[strongSwan-dev] DHCP plugin ignores "server" parameter

[Thomas Klute]

Hi everyone,

I'm running a Strongswan-based experimental system on a virtual machine
with two network interface, eth0 and eth1. Strongswan is supposed to
accept connections on eth0 (responder) and get virtual IP addresses for
the clients from a DHCP server which is listening on eth1 (on the same
host). The problem is that DHCP packets get sent over the wrong interface.

I have set
charon.plugins.dhcp.server = 10.50.1.255 # Broadcast address for eth1
and
charon.plugins.dhcp.force_server_address = yes
in strongswan.conf, but the DHCP discover packets are still sent on eth0
with a destination address of 255.255.255.255 (confirmed by packet
capture). Additionally, even if I configure the DHCP server to listen on
eth0, Strongswan ignores the DHCP offers.

Can anyone tell me how to resolve or further debug this? I have tried
both the Debian package from wheezy-backports (5.1.0-3~bpo70+1) and the
latest version from git (5.1.2dr2, built locally).

I have tried the patch at
https://lists.strongswan.org/pipermail/dev/2013-November/000948.html
and the "interface" parameter defined there, but without result.

Regards,
Thomas