[PATCH v8 1/4] drm: move drm based debugfs funcs to drm_debugfs.c

[PATCH v8 1/4] drm: move drm based debugfs funcs to drm_debugfs.c

[Sunil Khatri]

Requirement is to create per client-id based directories to
hold key debugging information and for that access to
root debugfs dentry is need which is not in one place
and that information cannot be stored in drm_device.

Move the debugfs functionality from drm_drv.c and drm_accel.c
to drm_debugfs.c This enables debugfs root node reference
directly drm_debugfs.c and hence enable to create per client-id
directory.

v8: Create drm_accel dentry only if it's config is enabled (Jeff, Hugo)
v8: Merge drm_drv and drm_accel debugfs patches (Koenig, Christian)

Suggested-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
---
 drivers/accel/drm_accel.c      | 16 ----------------
 drivers/gpu/drm/drm_debugfs.c  | 32 ++++++++++++++++++++++++++------
 drivers/gpu/drm/drm_drv.c      | 14 ++++----------
 drivers/gpu/drm/drm_internal.h |  6 ++----
 include/drm/drm_accel.h        |  5 -----
 include/drm/drm_drv.h          | 14 ++++++++++++--
 6 files changed, 44 insertions(+), 43 deletions(-)

diff --git a/drivers/accel/drm_accel.c b/drivers/accel/drm_accel.c
index aa826033b0ce..ca3357acd127 100644
--- a/drivers/accel/drm_accel.c
+++ b/drivers/accel/drm_accel.c
@@ -20,8 +20,6 @@
 
 DEFINE_XARRAY_ALLOC(accel_minors_xa);
 
-static struct dentry *accel_debugfs_root;
-
 static const struct device_type accel_sysfs_device_minor = {
 	.name = "accel_minor"
 };
@@ -73,17 +71,6 @@ static const struct drm_info_list accel_debugfs_list[] = {
 };
 #define ACCEL_DEBUGFS_ENTRIES ARRAY_SIZE(accel_debugfs_list)
 
-/**
- * accel_debugfs_init() - Initialize debugfs for device
- * @dev: Pointer to the device instance.
- *
- * This function creates a root directory for the device in debugfs.
- */
-void accel_debugfs_init(struct drm_device *dev)
-{
-	drm_debugfs_dev_init(dev, accel_debugfs_root);
-}
-
 /**
  * accel_debugfs_register() - Register debugfs for device
  * @dev: Pointer to the device instance.
@@ -194,7 +181,6 @@ static const struct file_operations accel_stub_fops = {
 void accel_core_exit(void)
 {
 	unregister_chrdev(ACCEL_MAJOR, "accel");
-	debugfs_remove(accel_debugfs_root);
 	accel_sysfs_destroy();
 	WARN_ON(!xa_empty(&accel_minors_xa));
 }
@@ -209,8 +195,6 @@ int __init accel_core_init(void)
 		goto error;
 	}
 
-	accel_debugfs_root = debugfs_create_dir("accel", NULL);
-
 	ret = register_chrdev(ACCEL_MAJOR, "accel", &accel_stub_fops);
 	if (ret < 0)
 		DRM_ERROR("Cannot register ACCEL major: %d\n", ret);
diff --git a/drivers/gpu/drm/drm_debugfs.c b/drivers/gpu/drm/drm_debugfs.c
index 2d43bda82887..04da7c27ff7c 100644
--- a/drivers/gpu/drm/drm_debugfs.c
+++ b/drivers/gpu/drm/drm_debugfs.c
@@ -44,6 +44,9 @@
 #include "drm_crtc_internal.h"
 #include "drm_internal.h"
 
+static struct dentry *accel_debugfs_root;
+static struct dentry *drm_debugfs_root;
+
 /***************************************************
  * Initialization, etc.
  **************************************************/
@@ -286,16 +289,34 @@ int drm_debugfs_remove_files(const struct drm_info_list *files, int count,
 }
 EXPORT_SYMBOL(drm_debugfs_remove_files);
 
+void drm_debugfs_init_root(void)
+{
+	drm_debugfs_root = debugfs_create_dir("dri", NULL);
+#if IS_ENABLED(CONFIG_DRM_ACCEL)
+	accel_debugfs_root = debugfs_create_dir("accel", NULL);
+#endif
+}
+
+void drm_debugfs_remove_root(void)
+{
+#if IS_ENABLED(CONFIG_DRM_ACCEL)
+	debugfs_remove(accel_debugfs_root);
+#endif
+	debugfs_remove(drm_debugfs_root);
+}
+
 /**
  * drm_debugfs_dev_init - create debugfs directory for the device
  * @dev: the device which we want to create the directory for
- * @root: the parent directory depending on the device type
  *
  * Creates the debugfs directory for the device under the given root directory.
  */
-void drm_debugfs_dev_init(struct drm_device *dev, struct dentry *root)
+void drm_debugfs_dev_init(struct drm_device *dev)
 {
-	dev->debugfs_root = debugfs_create_dir(dev->unique, root);
+	if (drm_core_check_feature(dev, DRIVER_COMPUTE_ACCEL))
+		dev->debugfs_root = debugfs_create_dir(dev->unique, accel_debugfs_root);
+	else
+		dev->debugfs_root = debugfs_create_dir(dev->unique, drm_debugfs_root);
 }
 
 /**
@@ -322,14 +343,13 @@ void drm_debugfs_dev_register(struct drm_device *dev)
 		drm_atomic_debugfs_init(dev);
 }
 
-int drm_debugfs_register(struct drm_minor *minor, int minor_id,
-			 struct dentry *root)
+int drm_debugfs_register(struct drm_minor *minor, int minor_id)
 {
 	struct drm_device *dev = minor->dev;
 	char name[64];
 
 	sprintf(name, "%d", minor_id);
-	minor->debugfs_symlink = debugfs_create_symlink(name, root,
+	minor->debugfs_symlink = debugfs_create_symlink(name, drm_debugfs_root,
 							dev->unique);
 
 	/* TODO: Only for compatibility with drivers */
diff --git a/drivers/gpu/drm/drm_drv.c b/drivers/gpu/drm/drm_drv.c
index 17fc5dc708f4..b9b5c9f953af 100644
--- a/drivers/gpu/drm/drm_drv.c
+++ b/drivers/gpu/drm/drm_drv.c
@@ -69,8 +69,6 @@ DEFINE_XARRAY_ALLOC(drm_minors_xa);
  */
 static bool drm_core_init_complete;
 
-static struct dentry *drm_debugfs_root;
-
 DEFINE_STATIC_SRCU(drm_unplug_srcu);
 
 /*
@@ -183,8 +181,7 @@ static int drm_minor_register(struct drm_device *dev, enum drm_minor_type type)
 		return 0;
 
 	if (minor->type != DRM_MINOR_ACCEL) {
-		ret = drm_debugfs_register(minor, minor->index,
-					   drm_debugfs_root);
+		ret = drm_debugfs_register(minor, minor->index);
 		if (ret) {
 			DRM_ERROR("DRM: Failed to initialize /sys/kernel/debug/dri.\n");
 			goto err_debugfs;
@@ -751,10 +748,7 @@ static int drm_dev_init(struct drm_device *dev,
 		goto err;
 	}
 
-	if (drm_core_check_feature(dev, DRIVER_COMPUTE_ACCEL))
-		accel_debugfs_init(dev);
-	else
-		drm_debugfs_dev_init(dev, drm_debugfs_root);
+	drm_debugfs_dev_init(dev);
 
 	return 0;
 
@@ -1168,7 +1162,7 @@ static void drm_core_exit(void)
 	drm_panic_exit();
 	accel_core_exit();
 	unregister_chrdev(DRM_MAJOR, "drm");
-	debugfs_remove(drm_debugfs_root);
+	drm_debugfs_remove_root();
 	drm_sysfs_destroy();
 	WARN_ON(!xa_empty(&drm_minors_xa));
 	drm_connector_ida_destroy();
@@ -1187,7 +1181,7 @@ static int __init drm_core_init(void)
 		goto error;
 	}
 
-	drm_debugfs_root = debugfs_create_dir("dri", NULL);
+	drm_debugfs_init_root();
 
 	ret = register_chrdev(DRM_MAJOR, "drm", &drm_stub_fops);
 	if (ret < 0)
diff --git a/drivers/gpu/drm/drm_internal.h b/drivers/gpu/drm/drm_internal.h
index b2b6a8e49dda..d2d8e72f32d9 100644
--- a/drivers/gpu/drm/drm_internal.h
+++ b/drivers/gpu/drm/drm_internal.h
@@ -186,8 +186,7 @@ void drm_gem_vunmap(struct drm_gem_object *obj, struct iosys_map *map);
 #if defined(CONFIG_DEBUG_FS)
 void drm_debugfs_dev_fini(struct drm_device *dev);
 void drm_debugfs_dev_register(struct drm_device *dev);
-int drm_debugfs_register(struct drm_minor *minor, int minor_id,
-			 struct dentry *root);
+int drm_debugfs_register(struct drm_minor *minor, int minor_id);
 void drm_debugfs_unregister(struct drm_minor *minor);
 void drm_debugfs_connector_add(struct drm_connector *connector);
 void drm_debugfs_connector_remove(struct drm_connector *connector);
@@ -205,8 +204,7 @@ static inline void drm_debugfs_dev_register(struct drm_device *dev)
 {
 }
 
-static inline int drm_debugfs_register(struct drm_minor *minor, int minor_id,
-				       struct dentry *root)
+static inline int drm_debugfs_register(struct drm_minor *minor, int minor_id)
 {
 	return 0;
 }
diff --git a/include/drm/drm_accel.h b/include/drm/drm_accel.h
index 038ccb02f9a3..20a665ec6f16 100644
--- a/include/drm/drm_accel.h
+++ b/include/drm/drm_accel.h
@@ -58,7 +58,6 @@ void accel_core_exit(void);
 int accel_core_init(void);
 void accel_set_device_instance_params(struct device *kdev, int index);
 int accel_open(struct inode *inode, struct file *filp);
-void accel_debugfs_init(struct drm_device *dev);
 void accel_debugfs_register(struct drm_device *dev);
 
 #else
@@ -77,10 +76,6 @@ static inline void accel_set_device_instance_params(struct device *kdev, int ind
 {
 }
 
-static inline void accel_debugfs_init(struct drm_device *dev)
-{
-}
-
 static inline void accel_debugfs_register(struct drm_device *dev)
 {
 }
diff --git a/include/drm/drm_drv.h b/include/drm/drm_drv.h
index a43d707b5f36..392baeaaa4a1 100644
--- a/include/drm/drm_drv.h
+++ b/include/drm/drm_drv.h
@@ -566,9 +566,19 @@ static inline bool drm_firmware_drivers_only(void)
 }
 
 #if defined(CONFIG_DEBUG_FS)
-void drm_debugfs_dev_init(struct drm_device *dev, struct dentry *root);
+void drm_debugfs_dev_init(struct drm_device *dev);
+void drm_debugfs_init_root(void);
+void drm_debugfs_remove_root(void);
 #else
-static inline void drm_debugfs_dev_init(struct drm_device *dev, struct dentry *root)
+static inline void drm_debugfs_dev_init(struct drm_device *dev)
+{
+}
+
+static inline void drm_debugfs_init_root(void)
+{
+}
+
+static inline void drm_debugfs_remove_root(void)
 {
 }
 #endif
-- 
2.34.1

[PATCH v8 2/4] drm: add debugfs support on per client-id basis

[Sunil Khatri]

add support to add a directory for each client-id
with root at the dri level. Since the clients are
unique and not just related to one single drm device,
so it makes more sense to add all the client based
nodes with root as dri.

Also create a debugfs file which show the process
information for the client and create a symlink back
to the parent drm device from each client.

Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
---
 drivers/gpu/drm/drm_debugfs.c | 81 +++++++++++++++++++++++++++++++++++
 drivers/gpu/drm/drm_file.c    |  9 ++++
 include/drm/drm_debugfs.h     | 11 +++++
 include/drm/drm_file.h        |  7 +++
 4 files changed, 108 insertions(+)

diff --git a/drivers/gpu/drm/drm_debugfs.c b/drivers/gpu/drm/drm_debugfs.c
index 04da7c27ff7c..6d1f1ca7ad9e 100644
--- a/drivers/gpu/drm/drm_debugfs.c
+++ b/drivers/gpu/drm/drm_debugfs.c
@@ -305,6 +305,87 @@ void drm_debugfs_remove_root(void)
 	debugfs_remove(drm_debugfs_root);
 }
 
+static int drm_debugfs_proc_info_show(struct seq_file *m, void *unused)
+{
+	struct pid *pid;
+	struct task_struct *task;
+	struct drm_file *file = m->private;
+
+	if (!file)
+		return -EINVAL;
+
+	rcu_read_lock();
+	pid = rcu_dereference(file->pid);
+	task = pid_task(pid, PIDTYPE_TGID);
+
+	seq_printf(m, "pid: %d\n", task ? task->pid : 0);
+	seq_printf(m, "comm: %s\n", task ? task->comm : "Unset");
+	rcu_read_unlock();
+	return 0;
+}
+
+static int drm_debufs_proc_info_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, drm_debugfs_proc_info_show, inode->i_private);
+}
+
+static const struct file_operations drm_debugfs_proc_info_fops = {
+	.owner = THIS_MODULE,
+	.open = drm_debufs_proc_info_open,
+	.read = seq_read,
+	.llseek = seq_lseek,
+	.release = single_release,
+};
+
+/**
+ * drm_debugfs_clients_add - Add a per client debugfs directory
+ * @file: drm_file for a client
+ *
+ * Create the debugfs directory for each client. This will be used to populate
+ * driver specific data for each client.
+ *
+ * Also add the process information debugfs file for each client to tag
+ * which client belongs to which process.
+ */
+void drm_debugfs_clients_add(struct drm_file *file)
+{
+	char *client;
+
+	client = kasprintf(GFP_KERNEL, "client-%llu", file->client_id);
+	if (!client)
+		return;
+
+	/* Create a debugfs directory for the client in root on drm debugfs */
+	file->debugfs_client = debugfs_create_dir(client, drm_debugfs_root);
+	kfree(client);
+
+	debugfs_create_file("proc_info", 0444, file->debugfs_client, file,
+			    &drm_debugfs_proc_info_fops);
+
+	client = kasprintf(GFP_KERNEL, "../%s", file->minor->dev->unique);
+	if (!client)
+		return;
+
+	/* Create a link from client_id to the drm device this client id belongs to */
+	debugfs_create_symlink("device", file->debugfs_client, client);
+	kfree(client);
+}
+
+/**
+ * drm_debugfs_clients_remove - removes all debugfs directories and files
+ * @file: drm_file for a client
+ *
+ * Removes the debugfs directories recursively from the client directory.
+ *
+ * There is also a possibility that debugfs files are open while the drm_file
+ * is released.
+ */
+void drm_debugfs_clients_remove(struct drm_file *file)
+{
+	debugfs_remove_recursive(file->debugfs_client);
+	file->debugfs_client = NULL;
+}
+
 /**
  * drm_debugfs_dev_init - create debugfs directory for the device
  * @dev: the device which we want to create the directory for
diff --git a/drivers/gpu/drm/drm_file.c b/drivers/gpu/drm/drm_file.c
index 06ba6dcbf5ae..3f4c52d00540 100644
--- a/drivers/gpu/drm/drm_file.c
+++ b/drivers/gpu/drm/drm_file.c
@@ -45,6 +45,7 @@
 #include <drm/drm_file.h>
 #include <drm/drm_gem.h>
 #include <drm/drm_print.h>
+#include <drm/drm_debugfs.h>
 
 #include "drm_crtc_internal.h"
 #include "drm_internal.h"
@@ -167,6 +168,9 @@ struct drm_file *drm_file_alloc(struct drm_minor *minor)
 
 	drm_prime_init_file_private(&file->prime);
 
+	if (!drm_core_check_feature(dev, DRIVER_COMPUTE_ACCEL))
+		drm_debugfs_clients_add(file);
+
 	if (dev->driver->open) {
 		ret = dev->driver->open(dev, file);
 		if (ret < 0)
@@ -181,6 +185,8 @@ struct drm_file *drm_file_alloc(struct drm_minor *minor)
 		drm_syncobj_release(file);
 	if (drm_core_check_feature(dev, DRIVER_GEM))
 		drm_gem_release(dev, file);
+
+	drm_debugfs_clients_remove(file);
 	put_pid(rcu_access_pointer(file->pid));
 	kfree(file);
 
@@ -235,6 +241,9 @@ void drm_file_free(struct drm_file *file)
 		     (long)old_encode_dev(file->minor->kdev->devt),
 		     atomic_read(&dev->open_count));
 
+	if (!drm_core_check_feature(dev, DRIVER_COMPUTE_ACCEL))
+		drm_debugfs_clients_remove(file);
+
 	drm_events_release(file);
 
 	if (drm_core_check_feature(dev, DRIVER_MODESET)) {
diff --git a/include/drm/drm_debugfs.h b/include/drm/drm_debugfs.h
index cf06cee4343f..ea8cba94208a 100644
--- a/include/drm/drm_debugfs.h
+++ b/include/drm/drm_debugfs.h
@@ -153,6 +153,9 @@ void drm_debugfs_add_files(struct drm_device *dev,
 
 int drm_debugfs_gpuva_info(struct seq_file *m,
 			   struct drm_gpuvm *gpuvm);
+
+void drm_debugfs_clients_add(struct drm_file *file);
+void drm_debugfs_clients_remove(struct drm_file *file);
 #else
 static inline void drm_debugfs_create_files(const struct drm_info_list *files,
 					    int count, struct dentry *root,
@@ -181,6 +184,14 @@ static inline int drm_debugfs_gpuva_info(struct seq_file *m,
 {
 	return 0;
 }
+
+static inline void drm_debugfs_clients_add(struct drm_file *file)
+{
+}
+
+static inline void drm_debugfs_clients_remove(struct drm_file *file)
+{
+}
 #endif
 
 #endif /* _DRM_DEBUGFS_H_ */
diff --git a/include/drm/drm_file.h b/include/drm/drm_file.h
index 5c3b2aa3e69d..eab7546aad79 100644
--- a/include/drm/drm_file.h
+++ b/include/drm/drm_file.h
@@ -400,6 +400,13 @@ struct drm_file {
 	 * @client_name_lock: Protects @client_name.
 	 */
 	struct mutex client_name_lock;
+
+	/**
+	 * @debugfs_client:
+	 *
+	 * debugfs directory for each client under a drm node.
+	 */
+	struct dentry *debugfs_client;
 };
 
 /**
-- 
2.34.1

[PATCH v8 3/4] drm/amdgpu: add debugfs support for VM pagetable per client

[Sunil Khatri]

Add a debugfs file under the client directory which shares
the root page table base address of the VM.

This address could be used to dump the pagetable for debug
memory issues.

Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 52 +++++++++++++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.h |  1 +
 drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c     |  2 +-
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c      |  4 +-
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h      |  4 +-
 5 files changed, 60 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
index f81608330a3d..6762dd11f00c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
@@ -2131,6 +2131,55 @@ int amdgpu_debugfs_init(struct amdgpu_device *adev)
 	return 0;
 }
 
+static int amdgpu_pt_info_read(struct seq_file *m, void *unused)
+{
+	struct drm_file *file;
+	struct amdgpu_fpriv *fpriv;
+	struct amdgpu_bo *root_bo;
+	int r;
+
+	file = m->private;
+	if (!file)
+		return -EINVAL;
+
+	fpriv = file->driver_priv;
+	if (!fpriv && !fpriv->vm.root.bo)
+		return -ENODEV;
+
+	root_bo = amdgpu_bo_ref(fpriv->vm.root.bo);
+	r = amdgpu_bo_reserve(root_bo, true);
+	if (r) {
+		amdgpu_bo_unref(&root_bo);
+		return -EINVAL;
+	}
+
+	seq_printf(m, "gpu_address: 0x%llx\n", amdgpu_bo_gpu_offset(fpriv->vm.root.bo));
+
+	amdgpu_bo_unreserve(root_bo);
+	amdgpu_bo_unref(&root_bo);
+
+	return 0;
+}
+
+static int amdgpu_pt_info_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, amdgpu_pt_info_read, inode->i_private);
+}
+
+static const struct file_operations amdgpu_pt_info_fops = {
+	.owner = THIS_MODULE,
+	.open = amdgpu_pt_info_open,
+	.read = seq_read,
+	.llseek = seq_lseek,
+	.release = single_release,
+};
+
+void amdgpu_debugfs_vm_init(struct drm_file *file)
+{
+	debugfs_create_file("vm_pagetable_info", 0444, file->debugfs_client, file,
+			    &amdgpu_pt_info_fops);
+}
+
 #else
 int amdgpu_debugfs_init(struct amdgpu_device *adev)
 {
@@ -2140,4 +2189,7 @@ int amdgpu_debugfs_regs_init(struct amdgpu_device *adev)
 {
 	return 0;
 }
+void amdgpu_debugfs_vm_init(struct drm_file *file)
+{
+}
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.h
index 0425432d8659..e7b3c38e5186 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.h
@@ -33,4 +33,5 @@ void amdgpu_debugfs_fence_init(struct amdgpu_device *adev);
 void amdgpu_debugfs_firmware_init(struct amdgpu_device *adev);
 void amdgpu_debugfs_gem_init(struct amdgpu_device *adev);
 void amdgpu_debugfs_mes_event_log_init(struct amdgpu_device *adev);
+void amdgpu_debugfs_vm_init(struct drm_file *file);
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
index 195ed81d39ff..27aa1b551dbf 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
@@ -1395,7 +1395,7 @@ int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv)
 	if (r)
 		goto error_pasid;
 
-	r = amdgpu_vm_init(adev, &fpriv->vm, fpriv->xcp_id);
+	r = amdgpu_vm_init(adev, &fpriv->vm, fpriv->xcp_id, file_priv);
 	if (r)
 		goto error_pasid;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index f042372d9f2e..7e31fb5f6f33 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -2527,6 +2527,7 @@ void amdgpu_vm_set_task_info(struct amdgpu_vm *vm)
  * @adev: amdgpu_device pointer
  * @vm: requested vm
  * @xcp_id: GPU partition selection id
+ * @file: drm_file
  *
  * Init @vm fields.
  *
@@ -2534,7 +2535,7 @@ void amdgpu_vm_set_task_info(struct amdgpu_vm *vm)
  * 0 for success, error for failure.
  */
 int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-		   int32_t xcp_id)
+		   int32_t xcp_id, struct drm_file *file)
 {
 	struct amdgpu_bo *root_bo;
 	struct amdgpu_bo_vm *root;
@@ -2610,6 +2611,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	if (r)
 		dev_dbg(adev->dev, "Failed to create task info for VM\n");
 
+	amdgpu_debugfs_vm_init(file);
 	amdgpu_bo_unreserve(vm->root.bo);
 	amdgpu_bo_unref(&root_bo);
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
index f3ad687125ad..555afaf867c4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
@@ -487,7 +487,9 @@ int amdgpu_vm_set_pasid(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 			u32 pasid);
 
 long amdgpu_vm_wait_idle(struct amdgpu_vm *vm, long timeout);
-int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, int32_t xcp_id);
+int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, int32_t xcp_id,
+		   struct drm_file *file);
+
 int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm);
 void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm);
 int amdgpu_vm_lock_pd(struct amdgpu_vm *vm, struct drm_exec *exec,
-- 
2.34.1

[PATCH v8 4/4] drm/amdgpu: add support of debugfs for mqd information

[Sunil Khatri]

Add debugfs support for mqd for each queue of the client.

The address exposed to debugfs could be used to dump
the mqd.

Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c | 52 +++++++++++++++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h |  1 +
 2 files changed, 53 insertions(+)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c
index 295e7186e156..115d53bc9a8d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c
@@ -318,6 +318,9 @@ amdgpu_userq_destroy(struct drm_file *filp, int queue_id)
 		amdgpu_bo_unreserve(queue->db_obj.obj);
 	}
 	amdgpu_bo_unref(&queue->db_obj.obj);
+
+	debugfs_remove_recursive(queue->debugfs_queue);
+
 	r = amdgpu_userq_unmap_helper(uq_mgr, queue);
 	amdgpu_userq_cleanup(uq_mgr, queue, queue_id);
 	mutex_unlock(&uq_mgr->userq_mutex);
@@ -343,6 +346,46 @@ static int amdgpu_userq_priority_permit(struct drm_file *filp,
 	return -EACCES;
 }
 
+#if defined(CONFIG_DEBUG_FS)
+static int amdgpu_mqd_info_read(struct seq_file *m, void *unused)
+{
+	struct amdgpu_usermode_queue *queue = m->private;
+	struct amdgpu_bo *bo;
+	int r;
+
+	if (!queue || !queue->mqd.obj)
+		return -EINVAL;
+
+	bo = amdgpu_bo_ref(queue->mqd.obj);
+	r = amdgpu_bo_reserve(bo, true);
+	if (r) {
+		amdgpu_bo_unref(&bo);
+		return -EINVAL;
+	}
+
+	seq_printf(m, "queue_type %d\n", queue->queue_type);
+	seq_printf(m, "mqd_gpu_address: 0x%llx\n", amdgpu_bo_gpu_offset(queue->mqd.obj));
+
+	amdgpu_bo_unreserve(bo);
+	amdgpu_bo_unref(&bo);
+
+	return 0;
+}
+
+static int amdgpu_mqd_info_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, amdgpu_mqd_info_read, inode->i_private);
+}
+
+static const struct file_operations amdgpu_mqd_info_fops = {
+	.owner = THIS_MODULE,
+	.open = amdgpu_mqd_info_open,
+	.read = seq_read,
+	.llseek = seq_lseek,
+	.release = single_release,
+};
+#endif
+
 static int
 amdgpu_userq_create(struct drm_file *filp, union drm_amdgpu_userq *args)
 {
@@ -352,6 +395,7 @@ amdgpu_userq_create(struct drm_file *filp, union drm_amdgpu_userq *args)
 	const struct amdgpu_userq_funcs *uq_funcs;
 	struct amdgpu_usermode_queue *queue;
 	struct amdgpu_db_info db_info;
+	char *queue_name;
 	bool skip_map_queue;
 	uint64_t index;
 	int qid, r = 0;
@@ -475,6 +519,14 @@ amdgpu_userq_create(struct drm_file *filp, union drm_amdgpu_userq *args)
 		}
 	}
 
+	queue_name = kasprintf(GFP_KERNEL, "queue-%d", qid);
+	if (!queue_name)
+		return -ENOMEM;
+
+	/* Queue dentry per client to hold MQD information   */
+	queue->debugfs_queue = debugfs_create_dir(queue_name, filp->debugfs_client);
+	debugfs_create_file("mqd_info", 0444, queue->debugfs_queue, queue, &amdgpu_mqd_info_fops);
+	kfree(queue_name);
 
 	args->out.queue_id = qid;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h
index ec040c2fd6c9..b1ca91b7cda4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h
@@ -65,6 +65,7 @@ struct amdgpu_usermode_queue {
 	struct dma_fence	*last_fence;
 	u32			xcp_id;
 	int			priority;
+	struct dentry		*debugfs_queue;
 };
 
 struct amdgpu_userq_funcs {
-- 
2.34.1

RE: [PATCH v8 1/4] drm: move drm based debugfs funcs to drm_debugfs.c

[Khatri, Sunil]

[AMD Official Use Only - AMD Internal Distribution Only]

-----Original Message-----
From: Sunil Khatri <sunil.khatri@amd.com>
Sent: Tuesday, July 1, 2025 12:03 PM
To: Koenig, Christian <Christian.Koenig@amd.com>; dri-devel@lists.freedesktop.org
Cc: amd-gfx@lists.freedesktop.org; simona@ffwll.ch; tzimmermann@suse.de; tursulin@ursulin.net; phasta@kernel.org; dakr@kernel.org; linux-kernel@vger.kernel.org; Oded Gabbay <ogabbay@kernel.org>; Khatri, Sunil <Sunil.Khatri@amd.com>
Subject: [PATCH v8 1/4] drm: move drm based debugfs funcs to drm_debugfs.c

Requirement is to create per client-id based directories to hold key debugging information and for that access to root debugfs dentry is need which is not in one place and that information cannot be stored in drm_device.

Move the debugfs functionality from drm_drv.c and drm_accel.c to drm_debugfs.c This enables debugfs root node reference directly drm_debugfs.c and hence enable to create per client-id directory.

v8: Create drm_accel dentry only if it's config is enabled (Jeff, Hugo)
v8: Merge drm_drv and drm_accel debugfs patches (Koenig, Christian)

Suggested-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
---
 drivers/accel/drm_accel.c      | 16 ----------------
 drivers/gpu/drm/drm_debugfs.c  | 32 ++++++++++++++++++++++++++------
 drivers/gpu/drm/drm_drv.c      | 14 ++++----------
 drivers/gpu/drm/drm_internal.h |  6 ++----
 include/drm/drm_accel.h        |  5 -----
 include/drm/drm_drv.h          | 14 ++++++++++++--
 6 files changed, 44 insertions(+), 43 deletions(-)

diff --git a/drivers/accel/drm_accel.c b/drivers/accel/drm_accel.c index aa826033b0ce..ca3357acd127 100644
--- a/drivers/accel/drm_accel.c
+++ b/drivers/accel/drm_accel.c
@@ -20,8 +20,6 @@

 DEFINE_XARRAY_ALLOC(accel_minors_xa);

-static struct dentry *accel_debugfs_root;
-
 static const struct device_type accel_sysfs_device_minor = {
        .name = "accel_minor"
 };
@@ -73,17 +71,6 @@ static const struct drm_info_list accel_debugfs_list[] = {  };  #define ACCEL_DEBUGFS_ENTRIES ARRAY_SIZE(accel_debugfs_list)

-/**
- * accel_debugfs_init() - Initialize debugfs for device
- * @dev: Pointer to the device instance.
- *
- * This function creates a root directory for the device in debugfs.
- */
-void accel_debugfs_init(struct drm_device *dev) -{
-       drm_debugfs_dev_init(dev, accel_debugfs_root);
-}
-
 /**
  * accel_debugfs_register() - Register debugfs for device
  * @dev: Pointer to the device instance.
@@ -194,7 +181,6 @@ static const struct file_operations accel_stub_fops = {  void accel_core_exit(void)  {
        unregister_chrdev(ACCEL_MAJOR, "accel");
-       debugfs_remove(accel_debugfs_root);
        accel_sysfs_destroy();
        WARN_ON(!xa_empty(&accel_minors_xa));
 }
@@ -209,8 +195,6 @@ int __init accel_core_init(void)
                goto error;
        }

-       accel_debugfs_root = debugfs_create_dir("accel", NULL);
-
        ret = register_chrdev(ACCEL_MAJOR, "accel", &accel_stub_fops);
        if (ret < 0)
                DRM_ERROR("Cannot register ACCEL major: %d\n", ret); diff --git a/drivers/gpu/drm/drm_debugfs.c b/drivers/gpu/drm/drm_debugfs.c index 2d43bda82887..04da7c27ff7c 100644
--- a/drivers/gpu/drm/drm_debugfs.c
+++ b/drivers/gpu/drm/drm_debugfs.c
@@ -44,6 +44,9 @@
 #include "drm_crtc_internal.h"
 #include "drm_internal.h"

+static struct dentry *accel_debugfs_root; static struct dentry
+*drm_debugfs_root;
+
 /***************************************************
  * Initialization, etc.
  **************************************************/
@@ -286,16 +289,34 @@ int drm_debugfs_remove_files(const struct drm_info_list *files, int count,  }  EXPORT_SYMBOL(drm_debugfs_remove_files);

+void drm_debugfs_init_root(void)
+{
+       drm_debugfs_root = debugfs_create_dir("dri", NULL); #if
+IS_ENABLED(CONFIG_DRM_ACCEL)
+       accel_debugfs_root = debugfs_create_dir("accel", NULL); #endif }
+
+void drm_debugfs_remove_root(void)
+{
+#if IS_ENABLED(CONFIG_DRM_ACCEL)
+       debugfs_remove(accel_debugfs_root);
+#endif
+       debugfs_remove(drm_debugfs_root);
+}
+
 /**
  * drm_debugfs_dev_init - create debugfs directory for the device
  * @dev: the device which we want to create the directory for
- * @root: the parent directory depending on the device type
  *
  * Creates the debugfs directory for the device under the given root directory.
  */
-void drm_debugfs_dev_init(struct drm_device *dev, struct dentry *root)
+void drm_debugfs_dev_init(struct drm_device *dev)
 {
-       dev->debugfs_root = debugfs_create_dir(dev->unique, root);
+       if (drm_core_check_feature(dev, DRIVER_COMPUTE_ACCEL))
+               dev->debugfs_root = debugfs_create_dir(dev->unique, accel_debugfs_root);
+       else
+               dev->debugfs_root = debugfs_create_dir(dev->unique,
+drm_debugfs_root);
 }

 /**
@@ -322,14 +343,13 @@ void drm_debugfs_dev_register(struct drm_device *dev)
                drm_atomic_debugfs_init(dev);
 }

-int drm_debugfs_register(struct drm_minor *minor, int minor_id,
-                        struct dentry *root)
+int drm_debugfs_register(struct drm_minor *minor, int minor_id)
 {
        struct drm_device *dev = minor->dev;
        char name[64];

        sprintf(name, "%d", minor_id);
-       minor->debugfs_symlink = debugfs_create_symlink(name, root,
+       minor->debugfs_symlink = debugfs_create_symlink(name,
+drm_debugfs_root,
                                                        dev->unique);

        /* TODO: Only for compatibility with drivers */ diff --git a/drivers/gpu/drm/drm_drv.c b/drivers/gpu/drm/drm_drv.c index 17fc5dc708f4..b9b5c9f953af 100644
--- a/drivers/gpu/drm/drm_drv.c
+++ b/drivers/gpu/drm/drm_drv.c
@@ -69,8 +69,6 @@ DEFINE_XARRAY_ALLOC(drm_minors_xa);
  */
 static bool drm_core_init_complete;

-static struct dentry *drm_debugfs_root;
-
 DEFINE_STATIC_SRCU(drm_unplug_srcu);

 /*
@@ -183,8 +181,7 @@ static int drm_minor_register(struct drm_device *dev, enum drm_minor_type type)
                return 0;

        if (minor->type != DRM_MINOR_ACCEL) {
-               ret = drm_debugfs_register(minor, minor->index,
-                                          drm_debugfs_root);
+               ret = drm_debugfs_register(minor, minor->index);
                if (ret) {
                        DRM_ERROR("DRM: Failed to initialize /sys/kernel/debug/dri.\n");
                        goto err_debugfs;
@@ -751,10 +748,7 @@ static int drm_dev_init(struct drm_device *dev,
                goto err;
        }

-       if (drm_core_check_feature(dev, DRIVER_COMPUTE_ACCEL))
-               accel_debugfs_init(dev);
-       else
-               drm_debugfs_dev_init(dev, drm_debugfs_root);
+       drm_debugfs_dev_init(dev);

        return 0;

@@ -1168,7 +1162,7 @@ static void drm_core_exit(void)
        drm_panic_exit();
        accel_core_exit();
        unregister_chrdev(DRM_MAJOR, "drm");
-       debugfs_remove(drm_debugfs_root);
+       drm_debugfs_remove_root();
        drm_sysfs_destroy();
        WARN_ON(!xa_empty(&drm_minors_xa));
        drm_connector_ida_destroy();
@@ -1187,7 +1181,7 @@ static int __init drm_core_init(void)
                goto error;
        }

-       drm_debugfs_root = debugfs_create_dir("dri", NULL);
+       drm_debugfs_init_root();

        ret = register_chrdev(DRM_MAJOR, "drm", &drm_stub_fops);
        if (ret < 0)
diff --git a/drivers/gpu/drm/drm_internal.h b/drivers/gpu/drm/drm_internal.h index b2b6a8e49dda..d2d8e72f32d9 100644
--- a/drivers/gpu/drm/drm_internal.h
+++ b/drivers/gpu/drm/drm_internal.h
@@ -186,8 +186,7 @@ void drm_gem_vunmap(struct drm_gem_object *obj, struct iosys_map *map);  #if defined(CONFIG_DEBUG_FS)  void drm_debugfs_dev_fini(struct drm_device *dev);  void drm_debugfs_dev_register(struct drm_device *dev); -int drm_debugfs_register(struct drm_minor *minor, int minor_id,
-                        struct dentry *root);
+int drm_debugfs_register(struct drm_minor *minor, int minor_id);
 void drm_debugfs_unregister(struct drm_minor *minor);  void drm_debugfs_connector_add(struct drm_connector *connector);  void drm_debugfs_connector_remove(struct drm_connector *connector); @@ -205,8 +204,7 @@ static inline void drm_debugfs_dev_register(struct drm_device *dev)  {  }

-static inline int drm_debugfs_register(struct drm_minor *minor, int minor_id,
-                                      struct dentry *root)
+static inline int drm_debugfs_register(struct drm_minor *minor, int
+minor_id)
 {
        return 0;
 }
diff --git a/include/drm/drm_accel.h b/include/drm/drm_accel.h index 038ccb02f9a3..20a665ec6f16 100644
--- a/include/drm/drm_accel.h
+++ b/include/drm/drm_accel.h
@@ -58,7 +58,6 @@ void accel_core_exit(void);  int accel_core_init(void);  void accel_set_device_instance_params(struct device *kdev, int index);  int accel_open(struct inode *inode, struct file *filp); -void accel_debugfs_init(struct drm_device *dev);  void accel_debugfs_register(struct drm_device *dev);

 #else
@@ -77,10 +76,6 @@ static inline void accel_set_device_instance_params(struct device *kdev, int ind  {  }

-static inline void accel_debugfs_init(struct drm_device *dev) -{ -}
-
 static inline void accel_debugfs_register(struct drm_device *dev)  {  } diff --git a/include/drm/drm_drv.h b/include/drm/drm_drv.h index a43d707b5f36..392baeaaa4a1 100644
--- a/include/drm/drm_drv.h
+++ b/include/drm/drm_drv.h
@@ -566,9 +566,19 @@ static inline bool drm_firmware_drivers_only(void)  }

 #if defined(CONFIG_DEBUG_FS)
-void drm_debugfs_dev_init(struct drm_device *dev, struct dentry *root);
+void drm_debugfs_dev_init(struct drm_device *dev); void
+drm_debugfs_init_root(void); void drm_debugfs_remove_root(void);
 #else
-static inline void drm_debugfs_dev_init(struct drm_device *dev, struct dentry *root)
+static inline void drm_debugfs_dev_init(struct drm_device *dev) { }
+
+static inline void drm_debugfs_init_root(void) { }
+
+static inline void drm_debugfs_remove_root(void)
 {
 }
 #endif
--
2.34.1

Re: [PATCH v8 1/4] drm: move drm based debugfs funcs to drm_debugfs.c

[Jeff Hugo]

On 7/1/2025 7:02 AM, Khatri, Sunil wrote:
> [AMD Official Use Only - AMD Internal Distribution Only]

I cannot review this message with this restriction.  In my opinion, your 
email client is not properly configured for interfacing with the community.

-Jeff

Re: [PATCH v8 1/4] drm: move drm based debugfs funcs to drm_debugfs.c

[Khatri, Sunil]

On 7/1/2025 8:50 PM, Jeff Hugo wrote:
> On 7/1/2025 7:02 AM, Khatri, Sunil wrote:
>> [AMD Official Use Only - AMD Internal Distribution Only]
>
> I cannot review this message with this restriction.  In my opinion, 
> your email client is not properly configured for interfacing with the 
> community.

Hi Hugo,

Sorry for the email client issue. Fixed that and sent v9 patch set with 
right format and also attached the cover letter too. Hope this clarifies.

Regards
Sunil Khatri

>
> -Jeff