On Thu, Sep 05, 2013 at 01:43:42PM +0200, Henning Rogge wrote:
> On 09/04/2013 04:26 PM, Ben Pfaff wrote:
> >It is probably a mistake to run the ovs database server (osvdb-server)
> >remotely. It is not a recommended configuration. ovsdb-server and
> >ovs-vswitchd should run on the same device.
>
> I know... its just that the device has only 4 MB of flash and 16 MB
> of RAM, so I am looking for ways to minimize the local footprint of
> openvswitch.
These size requirements may be a challenge. I wish you luck.
> >IPv6 is not currently supported, but it is not a design choice, it is
> >only a missing feature. We would gladly accept patches to add support
> >for IPv6.
>
> I have looked a little bit more into the OpenVSwitch 1.9.0 codebase
> for the IPv4/IPv6 thing.
>
> The main trouble I see is that the libs/stream*.[ch] abstraction is
> only able to work with IPv4 at the moment, both the variables and
> the functions directly use 32-bit integers for IP addresses.
I guess you're talking about these functions:
ovs_be32 stream_get_remote_ip(const struct stream *);
ovs_be16 stream_get_remote_port(const struct stream *);
ovs_be32 stream_get_local_ip(const struct stream *);
ovs_be16 stream_get_local_port(const struct stream *);
Those are only important for implementing in-band control. There
would be no need to tie adding support for IPv6 controllers to adding
*in-band* support for IPv6 controllers. You could ignore them and
just return 0.
_______________________________________________
discuss mailing list
discuss@openvswitch.org
http://openvswitch.org/mailman/listinfo/discuss
