On Tue, Apr 22, 2003 at 11:21:47AM +0200, Martin Husemann wrote:
> On Tue, Apr 22, 2003 at 10:49:10AM +0300, Juha K Kallio wrote:
> > Doesn't the in-kernel pppoe use LCP echo to detect dead connections?
> > If the connection dies silently, it just stays 
> > forever in 'network' state.
> 
> It does. And it logs it's failures like this:
> 
> Apr 22 04:39:54 nightfall /netbsd: pppoe0: LCP keepalive timed out, going to restart the connection
> 
> Maybe your peer is answering LCP echo requests, but IP routing does not work
> any more?
> 
> You can check by running "tcpdump -i pppoe0 -vvv" or something similar
> on the PPPoE interface and watch the regular (10 sec from memory) LCP echos.
> And then ping some remote host and check the IP packets.
> 
> Martin

No, it was not answering anymore. This is from a linux machine with the same peer:

Apr 21 23:46:48 porofarmi pppd[6868]: No response to 3 echo-requests
Apr 21 23:46:48 porofarmi pppd[6868]: Serial link appears to be disconnected.
Apr 21 23:46:54 porofarmi pppd[6868]: Connection terminated.

Now that you gave the message, it seems familiar. I've seen it with another NetBSD machine, with the same 
peer. The only difference is that I used PAP there, and the "magic" 0.0.0.0 and 0.0.0.1 IPs. With my own
installation, I'm using CHAP and the real IPs, though I can't see why that would change anything.