Date: Mon, 22 Jul 1996 13:06:39 -0700 From: Terry Lee <terryl@ienet.com> To: Kent Hamilton <kenth@HNS.St-Louis.Mo.US> Cc: Pius Fischer <pius@iago.ienet.com>, freebsd-hackers@FreeBSD.ORG Subject: Re: bizarre network problem Message-ID: <31F3DF4F.68DE@ienet.com> References: <199607210811.DAA06206@gwydion.hns.st-louis.mo.us>
next in thread | previous in thread | raw e-mail | index | archive | help
Kent Hamilton wrote: > > Just FYI I'm seeing the exact same problem on one of my machines > at work. The box is a P100 2.1-STABLE, with two Intel EtherExpress > Pro 100B cards, on fxp0 and fpx1. Our problem turned out to have nothing to do with FreeBSD. The problems turned out to be improper address table aging. Our LinkSwitch was expiring MAC addresses in 15 minutes, but our router ARP cache entries were set expire in 4 hours. Setting the router ARP cache expiration to <= the switch expiration fixed the problem. These switching hubs can be kind of tricky. > > We've been experiencing a strange network problem with a machine > > running FreeBSD 2.1-stable. Any help would be greatly appreciated. > > > > Here's out setup: > > > > ---------------------- > > | portia.ienet.com | ---------------------- ------------------- > > | Pentium 166 MHz |----|3Com LinkSwitch 1000|--|gate.ienet.com |--> UUNet > > | FreeBSD 2.1-STABLE | ---------------------- |Cisco 2500 series| > > ---------------------- ------------------- > > > > So, we've got a Cisco connecting us to the Internet via a T-1. The > > Cisco connects to our Ethernet hub (the LinkSwitch) via 10 Mb/s Ethernet > > and the LinkSwitch connects to the host portia via 100 Mb/s Ethernet. > > > > We've also got a few other hosts connected to the LinkSwitch. > > > > The problem is that every once in a while packets coming from the outside > > (coming in through our Cisco) don't make it to portia. This happens usually > > whenever there is little network traffic to portia for a period of time. > > traceroutes from the outside will get to the Cisco but not to portia. > > > > We put the Cisco into debugging mode (with 'debug ip packet') and it > > appears to correctly forward the packets to portia, but a tcpdump process > > running on portia never showed us any of those packets. > > > > This problem never occurs with packets sent from any of the other hosts > > connected to the Ethernet hub. In fact, if I ping or telnet to portia from > > one of these local hosts, portia appears to magically wake up and process > > all packets correctly. > > > > And if I'm on portia's console and do an 'arp -a', normally I get a quick > > response and see the arp cache. However, if this problem is happening, then > > doing an 'arp -a' takes about 3 or 4 seconds before showing me the cache > > and, again, portia wakes up and all packets are received and replied to. > > > > It appears that at the time the problem is happening, the arp cache entry > > for the Cisco is incomplete, but that is probably just because it expired > > and so shouldn't portia just reissue an arp query? Well, I guess it wouldn't > > if it's not receiving anything to prompt the query. > > > > So the question is, is this a problem with the kernel or with our LinkSwitch > > or with something else? What does 'arp -a' do that could cause the kernel to > > wake up? It never fails to receive local packets like RIP updates every 30 > > seconds and broadcast ARP queries, etc., but those, unlike the ping or telnet > > don't cause it to wake up to remote traffic. > > > > portia uses an Intel EtherExpress PRO/100 LAN Adapter and the fxp0 device > > driver. > > > > Any ideas? > > > > Thanks, > > Pius > > > > -- > Kent Hamilton Play: KentH@HNS.St-Louis.MO.US > NIC Handle: KH91 URL: http://www.icon-stl.net/~khamilto/ > Blessed Be.... Work: KHamilton@Hunter.COM -- Terry Lee Technical Director mailto:terryl@ienet.com I N T E R N E T D E S I G N G R O U P 213.488.6100 fax.488.6101 A Division of INTERNET EXPRESS NETWORK, INC http://www.mall.net
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?31F3DF4F.68DE>