[scyld-users] intel etherpro 100 problems

[Donald Becker]

On Wed, 1 Nov 2006, Barry, Garrett wrote:

> I am seeing MAC address corruption problems in the eeprom of the 82559. 
> We see the MAC address corruption after we modify the sleep bit of the 
> eeprom configuration. Anyone ever see issues simular to this ?? 

Which way do you modify the sleep bit -- are you clearing it?
(It's probably not related to this issue, just checking.)

> we are using the eepro100-diag.c stuff to modify the eeprom. Only a 
> single byte of the MAC address stored in the  eeprom is corrupted. We 
> see this after a system has been up and running for a very long time.
> All of a sudden can't communicate to the port and see a bads mac 
> address in the eeprom.

Hmm, that is curious.  That suggests that some other part of the system is 
writing into the NIC.

The EEPROM is used only when the driver is loaded.  The driver reads and 
remembers the important fields.  If the EEPROM word was just randomly 
being overwritten by electrical noise or internal firmware, you wouldn't 
notice a problem until the next time the driver was started.

Reading and writing the 82559 EEPROM is non-trivial.  The diagnostic
program does it the documented way -- generating a bit-serial data
stream by writing the EEPROM control register.  To generate the command
to write an EEPROM word requires over 50 register writes, so it's not
a single errant write causing the problem.

Which word of the EEPROM is being overwritten?  What is the changed value?

(The MAC address is stored in the first three words of the EEPROM.  Each 
word is 16 bits, so typically a pair of bytes would be changed at once.)

-- 
Donald Becker				becker at scyld.com
Scyld Software	 			Scyld Beowulf cluster systems
914 Bay Ridge Road, Suite 220		www.scyld.com
Annapolis MD 21403			410-990-9993