Re: [PATCH] drm/xe: Don't fall through on DRM_GPUVA_OP_DRIVER in op_check_svm_userptr

[Matthew Auld <matthew.auld@intel.com>]

Hi,

On 10/07/2026 15:12, Mika Kuoppala wrote:
> The DRM_GPUVA_OP_DRIVER operation is used by eudebug (and potentially
> other driver ops). Because the break was inside the #if GPUSVM block,
> we fell through and printed the "NOT POSSIBLE" warning.

I think throwing a warn with "NOT POSSIBLE" was maybe not the wrong 
thing wrt SVM and the current driver code, in the sense that there would 
be a bug somewhere if we reach here without GPUSVM support?

Are you triggering this only with upcoming eudebug? i.e there is a new 
DRM_GPUVA_OP_DRIVER user, which is not SVM?

> 
> Fixes: 7477c4bd20dc ("drm/xe/pt: unify xe_pt_svm_pre_commit with userptr")

I think maybe drop the fixes, and tweak the commit to make it clear that 
this is just prep work for a new user?

> Cc: Matthew Auld <matthew.auld@intel.com>
> Signed-off-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
> ---
>   drivers/gpu/drm/xe/xe_pt.c | 4 ++--
>   1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/gpu/drm/xe/xe_pt.c b/drivers/gpu/drm/xe/xe_pt.c
> index 7111b5d0f6aa..f5ce1fcf484c 100644
> --- a/drivers/gpu/drm/xe/xe_pt.c
> +++ b/drivers/gpu/drm/xe/xe_pt.c
> @@ -1547,8 +1547,8 @@ static int op_check_svm_userptr(struct xe_vm *vm, struct xe_vma_op *op,
>   			err = vma_check_userptr(vm, gpuva_to_vma(op->base.prefetch.va), pt_update);
>   		}
>   		break;
> -#if IS_ENABLED(CONFIG_DRM_XE_GPUSVM)
>   	case DRM_GPUVA_OP_DRIVER:
> +#if IS_ENABLED(CONFIG_DRM_XE_GPUSVM)
>   		if (op->subop == XE_VMA_SUBOP_MAP_RANGE) {
>   			struct xe_svm_range *range = op->map_range.range;
>   
> @@ -1561,8 +1561,8 @@ static int op_check_svm_userptr(struct xe_vm *vm, struct xe_vma_op *op,
>   				return -EAGAIN;
>   			}
>   		}
> -		break;
>   #endif
> +		break;
>   	default:
>   		drm_warn(&vm->xe->drm, "NOT POSSIBLE");
>   	}