Re: [acpi-jp 3117] RE: ACPI-CA 20040311 imported

From: Nate Lawson <nate_at_root.org>
Date: Thu, 18 Mar 2004 23:10:56 -0800 (PST)
> -----Original Message-----
> From: owner-acpi-jp_at_jp.FreeBSD.org [mailto:owner-acpi-jp_at_jp.FreeBSD.org]
> On Behalf Of Nate Lawson
> Sent: Thursday, March 18, 2004 10:47 AM
> To: current_at_freebsd.org; acpi-jp_at_jp.FreeBSD.org
> Subject: [acpi-jp 3116] ACPI-CA 20040311 imported
>
> See src/sys/contrib/dev/acpica/CHANGES.txt for specific changes.
>
> The main change is that we now support _OSI to announce we're compatible
> with all the NT-derived MS systems.  Also, we now serialize all method
> execution as some ASL depends on this behavior.  The MS interpreter
> doesn't support parallel execution, hence this matches their behavior.
>
> If there are problems with these features, please try the tunables:
>
> hw.acpi.osi_method
> hw.acpi.serialize_methods
>
> You can disable each feature by setting it to 0 at the loader prompt or
> loader.conf.

On Thu, 18 Mar 2004, Moore, Robert wrote:
> 1) If you serialize all methods by default, you will prohibit recursive
> methods.  That's why we made this an option for Linux, and the default
> is to allow reentrant methods.
>
> 2) We are not really sure about the MS interpreter.  They claim that
> they support reentrant methods and allow multiple threads to execute.
> But we see problems with the coding of reentrant ASL methods that imply
> that that multiple threads never execute control methods concurrently on
> Win*

Interesting.  Do you know of any ASL that uses recursive methods?  I
haven't ever found any like that.  If you don't have a counter-example,
I'm happy to let this sit in the tree for a little while to see if it
solves problems or breaks things for people.  If we don't turn it on by
default, it won't get the testing it needs.

-Nate
Received on Thu Mar 18 2004 - 22:10:53 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:37:48 UTC