One last try..

Robert G. Brown rgb@phy.duke.edu
Wed Nov 25 18:56:11 1998


On Wed, 25 Nov 1998, Maurice Hilarius wrote:

> I earlier posted this problem, which many of us are having with a
> particular implementation of the Digital 21143PC chip on the Samsung/API UX
> Alpha motherboards.
> We have tried running this with the tulip V.90, and the DE4X5 drivers, as
> modules, and in the kernels. In each case we get the same thing:
> On startup the NIC tries to auto-negotiate for media type, floods the
> network with data, and never succeeds in detecting media type .
> 
> When I posted this enqury a couple of weeks ago, the only repsonses were
> about how to "hard-wire" the media-detect. While a partial solution, it is
> not the "real" answer.
> 
> This is of a pretty high priority as we have 2 Beowulf clusters waiting for
> this fix. This totals about 100 machines sitting "dead in the water".
> 
> Has anyone got any suggestions? We could sure use one.

What are the tulips connected to?  I believe (from bitter experience)
that media detection problems are as often the fault of the switch or
hub as the failure of negotiation on the part of the card/drivers.
That's probably charitable -- I suspect that most of the time the
problem is on the other end and not the fault of the card/driver.

The fact that the problem persists over both the tulip and the de4x5
driver suggests strongly that either the interfaces are somehow
"different" on this particular motherboard (in which case you'll either
have to start working on the device driver code from the
interface/motherboard specs or rely on Don Becker to help you directly)
or, in my opinion more likely (but far from certain) the cards are
working fine but the switch you have them plugged into is conducting a
nonstandard negotiation sequence.

    rgb

Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb@phy.duke.edu