On Wed, 8 Jun 2005, Jason Lunz wrote: > If many clients are connected to a server-mode openvpn instance running > in bridge mode, any client can inject arbitrary ethernet frames onto the > VPN. By arbitrary, I mean that clients can change source mac and source > IP at will. This is a good thing; it's flexible. > > However, this makes it useless to attempt user-based firewalling on the > vpn server on the packets arriving on the tap device. By the time a > given packet arrives at firewall rules on the tap device, there's no > non-spoofable way to determine which user sent the packet. > > How difficult would it be to add an option to enforce that packets > arriving from a particular user be discarded by openvpn if they do not > match a given mac address (and/or IP)? That way, firewall rules could > be written on the tap device with per-user granularity, and openvpn > would prevent any possibility of spoofing.
This has been discussed before on openvpn-users. See the "learn-address" script callback as described in the man page. This can be used to set firewall rules by client in a bridging configuration. James