From: Jason Thorpe <thorpej@nas.nasa.gov>
> On Tue, 11 May 1999 19:20:07 -0700
>  "Scott Ellis" <scotte@warped.com> wrote:
>
>  > Name  Mtu   Network       Address              Ipkts Ierrs    Opkts
Oerrs
>  > Colls
>  > ex0   1500  <Link>        00:60:97:37:b9:74    48014     0    58806
859
>  > 692
>  > ex0   1500  192.168.1     192.168.1.1          48014     0    58806
859
>  > 692
>  >
>  > Neither of the other two machines used in the test show any Ierrs or
Oerrs
>  > (and they both report being in Half-duplex mode, like they should be),
which
>  > brings me back to thinking it's something specific to the ex0
interface.
>  >
>  > Any other ideas on what could be making this weirdness happen?
>
> I see the problem... ex_getstats() counts deferrals as oerrors, which
> is incorrect.  Deferrals are a normal part of Ethernet transmissions,
> and should only be counted as errors if an excessive number of them
> occurs (which causes the packet transmission to fail).
>
> The bug in the driver is harmless; it just skews the stats.

For the sake of completeness, I tried the interface with a crossover cable,
bypassing the (potentially hosed) hub, and got the same results, so I
imagine the ex_getstats() bug was the problem.

Thanks for the help!

    Scott