On Monday, 25 February 2019 15:06:21 CET umut deniz wrote: [...]
Thank your for your swift reply and advice. As you said TI supports limited number of peers with its driver and hardware in mesh mode. According to TI documentation it is 32 nodes.
I think this is the number of nodes in a complete 802.11s mesh. This doesn't mean that it supports these number of direct peers the driver can't even represent more than 16 direct peers/links in its structures. I heard something of 10 peers supported by the firmware when running in AP mode. Most likely lower or equal in meshpoint mode.
And maybe it also needs some special peers to work correctly. Wouldn't be the first firmware which needs such kind of hacks.
The question bogging my mind is how could the < :#iw dev adhoc0 station dump > command return 6 active and plink established nodes (with inactive times around 80ms) while batman-adv's last-seen is increasing till timeout occurs on all neighbors and after timeout iw station dump still reports the same active stations (with inactive times around 80ms). Am I missing something to debug or a log to look in to or a configuration tied with batman-adv needed to be made.
If the last time increases, this means that the lower link doesn't do correct broadcast anymore between the peers. The numbers you see in iw can be from any kind of frame (there might be exceptions - check the ieee80211 subsystem for details).
This can either be a bug in the TI firmware or driver or maybe this is their expected behavior when going over a specific number of entries. But there is nothing batman-adv can do when the broadcast doesn't work correctly. B.A.T.M.A.N. IV's metric is build around the broadcast frames and also B.A.T.M.A.N. V needs them for local discovery.
Kind regards, Sven