From scottd@cloud9.net  Mon Jun  1 20:01:06 1998
Received: from russian-caravan.cloud9.net (russian-caravan.cloud9.net [168.100.1.4])
          by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id UAA18763
          for <FreeBSD-gnats-submit@freebsd.org>; Mon, 1 Jun 1998 20:00:58 -0700 (PDT)
          (envelope-from scottd@cloud9.net)
Received: from earl-grey.cloud9.net (bF8sPDx6eV44O36ZR6BOchpRgD47fZA1@earl-grey.cloud9.net [168.100.1.1])
	by russian-caravan.cloud9.net (8.8.8/8.8.8) with ESMTP id XAA16695
	for <FreeBSD-gnats-submit@freebsd.org>; Mon, 1 Jun 1998 23:00:54 -0400 (EDT)
Received: (from scottd@localhost)
	by earl-grey.cloud9.net (8.8.8/8.8.8) id XAA04725;
	Mon, 1 Jun 1998 23:00:54 -0400 (EDT)
Message-Id: <199806020300.XAA04725@earl-grey.cloud9.net>
Date: Mon, 1 Jun 1998 23:00:54 -0400 (EDT)
From: Scott Drassinower <scottd@cloud9.net>
Reply-To: scottd@cloud9.net
To: FreeBSD-gnats-submit@freebsd.org
Subject: Intel EtherExpress 100+, 2.2.6 NFS troubles
X-Send-Pr-Version: 3.2

>Number:         6824
>Category:       misc
>Synopsis:       Intel EtherExpress 100+, 2.2.6 NFS troubles
>Confidential:   no
>Severity:       serious
>Priority:       high
>Responsible:    peter
>State:          closed
>Quarter:        
>Keywords:       
>Date-Required:  
>Class:          sw-bug
>Submitter-Id:   current-users
>Arrival-Date:   Mon Jun  1 20:10:01 PDT 1998
>Closed-Date:    Sun Dec 12 21:23:15 PST 1999
>Last-Modified:  Sun Dec 12 21:24:22 PST 1999
>Originator:     Scott Drassinower
>Release:        FreeBSD 2.2.6-RELEASE i386
>Organization:
Cloud 9 Consulting, Inc.
>Environment:

	A network of 2.2.6 machines, all with Intel EtherExpress 100+ cards
	running at 100 megabit/full duplex, attached to a Cisco Catalyst 2900XL
	Ethernet switch.  Various directories are shared via nfs between the
	machines.

>Description:

	Some operations across nfs (like tail -f, or grep) with large files
	that can be updated while the process is running (like a log file) will
	eventually hang and become stuck in disk wait on the client.
	

>How-To-Repeat:

	tail -f / grep / sz a large file that is getting regularly updated on the
	server (maillog, access_log, etc).  When running at 100 megabit/full duplex
	on the switch, the process will eventually hang.  Other operations on the
	mount are fine.  Running the network at 10 megabit/half duplex (on a hub)
	won't cause this problem.

>Fix:
	
	There is no workaround except for running at 10 megabits/half duplex.
	I have tried various combinations of read/write data sizes with mount_nfs
	(-r and -w on the client, at 1024 and 4096), as well as TCP transport
	(-T) and still can't eliminate this problem.



>Release-Note:
>Audit-Trail:
Responsible-Changed-From-To: freebsd-bugs->peter 
Responsible-Changed-By: phk 
Responsible-Changed-When: Tue Jun 2 04:19:40 PDT 1998 
Responsible-Changed-Why:  
Mr nfs... 
State-Changed-From-To: open->closed 
State-Changed-By: dillon 
State-Changed-When: Sun Dec 12 21:23:15 PST 1999 
State-Changed-Why:  
All known client lockups on NFS files changed out from under them by the  
server have been fixed, though there are almost certainly still going to 
be cache coherency issues.  The fixes are mainly in 3.x and 4.x kernels. 
>Unformatted:
