On Wed, Aug 29, 2001 at 10:39:02AM +0100, Dr R.S. Brooks wrote:
> On Tue, 28 Aug 2001, Thor Lancelot Simon wrote:
> 
> >On Tue, Aug 28, 2001 at 01:30:30PM +1000, matthew green wrote:
> >>
> >>     > The problem is the hme does not have its MAC address on it.
> >>     > (Current driver uses MAC address in sparc NVRAM)
> >>     > How should we handle it on other ports?
> >>
> >>    What about HME add-in boards?  It's gotta be in there somewhere.
> >>
> >>
> >> i spent quite some time attempting to find the ethernet address on my
> >> PCI hme card.  i failed.  i'm willing to test code anyone happens to
> >> have, though.
> >
> >Unfortunately, it doesn't "gotta be in there somewhere".  Since the
> >ethernet address is technically a *node* address, not an *interface*
> >address, some workstation vendors (Sun included) use the same MAC addr on
> >every interface on the box.
> 
> I remember looking at this when I was doing some rearrangement of the
> rc scripts under SunOS 4.1.X (because we had a number of machines where
> the primary network interface was FDDI, which was started too late the
> way Sun did it).
> 
> AFAIR, there was one MAC address stored in the EEPROM, and one in each
> network interface (or at least le and nf).  At initialisation, the
> OpenBoot firmware overwrote the address in each interface with the
> EEPROM address.  However, there was some way (maybe an option to
> ifconfig) which would read the interface's own MAC address, which you
> could then set back on the interface.

There is also an openprom boolean that you can set to true to have one
different MAC addr per interface.

--
Manuel Bouyer, LIP6, Universite Paris VI.           Manuel.Bouyer@lip6.fr
--