From nobody@FreeBSD.ORG Mon May 24 04:53:08 1999
Return-Path: <nobody@FreeBSD.ORG>
Received: by hub.freebsd.org (Postfix, from userid 32767)
	id 2F08914D90; Mon, 24 May 1999 04:53:08 -0700 (PDT)
Message-Id: <19990524115308.2F08914D90@hub.freebsd.org>
Date: Mon, 24 May 1999 04:53:08 -0700 (PDT)
From: delaitt@cpc.wmin.ac.uk
Sender: nobody@FreeBSD.ORG
To: freebsd-gnats-submit@freebsd.org
Subject: Network hangging due to xl0: tx underrun 
X-Send-Pr-Version: www-1.0

>Number:         11869
>Category:       kern
>Synopsis:       Network hangging due to xl0: tx underrun
>Confidential:   no
>Severity:       critical
>Priority:       high
>Responsible:    wpaul
>State:          closed
>Quarter:        
>Keywords:       
>Date-Required:  
>Class:          sw-bug
>Submitter-Id:   current-users
>Arrival-Date:   Mon May 24 05:00:01 PDT 1999
>Closed-Date:    Thu Jan 10 10:11:06 PST 2002
>Last-Modified:  Tue Nov 23 18:40:32 GMT 2004
>Originator:     Thierry Delaitre
>Release:        3.2-19990522-STABLE
>Organization:
Centre for Parallel Computing
>Environment:
FreeBSD seth.cpc.wmin.ac.uk 3.2-19990522-STABLE FreeBSD 3.2-19990522-STABLE #1: Sun May 23 13:19:23 BST 1999     root@gnu.cpc.wmin.ac.uk:/usr/src/sys/compile/SETH  i386
>Description:
The problem is that I keep geeting the following messages:

xl0: transmission error: 90
xl0: tx underrun, increasing tx start threshold to 120 bytes

and sometimes the network hangs and I have to reboot the machine.
The FreeBSD PC is mounting some NFS volumes and I get these messages
when I write a large file over NFS.

I'm using the following Ethernet Card:

xl0: <3Com 3c905-TX Fast Etherlink XL> rev 0x00 int a irq 17 on pci0.11.0
xl0: Ethernet address: 00:60:97:b8:65:cf
xl0: autoneg complete, link status good (full-duplex, 100Mbps)


>How-To-Repeat:
Write large files over NFS mounted volumes.

>Fix:


>Release-Note:
>Audit-Trail:
Responsible-Changed-From-To: freebsd-bugs->wpaul 
Responsible-Changed-By: n_hibma 
Responsible-Changed-When: Thu Jun 24 11:53:42 PDT 1999 
Responsible-Changed-Why:  
the xl driver is his baby. 

From: Nick Hibma <nick.hibma@jrc.it>
To: freebsd-gnats-submit@freebsd.org, delaitt@cpc.wmin.ac.uk
Cc:  
Subject: Re: kern/11869: Network hangging due to xl0: tx underrun
Date: Thu, 24 Jun 1999 20:53:12 +0200

 Could you check which version of if_xl.c you have?
 
 	 grep Id  /usr/src/sys/pci/if_xl.c
 
 
 The following sounds like a fix for your problem.
 
 RCS file: /home/ncvs/src/sys/pci/if_xl.c,v
 ...
 revision 1.36
 date: 1999/05/05 17:05:06;  author: wpaul;  state: Exp;  lines: +12 -4
 Dynamically increase TX start threshold if TX underruns are detected.
 
 
 Either cvsup the newest version of the source or ask me for a copy of
 the file if_xl.c .
 
 Cheers,
 
 Nick
 

From: Thierry Delaitre <delaitt@cpc.westminster.ac.uk>
To: Nick Hibma <nick.hibma@jrc.it>
Cc: freebsd-gnats-submit@freebsd.org
Subject: Re: kern/11869: Network hangging due to xl0: tx underrun
Date: Mon, 28 Jun 1999 18:01:43 +0100 (BST)

 Thanks. I have now increased XL_MIN_FRAMELEN to 300 and it works ok.
 
 Cheers,
 
 Thierry.
 
 On Thu, 24 Jun 1999, Nick Hibma wrote:
 
 > Could you check which version of if_xl.c you have?
 > 
 > 	 grep Id  /usr/src/sys/pci/if_xl.c
 > 
 > 
 > The following sounds like a fix for your problem.
 > 
 > RCS file: /home/ncvs/src/sys/pci/if_xl.c,v
 > ...
 > revision 1.36
 > date: 1999/05/05 17:05:06;  author: wpaul;  state: Exp;  lines: +12 -4
 > Dynamically increase TX start threshold if TX underruns are detected.
 > 
 > 
 > Either cvsup the newest version of the source or ask me for a copy of
 > the file if_xl.c .
 > 
 > Cheers,
 > 
 > Nick
 
 

From: Reto Burkhalter <borki@siconline.ch>
To: freebsd-gnats-submit@FreeBSD.org, delaitt@cpc.wmin.ac.uk
Cc:  
Subject: Re: kern/11869: Network hangging due to xl0: tx underrun
Date: Sun, 16 Jan 2000 01:21:09 +0100

 Hi
 
 We have the same problem with 3.4-RELEASE (where the bug should be
 fixed, right?)
 
 Any hints?
 
 
 -Reto
 
 

From: Thierry Delaitre <delaitt@cpc.wmin.ac.uk>
To: Reto Burkhalter <borki@siconline.ch>
Cc: freebsd-gnats-submit@FreeBSD.org
Subject: Re: kern/11869: Network hangging due to xl0: tx underrun
Date: Mon, 17 Jan 2000 09:39:37 +0000 (GMT)

 On Sun, 16 Jan 2000, Reto Burkhalter wrote:
 
 > Hi
 > 
 > We have the same problem with 3.4-RELEASE (where the bug should be
 > fixed, right?)
 > 
 > Any hints?
 > 
 > 
 > -Reto
 
 Please find an answer from Bill Paul:
 
 > you can try to set the default transmit threshold in the driver
 > to something higher. Edit /sys/pci/if_xl.c and look in the xl_init()
 > routine where sc->xl_txthresh is initialized. By default it starts
 > out at XL_MIN_FRAMELEN (60). Try setting it to 256 or 512 or some
 > larger value.
 
 PS: I have not had any network hanging due to this when XL_MIN_FRAMELEN is
     set to its default value since I upgraded to 3.4RC.
 
 Thierry.
 
 
 

From: Arnaud Launay <asl@launay.org>
To: freebsd-gnats-submit@FreeBSD.org, delaitt@cpc.wmin.ac.uk
Cc:  
Subject: Re: kern/11869: Network hangging due to xl0: tx underrun
Date: Wed, 31 Oct 2001 13:58:38 +0100

 Hello,
 
 Could you try under a 4.x release if you always have this problem ?
 
 Thanks,
 	Arnaud.
 
State-Changed-From-To: open->feedback 
State-Changed-By: silby 
State-Changed-When: Sat Jan 5 16:41:15 PST 2002 
State-Changed-Why:  
Theirry, do you still need to up XL_MIN_FRAMELEN under 4.5-pre 
or later?  We're not being flooded with similar problems, 
so I suspect that this problem has gone away with faster 
machines + newer NICs. 

http://www.FreeBSD.org/cgi/query-pr.cgi?pr=11869 
State-Changed-From-To: feedback->closed 
State-Changed-By: silby 
State-Changed-When: Thu Jan 10 10:11:06 PST 2002 
State-Changed-Why:  
Thierry is still running 3.4.  Since there are no reports of this causing 
problems on 4.x, I'm going to go ahead and close this PR.  If the problem 
resurfaces on a 4.x system, we can reopen the PR. 

http://www.FreeBSD.org/cgi/query-pr.cgi?pr=11869 

From: "Iwan Bambang A." <iwe@rad.net.id>
To: freebsd-gnats-submit@FreeBSD.org, delaitt@cpc.wmin.ac.uk
Cc:  
Subject: Re: kern/11869: Network hangging due to xl0: tx underrun
Date: Mon, 06 May 2002 11:32:09 +0700

 Hi,
 
 I'm using 4.4-RELEASE on PIII-800, 128 MB of RAM and still have this kind 
 of error (although it only appears once in a while).
 
 just to let you know.
 
 thanks.
 
 -iwe-
 

From: Gary Bajaj <b04@interbaun.com>
To: freebsd-gnats-submit@FreeBSD.org, delaitt@cpc.wmin.ac.uk
Cc:  
Subject: Re: kern/11869: Network hangging due to xl0: tx underrun
Date: Tue, 23 Nov 2004 11:32:23 -0700

 Nov 22 19:30:04 xxxxxxxx /kernel: xl0: transmission error: 90
 Nov 22 19:30:04 xxxxxxxx /kernel: xl0: tx underrun, increasing tx start 
 threshold to 120 bytes
 Nov 22 23:50:04 xxxxxxxx /kernel: xl0: transmission error: 90
 Nov 22 23:50:04 xxxxxxxx /kernel: xl0: tx underrun, increasing tx start 
 threshold to 180 bytes
 
 Are these messages normal on FreeBSD 4.10 with 3CSOHO100-TX (xl0)?  
 Never seen this before in the 2+ years the same NIC has been through 
 versions 4.8 through 4.10 and others of the same model work ok on 5.2.x 
 as well.  There was a RX checksum issue (RXCSUM turned off in 4.10) that 
 has helped eliminate interface errors.  So while there are no errors, 
 the above appeared yesterday for the first time following an upgrade two 
 days earlier from an AMD760/VIA motherboard to an NVIDIA nForce1.  
 Previously the same NVIDIA nForce1 was running with NICs of the same 
 model, FreeBSD versions 4.8, 5.2 and 5.2.1.
 
>Unformatted:
