Duplex operation of BayNetworks FA310tx cards

[Donald Becker]

On Fri, 11 Sep 1998, Joel Jaeggli wrote:

> I would bet that some random sampling will allways autodetect the
> incorrect media type on an autosensing switch... In the tulip driver you
> can force media selection, so If you want to make shure that they all do,
> you could set it explicitly...
> 
> I used to see this way more often with the old 3c590's where they would
> detect the wrong media type about 10% of the time. Changes to the driver
> eventually fixed that...

That was with the driver doing all of the work of sensing the media type,
with little help from the hardware.  It's complicated by having media types
with no link beat (10base2 and AUI) and by the time limits on generating
100baseTx link beat.
The 3c590 series would have been much better if
    there was a way to detect 100baseTx link beat without generating it.
    there was an interrupt when one of the link beat detectors changed state.

Hardware autosensing is much more robust, although it cannot detect duplex
(it must assume half duplex) and might fall back to 10baseT in a way that
requires both machines to be restarted at once.

Autonegotiation, where both ends advertise their capabilities and exchange
handshakes, is *very* reliable: 
  http://cesdis.gsfc.nasa.gov/linux/misc/NWay.html

Donald Becker					  becker@cesdis.gsfc.nasa.gov
USRA-CESDIS, Center of Excellence in Space Data and Information Sciences.
Code 930.5, Goddard Space Flight Center,  Greenbelt, MD.  20771
301-286-0882	     http://cesdis.gsfc.nasa.gov/people/becker/whoiam.html