Skip to contentSkip to conversation

drm/i915: Fix races on fbdev

PATCH RESEND Lukas Wunner 3 participants 3 messages Mar 3, 2016 – Mar 4, 2016

lore
Share & tools
lore
#0 lore raw

Hi Damien, Hi Daniel,

I've submitted the patch below for the third time now in an attempt to get CI to test it, again to no avail. This time I didn't set the In-Reply-To header and only submitted it as a single patch instead of as a series because I expected this might confuse CI.

Nevertheless CI misclassified it under "Series" instead of "Patches", reports that the series consists of 0 patches and doesn't test it:
https://patchwork.freedesktop.org/series/4068/

Earlier attempts:
https://patchwork.freedesktop.org/series/3453/
https://patchwork.freedesktop.org/series/3126/

Sorry, I give up.

Best regards,

Lukas

Quoted 71 lines · Lukas Wunner

The ->lastclose callback invokes intel_fbdev_restore_mode() and has been witnessed to run before intel_fbdev_initial_config_async() has finished.

We might likewise receive hotplug events or be suspended before we've had a chance to fully set up the fbdev.

Fix by waiting for the asynchronous thread to finish.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=93580

---
 drivers/gpu/drm/i915/i915_dma.c    | 8 +++-----
 drivers/gpu/drm/i915/intel_fbdev.c | 4 ++++
 2 files changed, 7 insertions(+), 5 deletions(-)
drivers/gpu/drm/i915/i915_dma.c +3 -5
index 4aa3db6..9d76dfb 100644
@@ -430,11 +430,9 @@ static int i915_load_modeset_init(struct drm_device *dev)
 	 * Some ports require correctly set-up hpd registers for detection to
 	 * work properly (leading to ghost connected connector status), e.g. VGA
 	 * on gm45.  Hence we can only set up the initial fbdev config after hpd
-	 * irqs are fully enabled. Now we should scan for the initial config
-	 * only once hotplug handling is enabled, but due to screwed-up locking
-	 * around kms/fbdev init we can't protect the fdbev initial config
-	 * scanning against hotplug events. Hence do this first and ignore the
-	 * tiny window where we will loose hotplug notifactions.
+	 * irqs are fully enabled. We protect the fbdev initial config scanning
+	 * against hotplug events by waiting in intel_fbdev_output_poll_changed
+	 * until the asynchronous thread has finished.
 	 */
 	intel_fbdev_initial_config_async(dev);
 
drivers/gpu/drm/i915/intel_fbdev.c +4 -0
index ae9cf6f..936c3d7 100644
@@ -754,6 +754,7 @@ void intel_fbdev_set_suspend(struct drm_device *dev, int state, bool synchronous
 	struct intel_fbdev *ifbdev = dev_priv->fbdev;
 	struct fb_info *info;
 
+	async_synchronize_full();
 	if (!ifbdev)
 		return;
 
@@ -800,6 +801,8 @@ void intel_fbdev_set_suspend(struct drm_device *dev, int state, bool synchronous
 void intel_fbdev_output_poll_changed(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
+
+	async_synchronize_full();
 	if (dev_priv->fbdev)
 		drm_fb_helper_hotplug_event(&dev_priv->fbdev->helper);
 }
@@ -811,6 +814,7 @@ void intel_fbdev_restore_mode(struct drm_device *dev)
 	struct intel_fbdev *ifbdev = dev_priv->fbdev;
 	struct drm_fb_helper *fb_helper;
 
+	async_synchronize_full();
 	if (!ifbdev)
 		return;
 
Signature · 1 line
1.8.5.2 (Apple Git-48)

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#1 lore raw
Quoted 11 lines · Lukas Wunner

Hi Damien, Hi Daniel,

I've submitted the patch below for the third time now in an attempt to get CI to test it, again to no avail. This time I didn't set the In-Reply-To header and only submitted it as a single patch instead of as a series because I expected this might confuse CI.

Nevertheless CI misclassified it under "Series" instead of "Patches", reports that the series consists of 0 patches and doesn't test it:
https://patchwork.freedesktop.org/series/4068/

Quoted 5 lines

Earlier attempts:
https://patchwork.freedesktop.org/series/3453/
https://patchwork.freedesktop.org/series/3126/

Sorry, I give up.

git send-email is all it takes.

BR,
Jani.

Quoted 81 lines

Best regards,

Lukas

Quoted 71 lines · Lukas Wunner

The ->lastclose callback invokes intel_fbdev_restore_mode() and has been witnessed to run before intel_fbdev_initial_config_async() has finished.

We might likewise receive hotplug events or be suspended before we've had a chance to fully set up the fbdev.

Fix by waiting for the asynchronous thread to finish.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=93580

---
 drivers/gpu/drm/i915/i915_dma.c    | 8 +++-----
 drivers/gpu/drm/i915/intel_fbdev.c | 4 ++++
 2 files changed, 7 insertions(+), 5 deletions(-)
drivers/gpu/drm/i915/i915_dma.c +3 -5
index 4aa3db6..9d76dfb 100644
@@ -430,11 +430,9 @@ static int i915_load_modeset_init(struct drm_device *dev)
 	 * Some ports require correctly set-up hpd registers for detection to
 	 * work properly (leading to ghost connected connector status), e.g. VGA
 	 * on gm45.  Hence we can only set up the initial fbdev config after hpd
-	 * irqs are fully enabled. Now we should scan for the initial config
-	 * only once hotplug handling is enabled, but due to screwed-up locking
-	 * around kms/fbdev init we can't protect the fdbev initial config
-	 * scanning against hotplug events. Hence do this first and ignore the
-	 * tiny window where we will loose hotplug notifactions.
+	 * irqs are fully enabled. We protect the fbdev initial config scanning
+	 * against hotplug events by waiting in intel_fbdev_output_poll_changed
+	 * until the asynchronous thread has finished.
 	 */
 	intel_fbdev_initial_config_async(dev);
 
drivers/gpu/drm/i915/intel_fbdev.c +4 -0
index ae9cf6f..936c3d7 100644
@@ -754,6 +754,7 @@ void intel_fbdev_set_suspend(struct drm_device *dev, int state, bool synchronous
 	struct intel_fbdev *ifbdev = dev_priv->fbdev;
 	struct fb_info *info;
 
+	async_synchronize_full();
 	if (!ifbdev)
 		return;
 
@@ -800,6 +801,8 @@ void intel_fbdev_set_suspend(struct drm_device *dev, int state, bool synchronous
 void intel_fbdev_output_poll_changed(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
+
+	async_synchronize_full();
 	if (dev_priv->fbdev)
 		drm_fb_helper_hotplug_event(&dev_priv->fbdev->helper);
 }
@@ -811,6 +814,7 @@ void intel_fbdev_restore_mode(struct drm_device *dev)
 	struct intel_fbdev *ifbdev = dev_priv->fbdev;
 	struct drm_fb_helper *fb_helper;
 
+	async_synchronize_full();
 	if (!ifbdev)
 		return;
 
Signature · 1 line
1.8.5.2 (Apple Git-48)

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Signature · 5 lines
Jani Nikula, Intel Open Source Technology Center
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#2 lore raw
In reply to Lukas Wunner

Hi Lukas,

I'm sorry I haven't reacted sooner. I've enabled the option to only consider git send-email patches on intel-gfx as we were having a lot of false positives (ie patchwork considering emails were people were making suggestions with diffs as patches to test).

I'm not sure how you send your emails, but I can tell why that it didn't pass the "is this a git send-email patch?" question. We're looking at two things:

- The X-Mailer header
- The Message-Id (git send-email has a special way to craft its message IDs)

I've replayed your email by hand with an extra X-Mailer header and patchwork picked it up (and hopefully we should have a run of our basic acceptance tests done on it):

https://patchwork.freedesktop.org/series/4068/

You can keep whatever you're doing now and trick patchwork to take your patches by added this extra header to your emails:

X-Mailer: git-send-email 2.4.3

Or use git send-email.

HTH,

Signature · 1 line
Damien
Quoted 86 lines · Lukas Wunner

Hi Damien, Hi Daniel,

I've submitted the patch below for the third time now in an attempt to get CI to test it, again to no avail. This time I didn't set the In-Reply-To header and only submitted it as a single patch instead of as a series because I expected this might confuse CI.

Nevertheless CI misclassified it under "Series" instead of "Patches", reports that the series consists of 0 patches and doesn't test it:
https://patchwork.freedesktop.org/series/4068/

Earlier attempts:
https://patchwork.freedesktop.org/series/3453/
https://patchwork.freedesktop.org/series/3126/

Sorry, I give up.

Best regards,

Lukas

Quoted 64 lines · Lukas Wunner

The ->lastclose callback invokes intel_fbdev_restore_mode() and has been witnessed to run before intel_fbdev_initial_config_async() has finished.

We might likewise receive hotplug events or be suspended before we've had a chance to fully set up the fbdev.

Fix by waiting for the asynchronous thread to finish.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=93580

---
 drivers/gpu/drm/i915/i915_dma.c    | 8 +++-----
 drivers/gpu/drm/i915/intel_fbdev.c | 4 ++++
 2 files changed, 7 insertions(+), 5 deletions(-)
drivers/gpu/drm/i915/i915_dma.c +3 -5
index 4aa3db6..9d76dfb 100644
@@ -430,11 +430,9 @@ static int i915_load_modeset_init(struct drm_device *dev)
 	 * Some ports require correctly set-up hpd registers for detection to
 	 * work properly (leading to ghost connected connector status), e.g. VGA
 	 * on gm45.  Hence we can only set up the initial fbdev config after hpd
-	 * irqs are fully enabled. Now we should scan for the initial config
-	 * only once hotplug handling is enabled, but due to screwed-up locking
-	 * around kms/fbdev init we can't protect the fdbev initial config
-	 * scanning against hotplug events. Hence do this first and ignore the
-	 * tiny window where we will loose hotplug notifactions.
+	 * irqs are fully enabled. We protect the fbdev initial config scanning
+	 * against hotplug events by waiting in intel_fbdev_output_poll_changed
+	 * until the asynchronous thread has finished.
 	 */
 	intel_fbdev_initial_config_async(dev);
 
drivers/gpu/drm/i915/intel_fbdev.c +3 -0
index ae9cf6f..936c3d7 100644
@@ -754,6 +754,7 @@ void intel_fbdev_set_suspend(struct drm_device *dev, int state, bool synchronous
 	struct intel_fbdev *ifbdev = dev_priv->fbdev;
 	struct fb_info *info;
 
+	async_synchronize_full();
 	if (!ifbdev)
 		return;
 
@@ -800,6 +801,8 @@ void intel_fbdev_set_suspend(struct drm_device *dev, int state, bool synchronous
 void intel_fbdev_output_poll_changed(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
+
+	async_synchronize_full();
 	if (dev_priv->fbdev)
 		drm_fb_helper_hotplug_event(&dev_priv->fbdev->helper);
 }
@@ -811,6 +814,7 @@ void intel_fbdev_restore_mode(struct drm_device *dev)
 	struct intel_fbdev *ifbdev = dev_priv->fbdev;
 	struct drm_fb_helper *fb_helper;
 
Quoted 7 lines
Quoted 7 lines
drivers/gpu/drm/i915/intel_fbdev.c +1 -0
+	async_synchronize_full();
 	if (!ifbdev)
 		return;
 
Signature · 1 line
1.8.5.2 (Apple Git-48)

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx