> From: Liu, Yi L <yi.l.liu@xxxxxxxxx>
> Sent: Wednesday, February 22, 2023 3:44 PM
>
> > From: Tian, Kevin <kevin.tian@xxxxxxxxx>
> > Sent: Wednesday, February 22, 2023 3:40 PM
> >
> > > From: Liu, Yi L <yi.l.liu@xxxxxxxxx>
> > > Sent: Tuesday, February 21, 2023 11:48 AM
> > >
> > > +
> > > +void vfio_device_cdev_close(struct vfio_device_file *df)
> > > +{
> > > + struct vfio_device *device = df->device;
> > > +
> > > + mutex_lock(&device->dev_set->lock);
> > > + if (!smp_load_acquire(&df->access_granted)) {
> >
> > there is no contention with another one changing this flag at this
> > point so directly accessing it is fine.
>
> make sense.
>
> > but actually should check device->open_count as v3 does. Otherwise
> > the last error on copy_to_user() in ioctl_bind_iommufd() simply returns
> > here given df->access_granted hasn't been set but .open_device()
> > has been completed.
>
> If copy_to_user() failed, vfio_device_close() would be called in the
> error path. Then this close function just returns.
>
yeah, I misread it.
