[OLPC-devel] Why ACPI DSDT tables?

[David Woodhouse]

On Mon, 2006-07-10 at 15:44 -0400, Jim Gettys wrote:
> A lot of the power management stuff (battery charging, on AC power,
> wakeup of the Geode) is done in the embedded controller (EC).  We have
> to have some sort of interface between it and the rest of the system.  
> 
> And yes, on the iPAQ (and maybe on the Nokia 770), we emulated APM, just
> for simplicity's sake.
> 
> If you are seriously suggesting that we do something other than ACPI for
> the internal interfaces from the EC and the Geode, (something that
> Quanta is already familiar with), then you need to make the case now in
> concert with Ray Tseng, while the code is still being written (by Ray
> and his team).

That makes no sense to me -- you're talking about a _hardware_
interface. That doesn't seem particularly relevant to the question of
what we run on the host CPU -- whether it's proper open source drivers
in the kernel or whether the host CPU should run the baroque AML stuff
in an interpreter instead.

> I'm very reluctant to see is another, different set of user space
> interfaces happen than those provided by the widely used ACPI and APM
> interfaces: there lies madness for people like David, and others working
> in user space. 

The best solution all round is to take this opportunity to _fix_ the
inconsistent kernel<->user interfaces, so that _all_ kernel
implementations would appear the same way in sysfs.

Emulating /proc/apm like everyone else already does would also work and
be slightly less effort. 

-- 
dwmw2