drm/probe_helper: Don't bother probing when connectors are forced off
Subsystems (2)
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
@@ -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
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
Signed-off-by Lyude Paul <[email protected]>--- drivers/gpu/drm/drm_probe_helper.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-)drivers/gpu/drm/drm_probe_helper.c +0
-0
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
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
Signed-off-by Lyude Paul <[email protected]>--- drivers/gpu/drm/drm_probe_helper.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-)drivers/gpu/drm/drm_probe_helper.c +0
-0drivers/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->forceis 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
-0else 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
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
Signed-off-by Lyude Paul <[email protected]>--- drivers/gpu/drm/drm_probe_helper.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-)drivers/gpu/drm/drm_probe_helper.c +0
-0drivers/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->forceis 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
-0else 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
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 disconnectedThat 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
@@ -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
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
bbd17813a7c7d0210c619365707044d0fb29e3f0Author: Ville Syrjälä <[email protected]> Date: Mon Jun 10 15:28:55 2013 +0300drm: 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 whendrm_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().Signed-off-by Ville Syrjälä <[email protected]>drivers/gpu/drm/drm_crtc_helper.c +4
-0@@ -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
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
bbd17813a7c7d0210c619365707044d0fb29e3f0Author: Ville Syrjälä <[email protected]> Date: Mon Jun 10 15:28:55 2013 +0300drm: 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 whendrm_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().Signed-off-by Ville Syrjälä <[email protected]>drivers/gpu/drm/drm_crtc_helper.c +4
-0@@ -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 intodrm_helper_probe_detect()anddrm_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