Re: [PATCH V2 1/2] drm/amdgpu/userq: only accept doorbell BOs as queue doorbell

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

On 9/30/26 11:38, Jesse Zhang wrote:
> amdgpu_userq_get_doorbell_index() pins any BO passed as doorbell_handle
> into the doorbell domain. For a regular BO, TTM moves it with
> ttm_bo_move_null() and its contents are lost. Since an importer on the
> same device gets the exporter's GEM object, a client can do this to a
> buffer another client shared with it.
> 
> Reject BOs that were not created in the doorbell domain.
> 
> v2: drop the error message, use a local abo variable (Christian)
> 
> Fixes: f09c1e6077ab ("drm/amdgpu: generate doorbell index for userqueue")
> Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>

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

> ---
>  drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c | 13 ++++++++++++-
>  1 file changed, 12 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c
> index 89e03d3a930c..5a922f283743 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c
> @@ -469,6 +469,7 @@ amdgpu_userq_get_doorbell_index(struct amdgpu_userq_mgr *uq_mgr,
>  	u64 doorbell_index;
>  	struct drm_gem_object *gobj;
>  	struct amdgpu_userq_obj *db_obj = db_info->db_obj;
> +	struct amdgpu_bo *abo;
>  	int r, db_size;
>  
>  	gobj = drm_gem_object_lookup(filp, db_info->doorbell_handle);
> @@ -477,7 +478,17 @@ amdgpu_userq_get_doorbell_index(struct amdgpu_userq_mgr *uq_mgr,
>  		return -EINVAL;
>  	}
>  
> -	db_obj->obj = amdgpu_bo_ref(gem_to_amdgpu_bo(gobj));
> +	/*
> +	 * Pinning a regular BO into the doorbell domain would discard its
> +	 * contents, possibly those of a buffer shared by another client.
> +	 */
> +	abo = gem_to_amdgpu_bo(gobj);
> +	if (!(abo->preferred_domains & AMDGPU_GEM_DOMAIN_DOORBELL)) {
> +		drm_gem_object_put(gobj);
> +		return -EINVAL;
> +	}
> +
> +	db_obj->obj = amdgpu_bo_ref(abo);
>  	drm_gem_object_put(gobj);
>  
>  	r = amdgpu_bo_reserve(db_obj->obj, true);