[eepro100] Speed and Duplexing, Once Again

[UNX Group]

Donald,

In response to the driver version question in your
response:

eepro100.c:v1.09j-t 9/29/99 Donald Becker
http://cesdis.gsfc.nasa.gov/linux/driv
ers/eepro100.html
eepro100.c: $Revision: 1.36 $ 2000/11/17 Modified by Andrey V. Savochkin
<saw@sa
w.sw.com.sg> and others
eth0: OEM i82557/i82558 10/100 Ethernet, 00:06:5B:04:34:11, IRQ 16.
  Forcing 100Mbs half-duplex operation.

This specific system is running RH 7.1 (2.4.9-12smp i686 optimized)

Thanks very much for your help!

Don

----- Original Message -----
From: "Donald Becker" <becker@scyld.com>
To: "UNX Group" <unxgroup@optonline.net>
Cc: <eepro100@scyld.com>
Sent: Thursday, November 29, 2001 8:04 AM
Subject: Re: [eepro100] Speed and Duplexing, Once Again


> On Wed, 28 Nov 2001, UNX Group wrote:
>
> > I have read now several times in your responses:
> >
> > "Forcing the duplex and speed is usually a mistake"
> >
> > Could you please expand on this for a newcomer to
> > Linux but experienced with Unix, although certainly
> > not to your level of expertise with networking and these
> > drivers.
>
> Read
>   http://scyld.com/expert/NWay.html
> for background.
>
> Autonegotiation is disabled when the speed and duplex is forced.
>
> Here's an obvious scenario of why disabling autonegotiation is wrong:
> Plug an unconfigured or non-configurable device into your network.
>      Does it work correctly?
>      Does it work at a near-optimal setting?
>
> Forcing the switch speed to 100Mbps will allow the device to work
> correctly, but performance features such as full duplex and flow control
> cannot be correctly enabled.  (Remember that the twisted pair Ethernet
> standard is half duplex.  Full duplex was only added to the standard
> with autonegotiation.  Forced full duplex is a common but non-standard
> extension. )
>
> Forcing the switch to full duplex mean that every new device must be
> configured by hand, and the configuration reversed when you move the
> device to standard networking equipment.  Forgetting either case results
> in difficult to track down performance problems.
>
> I consider forced-full-duplex to be an serious issue somewhere between
> "..and these cars have the brake pedal on the right" and "we decided to
> put the drinking water in the brown jugs, and the 'other' water in blue".
> You won't necessarily die right away, but it isn't healthy.
>
> > I recently inherited sa responsibility for a large number
> > of RedHat Linux 7.0 systems.  All machines have a variation
> > of Intel NIC, all utilizing e100 or eepro100.  All systems are
> > being "forced", via modules.conf, to 100Mbs, full-duplex.
> > I notice, via dmesg, that we recv the message:
> >
> > modules.conf specifies:
> > options eth0 options=0x20 full_duplex=1
>
> What driver version?
>
> > On "eepro100" systems, recv dmesg:
> > "Forcing 100Mbs half-duplex operation."
>
> Hmmm, this _is_ misleading.  I'll change it.  In the meantime try
>  options=0x200
>
> Other valid options are 0x100, 0x20, and 0x10.
> I hope the mapping of speed and duplex is obvious.
>
>
> Donald Becker becker@scyld.com
> Scyld Computing Corporation http://www.scyld.com
> 410 Severn Ave. Suite 210 Second Generation Beowulf Clusters
> Annapolis MD 21403 410-990-9993
>