[CRIU] Process Migration using Sockets - PATCH 1/2

[Cyrill Gorcunov]

On Sat, Oct 03, 2015 at 08:37:05PM +0100, Rodrigo Bruno wrote:
> Hi,
> 
> I will do the corrections you asked and resend both patches with protocol
> description as requested.
> 
> I added some comments on the mail below.
> 

Thanks a huge Rodrigo for detailed reply about the idea!
(And sorry for delay in response). You know I believe we
need some more general way for such kind of images
transport. I must admit I can't came in with some
already formed architecture but I guess the transport
should lay someway lower than pb_write functions and
be completely transparent to the CRIU code itself
(not sure though if it possible). After all the
transport might be completely non tcp based so
the transport driver would have to check the
consistency of the images, or maybe the transport
would provide crypted images so they could be
transferred via some non-secure environment.
In other words if we would have some pluggable
transport mechanism for image delivery this
would be just brilliant. But I fear I'm asking
for too much (because indeed making a pluggable
architecture for criu transports requires hell
knows how many men-hours), so as a first
iteration your approach should do the trick,
thanks!