On Mon, Oct 21, 2024 at 09:22:48AM -0300, Jason Gunthorpe wrote:
> On Sat, Oct 19, 2024 at 06:42:30PM -0700, Nicolin Chen wrote:
> > > But be mindful of this abort, it doesn't want to be inside the lock if
> > > it also gets the lock.. fail_nth should be updated to cover these new
> > > ioctls to look for tricky things like that
> >
> > I added an abort() beside destroy():
> >
> > +void iommufd_vdevice_abort(struct iommufd_object *obj)
> > +{
> > + struct iommufd_vdevice *old, *vdev =
> > + container_of(obj, struct iommufd_vdevice, obj);
> > + struct iommufd_viommu *viommu = vdev->viommu;
> > + struct iommufd_device *idev = vdev->idev;
> > +
> > + lockdep_assert_not_held(&idev->igroup->lock);
>
> ???
Oops. That's a typo from auto-completion.
> > +void iommufd_vdevice_destroy(struct iommufd_object *obj)
> > +{
> > + struct iommufd_vdevice *vdev =
> > + container_of(obj, struct iommufd_vdevice, obj);
> > +
> > + mutex_lock(&vdev->idev->igroup->lock);
> > + iommufd_vdevice_abort(obj);
>
> When we get it here??
LOCKDEP wasn't enabled when I tested it.. Just fixed and retested.
Thanks
Nicolin
