3c905b questions for 100Mb/full duplex

[Joel T. Thomas]

Hello All,

I have some questions regarding using the 3c905b-TX with a BayStack
450 Switch. We are able to get 100Mbps/half, and are attempting to get
100Mbps/full duplex.

We are running Red Hat Linux 5.1, kernel 2.0.34.

We have one master node with two 3c905b cards, one linked to the
external network, and one linked to the Baystack switch. We have
several slave systems attached to the Baystack, all with 3c905b
cards.

Firstly, using Donald Becker's 0.99E 3c59x.c, we were able to get
100Mbps/half duplex on all the systems. All the ports on the
BayStack were set to autonegotiate. We were passing "options=4" to the
network driver module.

My attempts to get the 'master' node and the slave nodes to full
duplex haven't worked though. We set "options=20", but the system went
back to 100Mbps/half.


When I used 0.99H 3c59x.c, even with the same options I was passing
before, the switch seemed to be showing 10Mbps/half duplex. So for
now, I've switched back to 0.99E . 



There are several things I want to verify.

1)
In the /etc/conf.modules file, is it not sufficient to have

options eth0 options=20

but rather do you need

options eth0 options=20 full_duplex=1

?


When using the full_duplex argument and 0.99E, vortex-diag showed that
it was finally full-duplex. The Baystack switch itself, however, still
showed 100/half.



2) What is the appropriate way to set 100 Mbps /full duplex if you're
compiling the driver into the kernel? My current way seemed to work
fine for 100Mbs/half. For my slave systems, I am making a boot image
for them, and using "ether=0,0,20,eth0" when I create the image. These
systems go to 100Mbps/half duplex fine, but do not go to 100 Mbps/full.


3)
Should all ports on the Baystack be set to autonegotiate? 

4) What else needs to be done to get 100mbps/full?

Any assistance would be greatly appreciated. I will definitely provide
more information upon request, so please let me know what further
tests or vortex-diag output I need to provide.


Thank you,
Joel


--
Joel T. Thomas, jthomas@aero.org