possible (kernel) bug with zebra

Roman Sidnev sirok at mail.ru
Thu Dec 25 15:09:34 PST 2003


>I don't know if my machine has some hardware problem, but I've noticed this
>strange behavior with zebra 0.93b_7 on 4.9-STABLE.
I think it not hardware problem, i have same problem
with gre and tun(vtund) interfaces and quagga. Panics occurs sometimes
when interface state changes (up/down). Sometimes kernel dont panics, but
zebra (quagga) dies. I don't know any solution. I had the same problem
with 4.8-STABLE.

>The first problem is zebra's inconsistent(?) handling of routing information,
>especialy when it comes to point-to-point interfaces (like tun) and ones
>handled by ppp(8).
>When ppp shuts down a link, it first deletes all routes, including the
>route to the remote host. Then it downs the interface. Zebra gets confused
>about this, because it gets the RTM_DELETE messages, but not the RTM_DELADDR
>message it seems to expect. (Which happens if you just do ifconfig -alias,
>there is a RTM_DELETE and then RTM_DELADDR)
>As a result, the zebra's routing table becomes bogus and the advertised
>routes are not correct. To fix this for now, I've put a script to do a
>ifconfig -alias which is run from ppp.linkdown.
>So far so good, but the kernel starts to panic :/
>Attached are the results from two consecutive panics

>I can provide more information/do more tests if someone finds this
>interesting :) Any help is appreciated, of course

Best regards,
 Roman                          mailto:sirok at mail.ru

More information about the freebsd-stable mailing list