Looking for an ASUS motherboard that net boots...

[Lennart Sorensen]

On Fri, Aug 15, 2008 at 10:13:24AM -0400, Lennart Sorensen wrote:
> On Thu, Aug 14, 2008 at 09:10:03PM -0400, Colin McGregor wrote:
> > I've previously noted on this mailing list the problems I have had
> > attempting to get an ASUS M2N-MX SE Plus motherboard to net boot
> > (bottom line, I have shipped the motherboard to ASUS twice, they have
> > shipped the motherboard back twice and I still have a motherboard with
> > a broadcast MAC address, grumble)...
> 
> So what exactly is the MAC addres that the board is sending when trying
> to network boot right now?
> 
> What MAC address does it have once in Linux if you boot a live cd or
> such?
> 
> If it sends a correct MAC address to get its IP and start booting but is
> only wrong once you hit linux, then you simply need a newer kernel, or
> to file a bug report because this is a known issue with the reverse
> engineered forcedeth driver and not actually a problem with the board.
> If the board is able to send a request to start booting from the network
> and load the kernel and start executing the kernel, then there is
> NOTHING wrong with your board, only with your kernel/network driver and
> there is nothing Asus can do to fix it for you.

In fact this has been an issue for a few years now, given I found people
discussing this issue when network booting.

The issue is:
The MAC address is stored backwards on nvidia boards (No one ever seems
to have explained who got that bright idea) so it has to be reversed
before using it.  This is done by reading it from the network chip and
reversing it and writing it back to the same location in the chip.

The linux driver knows to do this reverse thing, and also knows to
unreverse it on suspend and such so that it will be correct next time
the driver starts up again.

Of course the BIOS boot code also needs the MAC address correct, so it
too reverses it so it can use the port for network booting, but it does
NOT put it back when done network booting.  Apparently some board makers
have updated their PXE code to in fact reverse it again when PXE shuts
down, or at least so it seems.  I guess Asus has not done so on this
board.  I am wondering if the windows driver has been changed to deal
with this, or if they simply never tested network boot.

So the linux driver needs some way to detect if the address has already
been reversed or if it needs to be reversed.  How to do that reliably is
the big debate (or at least was 2 years ago).

Some people have even found the windows driver doesn't work after PXE
boot for the same reason.  Briliant move on the part of whoever got the
stupid reverse MAC address storage idea. :)

You can see the discussion I found here:
http://linux.derkeiler.com/Mailing-Lists/Kernel/2006-10/msg01345.html

-- 
Len Sorensen
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists