From owner-freebsd-questions Mon Nov 11 00:44:53 1996 Return-Path: owner-questions Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id AAA00408 for questions-outgoing; Mon, 11 Nov 1996 00:44:53 -0800 (PST) Received: from gatekeeper.barcode.co.il (gatekeeper.barcode.co.il [192.116.93.17]) by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id AAA00402 for ; Mon, 11 Nov 1996 00:44:48 -0800 (PST) Received: (from smap@localhost) by gatekeeper.barcode.co.il (8.7.5/8.6.12) id KAA16933; Mon, 11 Nov 1996 10:42:50 +0200 (IST) X-Authentication-Warning: gatekeeper.barcode.co.il: smap set sender to using -f Received: from localhost.barcode.co.il(127.0.0.1) by gatekeeper.barcode.co.il via smap (V1.3) id sma016931; Mon Nov 11 10:42:22 1996 Message-ID: <3286E708.13F5@barcode.co.il> Date: Mon, 11 Nov 1996 10:42:48 +0200 From: Nadav Eiron X-Mailer: Mozilla 2.02 (X11; I; SunOS 5.5 sun4m) MIME-Version: 1.0 To: David Kelly CC: questions@freebsd.org Subject: Re: Ethernet Collisions References: Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: owner-questions@freebsd.org X-Loop: FreeBSD.org Precedence: bulk David Kelly wrote: > > I noticed something recently that's bothering me. Not sure if its > a FreeBSD problem, or more likely its a problem with the way the > network is wired to my office: > > Name Mtu Network Address Ipkts Ierrs Opkts Oerrs Coll > ed0 1500 00.00.b4.23.d1.21 22911 0 8188 0 4080 > ed0 1500 140.165.210/2 PeeCee 22911 0 8188 0 4080 > lp0* 1500 0 0 0 0 0 > tun0* 1500 0 0 0 0 0 > sl0* 552 0 0 0 0 0 > lo0 16384 2 0 2 0 0 > lo0 16384 your-net localhost 2 0 2 0 0 > > The above output of "netstat -i" is from: > PeeCee: {1003} uname -a > FreeBSD PeeCee.tbe.com 3.0-CURRENT FreeBSD 3.0-CURRENT #0: Wed Nov 6 18:22:58 > CST 1996 dkelly@PeeCee.tbe.com:/usr/src/sys/ > compile/PEECEE i386 > > and captured shortly after a reboot and ftp download of a 30M file. > The concern is how most every packet transmitted first collided with > another before getting out. This is repeatable, with either NE2000 > or DEC-21040 ethernet cards. Doesn't matter if ftp or ncftp is used. > Doesn't matter if the machine was freshly rebooted or has been > running for several days. The remote system is an old SGI, ping > reports TTL of 253 so there are one or two gateways between us. > > The network is UTP with a Farallon EtherWave in my office splitting > it between Mac and PeeCee. The collision light on the Etherwave > frist brought this to my attention. > > The de0 card still results in 800k to 900k transfer rates. Should > I worry about this? These rates are very high for something that goes through two gateways anyhow. I'd suspect the statistics are not accurate. How large are the files? (anything that is less than about half a MB might cause, in my experience, bogus statistics). I don't know if you should be worried, but I would give it further investigation. The easiest thing would have been to use a LAN analyzer, but those are usualy hard to come by. Here is my poor man's troubleshooting startegy: First, does this happen only for ftp's or for any type of traffic (I would suspect the latter). Usually excessive collisions are a result of defective hardware. Make your tests against a machine that is on the same segment from the same repeater, and disconnect everything else. If this solves the problem try finding out which node was the one that caused the problem by connecting them one by one. That node probably has a defective card. If disconnecting everything doesn't solve the problem, replace not just the card on your machine, but also *all* of the cables between it and the repeater (use a different wall outlet, different port on the repeater with different patch cables). A cable can cause excessive collisions (especially in 100BaseT4 and 10Base2, but also in 10BaseT and 100BaseTX). If all this fails to locate the source of the problem, try checking statistics for other machines connected to the same hub and see if they have similar symptoms. You may then want to try replacing the hub. I know this seems to be a lot of work. All I can say is that in my experience network cabling/hardware problem are very hard to find and very easy to fix. > -- > David Kelly N4HHE, dkelly@tomcat1.tbe.com (wk), dkelly@hiwaay.net (hm) > ===================================================================== > The human mind ordinarily operates at only ten percent of its > capacity -- the rest is overhead for the operating system. Good luck Nadav