I have one PC with the ATX power supply always on (green wire cut and
joined to black).
So when you power off from operating system, last thing should be to
toggle the power control line (green wire). With this system, I find
that the fans go off. Furthermore the mainboard has one of those
hexadecimal status displays. It goes to 00, which manual says is
"unused". I would hope then that CPU has gone into low-power state.

Am 29.08.2023 schrieb Woozy Song <suzyw0ng@outlook.com>:

> I have one PC with the ATX power supply always on (green wire cut and
> joined to black).

Why that?

What is your goal?

ACPI motherboards and ATX power supplies are common for more than 20
years, almost nobody of the developers cares about AT systems that need
to be powered off manually.

Woozy Song <suzyw0ng@outlook.com> wrote:
> I have one PC with the ATX power supply always on (green wire cut and
> joined to black).
> So when you power off from operating system, last thing should be to
> toggle the power control line (green wire). With this system, I find
> that the fans go off. Furthermore the mainboard has one of those
> hexadecimal status displays. It goes to 00, which manual says is
> "unused". I would hope then that CPU has gone into low-power state.

It probably goes into either the HLT state:
https://en.wikipedia.org/wiki/HLT_(x86_instruction)

Or if there's a Linux ACPI driver enabled then maybe it will use
the ACPI C1 state:
https://en.wikipedia.org/wiki/ACPI#Processor_states

Either of these is likely to be a lower power state on relatively
recent x86/x86_64 CPUs. But I don't know conclusively how the
Linux kernel handles power-off, it just seems silly that it
wouldn't enter the CPU into one of those modes. Also in the past
I've measured a router, an x86_64 single-board-computer (with an
external always-on power supply), and Raspberry Pi Zero boards,
when powered off from Linux and all showed significantly reduced
power consumption. It _could_ be that this is just from disabling
peripheral chips/functions, but it would be stupid if it doesn't
put the CPU into a low power mode too.

Someone at Stack Overflow dug deeper than I'm willing to go and
found out which routines the kernel calls during power-off/halt:
https://unix.stackexchange.com/a/122667

--
__ __
#_ < |\| |< _#

At 31 Aug 2023 08:52:40 +1000 not@telling.you.invalid (Computer Nerd Kev) wrote:

>
> Woozy Song <suzyw0ng@outlook.com> wrote:
> > I have one PC with the ATX power supply always on (green wire cut and
> > joined to black).
> > So when you power off from operating system, last thing should be to
> > toggle the power control line (green wire). With this system, I find
> > that the fans go off. Furthermore the mainboard has one of those
> > hexadecimal status displays. It goes to 00, which manual says is
> > "unused". I would hope then that CPU has gone into low-power state.
>
> It probably goes into either the HLT state:
> https://en.wikipedia.org/wiki/HLT_(x86_instruction)
>
> Or if there's a Linux ACPI driver enabled then maybe it will use
> the ACPI C1 state:
> https://en.wikipedia.org/wiki/ACPI#Processor_states
>
> Either of these is likely to be a lower power state on relatively
> recent x86/x86_64 CPUs. But I don't know conclusively how the
> Linux kernel handles power-off, it just seems silly that it
> wouldn't enter the CPU into one of those modes. Also in the past
> I've measured a router, an x86_64 single-board-computer (with an
> external always-on power supply), and Raspberry Pi Zero boards,
> when powered off from Linux and all showed significantly reduced
> power consumption. It _could_ be that this is just from disabling
> peripheral chips/functions, but it would be stupid if it doesn't
> put the CPU into a low power mode too.
>
> Someone at Stack Overflow dug deeper than I'm willing to go and
> found out which routines the kernel calls during power-off/halt:
> https://unix.stackexchange.com/a/122667

Why guess? The Linux kernel source is available for download -- "Use The
Source, Luke"...

>

--
Robert Heller -- Cell: 413-658-7953 GV: 978-633-5364
Deepwoods Software -- Custom Software Services
http://www.deepsoft.com/ -- Linux Administration Services
heller@deepsoft.com -- Webhosting Services

Robert Heller <heller@deepsoft.com> wrote:
> At 31 Aug 2023 08:52:40 +1000 not@telling.you.invalid (Computer Nerd Kev) wrote:
>>
>> Either of these is likely to be a lower power state on relatively
>> recent x86/x86_64 CPUs. But I don't know conclusively how the
>> Linux kernel handles power-off, it just seems silly that it
>> wouldn't enter the CPU into one of those modes. Also in the past
>> I've measured a router, an x86_64 single-board-computer (with an
>> external always-on power supply), and Raspberry Pi Zero boards,
>> when powered off from Linux and all showed significantly reduced
>> power consumption. It _could_ be that this is just from disabling
>> peripheral chips/functions, but it would be stupid if it doesn't
>> put the CPU into a low power mode too.
>>
>> Someone at Stack Overflow dug deeper than I'm willing to go and
>> found out which routines the kernel calls during power-off/halt:
>> https://unix.stackexchange.com/a/122667
>
> Why guess? The Linux kernel source is available for download -- "Use The
> Source, Luke"...

If I needed to know then that's where I'd look. But past experience
suggests it might take many hours to find the right place,
especially for architecture-specific functionality like this.

My proverb: He who asketh the question can wasteth his own
afternoon finding a precise answer.

That last link is a good starting point.

--
__ __
#_ < |\| |< _#

CN> If I needed to know then that's where I'd look. But past experience
CN> suggests it might take many hours to find the right place,

Maybe you could try grep ing for it? Or you could try looking through pull
requests to find ones related to power-off, and check the files edited.

-----------------------------------
user is generated from /dev/urandom

candycane <candycane@f172.n1.z21.fsxnet> wrote:
> CN> If I needed to know then that's where I'd look. But past experience
> CN> suggests it might take many hours to find the right place,
>
> Maybe you could try grep ing for it? Or you could try looking through pull
> requests to find ones related to power-off, and check the files edited.

Cscope is usually a bit easier to use for that than Grep for large
projects. It can still get you lost when things turn into a twisty
maze of function pointers, structures, and macros though.

--
__ __
#_ < |\| |< _#

CN> Cscope is usually a bit easier to use for that than Grep for large
CN> projects. It can still get you lost when things turn into a twisty
CN> maze of function pointers, structures, and macros though.

Cscope?

-----------------------------------
user is generated from /dev/urandom

candycane <candycane@f172.n1.z21.fsxnet> wrote:
> CN> Cscope is usually a bit easier to use for that than Grep for large
> CN> projects. It can still get you lost when things turn into a twisty
> CN> maze of function pointers, structures, and macros though.
>
> Cscope?

From the readme:

"Cscope is a text screen based source browsing tool. Although it is
primarily designed to search C code (including lex and yacc files), it
can also be used for C++ code."

https://cscope.sourceforge.net/

Essential keystrokes:
TAB to switch cursor between query and result sections.
CTRL-D to exit.

The are/were also GUIs, which I haven't tried.

--
__ __
#_ < |\| |< _#

CN> "Cscope is a text screen based source browsing tool. Although it is
CN> primarily designed to search C code (including lex and yacc files), it
CN> can also be used for C++ code."

Neat, I'll have to check it out.

-----------------------------------
user is generated from /dev/urandom