On Wed, Jul 24, 2019 at 18:48:38 +0200, Vitaly Kuznetsov wrote: > Hi, > > when some Hyper-V enlightenments are added to the domain configuration, > e.g. > > <features> > ... > <hyperv> > .. > <vpindex state='on'/> > </hyperv> > ... > > I'm observing the following: > > # virsh create /etc/libvirt/qemu/win2k16.xml > error: Failed to create domain from /etc/libvirt/qemu/win2k16.xml > error: unsupported configuration: host doesn't support hyperv 'vpindex' feature > > the feature is, of course, supported in KVM/QEMU. > > I've bisected the problem to the particular commit: > > commit 63acb7bfd56f117309e4fcaf438639d4d7bc7dcb > Author: Jiri Denemark <jdenemar(a)redhat.com&gt; > Date: Mon Jun 17 23:38:46 2019 +0200 > > qemu_process: Prefer generic qemuMonitorGetGuestCPU > > Any ideas? Is this with QEMU 4.0 (or older) or a current git, i.e., almost 4.1?

On Thu, Jul 25, 2019 at 11:09:28 +0200, Vitaly Kuznetsov wrote: > Jiri Denemark <jdenemar(a)redhat.com&gt; writes: > > > On Wed, Jul 24, 2019 at 18:48:38 +0200, Vitaly Kuznetsov wrote: > >> Hi, > >> > >> when some Hyper-V enlightenments are added to the domain configuration, > >> e.g. > >> > >> <features> > >> ... > >> <hyperv> > >> .. > >> <vpindex state='on'/> > >> </hyperv> > >> ... > >> > >> I'm observing the following: > >> > >> # virsh create /etc/libvirt/qemu/win2k16.xml > >> error: Failed to create domain from /etc/libvirt/qemu/win2k16.xml > >> error: unsupported configuration: host doesn't support hyperv 'vpindex' feature > >> > >> the feature is, of course, supported in KVM/QEMU. > >> > >> I've bisected the problem to the particular commit: > >> > >> commit 63acb7bfd56f117309e4fcaf438639d4d7bc7dcb > >> Author: Jiri Denemark <jdenemar(a)redhat.com&gt; > >> Date: Mon Jun 17 23:38:46 2019 +0200 > >> > >> qemu_process: Prefer generic qemuMonitorGetGuestCPU > >> > >> Any ideas? > > > > Is this with QEMU 4.0 (or older) or a current git, i.e., almost 4.1? > > QEMU-4.0 is definitely affected. Hmm, QEMU 4.0 does not support unavailable-features CPU property, which means the mentioned commit is a no-op. And I can't reproduce your issue in this case. However, QEMU 4.1.0-rc1 supports unavailable-features and I can reproduce the bug. Looks like something is wrong with the new way of checking for enabled CPU features.