[OLSR-users] Running two insances of olsr

[Andreas Tønnesen]

Cyrille Chepelov wrote:
> The current problem arises from the fact that you're creating sockets
> bound to "any address on interface foo", which really means "all
> addresses on interface foo", which for an ipv6 socket includes v4
> sockets, and for a v4 sockets means that the next similar request for a
> v6 socket will choke.
> If instead, you're creating sockets bound to "this specific address", then 
> the original problem disappears (you still need to handle the selection of the 
> packet parser routine based on the socket type).
> 

Yeah - I'm moving a bit slow here today :)

To me it would seem that binding the sockets to the proper
broadcast/multicast addresse instead of INADDR_ANY could
solve the problem without needing to update more of the code.
But when testing that here I get som very strange results
dependent on what kernel is running on the
different boxes. On this box(running 2.6.3) it works fine
if i bind IPv4 sockets to INADDR_BROADCAST(which is supposed
to yield the same results as INADDR_ANY). If binding to the
broadcast address olsrd receives nothing... On a box running 2.4.23
it works if I bind IPv4 sockets to the actual broadcast address -
but the the socket actually receives the IPv6 traffic emitted
from the IPv6 oslrd instance on the same box(!) causing all
kinds of strange behaviour.

IPv6 sockets are bound to the multicast address in use - and this
seems to work fine.

Well, I'll look moore into it when I can find the time. If anybody
implements a solution let me know.

- Andreas

-- 
Andreas Tønnesen((spam-protected))
UniK University Graduation Center
University of Oslo
http://www.olsr.org