Paul, thanks a lot for your quick response. I checked the log. Indeed, the peer id is the native ip. We already had uniqueids=no but we are using PSK. According to the manpage this setting ist ignored when authby=secret is used. I should have mentioned, that the described behaviour belongs solely to windows clients (we think so, but did not check very hard).
Today, I am not able to create a new test case on the openswan server, will do tomorrow. Here you can find libreswan.log and my ipsec.conf: https://cloud.uni-koblenz-landau.de/s/52fmDgFX9dMG4yF Will upload an openswan.log as soon as we have it. > Am 24.01.2022 um 16:52 schrieb Paul Wouters <[email protected]>: > > On Mon, 24 Jan 2022, Christoph Litauer wrote: > >> I use libreswan 3.29 on an Ubuntu 20.04.3 LTS. >> >> Until now we had a l2tp-setup using openswan 2.6.49 on an ubuntu 16.04.7. No >> problems, we had hundreds of users in parallel. Some of these users by >> accident use the same local ip address. >> >> For security we have to upgrade to Ubuntu 20 and thus had to change to >> libreswan. Setup is nearly identical. But now we have a serious problem: - >> User A uses local ip 192.168.55.45 behind NAT. Public IP is 87.156.203.224 - >> User B uses local ip 192.168.55.45 behind NAT. Public IP is 185.66.195.84 >> >> A starts the tunnel. As soon es B starts his tunnel, the tunnel for A is >> terminated and vice versa. > > Do they use a unique ID or are these clients that pick their native IP > as ID? If so, they would both share the same ID 192.168.55.45 and > libreswan would assume this is a reconnect. The libreswan behaviour > depends a bit on whether certs or PSKs are used. > > You can try setting uniqueids=no in "config setup" because I believe > Windows clients do use their pre-NAT IP as the ID for l2tp. > > A more modern solution would be to migrate to IKEv2 and avoid the > whole problem of transport mode and virtual-private= and L2TP > protocol overhead. > > I would be interested in a full debug log of this event of a second > client connecting with both openswan and libreswan if you can, so > I can better understand the differece and if there is a libreswan > mitigation we can do. > > Paul -- Freundliche Grüße Christoph Litauer _________________________________________ Christoph Litauer Uni Koblenz, Rechenzentrum, Raum A 022 Postfach 201602, 56016 Koblenz Fon: +49 261 287-1311, Fax: -100 1311 _______________________________________________ Swan mailing list [email protected] https://lists.libreswan.org/mailman/listinfo/swan
