[Olsr-dev] Possible Bug

[elektra]

Hi -

I assume I came across a bug in the current developer version in CVS of  
2008-01-14 09:45. I'm testing several protocols in a 7x7 grid under lab 
conditions (physical hardware, not simulation). It happened to me twice, 
that all nodes in the routing table have a metric of 2, even if 'ping 
-R' shows severall hops. Even single hop links get metric 2:


(spam-protected):~# ip r
172.30.1.32 via 172.30.1.25 dev ath0  metric 2
172.30.1.33 via 172.30.1.25 dev ath0  metric 2
172.30.1.34 via 172.30.1.25 dev ath0  metric 2
172.30.1.35 via 172.30.1.25 dev ath0  metric 2
172.30.1.36 via 172.30.1.26 dev ath0  metric 2
172.30.1.37 via 172.30.1.26 dev ath0  metric 2
172.30.1.41 via 172.30.1.25 dev ath0  metric 2
172.30.1.42 via 172.30.1.25 dev ath0  metric 2
172.30.1.43 via 172.30.1.25 dev ath0  metric 2
172.30.1.44 via 172.30.1.25 dev ath0  metric 2
172.30.1.45 via 172.30.1.25 dev ath0  metric 2
172.30.1.46 via 172.30.1.25 dev ath0  metric 2
172.30.1.47 via 172.30.1.26 dev ath0  metric 2
172.30.1.51 via 172.30.1.25 dev ath0  metric 2
172.30.1.52 via 172.30.1.25 dev ath0  metric 2
172.30.1.53 via 172.30.1.25 dev ath0  metric 2
172.30.1.54 via 172.30.1.25 dev ath0  metric 2
172.30.1.55 via 172.30.1.26 dev ath0  metric 2
172.30.1.56 via 172.30.1.26 dev ath0  metric 2
172.30.1.57 via 172.30.1.25 dev ath0  metric 2
172.30.1.61 via 172.30.1.25 dev ath0  metric 2
172.30.1.62 via 172.30.1.25 dev ath0  metric 2
172.30.1.63 via 172.30.1.25 dev ath0  metric 2
172.30.1.11 via 172.30.1.25 dev ath0  metric 2
172.30.1.13 via 172.30.1.25 dev ath0  metric 2
172.30.1.12 via 172.30.1.25 dev ath0  metric 2
172.30.1.15 via 172.30.1.25 dev ath0  metric 2
172.30.1.14 via 172.30.1.25 dev ath0  metric 2
172.30.1.16 dev ath0  scope link  metric 2
172.30.1.21 via 172.30.1.25 dev ath0  metric 2
172.30.1.23 via 172.30.1.25 dev ath0  metric 2
172.30.1.22 via 172.30.1.25 dev ath0  metric 2
172.30.1.25 dev ath0  scope link  metric 2
172.30.1.24 via 172.30.1.25 dev ath0  metric 2
172.30.1.27 via 172.30.1.26 dev ath0  metric 2
172.30.1.26 dev ath0  scope link  metric 2
172.30.1.31 via 172.30.1.25 dev ath0  metric 2
172.30.1.67 via 172.30.1.26 dev ath0  metric 2
172.30.1.66 via 172.30.1.26 dev ath0  metric 2
172.30.1.65 via 172.30.1.26 dev ath0  metric 2
172.30.1.64 via 172.30.1.25 dev ath0  metric 2
172.30.1.71 via 172.30.1.25 dev ath0  metric 2
172.30.1.75 via 172.30.1.26 dev ath0  metric 2
172.30.1.74 via 172.30.1.26 dev ath0  metric 2
172.30.1.73 via 172.30.1.26 dev ath0  metric 2
172.30.1.72 via 172.30.1.26 dev ath0  metric 2
172.30.1.77 via 172.30.1.26 dev ath0  metric 2
172.30.1.76 via 172.30.1.26 dev ath0  metric 2
172.30.1.0/24 dev ath0  proto kernel  scope link  src 172.30.1.17
172.20.1.0/24 dev eth0  proto kernel  scope link  src 172.20.1.17



Also the routing results are bad, of course. CPU load is ~0. Sometimes 
it works fine, I get about 13% load without fisheye enabled and the 
routes that are chosen are fine. The wifi signal is intentionally very 
weak so the protocol must build routes with up to 7 hops when connecting 
from one corner of the grid to the opposite corner. I get 3% ping packet 
loss from corner to corner - which is very good (same as with 
Batman-Experimental, btw.)

I have a  dump (~520kByte compressed) of debuglevel 9 from one of the 
nodes. I guess I shouldn't send that by mail to the list...

I had to stop and restart some of the olsrd's - maybe that is the reason 
why it gets confused.

cu elektra