[Olsr-dev] Client Roaming Plugin

[Raphael]

sorry for having messed things up with mail threads....

On Thu, Jan 21, 2010 at 1:13 AM, Markus Kittenberger
<(spam-protected)> wrote:
> Am 21.01.2010, 00:16 Uhr, schrieb Raphael <(spam-protected)>:
>
>> On Wed, Jan 20, 2010 at 11:49 PM, Markus Kittenberger
>>>
>>> i think this is what u want to achieve/cure with "new olsrd" messages?
>>
>> Yes, exactly. Right now it takes about 10-20 seconds until the
>> connection continues.
>
> how do you "send" the new hna?
> and which timeouot do your hna messages have?
I add a new hna with "ip_prefix_list_add" and have a static timeout in
olsrd.conf of 10 seconds. When the, lets call the client "phone" now,
is no longer reachable, I call  "ip_prefix_list_remove".


>>
>>
>>> btw. does your plugin require a distinct interface or VAP or is your
>>> plugin
>>> capable of providing same functinoality on an alias ip of the adhoc
>>> interface (that olsrd is using aswell)?
>>
>> In my tests I use ath0(master) and ath1(ad-hoc) on wifi0. And there
>> comes the next evil trick: Everything has the same MAC-adress on ath0.
>
> ok very evil,.. this will not work with alias adresses, and VAPs will
> (probably) fail too.
>>
>> If you used an alias, the mac would change from station to station so
>> we would wait until the arp-cache for this IP times out and the
>
> imho just send out an gratuitous arp message!
> (in theory everyone that receives this should update its arp cache,.. but i
> did so far never verify this theory)
I have never heard of this thing, gratious arp. I will have to check
out about it.


>
> but this approach has at least one small drawback
> (i think it requires an 1:1 relation between client and gateway-ip)
>
>> connection continues. If we applied the same trick with using same
>> mac-adresses on an adhoc-network, bad things will happend, because in
>> this case, the same mac-adress would be present in the same network.
>
> no i think/hope this all can work well without too "evil" setups
>
This "evilness" has the advantage that the phone cannot distinguish
between stations and thinks it changed accespoints within a switched
network.

This gratious arp-thing might clean up things a bit. But it will
require that a gratious arp will definitley reach its target. I am
afraid it might be detected as arp-spoofing if we send a gratious arp
every time the "phone" changes between accespoints.

But when using adhoc-mode it might be worth a thought.

> i guess u already implemented some method to drop the packets on all masters
> that are not the actual "master" of this client, as otherwise we end up with
> duplicate packets with having multiple masters on same mac and ip.
>
I am not sure what you mean by this, because there is always only one
station with this mac&ip in reach. So lets give an example:
Accespoints have 10.0.0.1 and the phone has 10.0.0.101. It either
sends it out on ath0, because of 10.0.0.0/24 on this interface, or
there is a host route for this client and the packet will be
forwarded. 10.0.0.1, ath0, is no olsrd-interface.
So I dont see a way how duplicated may occur.


//Raphael