[haiku-gsoc] Re: [HPET] (Fwd:) I'm in a jam.

Hi Dustin

> The HPET ACPI data is stored in the DSDT in encoded AML;
> I believe that a full AML parser and DSDT Tree interpreter is out of
> the scope of the stage1 bootloader and would be unnecessary bloat; to
> do it in the kernel would be needless effort- and code-duplication
> between it and the ACPI bus_manager.

I thought there is enough data in the separate HPET table to do basic 
setup/use of the timer. This might not expose all available timers and 
such, but it may be enough for our needs? The idea being similar to 
that of the APIC table that provides boot code with enough MP 
information to configure/start a SMP system without a full blown AML 
parser.

> The only options I can think of are...
> 1) put it in the bootloader (bloat it and rely on code from acpica 
> and
> acpi bus manager),
> 2) likewise for the kernel except the bloat is less meaningful since
> the kernel is larger
> 3) have modules load earlier and use the ACPI module; could also make
> the timers pluggable modules if time permits
>
> Any ideas? I'm at a loss here :)

I was at pretty much the same place when doing the IOAPIC stuff. It's 
all nice and everything can be setup, but the quite important interrupt 
routing table that assigns PCI to global interrupts is only available 
in the DSDT. That's why IOAPICs are not in use right now, they would 
simply not work for any PCI devices...
Therefore I would like to see 3 as the best solution, as it could then 
be used in that code as well. Doing something specifically for HPET 
would probably be a waste of time, as there are a few other things that 
will require full ACPI anyway and having it available from the normal 
module instead of duplicating specific code makes the most sense to me.

Regards
Michael

Other related posts: