Hello Ming Ann,

we had quite some changes for the interface alternation which were released with batman-adv 2014.1.0 / kernel 3.14. Could you please try your scenario with the latest batman-adv to make sure that we are not talking about old implementations?

Here is a link which describes more technical details about the new implementation:

http://www.open-mesh.org/projects/batman-adv/wiki/Multi-link-optimize

Thank you! Simon

On Tuesday 07 October 2014 17:10:26 MingAnn Ng wrote:

Hello,

I am currently running some powerpc machines with buildroot environment. The Linux kernel version I am using is 3.8.13 and BATMAN-adv within the kernel without any modification. I am running interface alternating test with the topology below:

A========B=========C

There are two wireless links between A and B and two wireless links between B and C. The two wireless link is using two different channel and ESSID. TQ of the 4 links are vary from 250-255 at the initial state. I run udp iperf from A to C and observe tcpdump on node B for the two wireless interfaces (wlan0 and wlan1).

I observed that interface alternation is not happening always. Sometime the route will be using wlan0 or wlan1 all the way from A to C, as we observed at node B, only 1 interface has traffic on tcpdump. Sometime when interface alternation happened, tcpdump for wlan0 and wlan1 will show traffic.

I changed the topology to the below to test further:

A---------------B=========C

There is only 1 wireless link from A to B while there are two link from B to C. In this case, I observed that interface alternation will always happened, as I observed both wlan0 and wlan1 has traffic on tcpdump.

Could you please explain to me what has caused interface alternation not happening in the 1st topology and why the interface alternation will always happen in the second topology.

Thank you in advance, Ming Ann