[Openswan Users] PPP issues (was Re: openswan + l2tpd + iptables problem)

[mechanix at debian.org]

On Thu, Oct 19, 2006 at 07:30:25PM +0200, Paul Wouters wrote:
> On Thu, 19 Oct 2006, mechanix at debian.org wrote:
> 
> > I discovered something else: when I run tcpdump on the gateway, it seems
> > to `forget' how to reach the client - which seems to coincide with the
> > moment packets stop getting sent to the client - seeing that it performs
> > arp requests, which are never answered.
> >
> > Does anyone know what could be up with that?
> 
> Michael suggests this could be because you are losing arp packets.
> (then i looked at the log below, and indeed, it looks like that is the case)
> 
> > 22:57:44.146242 IP GW.IP.ADDR.ESS.4500 > REM.OTE.ADDR.ESS.4500: UDP-encap: ESP(spi=0xb1f325b8,seq=0x10), length 52
> > 22:57:46.267167 arp who-has REM.OTE.ADDR.ESS tell GW.IP.ADDR.ESS
> > 22:57:48.969962 IP REM.OTE.ADDR.ESS.4500 > GW.IP.ADDR.ESS.4500: UDP-encap: ESP(spi=0x537cb518,seq=0x17), length 76
> > 22:57:48.975399 arp who-has REM.OTE.ADDR.ESS tell GW.IP.ADDR.ESS
> > 22:57:49.975476 arp who-has REM.OTE.ADDR.ESS tell GW.IP.ADDR.ESS
> > 22:57:50.975562 arp who-has REM.OTE.ADDR.ESS tell GW.IP.ADDR.ESS
> 
> > Oct 19 22:57:42 scotos l2tpd[20820]: ourtid = 12824, entropy_buf = 3218
> 
> Does the problem go away with xl2tpd ? We did quite some fixing of the l2tpd code.

Unfortunately, not. Well, actually I seem to run into a different problem
again -- with xl2tpd I now get this in /var/log/debug (and nothing more):

Oct 23 16:04:45 scotos l2tpd[27457]: control_finish: Peer requested tunnel 4 twice, ignoring second one.
Oct 23 16:04:50 scotos last message repeated 2 times
Oct 23 16:04:55 scotos l2tpd[27457]: Unable to deliver closing message for tunnel 18614. Destroying anyway.
Oct 23 16:05:08 scotos l2tpd[27457]: control_finish: Peer requested tunnel 4 twice, ignoring second one.
Oct 23 16:05:10 scotos l2tpd[27457]: Unable to deliver closing message for tunnel 62456. Destroying anyway.

pppd never gets started this time.

Running tcpdump at the same time shows that arp packets are still lost,
I see the same arp who-has queries as before.

I also tried with not iptables rules at all just to make sure nothing was
in the way there but the results were the same.


Regards,

Filip


http://www.evonet.be/~filipvr/