Mysterious problem: can't backtrack an unwise router installation

[Frank Reichenbacher]

I have a small home network which ran wonderfully for several years with an
RH 7.0 box as the network gateway and firewall (using pmfirewall). The home
LAN is all on static IPs: 192.168.1.1-5 and the IP assigned to me by my ISP
is 66.93.153.62, with a gateway of 66.93.153.1. Pmfirewall masqued
everything internally to the outside network. I have an apache webserver on
the RH 7.0 box with about a dozen virtual servers configured in httpd.conf.
I make a few bucks a month on a ezmlm email list using qmail also on the RH
7.0 box. The domain, mollynet.com is DNSed at Zoneedit.com. 

I've been using a separate modem and dial-up connection with MS Remote
Desktop from my XP workstation to connect to my office network. This is very
slow and unsatisfactory, so I installed a MultiTech RouterFinder 560 at the
office on our W2K server/XP workstation LAN to help facilitate remote
connections (I have other remote users to the office LAN using XP). 

My computer consultant then talked me into installing the same router on my
home LAN and disabling the RH 7.0 box as the LAN gateway so that I could use
the speed of my DSL connection to connect to my office LAN. This was
necessary because of my inability to configure pmfirewall to let me get
through the RH 7.0 box using MS Remote Desktop from my home LAN XP
workstation. I figured it would be much easier to accomplish this task with
the same VPN router on both ends of the conection.

This involved the following steps:

1. Removed the ethernet cable connecting the DSL modem from eth0 (3com 3c900
Combo, 3c59x driver) on the RH 7.0 box and plugged it into the SOHO Router
outlet labelled "WAN".

2. Used netcfg to deactivate eth0

3. Used netcfg to change the gateway on the RH 7.0 box to 192.168.1.100 (the
SOHO Router) and make eth1 (3com 3c905C-TX Fast Etherlink, 3c90x driver, the
internal LAN device) as the gateway device. All I did on the XP workstations
was to change the gateway from 66.93.152.62 to 192.168.1.100. 

4. Used the handy-dandy browser configuration program that came with the
router to redirect ports 80, 110, 25, etc. to 192.168.1.1 (the RH 7.0 box)
so that my web and email server would still function.

3. Ran /etc/rc.d/init.d/pmfirewall stop to shut down the firewall.

It worked.

...sort of.

I have no idea why, but my virtual webs are ignored and all http requests
come to the root web at www.mollynet.com. The http logs (which are set up
for each virtual server in httpd.conf) show no activity to any of the
virtual servers and all activity is now directed to transfer log. And I am
unable to pop3 to local qmail email accounts from any of the LAN
workstations and I can't use the qmail smtp sever either.

This is obviously no good, but what is worse is that I now find I cannot
return the system to its original configuration at all. It is as if using
netcfg to deactivate eth0 has actually permanently deactivated it.
Retracking my steps results in a non-functioning network that can't ping to
the WAN at all. So my question to the group really is about this: why does
using netcfg to reactivate eth0 not work? How can I diagnose this problem to
determine whether eth0 is totally dead for some reason?

Frank