[OLPC-devel] Why ACPI DSDT tables?

[Eric W. Biederman]

David Zeuthen <davidz at redhat.com> writes:

> On Mon, 2006-07-10 at 14:10 +0100, David Woodhouse wrote:
>> On Sun, 2006-07-09 at 17:30 -0600, Eric W. Biederman wrote:
>> > 
>> > Skimming the developers list I caught a reference to ACPI DSDT tables.  
>> > Unless my memory fails me the DSDT tables require interpretation of AML,
>> > and the AML interpreter in the kernel is big (64K+ last time I
>> > checked).  With 128MB of RAM it is a cost that can be paid but at
>> > the same time I think it is expensive enough that if we can avoid them
>> > it gives more memory to other parts of the system.
>> > 
>> > So what is the reason for wanting an ACPI DSDT? 
>> 
>> I can't think of any reason why we'd want something like ACPI. We have
>> proper hardware documentation -- we don't have to be reduced to that
>> kind of abomination.
>
> How do you expect to export things like battery and "on ac power?"
> information to userspace? Frankly, the interfaces the Linux kernel
> exports for these all suck (/proc/acpi, apm, pmu, etc.) but we work
> around that in HAL... and now you want to add *another* interface?

No.  The desire is to provide a better kernel<->firmware interface.

If when we look it is clear there is no kernel<->user space api
then that should get fixed as well. 

Eric