[PATCH] drm/xe: check mask in xe_wait_user_fence_ioctl()

[Francois Dugast]

On Fri, Dec 08, 2023 at 12:34:23PM +0000, Matthew Brost wrote:
> On Fri, Dec 08, 2023 at 02:34:27PM +0000, Francois Dugast wrote:
> > Check the drm_xe_wait_user_fence mask is one of the allowed values
> > DRM_XE_UFENCE_WAIT_MASK_* provided in xe_drm.h.
> > 
> > Signed-off-by: Francois Dugast <francois.dugast at intel.com>
> > ---
> >  drivers/gpu/drm/xe/xe_wait_user_fence.c | 6 ++++++
> >  1 file changed, 6 insertions(+)
> > 
> > diff --git a/drivers/gpu/drm/xe/xe_wait_user_fence.c b/drivers/gpu/drm/xe/xe_wait_user_fence.c
> > index 4d5c2555ce41..a73139fe3fe0 100644
> > --- a/drivers/gpu/drm/xe/xe_wait_user_fence.c
> > +++ b/drivers/gpu/drm/xe/xe_wait_user_fence.c
> > @@ -164,6 +164,12 @@ int xe_wait_user_fence_ioctl(struct drm_device *dev, void *data,
> >  	if (XE_IOCTL_DBG(xe, args->num_engines > XE_HW_ENGINE_MAX_INSTANCE))
> >  		return -EINVAL;
> >  
> > +	if (XE_IOCTL_DBG(xe, args->mask != DRM_XE_UFENCE_WAIT_MASK_U8 &&
> > +			 args->mask != DRM_XE_UFENCE_WAIT_MASK_U16 &&
> > +			 args->mask != DRM_XE_UFENCE_WAIT_MASK_U32 &&
> > +			 args->mask != DRM_XE_UFENCE_WAIT_MASK_U64))
> > +		return -EINVAL;
> > +
> 
> Why do we need this check? Looking at do_compare() I believe any mask
> value actually works. Would it be better to just delete
> DRM_XE_UFENCE_WAIT_MASK* from the uAPI?

Sure let's do this.

> 
> Matt
> 
> >  	if (!no_engines) {
> >  		err = copy_from_user(eci, user_eci,
> >  				     sizeof(struct drm_xe_engine_class_instance) *
> > -- 
> > 2.34.1
> >