[eepro100] Problem with eepro100 and i82559 NIC

Donald Becker becker@scyld.com
Sat, 24 Jun 2000 17:52:42 -0400 (EDT)


On Sat, 24 Jun 2000, David [iso-8859-1] Müller wrote:

> I have a problem to bring my Strongarm (SA-110) based 
                               ^^^^^^^^^^^^^^^^^^
> Linux (2.3.39) system and an Intel 82559 based NIC together.

What hardware are you using?

The StrongARM is problematic, since it's not cache coherent with PCI bus
transactions.  All bus-master device drivers must be modified to work with
it.

We had been talking to the network device side of Intel (the people that
make switches, not the people that make the network adapters) about doing
the StrongARM support as part of a larger project, but that work seems to
have fallen through.  Thus I don't have any hardware to test StrongARM
specific driver modifications.

> eepro100.c:v1.09j 7/27/99 Donald Becker http://cesdis.gsfc.nasa.gov/linux/drivers/eepro100.html
> eth0: Intel PCI EtherExpress Pro100 at 0xc1800000, 00:D0:B7:4C:36:56, IRQ 24.
>   Receiver lock-up bug exists -- enabling work-around.
>   Board assembly 721383-008, Physical connectors present: RJ45
>   Primary interface chip i82555 PHY #1.
> Self test failed, status ffffffff:
>  Failure to initialize the i82557.

This is the primary problem.  It may be that the self-test worked but the
chip is reading the cache line with the old value.
You must put in cache coherency calls before reading any cache line written
by the bus master.

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