AODV Ipv6: Disable NDP

[Shashidhar B.L]

Hi,

The AODV protocol establishes neighbour information through hello messages. I think, the Neighbour Discovery Protocol (NDP) is interfering with the functioning of AODV neighbour discovery. 

The reason why I think so is, the neighbour solicitation and advertisement messages of NDP are being sent (probably with a higher priority over AODV) which is deferring my hello messages (of AODV). Meanwhile, the route which existed when this Hello message was sent, has expired due to the large number of NS messages sent by ICMPV6. (I may have mistaken, I have to admit am slightly confused). Therefore, I would like to know if the NDP can be disabled and I rely solely on AODV for establishing the neighbour information? 

If I could disable, could you please let me know how I could do it?

If I should not disable it, then what could be the reason that my hello messages are incorrectly processed? (Tough to say without the code, but any slight hints would be helpful)

Thanks in advance

Regards

Shashidhar B.L

[Tommaso Pecorella]

Hi,

the RFCs should tell you exactly what to do. If you must disable NDP, there should be a clear indication in the RFC. If the RFC doesn't say anything, NDP must be kept as it is.

By RFCs I mean:

https://tools.ietf.org/html/draft-perkins-aodv6-01

and

https://www.ietf.org/rfc/rfc3561.txt

This is the introduction. Now the analysis.

NDP in ad-hoc, multihop networks is a b*tch, and its working mode is all but obvious.

It is used for DADs and for neighbor discovery, and it shouldn't be suppressed, as it's the only way to map IP and MAC addresses.

I know that AODV uses (also) Hello messages, but they're for other purposes, and they're not useful to fill the NDISC cache.

About your specific issue, it could have a number of causes. As an example, an NDP flood could be caused by a sync in application start (NDP itself is randomized, so it should be that).

Or you could have a too packed network, leading to collisions.

Or... or it could be not the NDP.

My suggestion is to reduce the network to a 2-nodes, then increase the size of the net. In this way it will be easier to find the problem.

have fun,

T.