Ben,

These problems show up on both our IBM RS/6000 F50 and IBM RS/6000 B50.
In the case of the F50, the problem is apparent on the add-in card.  (I
believe the built in Ethernet only supports 10mbit)  It doesn't happen in
10mbit mode, however... only 100mbit.

Unfortunately, I have never worked with an Ethernet driver so I have no
idea where to start playing with parameters.  :)

Stephen

Terra Soft Solutions, Inc.
   Yellow Dog Linux
   "The Ultimate Companion for a Dedicated Server"
   http://www.yellowdoglinux.com/


On Wed, 15 Dec 1999, Benjamin Herrenschmidt wrote:

On Wed, Dec 15, 1999, Stephen Edie [off-list ref] wrote:

quoted

When running in 100mbit mode, the PCNET32 driver reports the following
error message:

eth0:TX FIFO error! Status 02e3

With this messages present, we have seen a number of problems from
TCP connections being closed after a few bytes of transfer to very poor
performance (1-mbit throughput) to not working at all.  Is there a quick
way to fix this problem perhaps?  We would like to get some of our
in-house CHRP machines running at the full 100mbit and we have some
customers who require 100mbit throughput as well.

Which pcnet32 chip are you using ? There are some issues with
auto-negociation on some integreated-PHY models, and you may have to
adjust a few of the chip parameters. I used the pcnet32 driver with an
AMD 973 on an ARM platform and it worked just fine (after I fixed the
cache coherency issues but those won't happen on PPC).

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/