Re: pchb@acpi support for ia64

[Jukka Ruohonen <jruohonen%iki.fi@localhost>]

On Tue, Jun 29, 2010 at 03:27:31PM +0900, KIYOHARA Takashi wrote:
> I will commit pchb@acpi on the next weekend if you do not have the
> problem. With the pci support for ia64.

I don't have objections about the code as such. But few points.

1. What is gained from this? From a quick look, only the bus number is used. 
   Generally, I think we want to move away from these ad hoc wrappers (cf.
   ISA devices at ACPI).

2. At minimum, this should be done so that the main ACPI architecture, x86,
   would benefit from it (sys/dev/acpi is MI, after all).

3. If this (MI) pchb_acpi.c is used, I would like to see the _PRT routing in
   there too, if available (instead of the MD mpacpi.c on x86).

4. Bear in mind that the surrounding big questions are all pretty much
   unresolved. Should conventional devices or buses attach to ACPI at all?
   Or should devices just query ACPI for the required information? 
   Autoconfiguration or something else? And so on.

I think a discussion particularly about (4) should occur before we start to
attach pci(4) to acpi(4) on any given architecture.

> By the way, The debug message of ACPI is very noisy.  Can I start
> more quietly?  X-<
> 
>     :
>   [ACPI Debug]  Integer 0x00000000
>   [ACPI Debug]  String [0x21] "PARS.GPTR: Getting SERIAL DEV Ptr"
>   [ACPI Debug]  String [0x1F] "PARS.GUID: evaluating for COMP1"
>     :
>   [ACPI Debug] ...

This is a new debug-feature introduced in the recently imported ACPICA.
I will silence it soon.

- Jukka.