2.4.20 gets duplex wrong on NIC

2.4.20 gets duplex wrong on NIC

[Bill Davidsen]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 388 bytes --]

In spite of modules.conf the system boots with the NIC in half duplex. I 
verified this with the mii-tool, I can set it full with mii-tool and it 
works right (copied a CD image 650MB), and the blade in the switch has 
been set either full or auto without gain. Yes, I tried the e100 driver as 
well.

Info I think shows this attached to prevent munging, let me know if more 
is needed.


[-- Attachment #2: Type: TEXT/PLAIN, Size: 2797 bytes --]

Script started on Tue Dec  3 18:16:21 2002
Common profile 1.9 hacked lastmod 2002-11-14 21:55:45-04
No common directory available
Session time 18:16:21 on 12/03/02
oddball:root> tail -20\b\b\b\b\b\b\b\b^[[Kdmesg | tail -20
EXT3 FS 2.4-0.9.19, 19 August 2002 on ide0(3,67), internal journal
EXT3-fs: mounted filesystem with ordered data mode.
hdc: DMA disabled
Serial driver version 5.05c (2001-07-08) with MANY_PORTS SHARE_IRQ SERIAL_PCI ISAPNP enabled
ttyS00 at 0x03f8 (irq = 4) is a 16550A
eepro100.c:v1.09j-t 9/29/99 Donald Becker http://www.scyld.com/network/eepro100.html
eepro100.c: $Revision: 1.36 $ 2000/11/17 Modified by Andrey V. Savochkin <saw@saw.sw.com.sg> and others
PCI: Found IRQ 9 for device 00:06.0
PCI: Sharing IRQ 9 with 00:07.2
eth0: Intel Corp. 82557/8/9 [Ethernet Pro 100], 00:90:27:07:F2:9E, IRQ 9.
  Receiver lock-up bug exists -- enabling work-around.
  Board assembly 713807-002, Physical connectors present: RJ45
  Primary interface chip i82555 PHY #1.
  General self-test: passed.
  Serial sub-system self-test: passed.
  Internal registers self-test: passed.
  ROM checksum self-test: passed (0x24c9f043).
  Receiver lock-up workaround activated.
apm: BIOS version 1.2 Flags 0x03 (Driver version 1.16)
lp: driver loaded but no devices found
oddball:root> mii-tool 
eth0: no autonegotiation, 10baseT-HD, link ok
oddball:root> ifconfig eth0
eth0      Link encap:Ethernet  HWaddr 00:90:27:07:F2:9E  
          inet addr:192.168.192.242  Bcast:192.168.192.255  Mask:255.255.255.0
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:253 errors:0 dropped:0 overruns:0 frame:0
          TX packets:29 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:100 
          RX bytes:28278 (27.6 Kb)  TX bytes:2192 (2.1 Kb)
          Interrupt:9 Base address:0x4000 

oddball:root> mii-tool -F 10baseT-FD /dev/eth0\b \b\b \b\b \b\b \b\b \b\b \b\b \b\b \b\b \beth0
oddball:root> mii-tool 
eth0: 10 Mbit, full duplex, link ok
oddball:root> cat /etc/modules.conf
# $Id: modules.conf 1.3 2002/11/13 20:37:15 root Exp $

# parallel port
alias parport_lowlevel parport_pc

# network card (e100?)
alias eth0 eepro100
options eth0 full_duplex=1

# try to force load sound
#alias sound-slot-0 ac97

# USB info
alias usb-controller	usb-uhci
oddball:root> lsmod
Module                  Size  Used by    Not tainted
lp                      8544   0  (autoclean)
parport                32928   0  (autoclean) [lp]
apm                    11996   1 
eepro100               20940   1 
mii                     3492   0  [eepro100]
serial                 47712   0  (autoclean)
oddball:root> exit
exit

Script done on Tue Dec  3 18:17:46 2002

Re: 2.4.20 gets duplex wrong on NIC

[Jeff Garzik]

Bill Davidsen wrote:
> In spite of modules.conf the system boots with the NIC in half duplex. I 
> verified this with the mii-tool, I can set it full with mii-tool and it 
> works right (copied a CD image 650MB), and the blade in the switch has 
> been set either full or auto without gain. Yes, I tried the e100 driver as 
> well.
> 
> Info I think shows this attached to prevent munging, let me know if more 
> is needed.


Lots of feedback/questions/response:

When you're on a network, more is always needed :)

Please give _plenty_ of details about what is on the other side of the 
cable: hub? switch? vendor of hub/switch?  crossover to another NIC? 
what is the port configuration and what are the capabilities of the 
other end?  is it set to autonegotiate (on the other end)?

Why do you force full duplex?  It is often the wrong thing to do.

For eepro100, you should use module option 'options' to specify 
10baseT-FD... full_duplex appears to be somewhat redundant in the 
context of your problem.

For e100, you should use 'e100_speed_duplex' module option to specify media.

Finally, I would be very interested to know the results of using ethtool 
to set, and get, your media settings.  It's in every distro these days, 
plus you can d/l it from http://sf.net/projects/gkernel/

	Jeff

Re: 2.4.20 gets duplex wrong on NIC

[Bill Davidsen]

On Tue, 3 Dec 2002, Jeff Garzik wrote:

> Bill Davidsen wrote:
> > In spite of modules.conf the system boots with the NIC in half duplex. I 
> > verified this with the mii-tool, I can set it full with mii-tool and it 
> > works right (copied a CD image 650MB), and the blade in the switch has 
> > been set either full or auto without gain. Yes, I tried the e100 driver as 
> > well.
> > 
> > Info I think shows this attached to prevent munging, let me know if more 
> > is needed.
> 
> 
> Lots of feedback/questions/response:
> 
> When you're on a network, more is always needed :)
> 
> Please give _plenty_ of details about what is on the other side of the 
> cable: hub? switch? vendor of hub/switch?  crossover to another NIC? 
> what is the port configuration and what are the capabilities of the 
> other end?  is it set to autonegotiate (on the other end)?

Cisco 5509 set auto.
> 
> Why do you force full duplex?  It is often the wrong thing to do.

It gives about 4x throughput...

The network folks say they have the same problem with most NICs and OS, so
they have to teach Windows users to diddle the NIC. At least I came to
them with the workaround. It's possible a known problem with those 10baseT
blades.

> 
> For eepro100, you should use module option 'options' to specify 
> 10baseT-FD... full_duplex appears to be somewhat redundant in the 
> context of your problem.

Okay, I assumed that since a grep showed:
  MODULE_PARM(full_duplex, "1-" __MODULE_STRING(8) "i");
that the option was supported. What does it do, if I need to diddle bits
in the options?

> For e100, you should use 'e100_speed_duplex' module option to specify media.
> 
> Finally, I would be very interested to know the results of using ethtool 
> to set, and get, your media settings.  It's in every distro these days, 
> plus you can d/l it from http://sf.net/projects/gkernel/

Well, it's not installed at any rate, so I'll have to download it. Timing
uncertain, I'm "off" for the next five days and leaving the site for home
shortly because I have a 131 mile commute and heavy snow. Perhaps from
home tonight.

-- 
bill davidsen <davidsen@tmr.com>
  CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.

Re: 2.4.20 gets duplex wrong on NIC

[Mr. James W. Laferriere]

	Hello Bill , Severly clipped original message .
	If that 5509 is running slightly older code (and maybe even
	newer) there is(was?) a difficulty with auto-negotiation between
	the cisco & many adapters .  This btw is(was?) a known issue with
	cisco .  Also the last item I had concerning the matter ,  cisco
	recommeded static duplex & rate settings .  Hth ,  JimL

On Thu, 5 Dec 2002, Bill Davidsen wrote:
 ...
> > Please give _plenty_ of details about what is on the other side of the
> > cable: hub? switch? vendor of hub/switch?  crossover to another NIC?
> > what is the port configuration and what are the capabilities of the
> > other end?  is it set to autonegotiate (on the other end)?
>
> Cisco 5509 set auto.
> >
> > Why do you force full duplex?  It is often the wrong thing to do.
>
> It gives about 4x throughput...
...
-- 
       +------------------------------------------------------------------+
       | James   W.   Laferriere | System    Techniques | Give me VMS     |
       | Network        Engineer |     P.O. Box 854     |  Give me Linux  |
       | babydr@baby-dragons.com | Coudersport PA 16915 |   only  on  AXP |
       +------------------------------------------------------------------+

Re: 2.4.20 gets duplex wrong on NIC

[Bill Davidsen]

On Thu, 5 Dec 2002, Mr. James W. Laferriere wrote:

> 
> 	Hello Bill , Severly clipped original message .
> 	If that 5509 is running slightly older code (and maybe even
> 	newer) there is(was?) a difficulty with auto-negotiation between
> 	the cisco & many adapters .  This btw is(was?) a known issue with
> 	cisco .  Also the last item I had concerning the matter ,  cisco
> 	recommeded static duplex & rate settings .  Hth ,  JimL

Yeah, I was really trying that;-) I'm working for the moment, there's a
6500 waiting to be deployed, so no effort will go into upgrading the old
machine.

Thanks for the note, I'm not clear why full_duplex doesn't work, but I
have it working now and hopefully the new router will be better.

-- 
bill davidsen <davidsen@tmr.com>
  CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.