[eepro100] wait_for_cmd_timeout (and other errors) on 1.20.2.10

[Derek Glidden]

jpranevich@lycos-inc.com wrote:
> 
> Hello,
> 
> I'm attempting to switch from Intel's eepro100 driver to the stock Linux driver
> in order to facilitate an eventual move to Linux 2.4 on one of my products. I'm
>
> [....]
> 
> Is there a newer driver that you would recommend that I try? Is this a known
> issue with this driver? Is it something odd with the hardware? (Doubtful because
> the Intel driver comes up fine.)

IM(ns)HO, the "stock" Linux driver is hopelessly buggy.  I've seen these
sorts of errors quite a number of times, as well as innumerable "Card
reports no resources" errors.  I can generally make any Intel eepro100
chipset fail with one of these error conditions by bringing the
interface up and down a half-dozen times, and can usually make the box
lock hard by trying to bring the interface up and down a few more times
once the driver has started to complain.  The errors seem to occur
during card/interface initialization, so if you can bring the box up
without the driver complaining, you have a pretty good shot of the
driver working correctly for the life of the box or until/unless you
bring up aliases or down/up the interface for some reason.

For example, this script will quickly start generating errors and has
eventually crashed *any* machine I've run it on using the stock eepro100
driver:

#!/bin/sh
while true
	do
		/sbin/ifup eth0
		ping -c 5 slashdot.org
		sleep 3

		/sbin/ifconfig eth0 down
		/sbin/rmmod eepro100
		sleep 3
	done

I've brought the subject up of these problems on this list a couple of
times recently, usually to an answer of something along the lines of "If
the driver doesn't work for you, install a different one."  I can
recreate the errors at will, but they don't seem to be "true bugs" for
some reason or another.  I've given up trying to find a "real" solution
to the problem, no offense to anyone involved, we just don't seem to be
meet a common ground where I believe that having a machine-crashing
buggy driver in the stock kernel is a big deal, but others don't...

Using either the Intel driver, or Donald Becker's driver from
www.scyld.com work just fine for me.  I personally picked to roll out
Donald's driver on all our eepro100-using boxes.  (Fortunately, it's an
easy install to replace the existing module.)  I don't think that either
of these drivers are designed to work with 2.4, though.  The fact that I
need to use one or the other of them for recent 2.2 kernels to work even
vaguely reliably for us makes that an entirely moot point, as far as I'm
concerned...

-- 
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
With Microsoft products, failure is not           Derek Glidden
an option - it's a standard component.      http://3dlinux.org/
Choose your life.  Choose your            http://www.tbcpc.org/
future.  Choose Linux.              http://www.illusionary.com/