[prev in list] [next in list] [prev in thread] [next in thread] 

List:       netbsd-current-users
Subject:    Re: ACPI0007 ?   Anyone know what it is?
From:       Jared McNeill <jmcneill () invisible ! ca>
Date:       2022-11-06 13:54:35
Message-ID: ce3637dd-acc1-5e0-a369-4871d77e765c () invisible ! ca
[Download RAW message or body]

ACPI0007 is the hardware ID of a device node that represents a processor 
device (there are two ways to do this -- either by declaring a Processor 
node in the device tree, or by declaring a standard device with HID 
ACPI0007).

It's not used to discover processors by OSes because typically that is 
done well before the AML interpreter is brought online. Instead it is used 
to describe how to control processor performance and thermal management 
(what ACPI calls C-states, P-states, and T-states on older systems, or 
CPPC on newer systems). How processors are discovered is platform 
dependent, but at least for x86/Arm/RISC-V you would parse the ACPI MADT 
to find these early at boot. Each processor has a unique ID in firmware 
that lets you associate an (optional) processor device node with the 
processor definition in the MADT.

The error message you are seeing comes from an (ill conceived IMHO) option 
ACPI_ACTIVATE_DEV which is not enabled by default in any kernels we ship. 
Each device node is going to have a _STA method that tells the OS whether 
that device is enabled. Because updating AML dynamically can be difficult, 
it's not uncommon for firmware to create templates for all possible 
processors and the _STA method will inspect some variable set early at 
boot to determine whether or not a given processor device is present.

So tl;dr:

  - Your firmware says "There is no processor device here"
  - Your kernel config says "Let me try anyway" and asks firmware to enable
    the device.
  - Firmware says no.
  - You get an error message logged.

HTH,
Jared


On Sat, 5 Nov 2022, Robert Elz wrote:

> This is just (well mostly) curiosity, it doesn't seem to
> affect anything that I know about, but during autoconfig,
> the kernel (any 9.99.x I have tried I think) prints
>
>   acpi0: autoconfiguration error: failed to activate ACPI0007
>
> 64 times during the boot sequence (this one is from 9.99.105)
>
> I was just wondering what I apparently have so much of that
> this would appear so often (the other "failed to activate"
> messages I get seem to just occur once).
>
> The only thing that I'm aware of that I have 64 of is 64GB ram,
> but that's on just 2 DIMMs, and it is hard to believe that something
> is existing for each individual GB.
>
> Before I actually counted, when I just watched them scroll past (very
> quickly) during the boot, I guessed that it was perhaps once per CPU,
> but 64 CPUs, no matter how you count them, is way more than I have,
> so I doubt it is that.
>
> Anyone know ACPI well enough to know what that is?
>
> kre
>
>
>
[prev in list] [next in list] [prev in thread] [next in thread]