Re: [RFC 03/12] vfio: Accept vfio device file in the driver facing kAPI

[Jason Gunthorpe <jgg@nvidia.com>]

On Wed, Dec 21, 2022 at 04:07:42AM +0000, Tian, Kevin wrote:
> > From: Liu, Yi L <yi.l.liu@intel.com>
> > Sent: Monday, December 19, 2022 4:47 PM
> > 
> > +static bool vfio_device_enforced_coherent(struct vfio_device *device)
> > +{
> > +	bool ret;
> > +
> > +	if (!vfio_device_try_get_registration(device))
> > +		return true;
> > +
> > +	ret = device_iommu_capable(device->dev,
> > +
> > IOMMU_CAP_ENFORCE_CACHE_COHERENCY);
> > +
> > +	vfio_device_put_registration(device);
> > +	return ret;
> > +}
> 
> This probably needs an explanation that recounting is required because
> this might be called before vfio_device_open() is called to hold the
> count.

How does that happen? The only call site already has a struct file and
the struct file implicitly holds the registration. This should be
removed.

Just inline the whole thing and make it clear:

       device = vfio_device_from_file(file);
       if (device)
       	  return device_iommu_capable(device->dev,
                                  IOMMU_CAP_ENFORCE_CACHE_COHERENCY);

Jason