SUMMARS #2: Two ethernet interfaces in one subnet ?

From: Thomas Leitner (
Date: Wed May 27 1998 - 02:49:21 CDT


After my first summary, I got a couple of e-mails with interesting
comments. The consensus is still: It is possible, but don't do it!

Thanks to: Eugene Kramer <>
           Jochen Bern <>
           NA/LMD/T/IA Roar Smith <>
           Parks Fields <>
           Seth Rothenberg <>
           Bismark Espinoza <bismark@alta.Jpl.Nasa.Gov>
           Chris Liljenstolpe <>

There are their replies.


From: Eugene Kramer <>

On Mon, 25 May 1998, Thomas Leitner wrote:

:The answer this: This is not a good idea because the hostid of the
:machine is used for the interface MAC address and both interfaces
:thus would have the same MAC address leading to much confusion
:with ARP. Since I really want to have all IP addresses in the same
:subnet, all I can do is to retire the le0 interface and stick with

Even though that this is *almost* true (the same MAC address is assigned
to all interfaces on the boot), you can overwrite that using ifconfig.
See a description of 'ether' address family in ifconfig's man pages.

I the suggestion of assigning reserved IP addresses (RFC 1918).


From: Jochen Bern <>

This is *WAY* too close to creating a Loop for my Tastes. If you try,
TRIPLEcheck that you turn off "IP Forwarding" on the SPARCstation

IMHO, if the Amount of Data to transfer warrants a 100BaseTX Link
to the 10/100 Switch, routing the remaining 10 Mbps of the le0
Interface through the Switch as well doesn't matter anymore.

Another Possibility would be to use the SPARC as Gateway/Firewall/
SomethingToThatEffect, and cut the Link between the two Switches


From: NA/LMD/T/IA Roar Smith <>


here is my late contribution to your question and summary.

You *can* have two interfaces on the same network, but then you must change
the MAC address of one of the interfaces in order not to violate Ethernet
standard and confuse the ARP tables.

You can set the MAC address of an interface in the ifconfig(1m) command.

Lets assume your hosts MAC address is 8:0:20:86:52:a0 and you use that on
hme0, then you can set the MAC address of le0 to a unique address on that

ifconfig le0 ether 48:0:20:86:52:a0

This must be done at each boot, so put it into a start script in /etc/rcS.d

I don't remember exactly where I found it, but the bit changed from 0 to 1
in the above MAC address substitution should be valid and unique according
to the Ethernet specifications - I used it on a server myself some time ago.

You probably won't get much effect from having two interfaces on the
same net though, and you should take care that le0 is not being used
as the default interface for all outgoing packets.


Roar Smith, M.Sc.E.E.
IT Architect L. M. Ericsson A/S
Phone: +45 3388 3577 Sluseholmen 8
FAX: +45 3388 3139 DK-1790 Copenhagen V

From: Parks Fields <>


using ifconfig you can assign a second MAC (ether) address also.
that way both cards have different IP and MAC addressed.


From: Seth Rothenberg <>

My system has two connections to the same
physical network. The main purpose for me is to
have one static interface for sysadm, and a
"public" IP address that can be moved to my
backup server using High Availability software.

In any case, the effect is the same as you need.
Your's might be a little different if you don't
have a flat network, but it may not matter.

We must have different IP addresses on our two
ethernet interfaces. This is so remote hosts
can tell the paths apart. We appear as two
separate hosts on the network. The two
(virtual) networks must have different network
numbers, so I use a different netmask on the
second network. I configure it manually, NOT
with a hostname.* file.

le0: - default netmask net# - see script, attached.

Once this is done, all inbound traffic uses
le1, and outbound to 110.31.11.* uses le1 also.

Everything else is le0. I don't care about it,
but I imagine you could add a route to force
110.31.11.* traffic to go OUT over le1 also.

Please let me know if I can help with other info.


  [ Part 2: "Attached Text" ]

# script configure le1
# /etc/rc2.d/S71configle1
# testdg IP address is
# With netmask ==> network #

if [ ! -d /usr/bin ]
then # /usr not mounted

echo "IN S71configle0" >> /tmp/success

date > /tmp/success
ifconfig -a >> /tmp/success 2>&1

echo "setting address for le1" >> /tmp/success 2>&1
ifconfig le1 plumb >> /tmp/success 2>&1
ifconfig le1 ether aa:0:04:0:2:50 >> /tmp/success 2>&1
ifconfig le1 inet testdg1 >> /tmp/success 2>&1
echo "setting netmask for le1" >> /tmp/success
ifconfig le1 netmask class_c broadcast + up >> /tmp/success 2>&1
ifconfig -a >> /tmp/success 2>&1


netstat -r >> /tmp/success

From: Bismark Espinoza <bismark@alta.Jpl.Nasa.Gov>

It would be better if you only connect the hme0 to
the 100 switch and leave le0 unconnected.

From: Chris Liljenstolpe <>


        If you are going to do this config, what does maintaining the
10Mb connection buy you (unless there is an over subscription problem
on the Dept 10Mb to 100Mb link)? As far as two interfaces on the same
subnet under solaris 2.5 is concerned, I am pretty sure you can't do
it, and would be very afraid, even if it was technically possible. In
short, don't.


------------------------- original posting ----------------------------

On Mon, 25 May 1998, Thomas Leitner wrote:

> We're in the process of upgrading our workgroup to 100MBit. So I've
> ordered a hme interface for our SparcStation 20 running Solaris 2.5.
> Now what I have in mind is this:
> to campus backbone
> !
> -----------------------
> ! Dept. 10Mbit Switch !
> -----------------------
> ! ! le0
> ---------------------------- -----------------
> ! 100Mbit Workgroup Switch ! ! Sparc 20 !
> ---------------------------- -----------------
> ! ! ! ! ! ! ! ! ! ! ! ! ! hme0
> ! ! ! ! ! ! ! ! ! ! ! !--------------
> Machines in our workgroup (new cable)
> (big workgroup server)
> The Sparc 20 actually belongs to our workgroup but due to space
> constaints it is located in another part of the building and
> connected to the 10MBit department switch.
> What I have in mind is to run a new cable between our workgroup switch
> and the Sparc 20 to the new hme0 interface.
> The question is: Does this make sense? Can I configure the Solaris box
> in a way that it runs all traffic to the Internet and other campus
> machines over the le0 interface but directs all traffic with our
> workgroup over the hme0 interface?
> All interfaces need to be in the same subnet, though and the main
> traffic is NFS where the Sun is an NFS client?
> Would a static routing entry like:
> route add host <big-workgroup-server> <name-of-hme-interface>
> do this?

T o m L e i t n e r Dept. of Communications
                                            Graz University of Technology,
e-mail : Inffeldgasse 12
Phone : +43-316-873-7455 A-8010 Graz / Austria / Europe
Fax : +43-316-463-697
Home page :
PGP public key on : or send
mail with subject "get Thomas Leitner" to
    Before we have the paperless office, we have the paperless toilet!

This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:12:40 CDT