[Beowulf] RE: S2466 Wake on Lan working, anyone?

[David Mathog]

After messing around with this for a couple of days, with some
very helpful messages from Donald Becker, I still had not
managed to make WOL work on the S2466.  Using pci-config it
was possible to put the onboard NIC into the D3 state and
to observe that the correct bit was set in the NIC when
a magic packet hit it. But no matter how late into the
poweroff sequence it was put into this state, once
powered off (poweroff or shutdown -h), it wouldn't power
back on after the magic packet arrived.

I had emailed Tyan support but they never sent anything back.
So today I called them, and the word came down that:
(The following quotes may not be exact but are as close as I can
remember.)

  "WOL only works on the AMD MPX boards from the S1 state".

S1 state has the motherboard fully powered but the disks
may be spun down.

The rationale given for this appalling design decision
(or more likely, cover up for the design error) was that
 
  "Most people who care about such things use a remote
   management card in the machine, so WOL would have
   been redundant". 

Sure, and that's why there are so many posts from people
trying to make WOL work on the S2466 mobos!

Well, we can all stop asking because apparently you can't
get there from here.  It's possible this is still a BIOS
problem that they could fix but I'm leaning more towards the
hypothesis that they neglected to put in the hardware support
required for the NIC to actually trigger a power on
event.

The more I learn about Tyan the less I like them. Surely they've
known all there is to know about the broken WOL for at least
3 years, yet they didn't ever post the reason for the lack of WOL
function in their FAQ section for this motherboard.  Could it
possibly be that they didn't want to lose sales by admitting
the board couldn't do WOL in a manner that was of any use
to anybody?

Regards,

David Mathog
mathog at caltech.edu
Manager, Sequence Analysis Facility, Biology Division, Caltech