Skip to contentSkip to conversation

drm/client: Send hotplug event after registering a client

PATCH In review Thomas Zimmermann 5 participants 8 messages Jul 10, 2023 – Jul 11, 2023

Subsystems (14)
lore
Share & tools
lore

Thomas Zimmermann

author maintainer
#0 lore raw

PATCH 1/1 drm/client: Send hotplug event after registering a client

Generate a hotplug event after registering a client to allow the client to configure its display. Remove the hotplug calls from the existing clients for fbdev emulation. This change fixes a concurrency bug between registering a client and receiving events from the DRM core. The bug is present in the fbdev emulation of all drivers.

The fbdev emulation currently generates a hotplug event before registering the client to the device. For each new output, the DRM core sends an additional hotplug event to each registered client.

If the DRM core detects first output between sending the artificial hotplug and registering the device, the output's hotplug event gets lost. If this is the first output, the fbdev console display remains dark. This has been observed with amdgpu and fbdev-generic.

Fix this by adding hotplug generation directly to the client's register helper drm_client_register(). Registering the client and receiving events are serialized by struct drm_device.clientlist_mutex. So an output is either configured by the initial hotplug event, or the client has already been registered.

The bug was originally added in commit 6e3f17ee73f7 ("drm/fb-helper:
generic: Call drm_client_add() after setup is done"), in which adding a client and receiving a hotplug event switched order. It was hidden, as most hardware and drivers have at least on static output configured. Other drivers didn't use the internal DRM client or still had struct drm_mode_config_funcs.output_poll_changed set. That callback handled hotplug events as well. After not setting the callback in amdgpu in commit 0e3172bac3f4 ("drm/amdgpu: Don't set struct drm_driver.output_poll_changed"), amdgpu did not show a framebuffer console if output events got lost. The bug got copy-pasted from fbdev-generic into the other fbdev emulation.

Fixes 6e3f17ee73f7 ("drm/fb-helper: generic: Call drm_client_add() after setup is done")
Fixes 8ab59da26bc0 ("drm/fb-helper: Move generic fbdev emulation into separate source file")
Fixes b79fe9abd58b ("drm/fbdev-dma: Implement fbdev emulation for GEM DMA helpers")
Fixes 63c381552f69 ("drm/armada: Implement fbdev emulation as in-kernel client")
Fixes 49953b70e7d3 ("drm/exynos: Implement fbdev emulation as in-kernel client")
Fixes 8f1aaccb04b7 ("drm/gma500: Implement client-based fbdev emulation")
Fixes 940b869c2f2f ("drm/msm: Implement fbdev emulation as in-kernel client")
Fixes 9e69bcd88e45 ("drm/omapdrm: Implement fbdev emulation as in-kernel client")
Fixes e317a69fe891 ("drm/radeon: Implement client-based fbdev emulation")
Fixes 71ec16f45ef8 ("drm/tegra: Implement fbdev emulation as in-kernel client")
Cc Daniel Vetter <[email protected]>
Cc David Airlie <[email protected]>
Cc Noralf Trønnes <[email protected]>
Cc Maarten Lankhorst <[email protected]>
Cc Maxime Ripard <[email protected]>
Cc Javier Martinez Canillas <[email protected]>
Cc Russell King <[email protected]>
Cc Inki Dae <[email protected]>
Cc Seung-Woo Kim <[email protected]>
Cc Kyungmin Park <[email protected]>
Cc Krzysztof Kozlowski <[email protected]>
Cc Patrik Jakobsson <[email protected]>
Cc Rob Clark <[email protected]>
Cc Abhinav Kumar <[email protected]>
Cc Dmitry Baryshkov <[email protected]>
Cc Tomi Valkeinen <[email protected]>
Cc Alex Deucher <[email protected]>
Cc "Christian König" <[email protected]>
Cc "Pan, Xinhui" <[email protected]>
Cc Thierry Reding <[email protected]>
Cc Mikko Perttunen <[email protected]>
Cc <[email protected]> # v5.2+
---
 drivers/gpu/drm/armada/armada_fbdev.c     |  4 ----
 drivers/gpu/drm/drm_client.c              | 21 +++++++++++++++++++++
 drivers/gpu/drm/drm_fbdev_dma.c           |  4 ----
 drivers/gpu/drm/drm_fbdev_generic.c       |  4 ----
 drivers/gpu/drm/exynos/exynos_drm_fbdev.c |  4 ----
 drivers/gpu/drm/gma500/fbdev.c            |  4 ----
 drivers/gpu/drm/msm/msm_fbdev.c           |  4 ----
 drivers/gpu/drm/omapdrm/omap_fbdev.c      |  4 ----
 drivers/gpu/drm/radeon/radeon_fbdev.c     |  4 ----
 drivers/gpu/drm/tegra/fbdev.c             |  4 ----
 10 files changed, 21 insertions(+), 36 deletions(-)
drivers/gpu/drm/armada/armada_fbdev.c +0 -4
index 3943e89cc06c..e40a95e51785 100644
@@ -209,10 +209,6 @@ void armada_fbdev_setup(struct drm_device *dev)
 		goto err_drm_client_init;
 	}
 
-	ret = armada_fbdev_client_hotplug(&fbh->client);
-	if (ret)
-		drm_dbg_kms(dev, "client hotplug ret=%d\n", ret);
-
 	drm_client_register(&fbh->client);
 
 	return;
drivers/gpu/drm/drm_client.c +21 -0
index f6292ba0e6fc..037e36f2049c 100644
@@ -122,13 +122,34 @@ EXPORT_SYMBOL(drm_client_init);
  * drm_client_register() it is no longer permissible to call drm_client_release()
  * directly (outside the unregister callback), instead cleanup will happen
  * automatically on driver unload.
+ *
+ * Registering a client generates a hotplug event that allows the client
+ * to set up its display from pre-existing outputs. The client must have
+ * initialized its state to able to handle the hotplug event successfully.
  */
 void drm_client_register(struct drm_client_dev *client)
 {
 	struct drm_device *dev = client->dev;
+	int ret;
 
 	mutex_lock(&dev->clientlist_mutex);
 	list_add(&client->list, &dev->clientlist);
+
+	if (client->funcs && client->funcs->hotplug) {
+		/*
+		 * Perform an initial hotplug event to pick up the
+		 * display configuration for the client. This step
+		 * has to be performed *after* registering the client
+		 * in the list of clients, or a concurrent hotplug
+		 * event might be lost; leaving the display off.
+		 *
+		 * Hold the clientlist_mutex as for a regular hotplug
+		 * event.
+		 */
+		ret = client->funcs->hotplug(client);
+		if (ret)
+			drm_dbg_kms(dev, "client hotplug ret=%d\n", ret);
+	}
 	mutex_unlock(&dev->clientlist_mutex);
 }
 EXPORT_SYMBOL(drm_client_register);
drivers/gpu/drm/drm_fbdev_dma.c +0 -4
index 8217f1ddc007..f9b1f7cd31b7 100644
@@ -248,10 +248,6 @@ void drm_fbdev_dma_setup(struct drm_device *dev, unsigned int preferred_bpp)
 		goto err_drm_client_init;
 	}
 
-	ret = drm_fbdev_dma_client_hotplug(&fb_helper->client);
-	if (ret)
-		drm_dbg_kms(dev, "client hotplug ret=%d\n", ret);
-
 	drm_client_register(&fb_helper->client);
 
 	return;
drivers/gpu/drm/drm_fbdev_generic.c +0 -4
index 98ae703848a0..b9343fb6cf13 100644
@@ -339,10 +339,6 @@ void drm_fbdev_generic_setup(struct drm_device *dev, unsigned int preferred_bpp)
 		goto err_drm_client_init;
 	}
 
-	ret = drm_fbdev_generic_client_hotplug(&fb_helper->client);
-	if (ret)
-		drm_dbg_kms(dev, "client hotplug ret=%d\n", ret);
-
 	drm_client_register(&fb_helper->client);
 
 	return;
drivers/gpu/drm/exynos/exynos_drm_fbdev.c +0 -4
index fdf65587f1fe..226310c765d8 100644
@@ -215,10 +215,6 @@ void exynos_drm_fbdev_setup(struct drm_device *dev)
 	if (ret)
 		goto err_drm_client_init;
 
-	ret = exynos_drm_fbdev_client_hotplug(&fb_helper->client);
-	if (ret)
-		drm_dbg_kms(dev, "client hotplug ret=%d\n", ret);
-
 	drm_client_register(&fb_helper->client);
 
 	return;
drivers/gpu/drm/gma500/fbdev.c +0 -4
index 955cbe9f05a7..054426549fc6 100644
@@ -328,10 +328,6 @@ void psb_fbdev_setup(struct drm_psb_private *dev_priv)
 		goto err_drm_fb_helper_unprepare;
 	}
 
-	ret = psb_fbdev_client_hotplug(&fb_helper->client);
-	if (ret)
-		drm_dbg_kms(dev, "client hotplug ret=%d\n", ret);
-
 	drm_client_register(&fb_helper->client);
 
 	return;
drivers/gpu/drm/msm/msm_fbdev.c +0 -4
index b933a85420f6..bf1e17dc4550 100644
@@ -246,10 +246,6 @@ void msm_fbdev_setup(struct drm_device *dev)
 		goto err_drm_fb_helper_unprepare;
 	}
 
-	ret = msm_fbdev_client_hotplug(&helper->client);
-	if (ret)
-		drm_dbg_kms(dev, "client hotplug ret=%d\n", ret);
-
 	drm_client_register(&helper->client);
 
 	return;
drivers/gpu/drm/omapdrm/omap_fbdev.c +0 -4
index b7ccce0704a3..fe6639c1cdf3 100644
@@ -318,10 +318,6 @@ void omap_fbdev_setup(struct drm_device *dev)
 
 	INIT_WORK(&fbdev->work, pan_worker);
 
-	ret = omap_fbdev_client_hotplug(&helper->client);
-	if (ret)
-		drm_dbg_kms(dev, "client hotplug ret=%d\n", ret);
-
 	drm_client_register(&helper->client);
 
 	return;
drivers/gpu/drm/radeon/radeon_fbdev.c +0 -4
index ab9c1abbac97..f941e2e7cae6 100644
@@ -383,10 +383,6 @@ void radeon_fbdev_setup(struct radeon_device *rdev)
 		goto err_drm_client_init;
 	}
 
-	ret = radeon_fbdev_client_hotplug(&fb_helper->client);
-	if (ret)
-		drm_dbg_kms(rdev->ddev, "client hotplug ret=%d\n", ret);
-
 	drm_client_register(&fb_helper->client);
 
 	return;
drivers/gpu/drm/tegra/fbdev.c +0 -4
index e74d9be981c7..d042234e1807 100644
@@ -225,10 +225,6 @@ void tegra_fbdev_setup(struct drm_device *dev)
 	if (ret)
 		goto err_drm_client_init;
 
-	ret = tegra_fbdev_client_hotplug(&helper->client);
-	if (ret)
-		drm_dbg_kms(dev, "client hotplug ret=%d\n", ret);
-
 	drm_client_register(&helper->client);
 
 	return;
Signature · 1 line
2.41.0
Reviewed-by on v1 1/1
#1 lore raw

Thomas Zimmermann <[email protected]> writes:

Hello Thomas,

Quoted 33 lines

Generate a hotplug event after registering a client to allow the client to configure its display. Remove the hotplug calls from the existing clients for fbdev emulation. This change fixes a concurrency bug between registering a client and receiving events from the DRM core. The bug is present in the fbdev emulation of all drivers.

The fbdev emulation currently generates a hotplug event before registering the client to the device. For each new output, the DRM core sends an additional hotplug event to each registered client.

If the DRM core detects first output between sending the artificial hotplug and registering the device, the output's hotplug event gets lost. If this is the first output, the fbdev console display remains dark. This has been observed with amdgpu and fbdev-generic.

Fix this by adding hotplug generation directly to the client's register helper drm_client_register(). Registering the client and receiving events are serialized by struct drm_device.clientlist_mutex. So an output is either configured by the initial hotplug event, or the client has already been registered.

The bug was originally added in commit 6e3f17ee73f7 ("drm/fb-helper:
generic: Call drm_client_add() after setup is done"), in which adding a client and receiving a hotplug event switched order. It was hidden, as most hardware and drivers have at least on static output configured. Other drivers didn't use the internal DRM client or still had struct drm_mode_config_funcs.output_poll_changed set. That callback handled hotplug events as well. After not setting the callback in amdgpu in commit 0e3172bac3f4 ("drm/amdgpu: Don't set struct drm_driver.output_poll_changed"), amdgpu did not show a framebuffer console if output events got lost. The bug got copy-pasted from fbdev-generic into the other fbdev emulation.

Quoted 2 lines

Aren't you missing a Fixes: for 0e3172bac3f4 too? Since that's the commit that unmasked the bug for amdgpu, IMO that is the most important to list.

Quoted 37 lines
Fixes 6e3f17ee73f7 ("drm/fb-helper: generic: Call drm_client_add() after setup is done")
Fixes 8ab59da26bc0 ("drm/fb-helper: Move generic fbdev emulation into separate source file")
Fixes b79fe9abd58b ("drm/fbdev-dma: Implement fbdev emulation for GEM DMA helpers")
Fixes 63c381552f69 ("drm/armada: Implement fbdev emulation as in-kernel client")
Fixes 49953b70e7d3 ("drm/exynos: Implement fbdev emulation as in-kernel client")
Fixes 8f1aaccb04b7 ("drm/gma500: Implement client-based fbdev emulation")
Fixes 940b869c2f2f ("drm/msm: Implement fbdev emulation as in-kernel client")
Fixes 9e69bcd88e45 ("drm/omapdrm: Implement fbdev emulation as in-kernel client")
Fixes e317a69fe891 ("drm/radeon: Implement client-based fbdev emulation")
Fixes 71ec16f45ef8 ("drm/tegra: Implement fbdev emulation as in-kernel client")
Cc Daniel Vetter <[email protected]>
Cc David Airlie <[email protected]>
Cc Noralf Trønnes <[email protected]>
Cc Maarten Lankhorst <[email protected]>
Cc Maxime Ripard <[email protected]>
Cc Javier Martinez Canillas <[email protected]>
Cc Russell King <[email protected]>
Cc Inki Dae <[email protected]>
Cc Seung-Woo Kim <[email protected]>
Cc Kyungmin Park <[email protected]>
Cc Krzysztof Kozlowski <[email protected]>
Cc Patrik Jakobsson <[email protected]>
Cc Rob Clark <[email protected]>
Cc Abhinav Kumar <[email protected]>
Cc Dmitry Baryshkov <[email protected]>
Cc Tomi Valkeinen <[email protected]>
Cc Alex Deucher <[email protected]>
Cc "Christian König" <[email protected]>
Cc "Pan, Xinhui" <[email protected]>
Cc Thierry Reding <[email protected]>
Cc Mikko Perttunen <[email protected]>
Quoted 8 lines

While it's true that the but was introduced by commit 6e3f17ee73f7 and that landed in v5.2, I wonder if this patch could even be applied to such olders Linux versions. Probably in practice it would be at most backported to v6.2, which is the release that exposed the bug for the amdgpu driver.

Your explanation makes sense to me and the patch looks good.

Signature · 5 lines
Best regards,

Javier Martinez Canillas
Core Platforms
Red Hat
author maintainer
#2 lore raw

Hi

Am 10.07.23 um 11:52 schrieb Javier Martinez Canillas:

Quoted 37 lines

Thomas Zimmermann <[email protected]> writes:

Hello Thomas,

Quoted 33 lines

Generate a hotplug event after registering a client to allow the client to configure its display. Remove the hotplug calls from the existing clients for fbdev emulation. This change fixes a concurrency bug between registering a client and receiving events from the DRM core. The bug is present in the fbdev emulation of all drivers.

The fbdev emulation currently generates a hotplug event before registering the client to the device. For each new output, the DRM core sends an additional hotplug event to each registered client.

If the DRM core detects first output between sending the artificial hotplug and registering the device, the output's hotplug event gets lost. If this is the first output, the fbdev console display remains dark. This has been observed with amdgpu and fbdev-generic.

Fix this by adding hotplug generation directly to the client's register helper drm_client_register(). Registering the client and receiving events are serialized by struct drm_device.clientlist_mutex. So an output is either configured by the initial hotplug event, or the client has already been registered.

The bug was originally added in commit 6e3f17ee73f7 ("drm/fb-helper:
generic: Call drm_client_add() after setup is done"), in which adding a client and receiving a hotplug event switched order. It was hidden, as most hardware and drivers have at least on static output configured. Other drivers didn't use the internal DRM client or still had struct drm_mode_config_funcs.output_poll_changed set. That callback handled hotplug events as well. After not setting the callback in amdgpu in commit 0e3172bac3f4 ("drm/amdgpu: Don't set struct drm_driver.output_poll_changed"), amdgpu did not show a framebuffer console if output events got lost. The bug got copy-pasted from fbdev-generic into the other fbdev emulation.

Quoted 5 lines
Quoted 2 lines

Aren't you missing a Fixes: for 0e3172bac3f4 too? Since that's the commit that unmasked the bug for amdgpu, IMO that is the most important to list.

Well, OK.

Quoted 47 lines
Quoted 38 lines
Fixes 6e3f17ee73f7 ("drm/fb-helper: generic: Call drm_client_add() after setup is done")
Fixes 8ab59da26bc0 ("drm/fb-helper: Move generic fbdev emulation into separate source file")
Fixes b79fe9abd58b ("drm/fbdev-dma: Implement fbdev emulation for GEM DMA helpers")
Fixes 63c381552f69 ("drm/armada: Implement fbdev emulation as in-kernel client")
Fixes 49953b70e7d3 ("drm/exynos: Implement fbdev emulation as in-kernel client")
Fixes 8f1aaccb04b7 ("drm/gma500: Implement client-based fbdev emulation")
Fixes 940b869c2f2f ("drm/msm: Implement fbdev emulation as in-kernel client")
Fixes 9e69bcd88e45 ("drm/omapdrm: Implement fbdev emulation as in-kernel client")
Fixes e317a69fe891 ("drm/radeon: Implement client-based fbdev emulation")
Fixes 71ec16f45ef8 ("drm/tegra: Implement fbdev emulation as in-kernel client")
Cc Daniel Vetter <[email protected]>
Cc David Airlie <[email protected]>
Cc Noralf Trønnes <[email protected]>
Cc Maarten Lankhorst <[email protected]>
Cc Maxime Ripard <[email protected]>
Cc Javier Martinez Canillas <[email protected]>
Cc Russell King <[email protected]>
Cc Inki Dae <[email protected]>
Cc Seung-Woo Kim <[email protected]>
Cc Kyungmin Park <[email protected]>
Cc Krzysztof Kozlowski <[email protected]>
Cc Patrik Jakobsson <[email protected]>
Cc Rob Clark <[email protected]>
Cc Abhinav Kumar <[email protected]>
Cc Dmitry Baryshkov <[email protected]>
Cc Tomi Valkeinen <[email protected]>
Cc Alex Deucher <[email protected]>
Cc "Christian König" <[email protected]>
Cc "Pan, Xinhui" <[email protected]>
Cc Thierry Reding <[email protected]>
Cc Mikko Perttunen <[email protected]>
Quoted 8 lines
Quoted 4 lines

While it's true that the but was introduced by commit 6e3f17ee73f7 and that landed in v5.2, I wonder if this patch could even be applied to such olders Linux versions. Probably in practice it would be at most backported to v6.2, which is the release that exposed the bug for the amdgpu driver.

No idea. The fix looks simple enough, but a lot has changed in the surrounding code.

Best regards
Thomas

Quoted 5 lines

Your explanation makes sense to me and the patch looks good.

Signature · 6 lines
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Frankenstrasse 146, 90461 Nuernberg, Germany
GF: Ivo Totev, Andrew Myers, Andrew McDonald, Boudien Moerman
HRB 36809 (AG Nuernberg)
#3 lore raw

+regressions

Quoted 95 lines · Thomas Zimmermann

Hi

Am 10.07.23 um 11:52 schrieb Javier Martinez Canillas:

Quoted 37 lines

Thomas Zimmermann <[email protected]> writes:

Hello Thomas,

Quoted 33 lines

Generate a hotplug event after registering a client to allow the client to configure its display. Remove the hotplug calls from the existing clients for fbdev emulation. This change fixes a concurrency bug between registering a client and receiving events from the DRM core. The bug is present in the fbdev emulation of all drivers.

The fbdev emulation currently generates a hotplug event before registering the client to the device. For each new output, the DRM core sends an additional hotplug event to each registered client.

If the DRM core detects first output between sending the artificial hotplug and registering the device, the output's hotplug event gets lost. If this is the first output, the fbdev console display remains dark. This has been observed with amdgpu and fbdev-generic.

Fix this by adding hotplug generation directly to the client's register helper drm_client_register(). Registering the client and receiving events are serialized by struct drm_device.clientlist_mutex. So an output is either configured by the initial hotplug event, or the client has already been registered.

The bug was originally added in commit 6e3f17ee73f7 ("drm/fb-helper:
generic: Call drm_client_add() after setup is done"), in which adding a client and receiving a hotplug event switched order. It was hidden, as most hardware and drivers have at least on static output configured. Other drivers didn't use the internal DRM client or still had struct drm_mode_config_funcs.output_poll_changed set. That callback handled hotplug events as well. After not setting the callback in amdgpu in commit 0e3172bac3f4 ("drm/amdgpu: Don't set struct drm_driver.output_poll_changed"), amdgpu did not show a framebuffer console if output events got lost. The bug got copy-pasted from fbdev-generic into the other fbdev emulation.

Quoted 5 lines
Quoted 2 lines

Aren't you missing a Fixes: for 0e3172bac3f4 too? Since that's the commit that unmasked the bug for amdgpu, IMO that is the most important to list.

Well, OK.

Quoted 47 lines
Quoted 46 lines
Fixes 6e3f17ee73f7 ("drm/fb-helper: generic: Call drm_client_add() after setup is done")
Fixes 8ab59da26bc0 ("drm/fb-helper: Move generic fbdev emulation into separate source file")
Fixes b79fe9abd58b ("drm/fbdev-dma: Implement fbdev emulation for GEM DMA helpers")
Fixes 63c381552f69 ("drm/armada: Implement fbdev emulation as in-kernel client")
Fixes 49953b70e7d3 ("drm/exynos: Implement fbdev emulation as in-kernel client")
Fixes 8f1aaccb04b7 ("drm/gma500: Implement client-based fbdev emulation")
Fixes 940b869c2f2f ("drm/msm: Implement fbdev emulation as in-kernel client")
Fixes 9e69bcd88e45 ("drm/omapdrm: Implement fbdev emulation as in-kernel client")
Fixes e317a69fe891 ("drm/radeon: Implement client-based fbdev emulation")
Fixes 71ec16f45ef8 ("drm/tegra: Implement fbdev emulation as in-kernel client")
Cc Daniel Vetter <[email protected]>
Cc David Airlie <[email protected]>
Cc Noralf Trønnes <[email protected]>
Cc Maarten Lankhorst <[email protected]>
Cc Maxime Ripard <[email protected]>
Cc Javier Martinez Canillas <[email protected]>
Cc Russell King <[email protected]>
Cc Inki Dae <[email protected]>
Cc Seung-Woo Kim <[email protected]>
Cc Kyungmin Park <[email protected]>
Cc Krzysztof Kozlowski <[email protected]>
Cc Patrik Jakobsson <[email protected]>
Cc Rob Clark <[email protected]>
Cc Abhinav Kumar <[email protected]>
Cc Dmitry Baryshkov <[email protected]>
Cc Tomi Valkeinen <[email protected]>
Cc Alex Deucher <[email protected]>
Cc "Christian König" <[email protected]>
Cc "Pan, Xinhui" <[email protected]>
Cc Thierry Reding <[email protected]>
Cc Mikko Perttunen <[email protected]>
Cc <[email protected]> # v5.2+
Quoted 7 lines
Quoted 4 lines

While it's true that the but was introduced by commit 6e3f17ee73f7 and that landed in v5.2, I wonder if this patch could even be applied to such olders Linux versions. Probably in practice it would be at most backported to v6.2, which is the release that exposed the bug for the amdgpu driver.

No idea. The fix looks simple enough, but a lot has changed in the surrounding code.

Actually it needs to go to at least 6.1.y.

Moritz found it in 6.1.35 (not present in 6.1.34).

Quoted 9 lines

Best regards
Thomas

Quoted 5 lines

Your explanation makes sense to me and the patch looks good.

Reviewed-by on v1 1/1
#4 lore raw
In reply to Thomas Zimmermann
Quoted 80 lines · Thomas Zimmermann

Generate a hotplug event after registering a client to allow the client to configure its display. Remove the hotplug calls from the existing clients for fbdev emulation. This change fixes a concurrency bug between registering a client and receiving events from the DRM core. The bug is present in the fbdev emulation of all drivers.

The fbdev emulation currently generates a hotplug event before registering the client to the device. For each new output, the DRM core sends an additional hotplug event to each registered client.

If the DRM core detects first output between sending the artificial hotplug and registering the device, the output's hotplug event gets lost. If this is the first output, the fbdev console display remains dark. This has been observed with amdgpu and fbdev-generic.

Fix this by adding hotplug generation directly to the client's register helper drm_client_register(). Registering the client and receiving events are serialized by struct drm_device.clientlist_mutex. So an output is either configured by the initial hotplug event, or the client has already been registered.

The bug was originally added in commit 6e3f17ee73f7 ("drm/fb-helper:
generic: Call drm_client_add() after setup is done"), in which adding a client and receiving a hotplug event switched order. It was hidden, as most hardware and drivers have at least on static output configured. Other drivers didn't use the internal DRM client or still had struct drm_mode_config_funcs.output_poll_changed set. That callback handled hotplug events as well. After not setting the callback in amdgpu in commit 0e3172bac3f4 ("drm/amdgpu: Don't set struct drm_driver.output_poll_changed"), amdgpu did not show a framebuffer console if output events got lost. The bug got copy-pasted from fbdev-generic into the other fbdev emulation.

Fixes 6e3f17ee73f7 ("drm/fb-helper: generic: Call drm_client_add() after setup is done")
Fixes 8ab59da26bc0 ("drm/fb-helper: Move generic fbdev emulation into separate source file")
Fixes b79fe9abd58b ("drm/fbdev-dma: Implement fbdev emulation for GEM DMA helpers")
Fixes 63c381552f69 ("drm/armada: Implement fbdev emulation as in-kernel client")
Fixes 49953b70e7d3 ("drm/exynos: Implement fbdev emulation as in-kernel client")
Fixes 8f1aaccb04b7 ("drm/gma500: Implement client-based fbdev emulation")
Fixes 940b869c2f2f ("drm/msm: Implement fbdev emulation as in-kernel client")
Fixes 9e69bcd88e45 ("drm/omapdrm: Implement fbdev emulation as in-kernel client")
Fixes e317a69fe891 ("drm/radeon: Implement client-based fbdev emulation")
Fixes 71ec16f45ef8 ("drm/tegra: Implement fbdev emulation as in-kernel client")
Cc Daniel Vetter <[email protected]>
Cc David Airlie <[email protected]>
Cc Noralf Trønnes <[email protected]>
Cc Maarten Lankhorst <[email protected]>
Cc Maxime Ripard <[email protected]>
Cc Javier Martinez Canillas <[email protected]>
Cc Russell King <[email protected]>
Cc Inki Dae <[email protected]>
Cc Seung-Woo Kim <[email protected]>
Cc Kyungmin Park <[email protected]>
Cc Krzysztof Kozlowski <[email protected]>
Cc Patrik Jakobsson <[email protected]>
Cc Rob Clark <[email protected]>
Cc Abhinav Kumar <[email protected]>
Cc Dmitry Baryshkov <[email protected]>
Cc Tomi Valkeinen <[email protected]>
Cc Alex Deucher <[email protected]>
Cc "Christian König" <[email protected]>
Cc "Pan, Xinhui" <[email protected]>
Cc Thierry Reding <[email protected]>
Cc Mikko Perttunen <[email protected]>
Cc <[email protected]> # v5.2+
Quoted 8 lines
---
  drivers/gpu/drm/armada/armada_fbdev.c     |  4 ----
  drivers/gpu/drm/drm_client.c              | 21 +++++++++++++++++++++
  drivers/gpu/drm/drm_fbdev_dma.c           |  4 ----
  drivers/gpu/drm/drm_fbdev_generic.c       |  4 ----
  drivers/gpu/drm/exynos/exynos_drm_fbdev.c |  4 ----
  drivers/gpu/drm/gma500/fbdev.c            |  4 ----
  drivers/gpu/drm/msm/msm_fbdev.c           |  4 ----
Quoted 4 lines
  drivers/gpu/drm/omapdrm/omap_fbdev.c      |  4 ----
  drivers/gpu/drm/radeon/radeon_fbdev.c     |  4 ----
  drivers/gpu/drm/tegra/fbdev.c             |  4 ----
  10 files changed, 21 insertions(+), 36 deletions(-)

BTW: As you have been clearing this area. I see that significant amount of DRM drivers use exactly the same code for msm_fbdev_client_funcs and for the significant part of foo_fbdev_setup(). Do you have any plans for moving that into a library / generic code? If not, I can take a look at crafting the patch.

Signature · 2 lines
With best wishes
Dmitry
author maintainer
#5 lore raw

Hi

Am 10.07.23 um 23:11 schrieb Dmitry Baryshkov:
[...]

Quoted 10 lines
Quoted 8 lines

---
  drivers/gpu/drm/armada/armada_fbdev.c     |  4 ---- drivers/gpu/drm/drm_client.c              | 21 +++++++++++++++++++++ drivers/gpu/drm/drm_fbdev_dma.c           |  4 ---- drivers/gpu/drm/drm_fbdev_generic.c       |  4 ---- drivers/gpu/drm/exynos/exynos_drm_fbdev.c |  4 ---- drivers/gpu/drm/gma500/fbdev.c            |  4 ---- drivers/gpu/drm/msm/msm_fbdev.c           |  4 ----

Thanks.

Quoted 8 lines
Quoted 5 lines

  drivers/gpu/drm/omapdrm/omap_fbdev.c      |  4 ---- drivers/gpu/drm/radeon/radeon_fbdev.c     |  4 ---- drivers/gpu/drm/tegra/fbdev.c             |  4 ----

  10 files changed, 21 insertions(+), 36 deletions(-)

BTW: As you have been clearing this area. I see that significant amount of DRM drivers use exactly the same code for msm_fbdev_client_funcs and for the significant part of foo_fbdev_setup(). Do you have any plans for moving that into a library / generic code? If not, I can take a look at crafting the patch.

You're not the first to ask. :) I've so far not attempted to address this duplication. I've been bitten by premature helperization before, so I wanted to wait a bit longer. A lot of the fbdev and client code is changing quite a bit. After things stabilized, I want to to try to do some more code sharing.

Best regards
Thomas

Signature · 6 lines
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Frankenstrasse 146, 90461 Nuernberg, Germany
GF: Ivo Totev, Andrew Myers, Andrew McDonald, Boudien Moerman
HRB 36809 (AG Nuernberg)
#6 lore raw
Quoted 23 lines · Thomas Zimmermann

Hi

Am 10.07.23 um 23:11 schrieb Dmitry Baryshkov:
[...]

Quoted 10 lines
Quoted 8 lines

---
  drivers/gpu/drm/armada/armada_fbdev.c     |  4 ---- drivers/gpu/drm/drm_client.c              | 21 +++++++++++++++++++++ drivers/gpu/drm/drm_fbdev_dma.c           |  4 ---- drivers/gpu/drm/drm_fbdev_generic.c       |  4 ---- drivers/gpu/drm/exynos/exynos_drm_fbdev.c |  4 ---- drivers/gpu/drm/gma500/fbdev.c            |  4 ---- drivers/gpu/drm/msm/msm_fbdev.c           |  4 ----

Thanks.

Quoted 6 lines
Quoted 5 lines

  drivers/gpu/drm/omapdrm/omap_fbdev.c      |  4 ---- drivers/gpu/drm/radeon/radeon_fbdev.c     |  4 ---- drivers/gpu/drm/tegra/fbdev.c             |  4 ----

  10 files changed, 21 insertions(+), 36 deletions(-)
Quoted 4 lines
Quoted 2 lines

BTW: As you have been clearing this area. I see that significant amount of DRM drivers use exactly the same code for msm_fbdev_client_funcs and for the significant part of foo_fbdev_setup(). Do you have any plans for moving that into a library / generic code? If not, I can take a look at crafting the patch.

You're not the first to ask. :) I've so far not attempted to address this duplication. I've been bitten by premature helperization before, so I wanted to wait a bit longer. A lot of the fbdev and client code is changing quite a bit. After things stabilized, I want to to try to do some more code sharing.

Ack, thank you for sharing this.

Signature · 2 lines
With best wishes
Dmitry
Reviewed-by on v1 1/1
#7 lore raw
In reply to Thomas Zimmermann
Quoted 80 lines · Thomas Zimmermann

Generate a hotplug event after registering a client to allow the client to configure its display. Remove the hotplug calls from the existing clients for fbdev emulation. This change fixes a concurrency bug between registering a client and receiving events from the DRM core. The bug is present in the fbdev emulation of all drivers.

The fbdev emulation currently generates a hotplug event before registering the client to the device. For each new output, the DRM core sends an additional hotplug event to each registered client.

If the DRM core detects first output between sending the artificial hotplug and registering the device, the output's hotplug event gets lost. If this is the first output, the fbdev console display remains dark. This has been observed with amdgpu and fbdev-generic.

Fix this by adding hotplug generation directly to the client's register helper drm_client_register(). Registering the client and receiving events are serialized by struct drm_device.clientlist_mutex. So an output is either configured by the initial hotplug event, or the client has already been registered.

The bug was originally added in commit 6e3f17ee73f7 ("drm/fb-helper:
generic: Call drm_client_add() after setup is done"), in which adding a client and receiving a hotplug event switched order. It was hidden, as most hardware and drivers have at least on static output configured. Other drivers didn't use the internal DRM client or still had struct drm_mode_config_funcs.output_poll_changed set. That callback handled hotplug events as well. After not setting the callback in amdgpu in commit 0e3172bac3f4 ("drm/amdgpu: Don't set struct drm_driver.output_poll_changed"), amdgpu did not show a framebuffer console if output events got lost. The bug got copy-pasted from fbdev-generic into the other fbdev emulation.

Fixes 6e3f17ee73f7 ("drm/fb-helper: generic: Call drm_client_add() after setup is done")
Fixes 8ab59da26bc0 ("drm/fb-helper: Move generic fbdev emulation into separate source file")
Fixes b79fe9abd58b ("drm/fbdev-dma: Implement fbdev emulation for GEM DMA helpers")
Fixes 63c381552f69 ("drm/armada: Implement fbdev emulation as in-kernel client")
Fixes 49953b70e7d3 ("drm/exynos: Implement fbdev emulation as in-kernel client")
Fixes 8f1aaccb04b7 ("drm/gma500: Implement client-based fbdev emulation")
Fixes 940b869c2f2f ("drm/msm: Implement fbdev emulation as in-kernel client")
Fixes 9e69bcd88e45 ("drm/omapdrm: Implement fbdev emulation as in-kernel client")
Fixes e317a69fe891 ("drm/radeon: Implement client-based fbdev emulation")
Fixes 71ec16f45ef8 ("drm/tegra: Implement fbdev emulation as in-kernel client")
Cc Daniel Vetter <[email protected]>
Cc David Airlie <[email protected]>
Cc Noralf Trønnes <[email protected]>
Cc Maarten Lankhorst <[email protected]>
Cc Maxime Ripard <[email protected]>
Cc Javier Martinez Canillas <[email protected]>
Cc Russell King <[email protected]>
Cc Inki Dae <[email protected]>
Cc Seung-Woo Kim <[email protected]>
Cc Kyungmin Park <[email protected]>
Cc Krzysztof Kozlowski <[email protected]>
Cc Patrik Jakobsson <[email protected]>
Cc Rob Clark <[email protected]>
Cc Abhinav Kumar <[email protected]>
Cc Dmitry Baryshkov <[email protected]>
Cc Tomi Valkeinen <[email protected]>
Cc Alex Deucher <[email protected]>
Cc "Christian König" <[email protected]>
Cc "Pan, Xinhui" <[email protected]>
Cc Thierry Reding <[email protected]>
Cc Mikko Perttunen <[email protected]>
Cc <[email protected]> # v5.2+
Quoted 13 lines
---
  drivers/gpu/drm/armada/armada_fbdev.c     |  4 ----
  drivers/gpu/drm/drm_client.c              | 21 +++++++++++++++++++++
  drivers/gpu/drm/drm_fbdev_dma.c           |  4 ----
  drivers/gpu/drm/drm_fbdev_generic.c       |  4 ----
  drivers/gpu/drm/exynos/exynos_drm_fbdev.c |  4 ----
  drivers/gpu/drm/gma500/fbdev.c            |  4 ----
  drivers/gpu/drm/msm/msm_fbdev.c           |  4 ----
  drivers/gpu/drm/omapdrm/omap_fbdev.c      |  4 ----
  drivers/gpu/drm/radeon/radeon_fbdev.c     |  4 ----
  drivers/gpu/drm/tegra/fbdev.c             |  4 ----
  10 files changed, 21 insertions(+), 36 deletions(-)