[CRIU] Bad commit

Pavel Emelyanov xemul at parallels.com
Thu Aug 14 07:05:51 PDT 2014


On 08/14/2014 05:50 PM, Cyrill Gorcunov wrote:
> On Thu, Aug 14, 2014 at 08:43:25AM -0500, Tycho Andersen wrote:
>> On Thu, Aug 14, 2014 at 05:42:18PM +0400, Cyrill Gorcunov wrote:
>>> On Thu, Aug 14, 2014 at 05:39:29PM +0400, Cyrill Gorcunov wrote:
>>>> On Thu, Aug 14, 2014 at 05:35:08PM +0400, Cyrill Gorcunov wrote:
>>>>>
>>>>> because we're restoring container in own pid namespace passing clone-parent
>>>>> triggers this condition and we're hitting einval. Ideas?
>>>>
>>>> Maybe something like
>>>
>>> False alarm. Really sorry for that, it's my custom kernel has such if statement.
>>
>> Ok, thanks. It worked fine for the stock 14.04 kernels, although I
>> haven't really tested on anything else.
> 
> Yeah, sorry for noise.

Wait a second, this code was in 3.12 and (probably) earlier:
commit 1f7f4dde5c945f41a7abc2285be43d918029ecc5 removed CLONE_PARENT
and it appeared in 

$ git describe 1f7f4dde5c945f41a7abc2285be43d918029ecc5
v3.13-rc1-130-g1f7f4dd

IOW -- --restore-detach with CLONE_NEWPID will fail when criu is run
on 3.11, which we recommend as the base kernel.

OTOH we need this CLONE_PARENT trick only when the root task has the
pdeath_sig set up. Can we relax the if () to only do the CLONE_PARENT
if the detach is requested and pdeath_sig is set up?

Thanks,
Pavel



More information about the CRIU mailing list