This might be intended behaviour but I'm unsure.
At least this issue should serve as documentation should someone else stumble upon this.
When a node has multiple neighbours with the same link local address (for example we had a wireguard tunnel server and were using a staticly configured LL address on the wireguard interface that was the same on all other nodes), OLSRD2 treats these neighbours as the same neighbour even though their advertised IP address and interface is different.
In our case this resulted in only a single route that was the IP address of the last node that was connected routed via the interface of the first node that was connected.
Changing the LL addresses of all nodes to be unique to each node solved this issue (having the same LL on all interfaces of one node is ok though).
For me this looks like a bug where the OLSRD2 does only look at the LL IP to identify a neighbour without looking at the interface name of this LL address (purely a guess).
I don't think that a fix for this has a high priority but awareness for this behaviour might be important ^^
This might be intended behaviour but I'm unsure. At least this issue should serve as documentation should someone else stumble upon this.
When a node has multiple neighbours with the same link local address (for example we had a wireguard tunnel server and were using a staticly configured LL address on the wireguard interface that was the same on all other nodes), OLSRD2 treats these neighbours as the same neighbour even though their advertised IP address and interface is different.
In our case this resulted in only a single route that was the IP address of the last node that was connected routed via the interface of the first node that was connected. Changing the LL addresses of all nodes to be unique to each node solved this issue (having the same LL on all interfaces of one node is ok though).
For me this looks like a bug where the OLSRD2 does only look at the LL IP to identify a neighbour without looking at the interface name of this LL address (purely a guess). I don't think that a fix for this has a high priority but awareness for this behaviour might be important ^^