[GIT PULL] ACPI fixes for v5.18-rc5

[GIT PULL] ACPI fixes for v5.18-rc5

[Rafael J. Wysocki]

Hi Linus,

Please pull from the tag

 git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm.git \
 acpi-5.18-rc5

with top-most commit 20e582e16af24b074e583f9551fad557882a3c9d

 Revert "ACPI: processor: idle: fix lockup regression on 32-bit ThinkPad T40"

on top of commit b2d229d4ddb17db541098b83524d901257e93845

 Linux 5.18-rc3

to receive ACPI fixes for 5.18-rc5.

These fix up the ACPI processor driver after a change made during
the 5.16 cycle that inadvertently broke falling back to shallower
C-states when C3 cannot be used.

Specifics:

 - Make the ACPI processor driver avoid falling back to C3 type of
   C-states when C3 cannot be requested (Ville Syrjälä).

 - Revert a quirk that is not necessary any more after fixing the
   underlying issue properly (Ville Syrjälä).

Thanks!


---------------

Ville Syrjälä (2):
      ACPI: processor: idle: Avoid falling back to C3 type C-states
      Revert "ACPI: processor: idle: fix lockup regression on 32-bit
ThinkPad T40"

---------------

 drivers/acpi/processor_idle.c | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)

Re: [GIT PULL] ACPI fixes for v5.18-rc5

[Linus Torvalds]

On Thu, Apr 28, 2022 at 8:30 AM Rafael J. Wysocki <rafael@kernel.org> wrote:
>
>  - Make the ACPI processor driver avoid falling back to C3 type of
>    C-states when C3 cannot be requested (Ville Syrjälä).

Hmm.

Doesn't this avoid C3 for _everybody_ rather than some conditional
"when C3 cannot be requested"?

Of course, it looks like the 'has_lpi' state still potentially allows
C3 even for busmastering sleeps. Maybe that is what you meant.

             Linus

Re: [GIT PULL] ACPI fixes for v5.18-rc5

[pr-tracker-bot]

The pull request you sent on Thu, 28 Apr 2022 17:30:15 +0200:

> git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm.git acpi-5.18-rc5

has been merged into torvalds/linux.git:
https://git.kernel.org/torvalds/c/f12d31c00b2e1d0774772695fd644b1c15e446ac

Thank you!

-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/prtracker.html

Re: [GIT PULL] ACPI fixes for v5.18-rc5

[Rafael J. Wysocki]

On Thu, Apr 28, 2022 at 8:50 PM Linus Torvalds
<torvalds@linux-foundation.org> wrote:
>
> On Thu, Apr 28, 2022 at 8:30 AM Rafael J. Wysocki <rafael@kernel.org> wrote:
> >
> >  - Make the ACPI processor driver avoid falling back to C3 type of
> >    C-states when C3 cannot be requested (Ville Syrjälä).
>
> Hmm.
>
> Doesn't this avoid C3 for _everybody_ rather than some conditional
> "when C3 cannot be requested"?

The bug was that if acpi_idle_bm_check() returned 1, it would "fall
back"  to the same state via safe_state_index.

> Of course, it looks like the 'has_lpi' state still potentially allows
> C3 even for busmastering sleeps. Maybe that is what you meant.

That too.

Thanks for pulling!

Rafael

Re: [GIT PULL] ACPI fixes for v5.18-rc5

[Linus Torvalds]

On Thu, Apr 28, 2022 at 12:50 PM Rafael J. Wysocki <rafael@kernel.org> wrote:
>
> The bug was that if acpi_idle_bm_check() returned 1, it would "fall
> back"  to the same state via safe_state_index.

That's actually quite hard to see in the source code.

Looking closer, the code calls "acpi_idle_enter_bm()" only when
"cx->type == ACPI_STATE_C3", but that is very non-obvious in the
context of that function (other than the comment).

It might be clearer if the function was actually called "enter_c3()"
rather than "enter_bm()". Particularly since it will continue to do
that "c3_cpu_count" stuff even if the actual power state it goes into
_isn't_ C3.

And the the C3 case doesn't seem to take that "demote C2 to C1" rule
into account, so now that code will go into "safe" C2 mode even if
ACPI_FADT_C2_MP_SUPPORTED isn't set.

It's all very confusing.

But whatever. I obviously already pulled the changes.

                Linus