Re: [PATCH v2 0/3] Implement CPU model driver for PowerPC

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Wed, Oct 17, 2012 at 4:22 PM, Michal Privoznik <mprivozn@xxxxxxxxxx> wrote:
> On 15.10.2012 11:07, Li Zhang wrote:
>> CPU model driver is to get host's CPU information and it also provides
>> one mechanism to expose host's CPU information to guests during migration.
>>
>> When migrating one guest from one machine to another machine, it will
>> compare the CPU information. If it is incomptible, it won't start the
>> guest.
>>
>> On x86, it uses CPUID instruction to get information. When migrating
>> the guest, if the CPU defined in guest XML file is incompatible with
>> host CPU, cpu-baseline will find out best features to expose to guest.
>> Refer to
>>  [http://berrange.com/posts/2010/02/15
>>   /guest-cpu-model-configuration-in-libvirt-with-qemukvm/].
>>
>> On PowerPC, it can get CPU version by mfpvr instruction. So, if PVR is
>>  different from definition in guest XML file by cpu-compare, it will fail
>> to start guest. The CPU infromation can be got by cpu-baseline. But only
>> CPU vendor and model can be shown, there are no features as x86.
>> So on PowrePC, it assumes that migration only occurs between the machines
>> with the same type CPUs on PowerPC.
>>
>> In this driver, there are definitions of models and PVRs for CPUs supported
>> on PowerPC. The relationship between models and PVRs are as the following:
>> For one specific CPU model, its PVR is unique. And also for one PVR code,
>> the PVR code is also unique.
>>
>> So from the PVR code of CPU data, it's easy to get model and vendor information.
>>
>> It is tested on my Power machine which CPU PVR is Power7_v2.3.
>>  * cap.xml has the same definition as the host.
>>    [root@ltckvmopal2 kvm-test]# /bin/virsh cpu-compare cap.xml
>>    CPU described in cap.xml is identical to host CPU
>>
>>    [root@ltckvmopal2 kvm-test]# /bin/virsh cpu-baseline cap.xml
>>    <cpu mode='custom' match='exact'>
>>     <model fallback='allow'>POWER7_v2.3</model>
>>     <vendor>IBM</vendor>
>>    </cpu>
>>
>>  * cap.xml has differnt model as the host.
>>    [root@ltckvmopal2 kvm-test]# /bin/virsh cpu-compare cap.xml
>>    CPU described in cap.xml is incompatible with host CPU
>>
>>    [root@ltckvmopal2 kvm-test]# /bin/virsh cpu-baseline cap.xml
>>    <cpu mode='custom' match='exact'>
>>      <model fallback='allow'>POWER7_v2.1</model>
>>      <vendor>IBM</vendor>
>>    </cpu>
>>
>>  *v1 -> v2:
>>    * Fix coding style's problems in several places.
>>    * Replace while with for loop in 2 places according to Michal Privoznik's suggestion.
>>    * Add powerpc.c to po/PORTFILES.in
>>
>> Li Zhang (3):
>>   libvirt: Add one file cpu_ppc_data.h to define CPU data for PPC
>>   libvirt: Implement CPU model driver for PowerPC
>>   Doc-fix for PowerPC CPU model driver
>>
>>  po/POTFILES.in         |    1 +
>>  src/cpu/cpu.h          |    3 +
>>  src/cpu/cpu_map.xml    |   14 ++
>>  src/cpu/cpu_powerpc.c  |  591 ++++++++++++++++++++++++++++++++++++++++++++++--
>>  src/cpu/cpu_ppc_data.h |   33 +++
>>  5 files changed, 624 insertions(+), 18 deletions(-)
>>  create mode 100644 src/cpu/cpu_ppc_data.h
>>
>
> I've reviewed second version too, ACKed all patches, applied small fixed
> I've found and pushed the whole series. Thanks!
>
Got it, thanks a lot. :)
> Michal



-- 

Best Regards
-Li

--
libvir-list mailing list
libvir-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/libvir-list


[Index of Archives]     [Virt Tools]     [Libvirt Users]     [Lib OS Info]     [Fedora Users]     [Fedora Desktop]     [Fedora SELinux]     [Big List of Linux Books]     [Yosemite News]     [KDE Users]     [Fedora Tools]