---------- Original Message ----------------------------------
From: patrick conlin <[EMAIL PROTECTED]>
Date:  Fri, 24 May 2002 16:02:02 -0400

>on 5/24/02 14:51, [EMAIL PROTECTED] wrote:
>
>> interface Ethernet0
>> ip address yyy.yyy.yyy.yyy 255.255.255.252
>
>> interface Ethernet1
>> ip address xxx.xxx.xxx.2 255.255.254.0
>
>> ip route 10.101.104.0 255.255.248.0 xxx.xxx.xxx.170
>> ip route 10.102.104.0 255.255.248.0 xxx.xxx.xxx.174
>> ip route 10.103.104.0 255.255.248.0 xxx.xxx.xxx.175
>> access-list 1 permit 10.101.104.0 0.0.7.255
>> access-list 1 permit 10.102.104.0 0.0.7.255
>> access-list 1 permit 10.103.104.0 0.0.7.255
> 
>> yyy.yyy.yyy.yyy is the upstream provider's IP for our router.
>> xxx.xxx.xxx addresses are our own public block of IP addresses.
>
>It looks like you've got the following setup (is this correct)?
>
>isp <-> cisco <-> prop router 01 (xxx.xxx.xxx.170) <-> 10.101.104.0
>              <-> prop router 02 (xxx.xxx.xxx.174) <-> 10.102.104.0
>              <-> prop router 03 (xxx.xxx.xxx.175) <-> 10.103.104.0
>
>All hosts on 10.101.104.0 gets natted to xxx.xxx.xxx.170
>All hosts on 10.102.104.0 gets natted to xxx.xxx.xxx.174
>All hosts on 10.103.104.0 gets natted to xxx.xxx.xxx.175

That's the layout.

>
>If your linux box has 4 free pci slots, fill 'em up with nics. Then you can
>eliminate the cisco box and the three routers.
>
>isp <-> eth0 <-> netfilter <-> eth1 <-> 10.101.104.0
>                           <-> eth2 <-> 10.102.104.0
>                           <-> eth3 <-> 10.103.104.0
>
>Then it's just a matter of standard SNAT and DNAT based on interface.

Well, in our case there is an added complexity.  The proprietary wireless 
routers/transmitters are actually sitting on towers, and we are using wireless bridges 
to communicate between them and to provide the backhaul for the NATed hosts.

>
>Why the funky netmasks on the 3 internal networks? Do they have subnetworks
>of their own?

Partially because we expect to have a large number of hosts on each subnet, and due to 
various limitations, each of the three networks must be on a different subnet.

As an endnote.  With the help of one of your posters, Joe Patterson, I have come up 
what appears to be the perfect solution.  I am using the 10.103.104.0/21 subnet as a 
testbed, since we don't have a tower installed yet, and the following line works just 
great!

iptables -t nat -A POSTROUTING -o eth0 -s 10.103.104.0/21 -j MASQUERADE

I have the FTP module loaded, and that works without a problem.

Only two things are really left to figure out right now.  We do have a couple of our 
public IP addresses mapped to NATed hosts via the "ip nat inside source static" and 
"ip nat outside source static" commands.  I will have to have this functionality.

The other thing I will need, though if I can't do it, I won't cry too many tears, is 
to give one of our machines on the public address side the ability to contact all the 
NAT hosts.

-- 
A. Clausen

Reply via email to