[GLLUG] RADVD configuration

Tim Woodall t at woodall.me.uk
Wed Apr 5 17:04:32 UTC 2017


On Wed, 5 Apr 2017, Chris Bell via GLLUG wrote:

> On Wednesday 05 Apr 2017 12:32:15 you wrote:
>> Hi,
>>
>> Think of the Global address as the one used between hosts (e.g. interface
>> eth0), and the site local as the loopback (e.g. interface lo). (i.e. not
>> route-able outside of the host)
>>
>> Kind Regards
>>
>> James
>>
> Thanks for the reply. There is sparse documentation for RADVD, so I have been
> ploughing through well over 100 RFC's trying to find information. An interface
> can have more than one IPv6 address, in theory as many as you wish. The fexx::
> series of address prefixes are available for site local use as specified in the
> RFC's, and I am trying to use one site-local prefix for each of the local
> networks, (I am required to generate a genuine random number for bits 9 to 56,
> and they have the same initial 62 bits with only bits 63 and 64 changed), to
> be used for site local connections only, and in addition all can eventually
> share the same allocated global prefix.
> According to the RFC's each network interface should receive the full
> information from RADVD and then choose the correct prefix to match the route.
> The shorewall6 interfaces should each have both a site local address prefix and
> the global allocated address prefix. I have not seen anything to suggest that
> the suffix for each address should be the same on any individual interface,
> there could be an automatically generated suffix based on the MAC as well as one
> or more manually configured. The fun thing is to persuade shorewall,
> shorewall6, rdnss, dnssl, abro, and radvd to actually run, and then work
> together, with a laborious time-consuming process of trying another
> configuration amendment, reading the error messages, and so on.
> It is not really necessary for my home network, but I think it is supposed to
> work.
>
>

I think you're possibly confused between unique-local and site-local.
(Or I am :-) )

site-local is (was) fec0::/10
unique-local is fc00::/7 (really fd00::/8)

Looks like you're trying to use the site-local prefix (first 8 bits)
with unique-local addressing. That may, or may not, be what is causing
some of your problems. If you're ending up with a link-local address
then it won't be able to route.

I've not tried to give multiple prefixes out via radvd. I have a global
/48 and link-local addressing only.

I have had problems with machines not picking up the correct prefixes
when I've been changing them. It can be really hard to make the machines
forget about what they saw previously/what they had configured
previously - especially if they're also configured to generate random
outgoing addresses which (obviously) they need to remember for a time
and especially if you're trying to configure one or more of them
remotely where you don't want to take down all the interfaces at once!

Finally, radvd seems to have problems with interfaces that don't exist
when it starts.
My /etc/ppp/ipv6-up.d/ has a 0000radvd script that generates a config
file for the interface that has just come up and then starts a radvd
instance for it. There are various config options in radvd for
interfaces that don't (yet) exist etc but I just couldn't ever get it to
work so that it would advertise over a vpn.

Tim.




More information about the GLLUG mailing list