On Fri, Oct 03, 2003 at 01:20:45PM +0200, Hauke Fath wrote:
> While we're on the issue of pppoe and ifwatchd... I noted that with a
> -current kernel on 1.6.1 userland, ifwatchd gets an "extra round" of
> up/down events during the daily disconnect [German Telekom]:

Something similar does sometimes happen with a -current ifwatchd too:

Oct  3 00:35:27 nightfall /usr/sbin/ifwatchd[149]: calling: /etc/ppp/ip-down pppoe0 /dev/null 9600 217.225.24.49 217.5.98.30 
Oct  3 00:35:30 nightfall /usr/sbin/ifwatchd[149]: calling: /etc/ppp/ip-up pppoe0 /dev/null 9600 0.0.0.0 0.0.0.1 

> This is somewhat annoying because the ip-{up,down} scripts communicate the
> ip address change to a nameserver, which then gets irritated by '0.0.0.?'
> ip addresses.

Yes, this is .. wrong.

> Is this a mismatch between the 1.6.1 ifwatchd and -current kernel
> interfaces, or a -current kernel issue?

No. Ifwatchd only uses the routing socket, and that API/ABI hasn't changed
in a long time (and should not change at all).

I've had no time to investigate, please file a PR.
It would be helpful if you could log the output of "route monitor" in parallel
when this happens the next time, and add an excerpt from that log to the PR.

Martin