[vfio-users] KVM + QEMU + GPU passthrough (Windows7 Guest)

[Eddie Yen]

Because vfio-pci.ids can make vfio directly access GPU with VID:PID. You
have to write another script to make VFIO mount GPU manually if using
pci-stub.

Also, on Fedora, directly using vfio-pci can let you install same GPUs and
work on host and guest at the same time.
For example, you can installed 2 x R9-285 on your host and let one working
in host and another one mount into guest.
It can't do this if using pci-stub only, because pci-stub only works on
VID:PID.

2015-09-27 19:07 GMT+08:00 José Ramón Muñoz <koalinux at gmail.com>:

>         Hi,
>
>         Actually it's more funny, it really works like you said, but, for
> any
> reason vfio-pci.ids doesn't prevent radeon and my xorg server to take the
> p-t
> gpu, and do their business. So using pci-stub.ids I can avoid ioh3420.
>
>         Then my question is, how vfio-pci.ids is different to
> pci-stub.ids, and why
> it works better with the latest?
>
>         Thanks!
>
>         José.
>
> On Saturday 26 September 2015 02:04:48 Blank Field wrote:
> > That messages about FM2 platform made me laugh.
> > My current W10+OVMF system outputs video until starting windows spinning
> > logo, then the video driver crashes.
> > The thing is: this behavior is dependent on OVMF version, guest driver
> > version and it is "phantom" - if the OS failed to show video output on
> the
> > first boot, a simple VM reboot will solve this.
> > Too bad i lack the skill to debug it further and can't afford building a
> > bunch of test configs.
> > On Sep 25, 2015 11:38 PM, "Alex Williamson" <alex.l.williamson at gmail.com
> >
> >
> > wrote:
> > > On Fri, Sep 25, 2015 at 12:48 PM, Jose Ramon Muñoz Pekkarinen <
> > >
> > > koalinux at gmail.com> wrote:
> > >>     Hi,
> > >>
> > >>     I'm sorry to report that, at least for FM2 platform, if there is
> no
> > >>
> > >> ioh3420, it doesn't work at all, w7 and w10 stalls in the windows
> logo on
> > >> 440fx. Well w10 doesn't work anyways, but w7 does.
> > >
> > > More and more it seems like if you get anything working on an FM2
> > > platform, you should count yourself lucky, but that doesn't mean that
> that
> > > combination of QEMU components were ever intended to be put together or
> > > make any kind of sense from a PCI perspective.
> > >
> > > _______________________________________________
> > > vfio-users mailing list
> > > vfio-users at redhat.com
> > > https://www.redhat.com/mailman/listinfo/vfio-users
>
>
> _______________________________________________
> vfio-users mailing list
> vfio-users at redhat.com
> https://www.redhat.com/mailman/listinfo/vfio-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/vfio-users/attachments/20150927/6bbb9c09/attachment.htm>