ip alias problems

ip alias problems

Post by Steve Alligo » Tue, 01 Oct 1996 04:00:00

I am having problems with the ip alias function in Linux (v.2.0.18)

My main ip works fine with everything.

When I alias multiple ips, the routes seem fine, but I can't access them
from most of my LAN or the WAN.  They ping fine, and the udp test is fine,
but they fail the TCP test (Invalid queue element).  I have a Macintosh
running open transport that works just fine with the aliased ip, but
nothing else will.  (Broken pipes in Netscape).

Any help would be appreciated.

-Steve Alligood

----                                                         *

(801)224-8555  voice                                       *   *
                           Free Speech Online                *
                           Blue Ribbon Campain             *   *
                   <http://www.eff.org/blueribbon.html>   *     *


1. IP aliasing Problem


I just stumbled over this problem while migrating from RH5.2 to RH6:

What we have :

Interface  private IP                             real IP (translated by

cisco (NAT)
eth0        <-------> a.b.c.d1
eth0:1       <-------> a.b.c.d2
eth0:2  192.168.1. 127      <-------> a.b.c.d3
eth0:3       <-------> a.b.c.d4

stuff added to rc.local :

# routing
route add default gw cisco-gw
route add -net netmask gw

# virtual Interface(s)
ifconfig eth0:1 netmask up
ifconfig eth0:2 netmask up
ifconfig eth0:3 netmask up

My machine has one physical Interface (one IP) with 3 virtual IPs like
this {eth0 eth0:1 eth0:2 eth0:3} bound to the same
Interface. We have a cisco Router which translates (NATs) the private
IPs bound to the eth's to real IPs. Default Gateway is the cisco
(cisco-gw).  With RH5.2 this works as expected  - _all_ 4 adresses are
global visible (Internet) via the NAT translations  - i.e  i can reach
each a.b.c.d?.

Now we moved to a SMP machine with RH6 (default Kernel Version
re-compiled to match processor etc.) and absolut
identical Configuration (copied from old machines rc.local to new ones
!) and ended up
with only the first adress 'global' visible (the eth0 !) and all 3
virtual adresses local visible (private LAN) only.

This means i can connect to all private IPs of the eth:x Interfaces
(192.168.1.x) but get no reply to the mapped ones
(a.b.c.d?). Can someone please explain why this doesn't work anymore and

what to do. To me it seems the default
gateway bound to the virtual IPs is broken or the configuration
procedure has changed.

PS Due to this 'bug'/whatever we moved back to the old machine until
     this problem is resolved (the old machine works
     fine as ever - so this is definitly a Linux problem !)

thanks in advance


UNIX SysAdmin - BCI GmbH

2. Possible bug in cc involving signed/unsigned chars

3. Linux 2.0.0 IP Aliasing Problem

4. Networking setup

5. IP Aliases problem with Redhat and Eth1:8


7. ip aliasing problem

8. Trying to compile 2.1.4

9. IP Aliasing problem

10. IP Alias problem

11. IP Aliasing problem

12. My machine isn't itself (IP Alias problem)