[eepro100] collision problem

[Donald Becker]

On Mon, 4 Mar 2002, Marco Di Folco wrote:

> I am having ethernet connection issues with a Dell PowerEdge 2450 server, it
> has two Intel NICs.
> The system is running Red Hat Linux 6.2 with all updates applied.
> The kernel version is 2.2.19 from Red Hat.
> The eepro100 driver version is $Revision: 1.20.2.10 $ 2000/05/31.
> Both NICs are connected to a 10/100 mbps Cisco switch.
> 
> Here are my problems:
> - the NICs seem to be running at 10 mbps, half-duplex.

The mii-diag output indicated that eth0 was forced to 10baseT.
Is that what you intended?

> - ifconfig reports a huge number of collisions for both NICs.

That's expected when you force 10baseT on a busy network.

> Also, close to this server, there is a Dell PowerEdge 1550 box, it also has
> two Intel NICs connected to the same Cisco switch, and they are running just
> fine at 100 mbps full-duplex with no collisions at all!
...
> How can I force the NICs to run at 100 mbps full-duplex?
> How can I get rid of the collisions?

You should really let the link autonegotiate.  This is the default
setting -- verify that you are not unintentionally passing an option to
the driver  (/etc/modules.conf or /etc/conf.modules)

If your Cisco switch is set to forced-full-duplex, which disables
autonegotiation, you can force the driver speed with an option as
described at
   http://www.scyld.com/network/eepro100.html
or do
   mii-diag -F 100baseTx-FDX


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