Skip to contentSkip to conversation

drm/msm: default separate_gpu_kms to auto selection

PATCH Superseded Mahadevan P 3 participants 4 messages Mar 17 – Mar 21

Subsystems (4)

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 drm/msm: default separate_gpu_kms to auto selection

On platforms with multiple display subsystems, such as SA8775P, the GPU binds to the first display subsystem that probes. This implicit binding prevents subsequent display subsystems from probing successfully, breaking multi-display support.

Use the tristate separate_gpu_kms module parameter with the default value set to auto (-1). In auto mode, the driver selects the binding behavior based on the number of GPUs and display subsystems. This allows display subsystems to probe independently when required, while preserving the existing single-card behavior on simpler systems.

The separate_gpu_kms module parameter has the following semantics:

-1 (auto, default):
Select the binding mode based on hardware topology. If exactly one GPU and one display subsystem are present, bind them together to form a single DRM device. Otherwise, expose the GPU and display subsystems as separate DRM devices.

0:
Always bind the GPU and display together to form a single DRM device.

1:
Always expose the GPU and display subsystems as separate DRM devices.

This ensures correct probing on multi-display platforms without affecting single-display, single-GPU systems.

---
Changes in v2:
- EDITME: describe what is new in this series revision.
- EDITME: use bulletpoints and terse descriptions.
- Link to v1: https://lore.kernel.org/r/[email protected]
---
Depends on:
https://lore.kernel.org/lkml/[email protected]/

When separate_gpu_kms is enabled, the GPU and display drivers are probed independently. In this configuration, the Adreno GPU driver may no longer be loaded implicitly via the display subsystem.

The referenced patch adds a MODULE_DEVICE_TABLE() entry for the Adreno GPU device, ensuring proper module autoloading based on device tree matching. This is required to guarantee that the GPU driver is loaded correctly when GPU and display probing are decoupled.

Changes in v2:

- Drop dependency on Lemans dual-DPU device tree changes as this patch
works independently (Dmitry)
- Switch separate_gpu_kms to tristate and default to auto mode (Rob)
- Rename msm_gpu_no_components() to msm_separate_gpu_kms_components() for clarity
- Link to v1: https://lore.kernel.org/r/[email protected]

---
 drivers/gpu/drm/msm/adreno/adreno_device.c |  2 +-
 drivers/gpu/drm/msm/msm_drv.c              | 52 ++++++++++++++++++++++++++----
 drivers/gpu/drm/msm/msm_drv.h              |  4 ++-
 drivers/gpu/drm/msm/msm_mdss.c             | 15 +++++++++
 4 files changed, 64 insertions(+), 9 deletions(-)
drivers/gpu/drm/msm/adreno/adreno_device.c +1 -1
index 4edfe80c5be7..e40648c05797 100644
@@ -272,7 +272,7 @@ static const struct component_ops a3xx_ops = {
 static int adreno_probe(struct platform_device *pdev)
 {
 	if (of_device_is_compatible(pdev->dev.of_node, "amd,imageon") ||
-	    msm_gpu_no_components())
+	    msm_separate_gpu_kms_components())
 		return msm_gpu_probe(pdev, &a3xx_ops);
 
 	return component_add(&pdev->dev, &a3xx_ops);
drivers/gpu/drm/msm/msm_drv.c +45 -7

75 lines not on this page · raw message

drivers/gpu/drm/msm/msm_drv.h +3 -1
index 6d847d593f1a..64a5ad35f7a2 100644
@@ -555,6 +555,8 @@ void msm_kms_shutdown(struct platform_device *pdev);
 
 bool msm_disp_drv_should_bind(struct device *dev, bool dpu_driver);
 
-bool msm_gpu_no_components(void);
+bool msm_separate_gpu_kms_components(void);
+
+int msm_mdss_count_masters(void);
 
 #endif /* __MSM_DRV_H__ */
drivers/gpu/drm/msm/msm_mdss.c +15 -0
index 9047e8d9ee89..00e3ac7dab3a 100644
@@ -73,6 +73,21 @@ static int msm_mdss_parse_data_bus_icc_path(struct device *dev,
 	return 0;
 }
 
+static const struct of_device_id mdss_dt_match[];
+
+int msm_mdss_count_masters(void)
+{
+	struct device_node *np;
+	int count = 0;
+
+	for_each_matching_node(np, mdss_dt_match) {
+		if (of_device_is_available(np))
+			count++;
+	}
+
+	return count;
+}
+
 static void msm_mdss_irq(struct irq_desc *desc)
 {
 	struct msm_mdss *msm_mdss = irq_desc_get_handler_data(desc);

---
base-commit: b84a0ebe421ca56995ff78b66307667b62b3a900
change-id: 20260316-separate_gpu_kms-04d2cf4d91e2

Best regards,

Signature · 1 line
Mahadevan P <[email protected]>
#1 lore raw
Quoted 30 lines · Mahadevan P

On platforms with multiple display subsystems, such as SA8775P, the GPU binds to the first display subsystem that probes. This implicit binding prevents subsequent display subsystems from probing successfully, breaking multi-display support.

Use the tristate separate_gpu_kms module parameter with the default value set to auto (-1). In auto mode, the driver selects the binding behavior based on the number of GPUs and display subsystems. This allows display subsystems to probe independently when required, while preserving the existing single-card behavior on simpler systems.

The separate_gpu_kms module parameter has the following semantics:

-1 (auto, default):
Select the binding mode based on hardware topology. If exactly one GPU and one display subsystem are present, bind them together to form a single DRM device. Otherwise, expose the GPU and display subsystems as separate DRM devices.

0:
Always bind the GPU and display together to form a single DRM device.

1:
Always expose the GPU and display subsystems as separate DRM devices.

This ensures correct probing on multi-display platforms without affecting single-display, single-GPU systems.

Quoted 6 lines

---
Changes in v2:
- EDITME: describe what is new in this series revision.
- EDITME: use bulletpoints and terse descriptions.
- Link to v1: https://lore.kernel.org/r/[email protected]

From coverletter this got added from b4 will make sure will be posted properly with subsequent patches.

Quoted 169 lines

---
Depends on:
https://lore.kernel.org/lkml/[email protected]/

When separate_gpu_kms is enabled, the GPU and display drivers are probed independently. In this configuration, the Adreno GPU driver may no longer be loaded implicitly via the display subsystem.

The referenced patch adds a MODULE_DEVICE_TABLE() entry for the Adreno GPU device, ensuring proper module autoloading based on device tree matching. This is required to guarantee that the GPU driver is loaded correctly when GPU and display probing are decoupled.

Changes in v2:

- Drop dependency on Lemans dual-DPU device tree changes as this patch
works independently (Dmitry)
- Switch separate_gpu_kms to tristate and default to auto mode (Rob)
- Rename msm_gpu_no_components() to msm_separate_gpu_kms_components() for clarity
- Link to v1: https://lore.kernel.org/r/[email protected]

---
  drivers/gpu/drm/msm/adreno/adreno_device.c |  2 +-
  drivers/gpu/drm/msm/msm_drv.c              | 52 ++++++++++++++++++++++++++----
  drivers/gpu/drm/msm/msm_drv.h              |  4 ++-
  drivers/gpu/drm/msm/msm_mdss.c             | 15 +++++++++
  4 files changed, 64 insertions(+), 9 deletions(-)
drivers/gpu/drm/msm/adreno/adreno_device.c +1 -1
index 4edfe80c5be7..e40648c05797 100644
@@ -272,7 +272,7 @@ static const struct component_ops a3xx_ops = {
  static int adreno_probe(struct platform_device *pdev)
  {
  	if (of_device_is_compatible(pdev->dev.of_node, "amd,imageon") ||
-	    msm_gpu_no_components())
+	    msm_separate_gpu_kms_components())
  		return msm_gpu_probe(pdev, &a3xx_ops);
 
  	return component_add(&pdev->dev, &a3xx_ops);
drivers/gpu/drm/msm/msm_drv.c +45 -7
index e5ab1e28851d..575d1aea7927 100644
@@ -54,16 +54,54 @@ static bool modeset = true;
  MODULE_PARM_DESC(modeset, "Use kernel modesetting [KMS] (1=on (default), 0=disable)");
  module_param(modeset, bool, 0600);
 
-static bool separate_gpu_kms;
-MODULE_PARM_DESC(separate_gpu_drm, "Use separate DRM device for the GPU (0=single DRM device for both GPU and display (default), 1=two DRM devices)");
-module_param(separate_gpu_kms, bool, 0400);
+/*
+ * separate_gpu_kms (tristate):
+ *   -1 (default): decide automatically based on hardware topology. Split devices
+ *                 if there is more than one GPU or more than one display master.
+ *    0: force single DRM device (bind display + GPU)
+ *    1: force separate DRM devices
+ */
+static int separate_gpu_kms = -1;
+MODULE_PARM_DESC(separate_gpu_kms,
+		 "Use separate DRM device for the GPU (-1=auto (default), 0=single DRM device, 1=separate DRM devices)");
+module_param(separate_gpu_kms, int, 0400);
 
  DECLARE_FAULT_ATTR(fail_gem_alloc);
  DECLARE_FAULT_ATTR(fail_gem_iova);
 
-bool msm_gpu_no_components(void)
+static const struct of_device_id msm_gpu_match[];
+static int msm_count_gpus(void)
+{
+	struct device_node *np;
+	int count = 0;
+
+	for_each_matching_node(np, msm_gpu_match) {
+		if (of_device_is_available(np) && adreno_has_gpu(np))
+			count++;
+	}
+
+	return count;
+}
+
+static bool msm_separate_gpu_kms_auto(void)
+{
+	int gpus = msm_count_gpus();
+	int mdss = msm_mdss_count_masters();
+
+	if (gpus <= 0 || mdss <= 0)
+		return false;
+
+	/* If exactly one GPU and one display subsystem single card */
+	return (gpus > 1) || (mdss > 1);
+}
+
+bool msm_separate_gpu_kms_components(void)
  {
-	return separate_gpu_kms;
+	if (separate_gpu_kms == 1)
+		return true;
+	if (separate_gpu_kms == 0)
+		return false;
+	return msm_separate_gpu_kms_auto();
  }
 
  static int msm_drm_uninit(struct device *dev, const struct component_ops *gpu_ops)
@@ -1030,7 +1068,7 @@ static int add_gpu_components(struct device *dev,
  static int msm_drm_bind(struct device *dev)
  {
  	return msm_drm_init(dev,
-			    msm_gpu_no_components() ?
+			    msm_separate_gpu_kms_components() ?
  				    &msm_kms_driver :
  				    &msm_driver,
  			    NULL);
@@ -1069,7 +1107,7 @@ int msm_drv_probe(struct device *master_dev,
  			return ret;
  	}
 
-	if (!msm_gpu_no_components()) {
+	if (!msm_separate_gpu_kms_components()) {
  		ret = add_gpu_components(master_dev, &match);
  		if (ret)
  			return ret;
drivers/gpu/drm/msm/msm_drv.h +3 -1
index 6d847d593f1a..64a5ad35f7a2 100644
@@ -555,6 +555,8 @@ void msm_kms_shutdown(struct platform_device *pdev);
 
  bool msm_disp_drv_should_bind(struct device *dev, bool dpu_driver);
 
-bool msm_gpu_no_components(void);
+bool msm_separate_gpu_kms_components(void);
+
+int msm_mdss_count_masters(void);
 
  #endif /* __MSM_DRV_H__ */
drivers/gpu/drm/msm/msm_mdss.c +15 -1
index 9047e8d9ee89..00e3ac7dab3a 100644
@@ -73,6 +73,21 @@ static int msm_mdss_parse_data_bus_icc_path(struct device *dev,
  	return 0;
  }
 
+static const struct of_device_id mdss_dt_match[];
+
+int msm_mdss_count_masters(void)
+{
+	struct device_node *np;
+	int count = 0;
+
+	for_each_matching_node(np, mdss_dt_match) {
+		if (of_device_is_available(np))
+			count++;
+	}
+
+	return count;
+}
+
  static void msm_mdss_irq(struct irq_desc *desc)
  {
  	struct msm_mdss *msm_mdss = irq_desc_get_handler_data(desc);
 
---

base-commit: b84a0ebe421ca56995ff78b66307667b62b3a900
change-id: 20260316-separate_gpu_kms-04d2cf4d91e2

Best regards,

#2 lore raw
Quoted 40 lines · Mahadevan P
Quoted 30 lines · Mahadevan P

On platforms with multiple display subsystems, such as SA8775P, the GPU binds to the first display subsystem that probes. This implicit binding prevents subsequent display subsystems from probing successfully, breaking multi-display support.

Use the tristate separate_gpu_kms module parameter with the default value set to auto (-1). In auto mode, the driver selects the binding behavior based on the number of GPUs and display subsystems. This allows display subsystems to probe independently when required, while preserving the existing single-card behavior on simpler systems.

The separate_gpu_kms module parameter has the following semantics:

-1 (auto, default):
Select the binding mode based on hardware topology. If exactly one GPU and one display subsystem are present, bind them together to form a single DRM device. Otherwise, expose the GPU and display subsystems as separate DRM devices.

0:
Always bind the GPU and display together to form a single DRM device.

1:
Always expose the GPU and display subsystems as separate DRM devices.

This ensures correct probing on multi-display platforms without affecting single-display, single-GPU systems.

Quoted 6 lines

---
Changes in v2:
- EDITME: describe what is new in this series revision.
- EDITME: use bulletpoints and terse descriptions.
- Link to v1: https://lore.kernel.org/r/[email protected]

Quoted 2 lines

From coverletter this got added from b4 will make sure will be posted properly with subsequent patches.

You know, you can just edit it in the b4's coverletter and drop your manual changelog...

Quoted 66 lines
Quoted 66 lines

---
Depends on:
https://lore.kernel.org/lkml/[email protected]/

When separate_gpu_kms is enabled, the GPU and display drivers are probed independently. In this configuration, the Adreno GPU driver may no longer be loaded implicitly via the display subsystem.

The referenced patch adds a MODULE_DEVICE_TABLE() entry for the Adreno GPU device, ensuring proper module autoloading based on device tree matching. This is required to guarantee that the GPU driver is loaded correctly when GPU and display probing are decoupled.

Changes in v2:

- Drop dependency on Lemans dual-DPU device tree changes as this patch
works independently (Dmitry)
- Switch separate_gpu_kms to tristate and default to auto mode (Rob)
- Rename msm_gpu_no_components() to msm_separate_gpu_kms_components() for clarity
- Link to v1: https://lore.kernel.org/r/[email protected]

---
  drivers/gpu/drm/msm/adreno/adreno_device.c |  2 +-
  drivers/gpu/drm/msm/msm_drv.c              | 52 ++++++++++++++++++++++++++----
  drivers/gpu/drm/msm/msm_drv.h              |  4 ++-
  drivers/gpu/drm/msm/msm_mdss.c             | 15 +++++++++
  4 files changed, 64 insertions(+), 9 deletions(-)
drivers/gpu/drm/msm/adreno/adreno_device.c +1 -1
index 4edfe80c5be7..e40648c05797 100644
@@ -272,7 +272,7 @@ static const struct component_ops a3xx_ops = {
  static int adreno_probe(struct platform_device *pdev)
  {
  	if (of_device_is_compatible(pdev->dev.of_node, "amd,imageon") ||
-	    msm_gpu_no_components())
+	    msm_separate_gpu_kms_components())
  		return msm_gpu_probe(pdev, &a3xx_ops);
  	return component_add(&pdev->dev, &a3xx_ops);
drivers/gpu/drm/msm/msm_drv.c +12 -4
index e5ab1e28851d..575d1aea7927 100644
@@ -54,16 +54,54 @@ static bool modeset = true;
  MODULE_PARM_DESC(modeset, "Use kernel modesetting [KMS] (1=on (default), 0=disable)");
  module_param(modeset, bool, 0600);
-static bool separate_gpu_kms;
-MODULE_PARM_DESC(separate_gpu_drm, "Use separate DRM device for the GPU (0=single DRM device for both GPU and display (default), 1=two DRM devices)");
-module_param(separate_gpu_kms, bool, 0400);
+/*
+ * separate_gpu_kms (tristate):
+ *   -1 (default): decide automatically based on hardware topology. Split devices
+ *                 if there is more than one GPU or more than one display master.
+ *    0: force single DRM device (bind display + GPU)
+ *    1: force separate DRM devices
+ */
+static int separate_gpu_kms = -1;
+MODULE_PARM_DESC(separate_gpu_kms,
+		 "Use separate DRM device for the GPU (-1=auto (default), 0=single DRM device, 1=separate DRM devices)");
+module_param(separate_gpu_kms, int, 0400);
  DECLARE_FAULT_ATTR(fail_gem_alloc);
  DECLARE_FAULT_ATTR(fail_gem_iova);
-bool msm_gpu_no_components(void)
+static const struct of_device_id msm_gpu_match[];
Quoted 8 lines

drivers/gpu/drm/msm/msm_drv.c static bool modeset = true; · line 80

Quoted 8 lines
drivers/gpu/drm/msm/msm_drv.c +8 -0
+static int msm_count_gpus(void)
+{
+	struct device_node *np;
+	int count = 0;
+
+	for_each_matching_node(np, msm_gpu_match) {
+		if (of_device_is_available(np) && adreno_has_gpu(np))
+			count++;

If !adreno_has_gpu() we definitely should be using separate binding.

Quoted 9 lines

drivers/gpu/drm/msm/msm_drv.c static bool modeset = true; · line 89

Quoted 9 lines
drivers/gpu/drm/msm/msm_drv.c +9 -0
+	}
+
+	return count;
+}
+
+static bool msm_separate_gpu_kms_auto(void)
+{
+	int gpus = msm_count_gpus();
+	int mdss = msm_mdss_count_masters();

This breaks MDP4 case.

Quoted 48 lines
Quoted 48 lines
drivers/gpu/drm/msm/msm_drv.c +16 -3
+
+	if (gpus <= 0 || mdss <= 0)
+		return false;
+
+	/* If exactly one GPU and one display subsystem single card */
+	return (gpus > 1) || (mdss > 1);
+}
+
+bool msm_separate_gpu_kms_components(void)
  {
-	return separate_gpu_kms;
+	if (separate_gpu_kms == 1)
+		return true;
+	if (separate_gpu_kms == 0)
+		return false;
+	return msm_separate_gpu_kms_auto();
  }
  static int msm_drm_uninit(struct device *dev, const struct component_ops *gpu_ops)
@@ -1030,7 +1068,7 @@ static int add_gpu_components(struct device *dev,
  static int msm_drm_bind(struct device *dev)
  {
  	return msm_drm_init(dev,
-			    msm_gpu_no_components() ?
+			    msm_separate_gpu_kms_components() ?
  				    &msm_kms_driver :
  				    &msm_driver,
  			    NULL);
@@ -1069,7 +1107,7 @@ int msm_drv_probe(struct device *master_dev,
  			return ret;
  	}
-	if (!msm_gpu_no_components()) {
+	if (!msm_separate_gpu_kms_components()) {
  		ret = add_gpu_components(master_dev, &match);
  		if (ret)
  			return ret;
drivers/gpu/drm/msm/msm_drv.h +3 -1
index 6d847d593f1a..64a5ad35f7a2 100644
@@ -555,6 +555,8 @@ void msm_kms_shutdown(struct platform_device *pdev);
  bool msm_disp_drv_should_bind(struct device *dev, bool dpu_driver);
-bool msm_gpu_no_components(void);
+bool msm_separate_gpu_kms_components(void);
+
+int msm_mdss_count_masters(void);
  #endif /* __MSM_DRV_H__ */
drivers/gpu/drm/msm/msm_mdss.c +0 -0
index 9047e8d9ee89..00e3ac7dab3a 100644
Quoted 8 lines

drivers/gpu/drm/msm/msm_mdss.c static int msm_mdss_parse_data_bus_icc_path(struct device *dev, · line 77

Quoted 8 lines
drivers/gpu/drm/msm/msm_mdss.c +3 -0
@@ -73,6 +73,21 @@ static int msm_mdss_parse_data_bus_icc_path(struct device *dev,
  	return 0;
  }
+static const struct of_device_id mdss_dt_match[];
+
+int msm_mdss_count_masters(void)

No masters please.

Quoted 21 lines
Quoted 21 lines
drivers/gpu/drm/msm/msm_mdss.c +12 -1
+{
+	struct device_node *np;
+	int count = 0;
+
+	for_each_matching_node(np, mdss_dt_match) {
+		if (of_device_is_available(np))
+			count++;
+	}
+
+	return count;
+}
+
  static void msm_mdss_irq(struct irq_desc *desc)
  {
  	struct msm_mdss *msm_mdss = irq_desc_get_handler_data(desc);
 
---

base-commit: b84a0ebe421ca56995ff78b66307667b62b3a900
change-id: 20260316-separate_gpu_kms-04d2cf4d91e2

Best regards,

Signature · 2 lines
With best wishes
Dmitry
#3 lore raw
In reply to Mahadevan P
kernel test robot noticed the following build errors:

Hi Mahadevan,

kernel test robot noticed the following build errors:

[auto build test ERROR on b84a0ebe421ca56995ff78b66307667b62b3a900]

url:    https://github.com/intel-lab-lkp/linux/commits/Mahadevan-P/drm-msm-default-separate_gpu_kms-to-auto-selection/20260318-063854
base:   b84a0ebe421ca56995ff78b66307667b62b3a900
patch link:    https://lore.kernel.org/r/20260317-separate_gpu_kms-v2-1-b027ca97b9fe%40oss.qualcomm.com
patch subject: [PATCH v2] drm/msm: default separate_gpu_kms to auto selection
config: loongarch-randconfig-002-20260321 (https://download.01.org/0day-ci/archive/20260321/[email protected]/config)
compiler: clang version 20.1.8 (https://github.com/llvm/llvm-project 87f0227cb60147a26a1eeb4fb06e3b505e9c7261)
rustc: rustc 1.88.0 (6b00bc388 2025-06-23)
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20260321/[email protected]/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <[email protected]>
| Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/

All errors (new ones prefixed by >>, old ones prefixed by <<):

>> ERROR: modpost: "msm_mdss_count_masters" [drivers/gpu/drm/msm/msm.ko] undefined!
Signature · 2 lines
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki