Hi !

Thanks Quentin for summarizing all this. I've been pretty busy 
elsewhere, so there is a while I havn't put my nose into acpi stuff. 
Yet, on my behalf, my basic need would be to access a userland variable 
capable of changing dynamically processor performance (C states and 
throttling).

> Ultimately this calls for an interface design, because there
> is much more than what the processor object provides.  Support for
> _PSV in the thermal zone objects needs some kind of a bus in order
> to control the processor, for example.

Question : the ACPI tree is already written in hardware. What you 
suggest is, according to what I see, like rebuilding the same tree at 
another level. Is that really efficient?

Vincent