On Wed, Feb 26, 2025 at 04:33:41PM +0100, Thomas Hellström wrote:
> If a userptr vma subject to prefetching was already invalidated
> or invalidated during the prefetch operation, the operation would
> repeatedly return -EAGAIN which would typically cause an infinite
> loop.
>
> Validate the userptr to ensure this doesn't happen.
>
> Fixes: 5bd24e78829a ("drm/xe/vm: Subclass userptr vmas")
> Fixes: 617eebb9c480 ("drm/xe: Fix array of binds")
> Cc: Matthew Brost <matthew.brost@xxxxxxxxx>
> Cc: <stable@xxxxxxxxxxxxxxx> # v6.9+
> Suggested-by: Matthew Brost <matthew.brost@xxxxxxxxx>
> Signed-off-by: Thomas Hellström <thomas.hellstrom@xxxxxxxxxxxxxxx>
> ---
> drivers/gpu/drm/xe/xe_vm.c | 9 ++++++++-
> 1 file changed, 8 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c
> index 996000f2424e..4c1ca47667ad 100644
> --- a/drivers/gpu/drm/xe/xe_vm.c
> +++ b/drivers/gpu/drm/xe/xe_vm.c
> @@ -2307,7 +2307,14 @@ static int vm_bind_ioctl_ops_parse(struct xe_vm *vm, struct drm_gpuva_ops *ops,
> }
> case DRM_GPUVA_OP_UNMAP:
> case DRM_GPUVA_OP_PREFETCH:
> - /* FIXME: Need to skip some prefetch ops */
The UNMAP case statement is falling through to pretech case which I
believe is not the intent.
So I think:
case DRM_GPUVA_OP_UNMAP:
xe_vma_ops_incr_pt_update_ops(vops, op->tile_mask);
break;
case DRM_GPUVA_OP_PREFETCH:
<new code>
Matt
> + vma = gpuva_to_vma(op->base.prefetch.va);
> +
> + if (xe_vma_is_userptr(vma)) {
> + err = xe_vma_userptr_pin_pages(to_userptr_vma(vma));
> + if (err)
> + return err;
> + }
> +
> xe_vma_ops_incr_pt_update_ops(vops, op->tile_mask);
> break;
> default:
> --
> 2.48.1
>
