It was brought to my attention recently that the ex0 interface on one of my
i386 machines has been accumulating an inordinate amount of Oerrs.  I'm
curious if anyone else is experiencing this, or just me. ;-) Here's the
skinny...

The ex0 card is a 3com 905 (not b), running on a 100baseTX network (hooked
to pair of other 100baseTX boxes, and a 10baseT box via a cheapo LinkSys
10/100 Hub), and is probed as follows:

ex0 at pci0 dev 11 function 0: 3Com 3c905-TX 10/100 Ethernet
ex0: interrupting at irq 9
ex0: MAC address 00:60:97:37:b9:74
nsphy0 at ex0 phy 24: DP83840 10/100 media interface, rev. 0
nsphy0: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, auto

The card is configured as follows:

ex0: flags=8863<UP,BROADCAST,NOTRAILERS,RUNNING,SIMPLEX,MULTICAST> mtu 1500
        media: Ethernet 100baseTX
        status: active
        inet 192.168.1.1 netmask 0xffffff00 broadcast 192.168.1.255

'netstat -i' shows the following:

Name  Mtu   Network       Address              Ipkts Ierrs    Opkts Oerrs
Colls
ex0   1500  <Link>        00:60:97:37:b9:74     9869     0    14509   164
22
ex0   1500  192.168.1     192.168.1.1           9869     0    14509   164
22

Note the inordinate amount of Oerrs versus Opkts.  The Oerrs increase at a
pretty consistent rate, maintaining the same Oerrs:Opkts ratio (on the
previous kernel, the Oerrs were on the order of 1mil by the time I decided
to try a new kernel).

The system was running April 22 -current (1.4_Beta), but is now running May
11 (1.4), with the April 22 userland...same results as with the first
kernel.  I tried switching cables, and ports on the hub, to no avail.  None
of the other machines on the network see this sort of behaviour (in fact,
they all have 0 Oerrs).  Something interesting is that if I switch the ex0
card into Full-Duplex mode, as expected, the transfer rate goes in the
toilet (the hub is Half-duplex only), but the Oerrs stop!  That really
confuses me.

This leaves me to conclude that either the card itself is on it's last legs
(which is possible, but I'm not sure jives with the problem), or that
there's some peculiarity in the ex* driver which is causing this.

Ideas?  Anyone else seeing this behaviour?

    Scott
    scotte@warped.com