[eepro100] Thin Client EEPRO100 issue.

[Donald Becker]

On Wed, 20 Feb 2002, Ben Greear wrote:

> Supposedly, the eepro100 driver in the 2.4.17-rc2 kernel
> works with 10bt in these cases.  I haven't had time to test it yet.

I've seen many claims of "I changed this <unrelated line of code> and
the problem went away."

The Sleep-State errata cannot be fixed by the driver at run-time.  The
only solution is to clear the EEPROM bit and power cycle the machine.
That's the only way to get the chip will to disable the broken feature.

> The e100 also seems to work.

The e100 driver clears the bit in the driver and then advises the user to
reboot the machine.  I decided not to take that risky approach, and
instead have the user explicitly run the eepro100-diag program to clear
the bit.

> Danilo Godec wrote:
> > On Wed, 20 Feb 2002, Kam Panesar wrote:
> > Unfortunately this is a hardware problem of the Intel ICH2 hub in relation
> > to the LAN controller (the problem is described in
> > ftp://download.intel.com/design/chipsets/specupdt/29824212.pdf).
> > 
> > Intel suggests a workaround by turning the feature 'Standby Enable' off. I
> > think this is also what Donald Becker's eepro100-diag utility suggests,
> > but I have tried it and it didn't help at all.

>From user reports, there are several eepro100 problems and the others
don't show up on the public errata lists.  (Although I do give Intel
credit for publishing the above URL.)

The Sleep-State bit seems to be more important than the document above
indicates -- it appears to cause problems with 100baseTx traffic as well.

> > What makes me worry is this sentence: "There are no plans to fix this
> > erratum."


Donald Becker				becker@scyld.com
Scyld Computing Corporation		http://www.scyld.com
410 Severn Ave. Suite 210		Second Generation Beowulf Clusters
Annapolis MD 21403			410-990-9993