kvm hangs on 2.6.32

[Rusty Burchfield]

Binary package hint: kvm

Works (vnc session shows installer and no errors flowing into kern.log):
$ kvm -cdrom ubuntu-10.04-rc-server-amd64.iso -vnc :0

Fails with black vnc session and errors in kern.log:
$ kvm -uuid `uuid` -cdrom ubuntu-10.04-rc-server-amd64.iso -vnc :0

This is the error from kern.log (happens about 40 times per second):
kernel: [592333.140692] handle_exception: unexpected, vectoring info 0x8000080e intr info 0x80000b0d

I have tried three different iso images to eliminate that as a factor (openfiler, 3ware firmware, ubuntu server).

I don't specifically want to use a uuid. However, managing vms with libvirt does so automatically. Therefore I am unable to use virt-manager until this is fixed.

$ cat /etc/lsb-release | tail -1
DISTRIB_DESCRIPTION="Ubuntu 10.04 LTS"
$ uname -a
Linux gicodefs 2.6.32-21-generic #32-Ubuntu SMP Fri Apr 16 08:09:38 UTC 2010 x86_64 GNU/Linux
$ dpkg --list qemu-kvm | tail -1
ii qemu-kvm 0.12.3+noroms-0ubuntu9 Full virtualization on i386 and amd64 hardware

[Rusty Burchfield (gicodewarrior)]

As a workaround, I have replaced the /usr/bin/kvm soft-link with the following ruby code and made it executable. I do not know what adverse side-effects it will have, but it works for now.
#!/usr/bin/ruby1.9

args = ARGV.dup
uuid_index = args.index('-uuid')
if (uuid_index)
  args.slice!(uuid_index, 2)
end

exec '/usr/bin/qemu-system-x86_64', *args

[Rusty Burchfield (gicodewarrior)]

OK, my workaround got me as far as the boot screen for the iso, but as soon as a choose an action it locks up and starts spewing the above errors into kern.log again.

So, it isn't the uuid option specifically, but that does trigger it sooner.

[Rusty Burchfield (gicodewarrior)]

BTW, this is on a VIA processor.

$ cat /proc/cpuinfo
processor : 0
vendor_id : CentaurHauls
cpu family : 6
model : 15
model name : VIA Nano processor U2250 (1.6GHz Capable)
stepping : 3
cpu MHz : 1599.796
cache size : 1024 KB
fpu : yes
fpu_exception : yes
cpuid level : 10
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat clflush acpi mmx fxsr sse sse2 ss tm pbe syscall nx lm constant_tsc up rep_good pni monitor vmx est tm2 ssse3 cx16 xtpr rng rng_en ace ace_en ace2 phe phe_en lahf_lm
bogomips : 3199.59
clflush size : 64
cache_alignment : 128
address sizes : 36 bits physical, 48 bits virtual
power management:

[Dustin Kirkland  (kirkland)]

Chuck, you can reproduce this?

[Chuck Short (zulcss)]

No but there should be enough information to reproduce it.

[Rusty Burchfield (gicodewarrior)]

It turns out this is a known-problem on VIA Nano processors.
http://<email address hidden>/msg33108.html

It doesn't sound like they are interested in the issue upstream right now. So, unless someone here is intrigued enough to look into it with me, I will just take the hit and use qemu.

[Dustin Kirkland  (kirkland)]

Thanks for the clarification, Chuck.

Thanks for the research, Rusty.

Marking triaged/wishlist, we'll probably await upstream's fix (if ever) for this.

Thanks.

[Frédéric Logier (fredix)]

Hi,

I'm using Ubuntu server 10.04 and a VIA nano and I'm very embarrased to can not use kvm. So I found a kernel thread about this and it seems that a redhat developer pushed a fix : http://<email address hidden>/msg11798.html

Maybe you could investigate about.

[Rusty Burchfield (gicodewarrior)]

Unfortunately, the message you reference is mis-threaded and unrelated to the VIA support discussion. :(

[Frédéric Logier (fredix)]

ouch sorry ...