Re: Unknown KVM internal error on 3.2.1

[Doug Goldstein <cardoe@xxxxxxxxxx>]

On Tue, Mar 27, 2012 at 5:53 AM, Avi Kivity <avi@xxxxxxxxxx> wrote:
> On 03/26/2012 10:12 PM, Doug Goldstein wrote:
>> >
>> > When it happens again, please keep the guest alive so we can examine it
>> > via qemu monitor commands.
>> >
>>
>> It happened again. Here's the output requested.
>>
>> Before resuming:
>>
>> # virsh qemu-monitor-command bb-rhel-5.3-i686 --hmp x/20i 0xf91b4748
>> 0x00000000f91b4748:  lock popf
>> 0x00000000f91b474a:  lock popf
>> 0x00000000f91b474c:  lock popf
>> 0x00000000f91b474e:  lock popf
>> 0x00000000f91b4750:  lock popf
>> 0x00000000f91b4752:  lock popf
>> 0x00000000f91b4754:  lock popf
>> 0x00000000f91b4756:  lock popf
>> 0x00000000f91b4758:  lock popf
>> 0x00000000f91b475a:  lock popf
>> 0x00000000f91b475c:  lock popf
>> 0x00000000f91b475e:  lock popf
>> 0x00000000f91b4760:  lock popf
>> 0x00000000f91b4762:  lock popf
>> 0x00000000f91b4764:  lock popf
>> 0x00000000f91b4766:  lock popf
>> 0x00000000f91b4768:  lock popf
>> 0x00000000f91b476a:  lock popf
>> 0x00000000f91b476c:  lock popf
>> 0x00000000f91b476e:  lock popf
>>
>> After resuming and immediately running the requested command:
>
> Did the guest run properly after resuming?

Yes. They always run fine after resuming. About 50% of the time
they'll hit the same issue within a few seconds/minute. And another
resume and they're off and running again for weeks.
>
>>
>> # virsh qemu-monitor-command bb-rhel-5.3-i686 --hmp x/20i 0xf91b4748
>> 0x00000000f91b4748:  mov    %eax,%cr0
>> 0x00000000f91b474b:  mov    %ebx,(%ecx)
>> 0x00000000f91b474d:  mov    %edx,%cr0
>> 0x00000000f91b4750:  mov    (%esi),%esi
>> 0x00000000f91b4752:  mov    (%esi),%eax
>> 0x00000000f91b4754:  prefetchnta (%eax)
>> 0x00000000f91b4757:  nop
>> 0x00000000f91b4758:  mov    0xf91db9f8,%eax
>> 0x00000000f91b475d:  add    $0x8,%eax
>> 0x00000000f91b4760:  cmp    %eax,%esi
>> 0x00000000f91b4762:  jne    0xf91b4738
>> 0x00000000f91b4764:  lock incl 0xf91da3b0
>> 0x00000000f91b476b:  pop    %ebx
>> 0x00000000f91b476c:  mov    $0x5,%edx
>> 0x00000000f91b4771:  pop    %esi
>> 0x00000000f91b4772:  mov    $0xf91da380,%eax
>> 0x00000000f91b4777:  jmp    0xc0432047
>> 0x00000000f91b477c:  push   %ebp
>> 0x00000000f91b477d:  push   %edi
>> 0x00000000f91b477e:  push   %esi
>>
>
> This is perfectly sane and expected code (unlike the previous dump).
>
> Are you running with transparent hugepages and/or ksm enabled?  You
> might try to disable them.

Yes. I'm running with both. I'll try disabling them both and see if
this reoccurs.

Thanks.

-- 
Doug Goldstein
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html