Re: [PATCH 20 of 20] n2 MSR handling and capability exposure

[Jeroen Groenewegen van der Weyden <groen692@grosc.com>]

Eddie, did I mention I'm eager and capable of testing anything you need.

mfg,
jeroen

Op 4-7-2011 10:58, Tim Deegan schreef:
> At 12:11 +0200 on 01 Jul (1309522280), Jeroen Groenewegen van der Weyden wrote:
>> Just a remark to avoid misunderstanding
>>
>> The first nested  KVM guest is running ok, for as long as i keep it running
>> When starting a second nested KVM guest (in the same Xen DOMU) that
>> is when the Xen DOMU becomes in-responsive. In DOM0 it it shown that
>> it is still running and consuming time.
>>
>>
>> The complete test is like this
>>
>> dom0
>> opensuse 11.4 x86
>> xen: unstable cs 23640
> You're testing on Intel hardware, right?
>
> Eddie, any comment?
>
> Tim.
>
>> domu1 (HVM)
>> sles11sp1
>> mem: 8 GB
>> vcpu: 4
>>
>> domu1 guests
>> 1) qemu-kvm: propetary OS
>> 2) qemu-kvm: propetary OS
>> 3) qemu-kvm: SLES10SP3
>> 4) qemu-kvm: SLES10SP3
>>
>> mfg,
>> Jeroen
>>
>> Op 1-7-2011 11:54, Christoph Egger schreef:
>>> KVM on Xen works for me. Also KVM on Xen on Xen.
>>>
>>> Christoph
>>>
>>>
>>> On 07/01/11 11:01, Tim Deegan wrote:
>>>> Eddie?  Did you test with KVM guests?  Is this expected to work?
>>>>
>>>> Tim.
>>>>
>>>> At 15:14 +0100 on 30 Jun (1309446872), Jeroen Groenewegen van
>>>> der Weyden wrote:
>>>>> Hi Tim/Eddi/Christoph,
>>>>>
>>>>> I just gave it a go again with cs23640. But the behaviour is still the
>>>>> same. so L1 becomes inresponsive after starting the second
>>>>> nested quest.
>>>>> I can deliver any trace input you want/need.
>>>>>
>>>>> mfg,
>>>>> Jeroen.
>>>>>
>>>>> Op 20-6-2011 8:31, Jeroen Groenewegen van der Weyden schreef:
>>>>>> Hi,
>>>>>>
>>>>>> Based on these patch series, I gave it go.
>>>>>> I noticed until now that the system becomes in-responsive a few second
>>>>>> after activating the second nested guest.
>>>>>> steps:
>>>>>> 1) dom0 creates domu (pass)
>>>>>> 2) domU creates first kvm guest (pass)
>>>>>> 3) domU creates second kvm guest (fail)
>>>>>> after the third step the domu becomes in-responsive. eg no network /
>>>>>> vnc console response. in dom0 is shows domu is stil running.
>>>>>>
>>>>>> I'm able to provide any input needed, just tell me what you need (and
>>>>>> how).
>>>>>>
>>>>>> mfg,
>>>>>> Jeroen
>>>>>>
>>>>>> My HW setup
>>>>>> mobo asus z8nr-d12
>>>>>> proc: 2 times e5645
>>>>>> mem: 96 GB
>>>>>>
>>>>>> My goal/test
>>>>>> in short: HW ->   dom0 ->   domU ->   4 time kvm guest
>>>>>>
>>>>>> dom0
>>>>>> opensuse 11.4 x86
>>>>>> xen: unstable cs 23553
>>>>>>
>>>>>> domu (HVM)
>>>>>> sles11sp1
>>>>>> mem: 8 GB
>>>>>> vcpu: 4
>>>>>>
>>>>>> domu guests
>>>>>> 1) qemu-kvm: propetary OS
>>>>>> 2) qemu-kvm: propetary OS
>>>>>> 3) qemu-kvm: SLES10SP3
>>>>>> 4) qemu-kvm: SLES10SP3
>>>>>>
>>>>>>
>>>>>>
>>>>>> Op 15-6-2011 14:45, Tim Deegan schreef:
>>>>>>> At 10:29 +0800 on 14 Jun (1308047377), Dong, Eddie wrote:
>>>>>>>>>> +    case MSR_IA32_VMX_MISC:
>>>>>>>>>> +    case MSR_IA32_VMX_CR0_FIXED0:
>>>>>>>>>> +    case MSR_IA32_VMX_CR0_FIXED1:
>>>>>>>>>> +    case MSR_IA32_VMX_CR4_FIXED0:
>>>>>>>>>> +    case MSR_IA32_VMX_CR4_FIXED1:
>>>>>>>>>> +    case MSR_IA32_VMX_VMCS_ENUM:
>>>>>>>>>> +        gdprintk(XENLOG_WARNING, "VMX MSR %x not fully supported
>>>>>>>>> yet.\n", msr);
>>>>>>>>>
>>>>>>>>> Are you planning to fix this before checking in this series?  I'm
>>>>>>>>> pretty
>>>>>>>>> sure that at least MSR_IA32_VMX_VMCS_ENUM should be
>>>>>>>>> trivial, since you
>>>>>>>>> define your own VMCS format.
>>>>>>>>>
>>>>>>>> Sure, the updated patch is attached.
>>>>>>> Thanks.  I've applied the full series.  I cleaned up a few things for
>>>>>>> the 32-bit build as well.
>>>>>>>
>>>>>>> Cheers,
>>>>>>>
>>>>>>> Tim.
>>>
>>
>> _______________________________________________
>> Xen-devel mailing list
>> Xen-devel@lists.xensource.com
>> http://lists.xensource.com/xen-devel