4

Some time ago it was noticed that Win10 1803 WM's would give high CPU usage on the host e.g. 25%, even though client CPU usage 0%. It was solved by having these settings in your virt XML file:

 <hyperv>
   <relaxed state='on'/>
   <vapic state='on'/>
   <spinlocks state='on' retries='8191'/>
   <synic state='on'/> 
   <stimer state='on'/>
 </hyperv>

 <clock offset='localtime'>
   <timer name='rtc' tickpolicy='catchup'/>
   <timer name='pit' tickpolicy='delay'/>
   <timer name='hpet' present='no'/>
   <timer name='hypervclock' present='yes'/>
 </clock>

But the problem is new back. I've moved from Debian stretch (QEMU 2.8.1 / Kernel 4.9) to Buster (QEMU 3.1.0 / Kernel 4.19) and Windows 10 1809.
The problem might be related to I can no longer use <synic state='on'/> on Buster which I could on Stretch on the same hardware.
When I try to start the VM i get this error

error: internal error: process exited while connecting to monitor: Hyper-V SynIC (requested by 'hv-synic' cpu flag) requires Hyper-V VP_INDEX ('hv-vpindex')
2019-08-06T13:29:14.114943Z qemu-system-x86_64: kvm_init_vcpu failed: Function not implemented

When I run lscpu I don't have any hv* flags at all. But again, it worked on Stretch on the same CPU.

1
  • Compated the XML with an old VM which support the <synic state='on'/>. The difference is machine='pc-i440fx-2.8' and machine='pc-q35-3.1'....need to do some more investigation.
    – MrCalvin
    Commented Aug 6, 2019 at 14:47

1 Answer 1

7

At least on QEMU 3.1.0 / Kernel 4.19 it seems QEMU machine types > pc-i440fx-3.0 doesn't support synIC, which is needed on Windows 10 guest build >= 1803 to not have issues with high host CPU usage.

I've tested following machine types:
pc-i440fx-2.8 (OK)
pc-i440fx-3.0 (OK)
pc-i440fx-3.1 (Fail)
pc-q35-3.1 (Fail)

The XML element in the Virt XML file:

</features>
   ...
   <hyperv>
      <relaxed state='on'/>
      <vapic state='on'/>
      <spinlocks state='on' retries='8191'/>
      <synic state='on'/>
      <stimer state='on'/>
   </hyperv>
   ...
</features>

<clock offset='localtime'>
   <timer name='rtc' tickpolicy='catchup'/>
   <timer name='pit' tickpolicy='delay'/>
   <timer name='hpet' present='no'/>
   <timer name='hypervclock' present='yes'/>
</clock>

UPDATE:
It turned out to be my XML settings that was the issue!
The following will also make your new QEMU machine types run smoothly without high host CPU usage:

<hyperv>
   <relaxed state='on'/>
   <vpindex state='on'/>
   <synic state='on'/>
   <stimer state='on'/>
</hyperv>

<clock offset='localtime'>
   <timer name='rtc' tickpolicy='catchup'/>
   <timer name='pit' tickpolicy='delay'/>
   <timer name='hpet' present='no'/>
   <timer name='hypervclock' present='yes'/>
</clock>

A big thank you to Vitaly Kuznetsov for being so patient with me and made me realize the misconfiguration.

2
  • Thank you for the above share - it was a massive improvement to some similar problems I was experiencing.
    – Mark
    Commented Feb 1, 2020 at 12:53
  • this is also working for machine: <type arch="x86_64" machine="pc-q35-rhel7.6.0">hvm</type>.
    – bebbo
    Commented Oct 16, 2020 at 10:31

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .