[OLSR-users] Re: Setting up OLSR for a complicated environment

Alexander Morlang (spam-protected)
Fri Feb 9 14:12:14 CET 2007



Sven-Ola Tuecke schrieb:
> Hi,
> 
> WDS and OLSR are somewhat incompat. The standard story is this:
> 
> - someone sets up a couple of WDS access points
> - after a while, the network grows (both tech + persons)
> - he/she realizes: WDS does not work beyond a certain AP count
> - WDS is replaced by OLSR or other routing protocols
> - the WEP/WPA encryption is sacrificed also
> 
> Why? The layer2 data path (aka. the way packets travel through the 
> different stations) is detected by a spanning tree protocol in WDS. 
> Links from A via B to C are fiddeld out in this way. The spanning tree 
> protocol chooses the least hop datapath. The spanning tree root node 
> needs to have good connectivity to all stations and this also introduces 
> a SPOF (single point of failure). A protocol like this runs fine with 
> every standard (wired) switch device you buy.

The usage of sanning tree is not a must, it is the way many people do.
They bridge all interfaces together and enable spanning tree.

> 
> OLSR is different. It is layer3. Routing as it's done on the internet to 
> interlink a bazillion of stations. A change from Switching to Routing 
> introduces a conceptual change. A network of equals. Of course, OLSR 
> does function with every link which can transport IP datagrams. This is 
> true for an AP to client link. For Ethernet. For VPN links. For pidgeons 
> (RFC 2549). And of course, this should be true for a couple of WDS APs. 
> But to profit from OLSR routing and the ETX/LQ optimized data path 
> calculation, all active stations/nodes needs to participate in the 
> routing process. Which requires Ad-Hoc or at least 
> Access-Points-to-client links with AP-Isolation=True if more than one 
> client.

not really, wds is a link between to APs, so giving out ip adresses to 
all interfaces of the wdslinks and adding them in the olsrd.conf should 
be possible. the, you do layer3 over wds, which _should_ do fine.

To get it dynamic, the is a need of a mechanism which adds new wdslinks 
to the olsr.

the advantage of this construct might be, that you still have AP 
funtionality, but for roaming would be some kind of layer2 vpn necessary.

I think, somebody did it that way, but for me, it purely theoretical as 
i work with ad-hoc.

> 
> To write a word on WEP/WPA: If everyone in a bigger net (more than dozen 
> people) knows the single secret (key), the layer2 encryption is useless. 
> For this reason, you switch from layer2 security to layer3 security. 
> Because a different key for each stationa-to-stationb link is not 
> supported/practical with those small boxes. Layer3 security involves 
> firewalling, IPSec and VPN tunnels - which is also very similar how this 
> stuff is handled on the internet.
> 
> Short PS: There are mixed concepts out there. E.g. the roofnet stuff. 
> But I personally do not know enough on the topological impact to write 
> down hints.
> 
> Keep playing!
> // Sven-Ola

Greets, Alex


> 
> ""Jon Smirl"" <(spam-protected)> schrieb im Newsbeitrag 
> news:(spam-protected)
>> On 2/8/07, Bernd Petrovitsch <(spam-protected)> wrote:
>>> On Thu, 2007-02-08 at 11:48 -0500, Jon Smirl wrote:
>>> > On 2/8/07, Bernd Petrovitsch <(spam-protected)> wrote:
>>> > > > Could I run my mesh nodes in Master mode and then use OLSR to 
>>> set > > > up
>>> > > > WDS links between them?
>>> > >
>>> > > I'm not a WDS expert. How are these links setup?
>>> >
>>> > WDS is a way of setting up links between stations running in Master
>>> > mode. My idea was to run all of my mesh nodes in Master mode and then
>>> > use OLSR to control the WDS links.
>>> >
>>> > http://en.wikipedia.org/wiki/Wireless_Distribution_System
>>>
>>> If I understand this correctly, WDS extends the functionality of a
>>> "normal" access point so that it can talk directly with another
>>> WDS-enabled access point.
>>> And with WDS you have to "route"[0] packets from one client to another.
>>> If I understand it correctly it - thus - implements "routing" based on
>>> MAC-adresses (of the access points and clients).
>>> If this is the case, WDS must be (conceptually) part of the MAC-layer.
>>>
>>> Since this OLSRD is an pure user-space program and routes on the IP
>>> layer, it is not usable for this that easy AFAICS.
>>>
>>> Of course, there is room for a hack (if someone finds the time): Given
>>> that the access points run Linux (or other OSs with such features), one
>>> can hook into the MAC layer (with ebtables or an extension thereof) and
>>> get the requests into the user-space. And then one can implement the
>>> third "protocol" (after IPv4 and IPv6) which can be used.
>>> The algorithms etc. are in place and perfectly usable to use MAC
>>> addresses instead of IP addresses.
>>
>> Access points are running Linux.
>>
>> That's what I'm looking for, someone who had already figured out how
>> to use OLSR to control how WDS nodes get linked and routed. I didn't
>> want to try building this again if someone has already tried and found
>> that it won't work.
>>
>> There are references to doing this in the mailing list archives but I
>> don't see any clear conclusions if anyone got it working or not.
>>
>>
>>
>> -- 
>> Jon Smirl
>> (spam-protected)
>>
>> _______________________________________________
>> olsr-users mailing list
>> (spam-protected)
>> https://www.olsr.org/mailman/listinfo/olsr-users 
> 
> 
> _______________________________________________
> olsr-users mailing list
> (spam-protected)
> https://www.olsr.org/mailman/listinfo/olsr-users




More information about the Olsr-users mailing list