linux-kernel - [PATCH v2 0/2] ACPI: platform_profile: fix legacy sysfs with multiple handlers

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-ID: <20250227153603.131046-1-lkml@antheas.dev>
Date: Thu, 27 Feb 2025 16:36:01 +0100
From: Antheas Kapenekakis <lkml@...heas.dev>
To: mario.limonciello@....com,
	mpearson-lenovo@...ebb.ca
Cc: ilpo.jarvinen@...ux.intel.com,
	lenb@...nel.org,
	linux-acpi@...r.kernel.org,
	linux-kernel@...r.kernel.org,
	platform-driver-x86@...r.kernel.org,
	rafael@...nel.org,
	hdegoede@...hat.com,
	me@...egospodneti.ch,
	luke@...nes.dev,
	Antheas Kapenekakis <lkml@...heas.dev>
Subject: [PATCH v2 0/2] ACPI: platform_profile: fix legacy sysfs with multiple
 handlers

On the Asus Z13 (2025), a device that would need the amd-pmf quirk that
was removed on the platform_profile refactor, we see the following output
from the sysfs platform profile:

$ cat /sys/firmware/acpi/platform_profile_choices
balanced performance

I.e., the quiet profile is missing. Which is a major regression in terms of
power efficiency and affects both tuned, and ppd (it also affected my
software but I fixed that on Saturday). This would affect any laptop that
loads both amd-pmf and asus-wmi (around 15 models give or take?) and only
get worse in 2025, as more laptops start to integrate amd-pmf.

The problem stems from the fact that wmi handlers use different
profiles than amd-pmf, which can load alongside them, and block their
choices. This patch series is a mitigation of this issue, by making pmf
accept all profiles through the legacy sysfs, and by making it a secondary
handler.

While we can argue about whether the secondary handler concept is
necessary, alternatives such as renaming profiles in current drivers will
break existing scripts that are tested for a particular manufacturer, and
allowing amd-pmf override options may cause unforseen regressions in
other wmi drivers.

+CC Luke

Changelog since V1:
    - merge patches 1 and 3 as per Rafael
    - simplify secondary comment about secondary and make it last

Behavior with this patch applied and asus-wmi, amd-pmf
(maintains interop with 6.13):

$ cat /sys/firmware/acpi/platform_profile_choices
quiet balanced performance

And writing quiet to it results in the profile being applied to both
platform profile handlers.

$ echo low-power > /sys/firmware/acpi/platform_profile
bash: echo: write error: Operation not supported
$ echo quiet > /sys/firmware/acpi/platform_profile
$ cat /sys/class/platform-profile/platform-profile-*/{name,profile}
asus-wmi
amd-pmf
quiet
quiet

Agreed ABI still works:
$ echo quiet > /sys/class/platform-profile/platform-profile-0/profile
$ echo quiet > /sys/class/platform-profile/platform-profile-1/profile
bash: echo: write error: Operation not supported
$ echo low-power > /sys/class/platform-profile/platform-profile-0/profile
bash: echo: write error: Operation not supported
$ echo low-power > /sys/class/platform-profile/platform-profile-1/profile

Antheas Kapenekakis (2):
  ACPI: platform_profile: Add handlers that do not occlude power options
  ACPI: platform_profile: make amd-pmf a secondary handler

 drivers/acpi/platform_profile.c    | 57 +++++++++++++++++++++++++-----
 drivers/platform/x86/amd/pmf/spc.c |  3 ++
 drivers/platform/x86/amd/pmf/sps.c |  8 +++++
 include/linux/platform_profile.h   |  5 +++
 4 files changed, 65 insertions(+), 8 deletions(-)

-- 
2.48.1