[Olsr-dev] segfault in -r3 reproduced

Markus Kittenberger (spam-protected)
Sat Dec 6 18:12:39 CET 2008


This bug also showed up yesterday afternoon (every some minutes), but as we
wanted to debug it in the evening everything was fine,..

today henning and me triggered it with doing ifdown eth1 && sleep 1 && ifup
eth1 serveral times

above is not very likely under real conditions,..

but on this machine there are many openvpn tunnels olsr runs on,.. which go
up and down quite often

so we also tried openvpn restarts but nothing happened,..

now i tried are more realistic approch, i blocked the openvpn traffic, to
get a typical tunnel timeout shutdown

and about nearly 50% of my tries the same segfault happened,.. so im quite
sure that this also triggered the problems yesterday (as there were some
instable tunnels around at his time, which were comlpetely offline in the
evening (when we couldn`t find a problem))

----

but during triggering this segfaults i also found some other bugs/strange
things

it seems that even if no segfault happens olsr tends to forget to update the
kernel routing table, under this test conditions,..

and once as there were stable ~ 90% wrong routes in the kernel (wrong meand
90% of the routes of txtinfo plugin where different (mostly different) or
missing in the kernel table)

as i looked around i found following Links table in txtinfo (httpinfo had
the same)

Table: Links
Local IP Remote IP Hyst. LQ NLQ Cost
0.0.0.0 2.0.0.0 0.00 0.000 0.000 1.000
107.105.101.7 2.0.0.0 0.00 0.000 0.000 1.860
0.0.0.0 2.0.0.0 0.00 0.000 0.000 1.016
0.0.0.0 2.0.0.0 0.00 0.000 0.000 1.020
107.105.101.7 2.0.0.0 0.00 0.000 0.000 1.000
0.0.0.0 2.0.0.0 0.00 0.000 0.000 3.735
0.0.0.0 2.0.0.0 0.00 0.000 0.000 1.000
0.0.0.0 100.0.0.0 0.00 0.000 0.000 1.000
0.0.0.0 100.0.0.0 0.00 0.000 0.000 1.000
107.105.101.7 100.0.0.0 0.00 0.000 0.000 1.000
0.0.0.0 2.0.0.0 0.00 0.000 0.000 3.609
107.105.101.6 100.0.0.0 0.00 0.000 0.000 1.000
0.0.0.0 100.0.0.0 0.00 0.000 0.000 1.000
0.0.0.0 100.0.0.0 0.00 0.000 0.000 1.000

usually there are 78.41.112.2 (instead of 0.0.0.0) and
193.238.159.100(instead of
107.105.101.7, which looks lik ea completely random number for me)

78.41.112.2 belongs to the eth1 devcie of this machine, which has been
ifupdowned several times (while olsr was starting up) but the interfaces
with 193.238.159.100 (openvpn tunnels) should have operated normally, ...

Markus
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.olsr.org/pipermail/olsr-dev/attachments/20081206/b2da37b0/attachment.html>


More information about the Olsr-dev mailing list