Re: [Etherboot-developers] problems with e1000

[Christopher Li <ch...@gn...>]

On Fri, 2 Aug 2002 al...@co... wrote:

> 
> Hi etherboot developers,
> 
> I have the following question: with the Intel e1000 etherboot driver
> everything works if we use a e1000 card connected to a 100mbit port on the
> switch and if we connect a 100mbit eepro100 card to the gigabit port.
> However, it does not work if we connect the e1000 card to the gigabit 
> port: no dhcp what-so-ever is coming through.. Does anyone have any ideas 
> how to solve this? Is it a etherboot problem at all?
I think it is a bug in the etherboot e1000 driver. I try to simplify
the intel linux driver due to the code size. It looks like I did not make it
right for the 1000M auto-negotiate. At that time I do not have a 1000M switch
so I can't test that part.

Can you compile e1000 with -DDEBUG=1 can tell me what is the output on
the screen?

Unfortunately I don't have that test environment any more.

I can try to send you some patch off the list for you to try it out, and you
can give me the feedback of the output. It might take several round to
fix it.


The first patch to try is:
========================================
diff -u -r1.1 e1000_phy.c
--- e1000_phy.c	2002/08/03 21:31:22	1.1
+++ e1000_phy.c	2002/08/03 21:31:35
@@ -537,7 +537,7 @@
 BOOLEAN PhySetupAutoNegAdvertisement(void)
  {
 
-#ifdef USE_COMPLEX_VERSION
+#ifdef 1
     UINT16 MiiAutoNegAdvertiseReg, Mii1000TCtrlReg;
 
     DEBUGFUNC("PhySetupAutoNegAdvertisement")
=======================================

See if it helps.

> 
> - Tried 5.0.7 and 5.0.6
> - E1000 is a copper gigabit card; I assume we must use e1000-cop
> - we tried a 3com  4924 switch and a 3com superstack 3 switch
> 
> 
> I read somewhere that it is possible that during phase2/phase3 the eepro's 
> MAC address can be pushed onto the e1000 (??) can this be a reason for our 
> problems?
I am not sure I understand you. But I think MAC address is not the problem.


Chris