Path: utzoo!attcan!uunet!snorkelwacker!usc!elroy.jpl.nasa.gov!decwrl!ucbvax!isis.u-strasbg.FR!pansiot
From: pansiot@isis.u-strasbg.FR (Jean-Jacques Pansiot Departement d'Informatique
 Universite Louis Pasteur Strasbourg FRANCE)
Newsgroups: comp.protocols.tcp-ip
Subject: broadcast storm
Message-ID: <9003121734.AA22356@ISIS.U-STRASBG.FR>
Date: 12 Mar 90 17:34:35 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 37

I know that there is a phenomenon called broadcast storm, and i thing  our
network has this kind of problem. Here is the situation:

We use a class B network (130.79) over ethernet.
>From time to time some stations (lets call them type A) send a broadcast,
with address 130.79.255.255 (I believe this is the correct Broadcast IP Address)
Currently such broadcasts are RWHO  or RIP.
Some stations (lets call them type B), retransmit the same packet(  as if they
were IP routers) on the same ethernet, with TTL decreased by one. Moreover,
they seem to receive their own retransmission, so the whole thing repeats
 itself,
until TTL is decreased from 15 to 0.

Some stations (lets call them type C), when receiving these packets,
send an ARP request for IP address
130.79.255.255. They do not recognize it as a broadcast address. I understand
that IP broadcast was formerly all 0. But why this ARP ? , as if they
wanted to route the packet  (as type B stations do ).

Fortunately, only one station (for now) is of type B. Still , for  1
(correct) broadcast, we have 15 * number of type C stations   packets.
At the end there is also some ICMP TTL Exceeded and ICMP Destination Unreach.
Very much for one broadcast. We could try to kill all rwhod, but I don't see
how to avoid RIP messages from our gateway if we wanto use routed on
our station?

Anybody to explain the behaviour of stations B and C?

Anything I can do? (Stations B and C
are mostly SunOS 3.5 , no SUNOS4.0). I do not , of course, have
the power to do upgrades on these machines, but I could recommend doing
a few changes (ifconfig....)

any help? may be this is a well known problem?

jean-jacques Pansiot, Universite Louis Pasteur, Strasbourg, France
pansiot@isis.u-strasbg.fr