[eepro100] Problems at 100fd with eepro100

[Donald Becker]

On Thu, 18 May 2000, Vernon McPherron wrote:
> On Thu, 18 May 2000, Philip Ronzone wrote:
> > The problem is not so odd. More than one Ethernet "chip" (NIC) works fine
> > with a hub, but has a variety of pronlems with a switch. In fact, I'm

Writing "repeater" is more precise than "hub".

A switch often can do full duplex, while a repeater never can.

The usual problem with a switch is caused by the admin forcing full duplex,
which must turn off autonegotiation, and then failing to update the settings
when the topology changes.

> > struggling with that very problem right now. Intel has admitted to problems
> > with the NT drivers (82559) and switches, and I'm seeing major problems with
> > a switch whenever the driver is set to something OTHER than "auto".

What kind of problems?

> > I'm using the Intel e100.c driver right now, but I'm told this is NOT a
> > driver problem, but a hardware issue.

Hmmm, very curious.
I could believe that Intel's  e100.c driver has bugs in common with their NT
driver.
Or they might mean that your Cisco or 3Com switch has broken autonegotiation.

> > From: Vernon McPherron [mailto:vernon@grooveauction.com]
..
> > I've got an odd problem, I don't know if anyone else is having the same
> > problem, but maybe I could get some help.
> > 
> > Not too long ago I was using a 100bt hub.  I had no problems.  But just
> > the other day I bought a 10/100 switch.  When I hooked it up, I found out
> > that I was having SERIOUS problems.  The module loaded fine, and I got a
...
> > patched it to 2.3.99-pre8 hoping that it'd have the latest driver.  After
...
> > of the pings did go through.  But I had about 95% packet loss.  Just to
> > test it, I booted into an older non-smp kernel, and yep...  worked
> > ok.  Unfortunately I don't want to use just one cpu.  Any ideas?

This really sounds like just a bug in the 2.3.99 driver.

Getting this driver working in all environments is difficult.  Those that
though that I was refusing to put in "the obvious fix" are now finding out
that putting the patch de jour into the distributed kernel doesn't result in
immediate convergence to a working driver.

Donald Becker				becker@scyld.com
Scyld Computing Corporation
410 Severn Ave. Suite 210
Annapolis MD 21403