After some uptime in the client, I have observed Windows7 going into a
CPU runaway lockup.
Host is Athlon x2 running Debian Lenny amd64
+ kernel.org 2.6.28 kernel
+ kvm-83
Client is Windows 7 beta 32-bit image, (UP or SMP) with virtio drivers.
This happens when running the guest either as UP or SMPx2.
When running as UP, only one of my host CPUs are affected.
When running as SMPx2, both of my host CPUs are affected.
This can be reproduced with reasonable reliability by either
a) Commanding a restart in the guest machine
b) Significant sustained disk IO traffic, e.g. 200+ MB/s
Invocation:
sudo $KVM -name kvm-windows7 -smp 2 -m 1024 -localtime \
-drive file=/dev/vm/kvm-windows7 \
-drive file=/dev/vm/usenet \
-net nic,macaddr=00:ff:3e:a4:f4:20,model=virtio -net tap \
-daemonize -vnc localhost:1,to=4 -usbdevice tablet
After lockup, here is the backtrace I am pulling from gdb
[Am I doing this right? I am a little rusty]
(gdb) attach 12717
(gdb) bt
at /usr/src/kvm-83/qemu/vl.c:3637
at /usr/src/kvm-83/qemu/qemu-kvm.c:600
argv=0x7fffba8f1f78, envp=<value optimized out>) at /usr/src/kvm-83/qemu/vl.c:3799
Detailed configuration information attached.
Nohing I can find in the logs I would consider relevant.
Host System Information
VirtIO drivers will not be supported in Windows 7 BETA -- only in Final.
-Alexey
I can reproduce this problem at will using -nic model=e1000
So this has nothing to do with virtio network drivers
I am attaching a gdb log showing backtrace of all 3 process threads
kvmstat after failure
Still present, and easily reproducible, on 2.6.29/kvm-84.