[radvd-devel-l] Wrongly picks and advertises /80 IPs

Roman Mamedov rm at romanrm.ru
Wed Mar 6 01:56:06 EST 2013


On Tue, 5 Mar 2013 22:38:45 -0800
Reuben Hawkins <reubenhwk at gmail.com> wrote:

> So it seems to me that RADVD is doing the correct thing, just not what you
> expect...
> 
> You have this for your network interface...
> 
>           inet6 addr: 2a02:2698:xxxx:xxxx:fff2::39/80 Scope:Global
>           inet6 addr: fd39:0:2::39/64 Scope:Global
>           inet6 addr: fe80::260:6eff:fe20:2599/64 Scope:Link
> 
> and these are the prefixes being advertised...
> 
>         prefix 2a02:2698:xxxx:xxxx:fff2::/80
>         prefix fd39:0:2::/64
> 
> The documentation for the prefix you specify in the config file, ::/64,
> specifies that all prefixes which are non-link-local will be advertised.
>  In this case, RADVD is advertising your two non-link-local prefixes
> correctly.

I would expect it to only pick prefixes from IPs that are configured as on
the interface /64, since autoconfiguration can only work with /64s (and I
specified ::/64).

On the other hand if I specified "::/80" in config file, then only /80
prefixes should be picked (not that it would make a while lot of sense).

> I decided a long time ago that these auto-selecting prefixes are bad and
> should be removed. I plan to remove the special ::/64, Base6Interface and
> Base6to4Interface.  A much better place to automatically gather prefixes to
> advertise is /proc/net/ipv6_route.

This is a really convenient feature, so please if you remove it, ensure there
is something that replaces it adequately. It is so nice currently to not have
to edit radvd.conf when my global IPv6 prefix changes. Only reconfigure IPs on
the actual interfaces and restart radvd, done.

-- 
With respect,
Roman
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: not available
URL: <http://lists.litech.org/pipermail/radvd-devel-l/attachments/20130306/87dee991/attachment.pgp>


More information about the radvd-devel-l mailing list