RE: routing woes|
Wed, 21 Nov 2001 00:10:51 +0100|
Could you send the list a sanitized copy of your option files so that we can
better help you? I was going to tell you how to automatically route the
network when the CIPE connection came up, but the error message you are
getting makes me think there is a deeper issue...
The basic idea is this: When CIPE brings up the point-to-point link, the
kernel adds a host route to the other end. That is all it can do because it
doesn't have knowledge about the net mask of the other end. After the
interface is created, CIPE then calls /etc/cipe/ip-up which you can modify
to create a net route to the other network. For example:
route add -net 192.168.2.0/24 gw $5 (I'm doing this from memory -- check
that $5 represents the IP address of the peer's p-t-p IP address).
> Greetings all,
> I have a small problem concerning the route and routing philosophy.
> Let's assume that I have the IP ranges 192.168.1-2 at my disposal and
> that they are routed "normally" in internet.
> 192.168.1.1 is a gateway (inet) & firewall
> 192.168.2.1 is a gateway (inet) & firewall
> other IP's are workstations (yes, they have a internet-visible IP) which
> are protected by the firewalls.
> a visualization:
> workstations gateway / fw gateway / fw workstations
> Now, I need to connect these two nets using CIPE.
> My problem is this:
> Once I create a CIPE interface in 1.1, it creates a route to 2.1 (the
> host, not the net 2.0).
> Now, when I try to create a net route to 2.0 through 2.1, it fails. I
> tried also to assign 2.2 to CIPE interface, but that failed also. Both
> tries ended up with CIPE error
> Nov 20 18:52:27 xyz kernel: cipcb0: looped route
> I'm not a routing guru and I'm a bit frustrated (well, a lot
> I understand that my 2nd solution fails when it has the gateway IP (2.2)
> in the net range (2.0) also, and both use the cipcb0 interface. (2.1
> route uses eth1, not cipcb0)
> The first should work, methinks.. future development or my stupidity?
> (please, my stupidity..)
> So, the philosophical question would be something like this: While CIPE
> is a point-to-point protocol, why couldn't packages be routed to it with
> a net target? or something, as the routing vocabulary is just starting
> to build up...