[Olsr-dev] olsrv2: observed routing loop when link goes down

[Vigneswaran R]

Hello,

I am running latest olsrv2 in the following topology.

vm1----------vm2---------vm3
| \                      |
|  \______               |
|         \              |
|          \             |
vm4----------vm5---------vm6

Fig.1  Six qemu virtual machines having four Ethernet NICs each

* Initially, route from vm1 to vm6 is via vm5 ("vm1-->vm5-->vm6")
* I started ping from vm1 to vm6
* While ping was in progress, I put vm5's interface which is connected 
to vm6, DOWN (just to simulate 'vm5-vm6' link break)

At that point, ping gave a couple of "Destination Net Unreachable" 
error. After that it started giving "Time to live exceeded". This 
persists until I make 'vm5-vm6' link up.

During that period, vm1 had route to vm6 via vm5, and vm5 had route to 
vm6 via vm1. That created a loop.

Note:
----
If I drop the packets using iptables to simulate the link break (instead 
of putting the interface down), olsrv2 is able to recover from the 
routing loop quickly (in 4 or 5 secs).


Regards,
Vignesh