Routing problem in RH5.1 (???) (Scheme redrawn)

["Paulo Afonso Graner Fessel @UOL" <pafessel uol com br>]

	Hi again.

	The scheme was unreadable, so here it goes again...

------- Forwarded Message Follows -------
From:           	Self <Single-user mode>
To:             	redhat-install-list redhat com
Subject:        	Routing problem in RH5.1 (???)
Send reply to:  	pafessel uol com br
Date sent:      	Thu, 29 Oct 1998 01:39:22 -0200

	Hi, guys. 

	PERHAPS an answer for this question has been posted somewhere else before.  
Anyway, I've searched thru the list archives and I haven't found a solution  
for the problem, so here it goes... 

	I have a Linux RH 5.1 machine which I'm setting up for use as a firewall.  
The logical topology of the network is as follows: 

	10.66.32.0/24                              10.66.16.0/24 
-----------------------------    -------------------------					 
      |     10.66.32.151   |      |	 10.66.16.151    | 10.66.16.150 
      |     (eth1)		+---------------+	    (eth0) +------------+ 
      |                |    RH 5.1     |           | Cisco 2501 | 
  +------+             | 2.0.35 Kernel |           +------------+ 
  |Win NT|             | (from RedHat) |
  | Wkst |             +---------------+	 
  +------+
10.66.32.240 

	The RH 5.1 has two NICs: one Intel EtherExpress Pro 100+ and one 3Com 3C905. 


	The case is as follows: from the RH5.1 machine, I can ping either the Cisco  
2501 and the Win NT Workstation machine, and this is the behavior I expect to 
 have. But from the NT, I can't ping the Cisco router. I can only reach the  
ethernet interface at 10.66.16.151; every other machine in the 10.66.16  
network remains inaccessible from the 10.66.32 network. In the 10.66.16 side, 
 I have also an AS/400 machine, and I can't ping it either. Also, if I do a  
"traceroute -i eth1 10.66.16.150" in the Linux box, I can't reach no machine  
at 10.66.16 network. 

	The routing tables are set up correctly: eth1 points to 10.66.32 network,  
eth0 points to 10.66.16 network and my default gateway as set in the RH  
machine is 10.66.16.150 (Cisco's ethernet). The Win NT workstation has  
10.66.32.151 as its default gateway (the eth1 network). 

	What can be wrong? I have enabled ip_forward both in the kernel and in  
/etc/sysconfig/network. When I make "echo x > /proc/sys/net/ipv4/ip_forward"  
with x = 1 or x = 0, I also can see in the syslog the messages "sysctl:  
ip_forwarding enabled" or "sysctl: ip_forwarding disabled" respectively (the  
messages aren't exactly these, but this is the meaning anyway). That is, the  
kernel is answering the sysctls to enable/disable ip forwarding. Finally,  
I've disabled all the blocking/forwarding rules and set the default policy to 
 "accept", with no results whatsoever. 

	I've thought about getting a pristine 2.0.35 kernel source (without the  
digital remastering done by RH) and compile it from scratch. Could this solve 
 the problem? 

	Also, is there a chance that this would be caused by a loop in the hubs or  
something like that? The site's physical installation is a little messy, and  
I'd like to know from you whether a loop in the hubs could also cause this. 

	TIA, 
		Paulo Fessel 


+---------------------------------------------------------------------------+
|       Paulo Afonso Graner Fessel, ou "Paulao" - pafessel uol com br       |
| E-mail no emprego: pafessel netsol com br | Tel domestico: 55-11-8262644  |
| "Isto não está certo. Aliás, isto nem mesmo está errado."(Wolfgang Pauli) |
+---------------------------------------------------------------------------+