Re: [PATCH V4 14/18] amdkfd: record kfd process id into kfd process_info

["Kuehling, Felix" <felix.kuehling@amd.com>]

On 2025-09-23 03:26, Zhu Lingshan wrote:
> This commit records the id of the owner
> kfd_process into a kfd process_info when
> create it.
>
> Signed-off-by: Zhu Lingshan <lingshan.zhu@amd.com>
> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h       | 2 ++
>   drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 4 ++++
>   2 files changed, 6 insertions(+)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
> index aa88bad7416b..d867984a68da 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
> @@ -146,6 +146,8 @@ struct amdkfd_process_info {
>   	/* MMU-notifier related fields */
>   	struct mutex notifier_lock;
>   	uint32_t evicted_bos;
> +	/* kfd process id */
> +	u16 process_id;

The name "process_id" is a bit misleading. I would prefer something like 
"context_id" or "secondary_id" to make it cleare that this identifies 
secondary contexts and has nothing to do with the PID. Maybe use the 
same name in struct kfd_process as well for clarity.

Regards,
   Felix


>   	struct delayed_work restore_userptr_work;
>   	struct pid *pid;
>   	bool block_mmu_notifications;
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
> index b16cce7c22c3..723d34921c12 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
> @@ -1382,8 +1382,10 @@ static int init_kfd_vm(struct amdgpu_vm *vm, void **process_info,
>   		       struct dma_fence **ef)
>   {
>   	struct amdkfd_process_info *info = NULL;
> +	struct kfd_process *process = NULL;
>   	int ret;
>   
> +	process = container_of(process_info, struct kfd_process, kgd_process_info);
>   	if (!*process_info) {
>   		info = kzalloc(sizeof(*info), GFP_KERNEL);
>   		if (!info)
> @@ -1410,6 +1412,8 @@ static int init_kfd_vm(struct amdgpu_vm *vm, void **process_info,
>   		INIT_DELAYED_WORK(&info->restore_userptr_work,
>   				  amdgpu_amdkfd_restore_userptr_worker);
>   
> +		info->process_id = process->id;
> +
>   		*process_info = info;
>   	}
>