21143 specific problems

David Sean Friedman dsf+@andrew.cmu.edu
Tue Sep 28 16:25:06 1999

Excerpts from tulip: 28-Sep-99 Re: 21143 specific problems by Donald
> What driver version?
I've tried several versions. The default (0.89H?) included with 2.2.12,
0.91, and even the test version 0.91g. All exhibited the same problem.

> Only v0.91 and later does autonegotiation correctly when using a 21143 with
> a SYM mode transceiver that requires specific output value on the
> general-purpose pins.
> (Yes, that's a very specific set of conditions.  Two years ago MII
> transceivers were the common connection type, and the older driver works
> correctly with them.  The changes in the newer 21143 chips and the lower
> cost of SYM mode transceivers has made 21143+SYM designs popular.)

The output of the driver does indicate the NIC is 21143+SYM.

> > However, if I use a 21140-based SMC NIC, the TCP transmit traffic is
> My guess is that the link partner is seeing the autonegotiation complete,
> but then isn't getting the 100Mbps link beat quickly enough.  It then drops
> to speed-sensing mode and ends up with a 100baseTx half duplex link.
> Getting the updated driver should fix this problem.

My first thoughts were also that it was a half duplex problem, since
I've seen similar results on some more inferior switches. However, the
fact that TCP traffic in the other direction works fine leads me to
believe the problem is elsewhere.

I've also tried using 'options=16' to force full duplex mode. With or
without this option, the driver reports 'Using NWay-set 100BaseTx-FD
media' at every timeout, and throughput is unaffected.

Is there any other way to diagnose this problem?


David Friedman