Firmware GPIO setup

[Jordan Crouse]

On 14/03/07 19:19 -1000, Mitch Bradley wrote:
> Analysis of differences:
>  PWR_BUT# might be a problem.  If it's not configured for input, you 
> can't use it to shutdown

As an aside, some time ago (B1 era?), I noticed that the 4 second fail
safe wasn't enabled in OFW - not having PWR_BUT# configured would go a long
way toward explaining that.  That might be something useful.

>  THERM_ALRM# is unknown.  Does the EC assert it?

THERM_ALRM# is an external IC that asserts the line when the temp gets
too high (I had asked for a nice SMBUS attached sensor that we could
play with, but apparently they are too expensive).  Its a very good idea
to enable THERM_ARLM# functionality in OFW - since it will protect the
machine if it gets booted (or resumed) in extreme tempratures, but its not
dangerous for B73.

>  Event enable on DCONIRQ - don't know

For B73, it shouldn't matter. OFW doesn't use any of the features that
DCONIRQ provides, and if it gets enabled in OFW, we run the risk
of spurious interrupts.  Its probably better for us to put that in the
kernel - but that means we'll have to have at least some of the GPIO
setup in the kernel driver - which I was hoping to avoid.

>  Event enable on SCI# - don't know - suspect not a problem?

Not a problem for B73.  And probably not useful for fastboot 
either - OFW will only be up for a few milliseconds
on both boot and resume - any SCI events that happen between
then and the kernel booting can probably be processed by the kernel.
EB_MODE or some wireless event would be the only typical SCI events
we would see during boot (most of the SCI events will be wake events).

>  Event enable on PWR_BUT_in - don't know - suspect not a problem?

Again, not dangerous for B73.   But an interesting question, none the
less.  Somebody boots (or resumes), and then immediately closes the
lid.  Do we boot into the kernel, process the lid event and turn around
and suspend?  Do we detect the event, set a special flag, and have OFW
suspend (knowing that when it resumes, that it should boot the kernel,
not resume to a subroutine?).  Do we ignore it all together? 

Jordan

-- 
Jordan Crouse
Senior Linux Engineer
Advanced Micro Devices, Inc.
<www.amd.com/embeddedprocessors>