21143:Asymmetric result

[M.Brands]

On Sun, Jul 11, 1999 at 03:32:44PM +0200, M.Brands allegedly wrote:
> On Sat, Jul 10, 1999 at 10:27:04PM -0400, Donald Becker allegedly wrote:
> > For almost all uses you shouldn't need to specify anything.  Just leave the
> > setting at the default at 0, "autonegotiate/autoselect".
> > 
> > The reasons for overriding the default media type:
> >   Some older cards didn't have a transceiver table in the serial ROM.
> >   A few current cards don't have correct tables.
> >   Older switches and 21140-SYM cards can't autonegotiate full duplex.
> How about cards connected by a crosslink cable? I've got two Digital
> DE500BA (21143-PC chipset) connected by a 1.2 meter CAT5 cross-link
> cable (IBM made). When I let them autonegotiate, they settle for
> 10baseT. If I force on of the cards to 10baseT-FD or 100baseTX, the
> other one also switches to 10baseT or 100baseTX. However, if I force
> 100baseTX-FD, I get a transferrate of about 130 kbyte/sec :(
> 
> If I swap one of the cards for an Intel EEPro 100, 100baseTX-FD does
> work correctly. (Doesn't matter which card.) I've used both the tulip
> driver (0.91e) and the DE4x5 driver. Note that autonegotiation still
> doesn't work.

I did some further checking. I appears 100baseTX doesn't work correctly
either. If I use 'options=3 debug=3', then my logfiles are filled with
the following entries:

Jul 11 15:49:49 bucephalus kernel: eth1: Using user-specified media 100baseTx.
Jul 11 15:49:52 bucephalus kernel: eth1: 21143 negotiation status 000000c6, 100bJul 11 15:49:53 bucephalus kernel: eth1: 21143 link status interrupt 000000c4, CJul 11 15:49:53 bucephalus kernel: eth1: 21143 100baseTx link beat good.
Jul 11 15:49:53 bucephalus kernel: eth1: 21143 link status interrupt 000000c6, CJul 11 15:49:53 bucephalus kernel: eth1: 21143 100baseTx link beat failed.
Jul 11 15:50:09 bucephalus kernel: eth1: 21143 link status interrupt 000000c4, CJul 11 15:50:09 bucephalus kernel: eth1: 21143 100baseTx link beat good.

Et cetera...

Mathijs