OFF)
+			ret = connector_status_disconnected;
+		else if (funcs->detect_ctx)
 			ret = funcs->detect_ctx(connector, &ctx, force);
 		else if (connector->funcs->detect)
 			ret = connector->funcs->detect(connector, force);
@@ -317,6 +319,9 @@ drm_helper_probe_detect(struct drm_connector *connector,
 	if (ret)
 		return ret;
 
+	if (connector->force == DRM_FORCE_OFF)
+		return connector_status_disconnected;
+
 	if (funcs->detect_ctx)
 		return funcs->detect_ctx(connector, ctx, force);
 	else if (connector->funcs->detect)
Signature · 1 line
2.17.1
#1 lore raw
Quoted 2 lines · Lyude Paul

Userspace asked them to be forced off, so why would we care about what a probe tells us?

I believe there should be force checks in the callers already. Or are we missing some?

Quoted 11 lines
---
 drivers/gpu/drm/drm_probe_helper.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
drivers/gpu/drm/drm_probe_helper.c +0 -0
index a1bb157bfdfa..56d2b5dd1f58 100644

drivers/gpu/drm/drm_probe_helper.c drm_helper_probe_detect_ctx(struct drm_connector *connector, bool force) · line 273

Quoted 6 lines
drivers/gpu/drm/drm_probe_helper.c +2 -1
@@ -269,7 +269,9 @@ drm_helper_probe_detect_ctx(struct drm_connector *connector, bool force)
 retry:
 	ret = drm_modeset_lock(&connector->dev->mode_config.connection_mutex, &ctx);
 	if (!ret) {
-		if (funcs->detect_ctx)
+		if (connector->force == DRM_FORCE_OFF)
+			ret = connector_status_disconnected;

connector->force is protected by mode_config.mutex IIRC.

Quoted 21 lines
drivers/gpu/drm/drm_probe_helper.c +4 -0
+		else if (funcs->detect_ctx)
 			ret = funcs->detect_ctx(connector, &ctx, force);
 		else if (connector->funcs->detect)
 			ret = connector->funcs->detect(connector, force);
@@ -317,6 +319,9 @@ drm_helper_probe_detect(struct drm_connector *connector,
 	if (ret)
 		return ret;
 
+	if (connector->force == DRM_FORCE_OFF)
+		return connector_status_disconnected;
+
 	if (funcs->detect_ctx)
 		return funcs->detect_ctx(connector, ctx, force);
 	else if (connector->funcs->detect)
Signature · 6 lines
2.17.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Signature · 2 lines
Ville Syrjälä
Intel
author
#2 lore raw
Quoted 6 lines · Ville Syrjälä
Quoted 2 lines · Lyude Paul

Userspace asked them to be forced off, so why would we care about what a probe tells us?

I believe there should be force checks in the callers already. Or are we missing some?

JFYI, this is to fix "DDC responded but no EDID" errors from nouveau, which presumably come from the fact that having a connector forced off disables reading it's EDID. It's possible we are missing something in nouveau_connector_detect(), but I'm confused as to why we would want to call ->probe() at all in the first place

Quoted 41 lines
Quoted 13 lines
---
 drivers/gpu/drm/drm_probe_helper.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
drivers/gpu/drm/drm_probe_helper.c +0 -0
index a1bb157bfdfa..56d2b5dd1f58 100644

drivers/gpu/drm/drm_probe_helper.c drm_helper_probe_detect_ctx(struct drm_connector · line 275

Quoted 8 lines
drivers/gpu/drm/drm_probe_helper.c +2 -1
@@ -269,7 +269,9 @@ drm_helper_probe_detect_ctx(struct drm_connector
 *connector, bool force)
 retry:
 	ret = drm_modeset_lock(&connector->dev->mode_config.connection_mutex,
 &ctx);
 	if (!ret) {
-		if (funcs->detect_ctx)
+		if (connector->force == DRM_FORCE_OFF)
+			ret = connector_status_disconnected;

connector->force is protected by mode_config.mutex IIRC.

Quoted 17 lines
drivers/gpu/drm/drm_probe_helper.c +1 -0
+		else if (funcs->detect_ctx)
 			ret = funcs->detect_ctx(connector, &ctx, force);
drivers/gpu/drm/drm_probe_helper.c +3 -0
 		else if (connector->funcs->detect)
 			ret = connector->funcs->detect(connector, force);
@@ -317,6 +319,9 @@ drm_helper_probe_detect(struct drm_connector *connector,
 	if (ret)
 		return ret;
 
+	if (connector->force == DRM_FORCE_OFF)
+		return connector_status_disconnected;
+
 	if (funcs->detect_ctx)
 		return funcs->detect_ctx(connector, ctx, force);
 	else if (connector->funcs->detect)
Signature · 1 line
2.17.1
Quoted 6 lines
Quoted 4 lines

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

author
#3 lore raw
In reply to Ville Syrjälä
Quoted 6 lines · Ville Syrjälä
Quoted 2 lines · Lyude Paul

Userspace asked them to be forced off, so why would we care about what a probe tells us?

I believe there should be force checks in the callers already. Or are we missing some?

JFYI, what triggered me to send this patch are these error messages that come from nouveau when a hotplug happens on a port that we've forced off:

[ 1903.918104] nouveau 0000:01:00.0: DRM: DDC responded, but no EDID for DP-2
[ 1903.918123] [drm:drm_helper_hpd_irq_event [drm_kms_helper]] [CONNECTOR:61:DP-2] status updated from disconnected to disconnected

That being said; I'm sure there are probably some checks missing, but I don't really see the purpose in calling the driver's probe functions at all if they're just supposed to return the status we forced.

Quoted 47 lines
Quoted 13 lines
---
 drivers/gpu/drm/drm_probe_helper.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
drivers/gpu/drm/drm_probe_helper.c +0 -0
index a1bb157bfdfa..56d2b5dd1f58 100644

drivers/gpu/drm/drm_probe_helper.c drm_helper_probe_detect_ctx(struct drm_connector · line 275

Quoted 8 lines
drivers/gpu/drm/drm_probe_helper.c +2 -1
@@ -269,7 +269,9 @@ drm_helper_probe_detect_ctx(struct drm_connector
 *connector, bool force)
 retry:
 	ret = drm_modeset_lock(&connector->dev->mode_config.connection_mutex,
 &ctx);
 	if (!ret) {
-		if (funcs->detect_ctx)
+		if (connector->force == DRM_FORCE_OFF)
+			ret = connector_status_disconnected;

connector->force is protected by mode_config.mutex IIRC.

Quoted 21 lines
drivers/gpu/drm/drm_probe_helper.c +1 -0
+		else if (funcs->detect_ctx)
 			ret = funcs->detect_ctx(connector, &ctx, force);
drivers/gpu/drm/drm_probe_helper.c +3 -0
 		else if (connector->funcs->detect)
 			ret = connector->funcs->detect(connector, force);
@@ -317,6 +319,9 @@ drm_helper_probe_detect(struct drm_connector *connector,
 	if (ret)
 		return ret;
 
+	if (connector->force == DRM_FORCE_OFF)
+		return connector_status_disconnected;
+
 	if (funcs->detect_ctx)
 		return funcs->detect_ctx(connector, ctx, force);
 	else if (connector->funcs->detect)
Signature · 6 lines
2.17.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#4 lore raw
Quoted 11 lines · Lyude Paul
Quoted 6 lines · Ville Syrjälä
Quoted 2 lines · Lyude Paul

Userspace asked them to be forced off, so why would we care about what a probe tells us?

I believe there should be force checks in the callers already. Or are we missing some?

JFYI, what triggered me to send this patch are these error messages that come from nouveau when a hotplug happens on a port that we've forced off:

Quoted 6 lines
[ 1903.918104] nouveau 0000:01:00.0: DRM: DDC responded, but no EDID for DP-2
[ 1903.918123] [drm:drm_helper_hpd_irq_event [drm_kms_helper]] [CONNECTOR:61:DP-2] status updated from disconnected to disconnected

That being said; I'm sure there are probably some checks missing, but I don't really see the purpose in calling the driver's probe functions at all if they're just supposed to return the status we forced.

Digging through my cobweb ridden local git repository I found this:

commit bbd17813a7c7d0210c619365707044d0fb29e3f0

Author: Ville Syrjälä <[email protected]>
Date:   Mon Jun 10 15:28:55 2013 +0300

drm: Ignore forced connectors in drm_helper_hpd_irq_event()

drm_helper_hpd_irq_event() calls the connector's .detect() function even for forced connectors. If the returned status doesn't match the forced status, we will send the hotplug event, causing userspace to re-probe all the connectors. Eventually we should end up back where we started when drm_helper_probe_single_connector_modes() overwrites the connector status with the forced status.

We can avoid all that pointles work if we just skip forced connectors in drm_helper_hpd_irq_event().

drivers/gpu/drm/drm_crtc_helper.c +4 -0
index ed1334e27c33..4fc2ad76c107 100644
@@ -1086,6 +1086,10 @@ void drm_helper_hpd_irq_event(struct drm_device *dev)
 	mutex_lock(&dev->mode_config.mutex);
 	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
 
+		/* Ignore forced connectors. */
+		if (connector->force)
+			continue;
+
 		/* Only handle HPD capable connectors. */
 		if (!(connector->polled & DRM_CONNECTOR_POLL_HPD))
 			continue;

I guess I never sent it out.

Signature · 2 lines
Ville Syrjälä
Intel
author
#5 lore raw
Quoted 62 lines · Ville Syrjälä
Quoted 15 lines · Lyude Paul
Quoted 6 lines · Ville Syrjälä
Quoted 2 lines · Lyude Paul

Userspace asked them to be forced off, so why would we care about what a probe tells us?

I believe there should be force checks in the callers already. Or are we missing some?

JFYI, what triggered me to send this patch are these error messages that come
from nouveau when a hotplug happens on a port that we've forced off:

[ 1903.918104] nouveau 0000:01:00.0: DRM: DDC responded, but no EDID for DP-
2
[ 1903.918123] [drm:drm_helper_hpd_irq_event [drm_kms_helper]]
Quoted 7 lines

[CONNECTOR:61:DP-2] status updated from disconnected to disconnected

That being said; I'm sure there are probably some checks missing, but I don't
really see the purpose in calling the driver's probe functions at all if they're
just supposed to return the status we forced.

Digging through my cobweb ridden local git repository I found this:

commit bbd17813a7c7d0210c619365707044d0fb29e3f0

Author: Ville Syrjälä <[email protected]>
Date:   Mon Jun 10 15:28:55 2013 +0300

drm: Ignore forced connectors in drm_helper_hpd_irq_event()

drm_helper_hpd_irq_event() calls the connector's .detect() function even for forced connectors. If the returned status doesn't match the forced status, we will send the hotplug event, causing userspace to re-probe all the connectors. Eventually we should end up back where we started when drm_helper_probe_single_connector_modes() overwrites the connector status with the forced status.

We can avoid all that pointles work if we just skip forced connectors in drm_helper_hpd_irq_event().

drivers/gpu/drm/drm_crtc_helper.c +4 -0
index ed1334e27c33..4fc2ad76c107 100644
@@ -1086,6 +1086,10 @@ void drm_helper_hpd_irq_event(struct drm_device *dev)
 	mutex_lock(&dev->mode_config.mutex);
 	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
 
+		/* Ignore forced connectors. */
+		if (connector->force)
+			continue;
+
 		/* Only handle HPD capable connectors. */
 		if (!(connector->polled & DRM_CONNECTOR_POLL_HPD))
 			continue;
Quoted 1 line

I guess I never sent it out.

Ahhh, to be honest though this patch isn't really enough. drm_helper_hpd_irq_event() isn't going to be used by all drivers (I may remove some usage of it in nouveau in the near future, even) so I still think it would be a better idea to just add this into drm_helper_probe_detect() and drm_helper_probe_detect_ctx() so everything gets covered

Quoted 1 line
#6 lore raw
Quoted 63 lines · Lyude Paul
Quoted 62 lines · Ville Syrjälä
Quoted 15 lines · Lyude Paul
Quoted 6 lines · Ville Syrjälä
Quoted 2 lines · Lyude Paul

Userspace asked them to be forced off, so why would we care about what a probe tells us?

I believe there should be force checks in the callers already. Or are we missing some?

JFYI, what triggered me to send this patch are these error messages that come
from nouveau when a hotplug happens on a port that we've forced off:

[ 1903.918104] nouveau 0000:01:00.0: DRM: DDC responded, but no EDID for DP-
2
[ 1903.918123] [drm:drm_helper_hpd_irq_event [drm_kms_helper]]
Quoted 7 lines

[CONNECTOR:61:DP-2] status updated from disconnected to disconnected

That being said; I'm sure there are probably some checks missing, but I don't
really see the purpose in calling the driver's probe functions at all if they're
just supposed to return the status we forced.

Digging through my cobweb ridden local git repository I found this:

commit bbd17813a7c7d0210c619365707044d0fb29e3f0

Author: Ville Syrjälä <[email protected]>
Date:   Mon Jun 10 15:28:55 2013 +0300

drm: Ignore forced connectors in drm_helper_hpd_irq_event()

drm_helper_hpd_irq_event() calls the connector's .detect() function even for forced connectors. If the returned status doesn't match the forced status, we will send the hotplug event, causing userspace to re-probe all the connectors. Eventually we should end up back where we started when drm_helper_probe_single_connector_modes() overwrites the connector status with the forced status.

We can avoid all that pointles work if we just skip forced connectors in drm_helper_hpd_irq_event().

drivers/gpu/drm/drm_crtc_helper.c +4 -0
index ed1334e27c33..4fc2ad76c107 100644
@@ -1086,6 +1086,10 @@ void drm_helper_hpd_irq_event(struct drm_device *dev)
 	mutex_lock(&dev->mode_config.mutex);
 	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
 
+		/* Ignore forced connectors. */
+		if (connector->force)
+			continue;
+
 		/* Only handle HPD capable connectors. */
 		if (!(connector->polled & DRM_CONNECTOR_POLL_HPD))
 			continue;
Quoted 7 lines
Quoted 1 line

I guess I never sent it out.

Ahhh, to be honest though this patch isn't really enough. drm_helper_hpd_irq_event() isn't going to be used by all drivers (I may remove some usage of it in nouveau in the near future, even) so I still think it would be a better idea to just add this into drm_helper_probe_detect() and drm_helper_probe_detect_ctx() so everything gets covered

Atm all connector->force handling is outside of drm_helper_probe_detect. I guess we could try to push (some) of it into this helper, if that's useful. There seems to be some duplication already. But adding a redundant check like you do here feels a bit funky. Maybe makes more sense in context of the nouveau stuff you're working on?
-Daniel

Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch