[Pkg-libvirt-maintainers] Bug#892471: Bug#892471: libvirt daemon on stretch not support AMD Epyc Processor

Mon Mar 12 11:12:44 UTC 2018

I did some tests with the buster packages; on my stretch I installed the 
following packages:

ipxe-qemu_1.0.0+git-20161027.b991c67-1_all.deb
libsdl2-2.0-0_2.0.8+dfsg1-1_amd64.deb
libvirt0_4.0.0-2_amd64.deb
libvirt-clients_4.0.0-2_amd64.deb
libvirt-daemon_4.0.0-2_amd64.deb
libvirt-daemon-system_4.0.0-2_amd64.deb
libvirt-glib-1.0-0_1.0.0-1_amd64.deb
libxss1_1.2.2-1+b2_amd64.deb
python-libvirt_4.0.0-1_amd64.deb
qemu-kvm_2.11+dfsg-1_amd64.deb
qemu-system-common_2.11+dfsg-1_amd64.deb
qemu-system-x86_2.11+dfsg-1_amd64.deb
qemu-utils_2.11+dfsg-1_amd64.deb

and now I can confirm that if in the host system if the CPU is 
configured "copy host CPU configuration" the system now recognizes the 
EPYC cpu (but if I look in the list the EPYC model from libvirt is not 
displayed, it does not matter).
Then, with the guest system started, the EPYC CPU model is displayed, 
while in the guest lscpu appears:

GUEST LSCPU

Architecture: x86_64
CPU op-mode(s):        32-bit, 64-bit
Byte Order:            Little Endian
CPU(s):                6
On-line CPU(s) list:   0-5
Thread(s) per core:    1
Core(s) per socket:    1
Socket(s):             6
NUMA node(s):          1
Vendor ID:             AuthenticAMD
CPU family:            23
Model:                 1
Model name:            AMD EPYC Processor
Stepping:              2
CPU MHz:               2199.996
BogoMIPS:              4399.99
Virtualization:        AMD-V
Hypervisor vendor:     KVM
Virtualization type:   full
L1d cache:             64K
L1i cache:             64K
L2 cache:              512K
NUMA node0 CPU(s):     0-5

GUESTDMESG

Mar 12 11:40:58 guest kernel: [    0.000000] kvm-clock: Using msrs 
4b564d01 and 4b564d00
Mar 12 11:40:58 guest kernel: [    0.000000] kvm-clock: cpu 0, msr 
1:9fff1001, primary cpu clock
Mar 12 11:40:58 guest kernel: [    0.000000] Zone ranges:
Mar 12 11:40:58 guest kernel: [    0.000000]   DMA      [mem 
0x00001000-0x00ffffff]
Mar 12 11:40:58 guest kernel: [    0.000000]   DMA32    [mem 
0x01000000-0xffffffff]
Mar 12 11:40:58 guest kernel: [    0.000000]   Normal   [mem 
0x100000000-0x19fffffff]
Mar 12 11:40:58 guest kernel: [    0.000000] Movable zone start for each 
node
Mar 12 11:40:58 guest kernel: [    0.000000] Early memory node ranges
Mar 12 11:40:58 guest kernel: [    0.000000]   node   0: [mem 
0x00001000-0x0009efff]
Mar 12 11:40:58 guest kernel: [    0.000000]   node   0: [mem 
0x00100000-0xdfff6fff]
Mar 12 11:40:58 guest kernel: [    0.000000]   node   0: [mem 
0x100000000-0x19fffffff]
Mar 12 11:40:58 guest kernel: [    0.000000] ACPI: PM-Timer IO Port: 0xb008
Mar 12 11:40:58 guest kernel: [    0.000000] KVM setup async PF for cpu 0
Mar 12 11:40:58 guest kernel: [    0.000000] kvm-stealtime: cpu 0, msr 
19fc0fc80
Mar 12 11:40:58 guest kernel: [    0.008000] kvm-clock: cpu 1, msr 
1:9fff1041, secondary cpu clock
Mar 12 11:40:58 guest kernel: [    0.049786] KVM setup async PF for cpu 1
Mar 12 11:40:58 guest kernel: [    0.049786] kvm-stealtime: cpu 1, msr 
19fc4fc80
Mar 12 11:40:58 guest kernel: [    0.049858]  #2
Mar 12 11:40:58 guest kernel: [    0.008000] kvm-clock: cpu 2, msr 
1:9fff1081, secondary cpu clock
Mar 12 11:40:58 guest kernel: [    0.068519]  #3
Mar 12 11:40:58 guest kernel: [    0.068442] KVM setup async PF for cpu 2
Mar 12 11:40:58 guest kernel: [    0.068442] kvm-stealtime: cpu 2, msr 
19fc8fc80
Mar 12 11:40:58 guest kernel: [    0.008000] kvm-clock: cpu 3, msr 
1:9fff10c1, secondary cpu clock
Mar 12 11:40:58 guest kernel: [    0.085417] KVM setup async PF for cpu 3
Mar 12 11:40:58 guest kernel: [    0.085417] kvm-stealtime: cpu 3, msr 
19fccfc80
Mar 12 11:40:58 guest kernel: [    0.085491]  #4
Mar 12 11:40:58 guest kernel: [    0.008000] kvm-clock: cpu 4, msr 
1:9fff1101, secondary cpu clock
Mar 12 11:40:58 guest kernel: [    0.102388] KVM setup async PF for cpu 4
Mar 12 11:40:58 guest kernel: [    0.102388] kvm-stealtime: cpu 4, msr 
19fd0fc80
Mar 12 11:40:58 guest kernel: [    0.102471]  #5
Mar 12 11:40:58 guest kernel: [    0.008000] kvm-clock: cpu 5, msr 
1:9fff1141, secondary cpu clock
Mar 12 11:40:58 guest kernel: [    0.119380] x86: Booted up 1 node, 6 CPUs
Mar 12 11:40:58 guest kernel: [    0.119382] smpboot: Total of 6 
processors activated (26399.95 BogoMIPS)
Mar 12 11:40:58 guest kernel: [    0.119369] KVM setup async PF for cpu 5
Mar 12 11:40:58 guest kernel: [    0.119369] kvm-stealtime: cpu 5, msr 
19fd4fc80

There remains only a defect on the host machine that once started 
appears this error, can you help me to understand what could be?

HOST DMESG

Mar 12 11:40:57 host kernel: [ 1529.452338] kvm [13298]: vcpu1, guest 
rIP: 0xffffffff81053b72 unhandled rdmsr: 0x3a
Mar 12 11:40:57 host kernel: [ 1529.452426] kvm [13298]: vcpu1, guest 
rIP: 0xffffffff81053b72 unhandled rdmsr: 0xd90

However, in the future this solution may not work because the gap 
between the buster and the stretch may be wider, so the dependency chain 
may break.

Diego

Il 12/03/2018 09:49, Raschi Diego ha scritto:
>
> I agree that the change should be made for both libvirt and qemu, in 
> fact I know that this escalation was made.
>
> https://bugzilla.redhat.com/show_bug.cgi?id=1445834
>
> I could use the emulation of another cpu to remedy the problem, but 
> instead of having a supercpu without introducing transcoding of the 
> processor I would have very degraded performance.
>
> Another solution would be that as you mentioned to use buster 
> packages, I try to install the new packages and I hope I do not run 
> into an endless series of installs for library dependencies.
>
> In the meantime, I thank you for your time and I will keep you updated.
> Diego
>
>
> Il 11/03/2018 12:55, Guido Günther ha scritto:
>> On Fri, Mar 09, 2018 at 03:17:57PM +0100, Raschi Diego wrote:
>>> I apologize for the inconvenience, I made a copy mistake
>>>
>>> https://bugzilla.redhat.com/show_bug.cgi?id=1481252
>> As far as I read this it's not related to the problem you're seeing and
>> if we add this qemu woul need support as well:
>>
>>      https://bugzilla.redhat.com/show_bug.cgi?id=1445834
>>
>> but is should work without it nevertheless using another CPU model.
>>
>> This:
>>
>>     kernel: [    2.623548] kvm: no hardware support
>>
>> means likely that virtualization is not enabled on your machine.  Can
>> you run accelerated VMs using only qemu but not libvirt? Does it work
>> with debian buster?
>> Cheers,
>>   -- Guido
>

AVVISO: Questo messaggio e' inviato da Banca Sammarinese di Investimento S.p.A. e puo' contenere informazioni di carattere estremamente riservato e confidenziale. Qualsivoglia utilizzo non autorizzato del contenuto di questo messaggio costituisce violazione dell'obbligo di non prendere cognizione della corrispondenza tra altri soggetti, salvo piu' grave illecito ed espone il responsabile alle relative conseguenze.
Qualora non foste i destinatari, vogliate immediatamente informarci con lo stesso mezzo ed eliminare il messaggio, con gli eventuali allegati, senza trattenerne copia. Grazie per la collaborazione.

DISCLAIMER: This e-mail contains confidential information, which is intended only for the use of the recipient(s) named above. If you have received this communication in error, please notify the sender immediately via e-mail and return the entire message. Thank you for your assistance.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.alioth.debian.org/pipermail/pkg-libvirt-maintainers/attachments/20180312/cbb06e57/attachment-0001.html>