[PATCH] Enable CHECKPOINT_RESTORE on power64le for CRIU feature

Laurent Dufour ldufour at linux.vnet.ibm.com
Mon Sep 7 07:45:34 UTC 2015


On 04/09/2015 19:21, Andrey Wagin wrote:
> 2015-09-04 20:15 GMT+03:00 Josh Boyer <jwboyer at fedoraproject.org>:
>> On Fri, Sep 4, 2015 at 1:08 PM, Andrey Wagin <avagin at gmail.com> wrote:
>>> 2015-09-04 20:00 GMT+03:00 Josh Boyer <jwboyer at fedoraproject.org>:
>>>> On Fri, Sep 4, 2015 at 12:49 PM, Andrey Vagin <avagin at gmail.com> wrote:
>>>>> Signed-off-by: Andrey Vagin <avagin at openvz.org>
>>>>> ---
>>>>>  config-powerpc64le | 3 +++
>>>>>  1 file changed, 3 insertions(+)
>>>>>
>>>>> diff --git a/config-powerpc64le b/config-powerpc64le
>>>>> index 48c8c0d..7d9f3fc 100644
>>>>> --- a/config-powerpc64le
>>>>> +++ b/config-powerpc64le
>>>>> @@ -1,3 +1,6 @@
>>>>>  CONFIG_CPU_LITTLE_ENDIAN=y
>>>>>
>>>>>  CONFIG_POWER7_CPU=y
>>>>> +
>>>>> +# https://fedoraproject.org/wiki/Features/Checkpoint_Restore
>>>>> +CONFIG_CHECKPOINT_RESTORE=y
>>>>
>>>> I'm fine with this, but why would you only enable this on ppc64le vs.
>>>> all of ppc64?
>>>
>>> Because the CRIU project supports ppc64le only.
>>
>> Because the format is endian-dependent or something?  Strange, but OK.
> 
> I added Laurent to CC, he ported CRIU to PowerPC. I think he can tell
> more about this question.

Currently C/R is only supported on ppc64le (not ppc64 BE). The reason is
in the CRIU's user space code where only the Little Endian's ABI is
supported.

For the kernel's point of view, there is no identified issue (at the
moment), but activating C/R for BE kernel seems useless until CRIU is
dealing with BE.

Laurent.



More information about the ppc mailing list