From root@blip.cetlink.net  Mon Jun 16 18:07:57 1997
Received: from blip.cetlink.net (blip.cetlink.net [209.54.62.5])
          by hub.freebsd.org (8.8.5/8.8.5) with ESMTP id SAA12840
          for <FreeBSD-gnats-submit@freebsd.org>; Mon, 16 Jun 1997 18:07:44 -0700 (PDT)
Received: (from root@localhost)
          by blip.cetlink.net (8.8.5/8.8.5) id VAA00867;
          Mon, 16 Jun 1997 21:07:07 -0400 (EDT)
Message-Id: <199706170107.VAA00867@blip.cetlink.net>
Date: Mon, 16 Jun 1997 21:07:07 -0400 (EDT)
From: Charlie Root <root@blip.cetlink.net>
Reply-To: root@blip.cetlink.net
To: FreeBSD-gnats-submit@freebsd.org
Subject: fxp problems
X-Send-Pr-Version: 3.2

>Number:         3887
>Category:       kern
>Synopsis:       fxp driver looses packets
>Confidential:   no
>Severity:       critical
>Priority:       high
>Responsible:    freebsd-bugs
>State:          closed
>Quarter:        
>Keywords:       
>Date-Required:  
>Class:          sw-bug
>Submitter-Id:   current-users
>Arrival-Date:   Mon Jun 16 18:10:01 PDT 1997
>Closed-Date:    Sun Mar 22 03:17:23 PST 1998
>Last-Modified:  Sun Mar 22 03:19:18 PST 1998
>Originator:     Jeff Wheat
>Release:        FreeBSD 2.2.1-RELEASE i386
>Organization:
Computer Enhancement Technologies, Inc
>Environment:
Server:
	CPU:			P5 166MHz
	Memory:			64MB DRAM
	Ethernet Card:		Intel EtherExpress 10/100B
	IDE Hard Drive:		Western Digital 2.1Gb

	This ethernet card has 68 aliases bound to it.

>Description:

	After running for anywhere from 12 hours to 2 days, the ethernet
card becomes unresponsive to any packets. The server continues to operate
though attempting to send or receive any packets fails. Netstat -nr shows
default route as well as the aliases. Netstat -na shows usually an address
bound to port 80 of one of the virtual webservers with a state of LAST_ACK.
Rebooting the machine returns all network functions to a normal state.

>How-To-Repeat:

	Probably sufficient to use an fxp device with multiple aliases
bound to the controller. This seems to exist now on two seperate machines.
The second machine ran without a single problem until I bound a couple of 
aliases. The symtoms are identical to the first server.

>Fix:
	
	Disabling aliases causes the problems to go away.

>Release-Note:
>Audit-Trail:

From: David Greenman <dg@root.com>
To: root@blip.cetlink.net
Cc: FreeBSD-gnats-submit@FreeBSD.ORG
Subject: Re: kern/3887: fxp problems 
Date: Mon, 16 Jun 1997 21:19:16 -0700

 >>Description:
 >
 >	After running for anywhere from 12 hours to 2 days, the ethernet
 >card becomes unresponsive to any packets. The server continues to operate
 >though attempting to send or receive any packets fails. Netstat -nr shows
 >default route as well as the aliases. Netstat -na shows usually an address
 >bound to port 80 of one of the virtual webservers with a state of LAST_ACK.
 >Rebooting the machine returns all network functions to a normal state.
 >
 >>How-To-Repeat:
 >
 >	Probably sufficient to use an fxp device with multiple aliases
 >bound to the controller. This seems to exist now on two seperate machines.
 >The second machine ran without a single problem until I bound a couple of 
 >aliases. The symtoms are identical to the first server.
 >
 >>Fix:
 >	
 >	Disabling aliases causes the problems to go away.
 
    The problem could be caused by a variety of things. One cause of something
 like this is running out of mbuf clusters. You might get a "Out of mbuf
 clusters - increase maxusers!" message in your /var/log/messages. You can
 increase the maximum number of mbuf clusters with the NMBCLUSTERS kernel
 option (see the mailing list archives). Another possible cause of this could
 be a bug in the 82557 NIC; the receiver goes dead when getting garbage on the
 wire. If you can "ifconfig fxp0 dowm; ifconfig fxp0 up" to clear the problem,
 then this might be the cause. Let me know if you determine anything further
 on this problem.
 
 -DG
 
 David Greenman
 Core-team/Principal Architect, The FreeBSD Project

From: Jeff Wheat <jeff@tad.cetlink.net>
To: dg@root.com
Cc: FreeBSD-gnats-submit@FreeBSD.ORG
Subject: Re: kern/3887: fxp problems
Date: Wed, 18 Jun 1997 09:39:24 -0400 (EDT)

 On 17-Jun-97 David Greenman wrote:
 >>>Description:
 >>
 >>      After running for anywhere from 12 hours to 2 days, the ethernet
 >>card becomes unresponsive to any packets. The server continues to operate
 >>though attempting to send or receive any packets fails. Netstat -nr shows
 >>default route as well as the aliases. Netstat -na shows usually an address
 >>bound to port 80 of one of the virtual webservers with a state of LAST_ACK.
 >>Rebooting the machine returns all network functions to a normal state.
 >>
 >>>How-To-Repeat:
 >>
 >>      Probably sufficient to use an fxp device with multiple aliases
 >>bound to the controller. This seems to exist now on two seperate machines.
 >>The second machine ran without a single problem until I bound a couple of 
 >>aliases. The symtoms are identical to the first server.
 >>
 >>>Fix:
 >>      
 >>      Disabling aliases causes the problems to go away.
 >
 >   The problem could be caused by a variety of things. One cause of something
 >like this is running out of mbuf clusters. You might get a "Out of mbuf
 >clusters - increase maxusers!" message in your /var/log/messages. You can
 >increase the maximum number of mbuf clusters with the NMBCLUSTERS kernel
 >option (see the mailing list archives). Another possible cause of this could
 >be a bug in the 82557 NIC; the receiver goes dead when getting garbage on the
 >wire. If you can "ifconfig fxp0 dowm; ifconfig fxp0 up" to clear the problem,
 >then this might be the cause. Let me know if you determine anything further
 >on this problem.
 >
 >-DG
 >
 >David Greenman
 >Core-team/Principal Architect, The FreeBSD Project
 
 David,
 
         Your suggestion of bringing the interface down and then back up worked.
 Besides replacing the cards with a different brand, is there anything that can 
 be down to help eliminate this problem? I'd hate to have to replace these cards
 as I have about 15 of them that were planned to be installed this month.
 
 Regards,
 Jeff
 
 
 -----------------------------------------------------------------------------
   Jeff Wheat                  jeff@cetlink.net             Senior Engineer
   CETLink.Net Inc.             South Carolina              +1.803.327.2754
 -----------------------------------------------------------------------------

From: David Greenman <dg@root.com>
To: Jeff Wheat <jeff@tad.cetlink.net>
Cc: FreeBSD-gnats-submit@FreeBSD.ORG
Subject: Re: kern/3887: fxp problems 
Date: Wed, 18 Jun 1997 07:01:42 -0700

 >        Your suggestion of bringing the interface down and then back up worked.
 >Besides replacing the cards with a different brand, is there anything that can 
 >be down to help eliminate this problem? I'd hate to have to replace these cards
 >as I have about 15 of them that were planned to be installed this month.
 
    The problem is caused by the cards seeing certain types of garbage -
 usually the kind that occurs when a hub or switch that it is connected to
 is power-cycled. The problem should otherwise never occur. In any case,
 the hardware bug is only present in "rev 1" parts (I'm refering to the
 FreeBSD PCI probe message for the fxp device). If you have a "rev 2",
 then there might be a different problem.
    There is a work-around for the bug which involves reprogramming the
 multicast filter after every few seconds of no receiver traffic, but I
 haven't attempted to implement this in the driver (the work around is
 disgusting, for one thing). I really should implement the work-around,
 but time has been in short supply recently. One could write, as an interrim
 solution, a few lines of shell code that does an "ifconfig fxp0 down;
 ifconfig fxp0 up" whenever it can't successfully ping the router...
    I'm really surprised that this is causing you so much trouble; you
 should probably have a good look at whatever hub/switch you're attaching
 the cards to and figure out why it is provoking this problem.
 
 -DG
 
 David Greenman
 Core-team/Principal Architect, The FreeBSD Project

From: tedm@toybox.placo.com
To: <freebsd-gnats-submit@freebsd.org>
Cc:  Subject: Re: kern/3887: fxp driver looses packets
Date: Fri, 25 Jul 97 00:42:21

 We also use these cards - and noticed similar symptoms under Windows 95.  The
 problem turned out to be a bad network card, we replaced it and the problem
 went away.  We also had a problem with corrupted packets on the 100BaseT
 network around the same time and traced the problem to an early-revision
 Intel 100baseT hub.  According to Intel, there is a known problem with their earlier
 100BaseT hubs corrupting packets under heavy load.  This same model hub is still
 being sold by Bay Networks.  Sites that have standardized on a single vendor (Intel)
 for their hubs and adapter cards should test for errors on 100BaseT networks under
 heavy load.  Intel also recommends checking cabling on all 100BaseT installations -
 pins 2 and 6 should be twisted together on UTP used for 100BaseT.

From: Studded <Studded@dal.net>
To: freebsd-gnats-submit@freebsd.org, root@blip.cetlink.net
Cc:  Subject: Re: kern/3887: fxp driver looses packets
Date: Sun, 22 Mar 1998 02:10:10 -0800

 The submitter indicated that the proposed solution solved his problem
 so this one can probably be closed.
 
 Doug
 -- 
 ***         Chief Operations Officer, DALnet IRC network       ***
 *** Proud operator, designer and maintainer of the world's largest
 *** Internet Relay Chat server.  5,328 clients and still growing.
 *** Try spider.dal.net on ports 6662-4    (Powered by FreeBSD)
State-Changed-From-To: open->closed 
State-Changed-By: danny 
State-Changed-When: Sun Mar 22 03:17:23 PST 1998 
State-Changed-Why:  
Originator says proposed solution works. 
>Unformatted:
