[CRIU] Towards an in-process libcriu

Ruslan Kuprieiev kupruser at gmail.com
Wed Jul 16 15:37:00 PDT 2014


Hi Tycho,

The only difference I see between existing libcriu and the one you want 
to make, is
running criu service. Criu service was designed to be used as system 
service, so
you could just make it run on start of your system using provided 
<http://criu.org/RPC#systemd> systemd
configs. Then, using libcriu is just as simple as #include criu.h and 
dump().

Indeed, creating libcriu in the form you suggested seems to me like a 
lot of work
and after that we will come up with libcriu with reduced functionality, 
as it wouldn't
be able to run as non-root, perform a self-dump etc.

But that's just my personal opinion. Maybe others would say smth else.

Thanks,
Ruslan

On 17.07.2014 00:50, Tycho Andersen wrote:
> Hi Ruslan,
>
> On Wed, Jul 16, 2014 at 11:38:29PM +0300, Ruslan Kuprieiev wrote:
>> Oh! I see. But there is another problem. I believe, criu doesn't
>> free all resources
>> after it is done its work.
> I understand that, but I suppose those are mostly bugs in criu that
> will be fixed. Supposing that I'm ok with the caveats (potential
> resource leaks, can't dump the tree with me in it, etc.), it would be
> nice not to have to fork() and set up a socket and all that, but
> rather just #include criu.h and dump(other_pid). If it is a lot of work
> or doesn't fit with the project, I'm happy to go with the fork model,
> but I think it shouldn't be that hard to get something minimal going.
> I'm just unsure of where the best place for it to live is.
>
> Tycho

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openvz.org/pipermail/criu/attachments/20140717/93fc8b10/attachment.html>


More information about the CRIU mailing list