[gentoo-server] Re: ipv6 + dhcpv6 + unique local addresses

["Halassy Zoltán" <zhalassy@loginet.hu>]

[-- Attachment #1: Type: text/plain, Size: 2463 bytes --]

To answer my own question, actually found the answer:

http://www.ietf.org/rfc/rfc2461.txt

On page 18:

Router Lifetime
                      [...] A
                      Lifetime of 0 indicates that the router is not a
                      default router and SHOULD NOT appear on the default
                      router list. [...]

So this needs to written in radvd.conf:

AdvDefaultLifetime 0;

Yay.

2013.04.01. 14:01 keltezéssel, Halassy Zoltán írta:
> Hello!
>
> Is there anyone who has experience with unique local addresses (fc00::/7)?
>
> I have experience with radvd and isc dhcp (in ipv6 mode too with the -6
> flag), I could already configure stateful configuration with global
> unicast (2000::/3) addresses with working default gateway.
>
> What I am trying to do now is to create a local IPv6 network space with
> a dhcpv6 server (amd64 gentoo), which is only reachable via VPN. The
> network does not have any router, it's isolated. IPv4 is not an option,
> and DHCPv6 is mandatory. The clients are mostly Windows Vista+ systems.
> What I am seeking is the proper way to do this. I could make it work,
> but I consider this a hack.
>
> I generated a random IPv6 address range, but I will use the
> fd00:2001:db8::/64 prefix in the description.
>
> Problem #1:
>
> DHCPv6 works fine, it pushes an IPv6 address to the client, but the
> client does not get the prefix information with it. Eg.: client gets
> fd00:2001:db8::ffff:fffe/128 as address, but missing the local route
> information for fd00:2001:db8::/64 through the interface.
>
> Problem #2:
>
> If I use radvd advertising the fd00:2001:db8::/64 prefix, the client
> configures that up, but it also configures a bogus default route too,
> which is definitely unwanted.
>
> Hack #1:
>
> Using dhcp and radvd together actually works (even though it's very
> ugly). It does not ruin an existing IPv6 connection, and does not cause
> problems when originally there is none. I just fear it *might*.
>
> Hack #2:
>
> It is possible to create static (even on-link) routes with netsh, but
> other than being ugly as well, it's not platform independent solution.
>
>
>
> What I would require is (if it's somehow possible), to make the
> platform-independent client do prefix discovery, find the prefix
> on-link, but do not configure routing information for that link. And to
> do it the proper way.
>
> Any ideas?
>



[-- Attachment #2: S/MIME kriptográfiai aláírás --]
[-- Type: application/pkcs7-signature, Size: 4462 bytes --]