From owner-freebsd-hackers  Mon Mar 18 10:32:01 1996
Return-Path: owner-hackers
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.3/8.7.3) id KAA11743
          for hackers-outgoing; Mon, 18 Mar 1996 10:32:01 -0800 (PST)
Received: from central.cis.upenn.edu (CENTRAL.CIS.UPENN.EDU [158.130.12.2])
          by freefall.freebsd.org (8.7.3/8.7.3) with ESMTP id KAA11710
          for <hackers@freebsd.org>; Mon, 18 Mar 1996 10:31:59 -0800 (PST)
Received: from gradine.cis.upenn.edu (GRADINE.CIS.UPENN.EDU [158.130.4.3]) by central.cis.upenn.edu (8.6.12/UPenn 1.4) with ESMTP 
	id NAA10057 for <hackers@freebsd.org>; Mon, 18 Mar 1996 13:31:43 -0500
Received: by gradine.cis.upenn.edu
	id NAA03117; Mon, 18 Mar 1996 13:31:42 -0500
From: dherbst@gradient.cis.upenn.edu (Darrel Herbst)
Posted-Date: Mon, 18 Mar 1996 13:31:42 -0500
Message-Id: <199603181831.NAA03117@gradine.cis.upenn.edu>
Subject: 3c509 mcast: host is down
To: hackers@freebsd.org
Date: Mon, 18 Mar 1996 13:31:42 -0500 (EST)
X-Mailer: ELM [version 2.4 PL23-upenn3.1]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: owner-hackers@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

I'm having trouble with the ep driver for the 3c509.  I'm opening up a
socket as AF_INET and SOCK_DGRAM and then sending out packets on a
multicast address... after about 100 packets or so, I start getting
sendto errors on my socket that the "host is down".

At first, I thought that by putting in the ether_addmulti calls that
are in the if_ed.c driver would solve my problem, but it hasn't.

I think what's happening is that when I add myself to the mcast group,
the driver should put an entry for the mcast ip in the arp table.
However, when this arp entry times out, it's not getting re-added to
the table, but the system is complaining that the host is down.  Is
this a correct assessment of what's happening?

Am I mistaken in thinking that the ether_addmulti call should add the
mcast ip to the arp table as a mcast ip -- one that shouldn't time-out
really?

Has anyone encountered this problem?

--dherbst@gradient.cis.upenn.edu