[PVE-User] Bug with cpu0 unhandled wrmsr

Mikhail m at plus-plus.su
Tue Jul 8 12:16:02 CEST 2014


Hello,

Proxmox node running pve-manager/3.1-21/93bf03d4 (kernel:
2.6.32-26-pve), CPU: Intel(R) Xeon(R) CPU E5-2620 v2 @ 2.10GHz.

Just rolled our new VM from latest Debian 7 installation ISO image. I
chose SandyBridge as a CPU for the VM. After VM boots the following
occurs in dmesg on host node:

kvm: 2984: cpu0 unhandled rdmsr: 0x345
kvm: 2984: cpu0 unhandled wrmsr: 0x680 data 0
kvm: 2984: cpu0 unhandled wrmsr: 0x6c0 data 0
kvm: 2984: cpu0 unhandled wrmsr: 0x681 data 0
kvm: 2984: cpu0 unhandled wrmsr: 0x6c1 data 0
kvm: 2984: cpu0 unhandled wrmsr: 0x682 data 0
kvm: 2984: cpu0 unhandled wrmsr: 0x6c2 data 0
kvm: 2984: cpu0 unhandled wrmsr: 0x683 data 0
kvm: 2984: cpu0 unhandled wrmsr: 0x6c3 data 0
kvm: 2984: cpu0 unhandled wrmsr: 0x684 data 0

Both host node and VM continue to run with no problems, however after
several hours Proxmox node locks itself and does not respond (resulting
all VMs to not respond as well) - no errors on the screen, nothing in
the logs after reboot.

My first thought was faulty hardware, but then I searched over other
mailing lists and Redhat's members also revealed similar issues.
Proxmox users also asked this in past:

http://forum.proxmox.com/threads/17291-cpu0-unhandled-wrmsr-0x680-data-0-What-does-it-mean

but nobody responded.

Anyway, I tried to reset the host node and it locked after several hours
again.
After changing VM CPU to "kvm64" the problem has gone so far (at least
I'm seeing 2 days uptime now).

Is there someone else seeing same behavior on SandyBridge VM CPU and
Intel as a host machine CPU?

Mikhail.



More information about the pve-user mailing list