Re: [PATCH v9 3/4] drm/amdgpu: add debugfs support for VM pagetable per client

["Christian König" <christian.koenig@amd.com>]

On 01.07.25 18:49, Sunil Khatri wrote:
> 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 4aab5e394ce2..d3f16a966c70 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
> @@ -1415,7 +1415,7 @@ 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);

Move that into the caller of amdgpu_vm_init(), this way amdgpu_vm_init() also doesn't need to get the drm_file as parameter.

With that done Reviewed-by: Christian König <christian.koenig@amd.com>.

If nobody objects I will push the first two patches to drm-misc-next now, so you only need to edit, rebase and send out again patch #3 and #4.

Regards,
Christian

>  	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,