Re: [RFC PATCH] PCI: Introduce new device binding path using pci_dev.driver_override

[Alex Williamson <alex.williamson-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>]

On Thu, 2014-04-10 at 20:52 +0000, Stuart Yoder wrote:
> > Another advantage to this approach is that we can specify a driver
> > override to force a specific binding or prevent any binding.  For
> > instance when an IOMMU group is exposed to userspace through VFIO
> > we require that all devices within that group are owned by VFIO.
> > However, devices can be hot-added into an IOMMU group, in which case
> > we want to prevent the device from binding to any driver (preferred
> > driver = "none") or perhaps have it automatically bind to vfio-pci.
> > With driver_override it's a simple matter for this field to be set
> > internally when the device is first discovered to prevent driver
> > matches.
> 
> Alex, I think the patch looks good, but did have a question about
> the hot-adding device scenario above.   How do you get the "field to
> be set internally when the device is first discovered"?   What
> hook is there to do that prior to the driver match processing?

Devices already need to be added to iommu groups prior to driver probe
which happens either by the iommu driver calling bus_register_notifier()
and looking for BUS_NOTIFY_ADD_DEVICE or they can register an add_device
callback in their iommu_ops and let the iommu subsystem register the
notifier.  When a device is then added to an iommu group, there is yet
another notifier.  vfio calls iommu_group_register_notifier() and gets
told via IOMMU_GROUP_NOTIFY_ADD_DEVICE when a device is added to a
group.  We currently have a WARN_ON when a device is added to a live
group - vfio_group_nb_add_dev().  I expect we'd just replace that with
detecting the bus_type of the device and allocating a string with the
proper vfio driver for the bus.  Thanks,

Alex