Re: updating Xen public headers

[Christoph Egger <Christoph_Egger%gmx.de@localhost>]

On 12/07/11 10:56, Manuel Bouyer wrote:
> On Tue, Dec 06, 2011 at 12:22:06PM +0100, Christoph Egger wrote:
> > New interfaces for PV drivers:
> >
> > - Xen transcedent memory
> > - USB IO
> > - SCSI IO
> >
> > PCI IO improvements for PV domU:
> >
> > - PCI MSI support
> > - PCI Express AER support
> >
> > New features:
> >
> > - xen honors flags to be placed into guest kernel available
> >    pte bits if enabled (for grant table)
> > - support for 128 vcpus
> >    (old interface is still present and supports up to 32 vcpus)
> > - PCI passthrough: new hypercalls to support SR-IOV
> > - new hypercall for physical cpu hotplugging
> > - new hypercall for physical page offlining
> > - fixes to compile with clang
> > - machine check recovery mechanism
>
> OK, looks fine. But we have to be carefull, when we'll start using
> these features, that a dom0/domU with them can still work with a domU/dom0
> that doesn't have them

The USB and SCSI IO interfaces need to be implemented in a frontend and
backend PV driver. The interaction with Xen is as minimal as with
the disk and network PV driver.

Hypercalls unknown to a xen kernel return ENOSYS. That means there are
three ways to handle the backward compatibility:

- at compile time by introducing #ifdef XEN4 ... #endif
- at runtime by checking the hypercall return value
- at runtime query the xen version we are on and enable/disable
  corresponding features.

Christoph