7:52 p.m.
Hi all,
after some puzzling issues with different DSDT on various system I would
like to know the correct/best way to setup ACPICA.
The core of the init sequence is:
AcpiEnableSubsystem(ACPI_FULL_INITIALIZATION);
AcpiInitializeObjects(ACPI_FULL_INITIALIZATION);
As part of the former call, the event subsystem is initialised,
including setting up the SCI. To process that request, the overrides in
the MADT have to be parsed and that in turn requires switching to APIC
mode.
This now leaves two primary options:
- Switch to APIC mode and parse the MADT from the
AcpiOsInstallInterruptHandler
- Defer the SCI setup until after the call.
The second is messy and can't be the intention of the code. The former
makes me wonder about the error resilience though. We have seen some
DSDTs in NetBSD that do not setup the _PRT correctly before the object
initialisations are done. I don't know yet, if the _INI change in the
last release fixed that. I am wondering though, if the MADT processing
can be affecting by the same issue. Are there any known cases of such
DSDTs?
The second issue is the question of reentrancy. Is ACPICA supposed to be
reentrant enough to allow this?
Joerg
7:58 p.m.
I'm not sure that I fully understand your question, but things like the MADT and ECDT
are typically parsed and consumed by the host OS *before* the ACPICA subsystem is
initialized. This is why the ACPICA Table Manager is available at early boot time, before
ACPICA is initialized, and even before dynamic memory is available.
Bob
-----Original Message-----
From: devel-bounces(a)acpica.org [mailto:[email protected]] On Behalf
Of Joerg Sonnenberger
Sent: Friday, September 04, 2009 12:52 PM
To: devel(a)acpica.org
Subject: [Devel] Correct ACPI setup sequence
Hi all,
after some puzzling issues with different DSDT on various system I would
like to know the correct/best way to setup ACPICA.
The core of the init sequence is:
AcpiEnableSubsystem(ACPI_FULL_INITIALIZATION);
AcpiInitializeObjects(ACPI_FULL_INITIALIZATION);
As part of the former call, the event subsystem is initialised,
including setting up the SCI. To process that request, the overrides in
the MADT have to be parsed and that in turn requires switching to APIC
mode.
This now leaves two primary options:
- Switch to APIC mode and parse the MADT from the
AcpiOsInstallInterruptHandler
- Defer the SCI setup until after the call.
The second is messy and can't be the intention of the code. The former
makes me wonder about the error resilience though. We have seen some
DSDTs in NetBSD that do not setup the _PRT correctly before the object
initialisations are done. I don't know yet, if the _INI change in the
last release fixed that. I am wondering though, if the MADT processing
can be affecting by the same issue. Are there any known cases of such
DSDTs?
The second issue is the question of reentrancy. Is ACPICA supposed to be
reentrant enough to allow this?
Joerg
_______________________________________________
Devel mailing list
Devel(a)acpica.org
http://lists.acpica.org/listinfo/devel
8:11 p.m.
On Fri, Sep 04, 2009 at 12:58:26PM -0700, Moore, Robert wrote:
I'm not sure that I fully understand your question, but things
like
the MADT and ECDT are typically parsed and consumed by the host OS
*before* the ACPICA subsystem is initialized. This is why the ACPICA
Table Manager is available at early boot time, before ACPICA is
initialized, and even before dynamic memory is available.
OK, I think that answers the question indirectly. Just to make sure:
- I can parse the MADT before anything else.
- If the MADT exists and contains IOAPICs, I can expect \_PIC to be
successful.
The second is important for the case of Interrupt Overrides for the SCI,
which is the problem here.
Joerg
PS: ACPI 4.0 has a bad cross-reference in 5.2.12.4, 5.2.10.8 no longer
exists. Who is the right person for that?
8:23 p.m.
-----Original Message-----
From: Joerg Sonnenberger [mailto:[email protected]]
Sent: Friday, September 04, 2009 1:11 PM
To: Moore, Robert
Cc: devel(a)acpica.org
Subject: Re: [Devel] Correct ACPI setup sequence
On Fri, Sep 04, 2009 at 12:58:26PM -0700, Moore, Robert wrote:
> I'm not sure that I fully understand your question, but things like
> the MADT and ECDT are typically parsed and consumed by the host OS
> *before* the ACPICA subsystem is initialized. This is why the ACPICA
> Table Manager is available at early boot time, before ACPICA is
> initialized, and even before dynamic memory is available.
OK, I think that answers the question indirectly. Just to make sure:
- I can parse the MADT before anything else.
Yes.
- If the MADT exists and contains IOAPICs, I can expect \_PIC to be
successful.
I don't know the answer to this.
The second is important for the case of Interrupt Overrides for the SCI,
which is the problem here.
Joerg
PS: ACPI 4.0 has a bad cross-reference in 5.2.12.4, 5.2.10.8 no longer
exists. Who is the right person for that?
I just forwarded this to the ACPI 4.0 editor. Thanks. You can send any such errors
directly to me. We own the ACPI specification here.
Bob
8:25 p.m.
On Fri, Sep 04, 2009 at 01:23:30PM -0700, Moore, Robert wrote:
>- If the MADT exists and contains IOAPICs, I can expect \_PIC to
be
> successful.
I don't know the answer to this.
What would be the most sensible behavior if it fails and the MADT listed
an entry for the SCI?
>PS: ACPI 4.0 has a bad cross-reference in 5.2.12.4, 5.2.10.8 no
longer
>exists. Who is the right person for that?
I just forwarded this to the ACPI 4.0 editor. Thanks. You can send any
such errors directly to me. We own the ACPI specification here.
Good to know, thanks.
Joerg
8:28 p.m.
Sorry, I'm not an expert on the interrupt routing and MADT stuff.
Perhaps someone else here can answer this.
-----Original Message-----
From: Joerg Sonnenberger [mailto:[email protected]]
Sent: Friday, September 04, 2009 1:25 PM
To: Moore, Robert
Cc: devel(a)acpica.org
Subject: Re: [Devel] Correct ACPI setup sequence
On Fri, Sep 04, 2009 at 01:23:30PM -0700, Moore, Robert wrote:
> >- If the MADT exists and contains IOAPICs, I can expect \_PIC to be
> > successful.
>
>
> I don't know the answer to this.
What would be the most sensible behavior if it fails and the MADT listed
an entry for the SCI?
> >PS: ACPI 4.0 has a bad cross-reference in 5.2.12.4, 5.2.10.8 no longer
> >exists. Who is the right person for that?
>
>
> I just forwarded this to the ACPI 4.0 editor. Thanks. You can send any
> such errors directly to me. We own the ACPI specification here.
Good to know, thanks.
Joerg
4728
days inactive
4728
days old
5 comments
2 participants
participants (2)
-
Joerg Sonnenberger -
Moore, Robert