I confirm that the vendor_id works great with libvirt, however all the other options included in that commit does not. Virsh will fail to start the host with runtime, vpindex and reset (though it does work with qemu command-line). Should that deserve a bug report?
2016-04-15 7:45 GMT+02:00 thibaut noah <[email protected]>: > Oops, i read too quickly... (it's late) > I run libvirt-git from the aur, i should have this commit thought it seems > i don't, will check that with arch people, thanks for the correct link. > > 2016-04-15 6:21 GMT+02:00 Alex Williamson <[email protected]>: > >> On Thu, Apr 14, 2016 at 9:07 PM, thibaut noah <[email protected]> >> wrote: >> >>> On the libvirt side, reading the documentation again i bumped into this : >>> >>> <vendor_id state='on' value='KVM Hv'/> >>> >>> >>> https://libvirt.org/formatdomain.html#elementsFeatures >>> >>> Anybody knowing why we have to rely on qemu since apparently libvirt is >>> suppose to support this since 2012 ? >>> https://www.redhat.com/archives/libvir-list/2012-June/msg00917.html >>> >>> I have to admit i'm curious. >>> >> >> Oh wow, but you picked the wrong patch, that one has to do with cpuid >> vendor, this is the one you're after: >> >> >> http://libvirt.org/git/?p=libvirt.git;a=commit;h=7068b56c8526667750d8cb91116e9b40382422f3 >> >> That's not even 3 weeks old, so you'll need a very, very new libvirt. >> This already exists in the fedora virt-preview repo and it works great. >> Thanks for the heads-up. >> > >
_______________________________________________ vfio-users mailing list [email protected] https://www.redhat.com/mailman/listinfo/vfio-users
