[support] About the Neighbor Unreachability Detection

[Angel Bartomeu Bonillo]

2009/5/10 Arnaud Ebalard <arno at natisbad.org>

> Hi,
>
> Angel Bartomeu Bonillo <angelbartomeu at gmail.com> writes:
>
> > How is the Neighbor Unreachability Detection implemented in the
> mip6daemon?
> > Does the daemon relay on the upper protocols to detect the two-ways
> > connection, does it uses Router Solicitation and Neighbor Solicitation or
> > does it use both methods as specified in the RFC 2461?
> >
> > More specifically my question could be: If a router send Unsolicited
> Router
> > Advertisement to a node running the mip6daemon but the router does not
> > implement any method to respond to Router Solicitation by means of a
> > Solicitated Router Advertisements, Does the node consider the router
> > reachable? Is there any problem in this scenario?
>
> I may be wrong (i.e. you'll have to test it to be sure) but I think what
> will happen is the following:
>
> UMIP will configure an address based on advertised prefix. If the
> router does not respond to solicited RS, this will obviously take some
> time depending on the interval configured between unsolicited RA.
>
> Then, because the neighbor cache is filled with the information on the
> router with the NUD_PROBE flag set (see call to neigh_add() in
> md_probe_router() in movement.c), UMIP will invalidate the router at
> some point if it does not receive a NA. It may use it during that time.
>

OK. Thanks.


>
> May I ask what kind of router/scenario you are trying to support?


Sure. In my scenario the mobile routers are not directly connected to the
Foreign Routers that announce the prefixes. Instead, the other mobile
routers that are directly connected to one Foreign Router forward the RAs
for them placing their own MAC as the source of the RA (Some kind of proxy
behavior). I made that question in the list to know if I needed to implement
in the transmitting nodes the responses to the Router Solicitation or if was
enough to resend the RA frequently enough.

I have actually tested it and it seems to work. And, as for the invalidation
of the retransmitting nodes, It doesn't seem to happen. I have been running
the test for a long time and It worked out.
On the other hand, In my test the MNNs were continuously sending transport
Data so what I really don't know is what would happen if the MNNs don't send
any data.

By the way. Is there any easy way to disable the node probe test?


>
> Cheers,
>
> a+
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://ml.nautilus6.org/pipermail/support/attachments/20090511/ca16566c/attachment.htm