[vfio-users] IGD Passthrough - does not work with intel kabylake

[nicolas prochazka]

Yes , intel driver, windows drivers, on 32bits / 64bits ( drivers igd
skylake , kabylake ) ,
it seems to be a bad news, we are a lot of guys, very happy with igd
passthrough for performance, and me for pc admin :)

Regards,
Nicolas

2017-02-17 18:43 GMT+01:00 Alex Williamson <alex.williamson at redhat.com>:

> On Fri, 17 Feb 2017 18:31:55 +0100
> nicolas prochazka <prochazka.nicolas at gmail.com> wrote:
>
> > Ok,
> > however, it works with linux , so it"s a windows drivel issue ?
>
> Linux working would seem to be yet more indication that Intel has
> broken their Windows driver, or at least hint that the virtualization
> stack is working.  Have you tried all versions of the driver that
> support this hardware?  Thanks,
>
> Alex
>
> > 2017-02-17 16:32 GMT+01:00 Alex Williamson <alex.williamson at redhat.com>:
> >
> > > On Fri, 17 Feb 2017 10:43:57 +0100
> > > nicolas prochazka <prochazka.nicolas at gmail.com> wrote:
> > >
> > > > hello again,
> > > > How can help you to try to analyse this failure ?  some debug option
> ?
> > >
> > > A Code 43 is the Intel driver finding some error or configuration that
> > > it doesn't like.  This is unexpected for IGD and I have no recipe for
> > > debugging it.  Theoretically anything Broadwell or newer should be
> > > supported in UPT mode (see docs/igd-assign.txt).  This involves running
> > > the assigned IGD device as secondary to a primary emulated graphics in
> > > the VM.  KVMGT, aka vGPUs, should also work with Broadwell and newer,
> > > it's possible Intel has broken some aspect of direct assignment in
> > > enabling this new functionality.  Thanks,
> > >
> > > Alex
> > >
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/vfio-users/attachments/20170217/ed0d6c2c/attachment.htm>