Hello everyone, in the deploy of DeltaLibre (a local WCN very similar to EigenNet (part of ninux) in Italy) i'm facing a dilemma: i'm currently using batman-adv gw_mode client in all nodes, and the ones that connect directly to internet through an ISP have gw_mode = server This way, clients have their DHCP requests transparently unicasted to the 'best' gateway, and get a corresponding IPv4 and default gw. Which is great.
But in IPv6 there's no equivalent functionality. If two routers on the same broadcast domain (batman-adv cloud) run radvd and announce two different prefixes, all hosts on the network autoconfigure 2 ips, one in each prefix. Then, AFAIU, it's up to the client to choose between the announced gateways, to send the traffic.
Are there any plans of implementing in batman-adv something analog to the mangling done on DHCPv4 packets, to RA packets as well? (the mechanism would be the inverse: instead of unicasting queries and replies, propagate to clients only 1 RA packet out of all incoming; decide which one to propagate based on gw_mode current logic)
Or, at least, extend the current mangling done in DHCPv4 packets to DHCPv6 packets as well?
I would *love* to simply show some code instead of asking 'wishlist' questions, but sadly i'm not a programmer :(
Thanks a lot,
Guido
pd. I'm pretty new to ipv6, so It's possible that i'm overlooking 'obvious' concepts in my questions: in that case you can politely tell me to RTFM :)
Hello Guido,
On Sun, Jun 10, 2012 at 08:14:27PM -0300, Guido Iribarren wrote:
Hello everyone, in the deploy of DeltaLibre (a local WCN very similar to EigenNet (part of ninux) in Italy) i'm facing a dilemma: i'm currently using batman-adv gw_mode client in all nodes, and the ones that connect directly to internet through an ISP have gw_mode = server This way, clients have their DHCP requests transparently unicasted to the 'best' gateway, and get a corresponding IPv4 and default gw. Which is great.
correct.
But in IPv6 there's no equivalent functionality. If two routers on the same broadcast domain (batman-adv cloud) run radvd and announce two different prefixes, all hosts on the network autoconfigure 2 ips, one in each prefix. Then, AFAIU, it's up to the client to choose between the announced gateways, to send the traffic.
yes. batman-adv does treat RA packets as normal broadcasts. They are neither filtered nor handled in any particular way.
Are there any plans of implementing in batman-adv something analog to the mangling done on DHCPv4 packets, to RA packets as well? (the mechanism would be the inverse: instead of unicasting queries and replies, propagate to clients only 1 RA packet out of all incoming; decide which one to propagate based on gw_mode current logic)
someone proposed the idea on the bugtraq, but no plans about implementation have been seen so far..
Or, at least, extend the current mangling done in DHCPv4 packets to DHCPv6 packets as well?
this is already implemented. Gw_mode supports snooping of both dhcpv4 and dhcpv6 packets. So you could use dhcpv6 instead of radvd (at least for now).
I would *love* to simply show some code instead of asking 'wishlist' questions, but sadly i'm not a programmer :(
you are contributing anyway :-)
Cheers,
On Monday 11 June 2012 01:23:00 Antonio Quartulli wrote:
Are there any plans of implementing in batman-adv something analog to the mangling done on DHCPv4 packets, to RA packets as well? (the mechanism would be the inverse: instead of unicasting queries and replies, propagate to clients only 1 RA packet out of all incoming; decide which one to propagate based on gw_mode current logic)
someone proposed the idea on the bugtraq, but no plans about implementation have been seen so far..
Just for completeness: http://www.open-mesh.org/issues/159
Kind regards, Sven
On Monday, June 11, 2012 07:14:27 Guido Iribarren wrote:
Are there any plans of implementing in batman-adv something analog to the mangling done on DHCPv4 packets, to RA packets as well? (the mechanism would be the inverse: instead of unicasting queries and replies, propagate to clients only 1 RA packet out of all incoming; decide which one to propagate based on gw_mode current logic)
Or, at least, extend the current mangling done in DHCPv4 packets to DHCPv6 packets as well?
What makes you think DHCPv6 is not working ? It should ...
Regards, Marek
On Sun, Jun 10, 2012 at 8:29 PM, Marek Lindner lindner_marek@yahoo.de wrote:
On Monday, June 11, 2012 07:14:27 Guido Iribarren wrote:
Or, at least, extend the current mangling done in DHCPv4 packets to DHCPv6 packets as well?
What makes you think DHCPv6 is not working ? It should ...
Nothing but my own unfounded prejudice... loving batman-adv every day a bit more!
one of these days i'm going to try "batctl solve life"
cheers!
b.a.t.m.a.n@lists.open-mesh.org