platform/x86: firmware_attributes_class: Add a high level API
From: | Kurt Borja <kuurtb-AT-gmail.com> | |
To: | Hans de Goede <hdegoede-AT-redhat.com>, Ilpo Järvinen <ilpo.jarvinen-AT-linux.intel.com>, Thomas Weißschuh <linux-AT-weissschuh.net>, Joshua Grisham <josh-AT-joshuagrisham.com>, Mark Pearson <mpearson-lenovo-AT-squebb.ca>, Armin Wolf <W_Armin-AT-gmx.de>, Mario Limonciello <mario.limonciello-AT-amd.com> | |
Subject: | [PATCH v4 0/6] platform/x86: firmware_attributes_class: Add a high level API | |
Date: | Mon, 30 Jun 2025 05:01:07 -0300 | |
Message-ID: | <20250630-fw-attrs-api-v4-0-1a04952b255f@gmail.com> | |
Cc: | Antheas Kapenekakis <lkml-AT-antheas.dev>, "Derek J. Clark" <derekjohn.clark-AT-gmail.com>, Prasanth Ksr <prasanth.ksr-AT-dell.com>, Jorge Lopez <jorge.lopez2-AT-hp.com>, platform-driver-x86-AT-vger.kernel.org, linux-kernel-AT-vger.kernel.org, Dell.Client.Kernel-AT-dell.com, Kurt Borja <kuurtb-AT-gmail.com> | |
Archive-link: | Article |
Hi all, After my discussion with Joshua on v2, I realized the API I made was not ergonomic at all and it didn't exactly respond to driver needs. In this version I tried a completely different approach and IMO it's much much better now. First of all I adopted standard sysfs terminology for everything. A "firmware attribute" is just an attribute_group under the attributes/ directory so everything related to this concept is just called "group" now. Everything refered as properties in the previous patch are now just plain "attributes". This new API revolves around the `fwat_{bool,enum,int,str}_data` structs. These hold all the metadata a "firmware_attribute" of that given type needs. These structs also hold `read` and `write` callbacks for the current_value attribute, because obviously that value is always dynamic. However the rest of attributes (default_value, display_name, min, max, etc) are constant. In the simple case this metadata structs can be defined statically with DEFINE_FWAT_{BOOL,ENUM,INT,STR}_GROUP() macros. However most users of this class obtain this values dynamically so you can also define this structs dynamically. In the end all groups (static and dynamic) will be created using fwat_create_group() after registering the class device. Let me know what you think, your feedback is very appreciated :) I do have one question for anyone interested. Should constraints over the current_value (such as min, max, increment, etc.) be enforced at the show/store level? i.e. before values reach read/write callbacks. Signed-off-by: Kurt Borja <kuurtb@gmail.com> --- Changes in v4: [Patch 1] - Embbed a device in fwat_device instead of a kobject. - Instead of an attrs_kobj root kobj, create a kset with the same name. [Patch 2] - Add a (*show_override) callback in fwat_group_data. - Instead of allocating and filling sysfs groups and attributes manually, I defined custom ktypes for each fwat type. All groups are now statically defined and added through default_groups. I think this is a BIG optimization in terms of memory at least. Also fwat_group memory is now managed by a kobject which is allocated one time. This is also a less impactful performance optimization (less individual allocations). - No changes to API :) (I take suggestions though) I might have lost some of the changelog. Sorry for that! - Link to v3: https://lore.kernel.org/r/20250621-fw-attrs-api-v3-0-3dd5... Changes in v3: [Patch 1] - Fixed UAF in fwat_device_unregister(). Device was unregistered after freeing fadev. [Patch 2] - Patch 2 was completely replaced. A new approach for the API is taken, based on Joshua's suggestions. - Link to v2: https://lore.kernel.org/r/20250517-fw-attrs-api-v2-0-fa1a... Changes in v2: [Patch 1] - Include kdev_t.h header [Patch 2] - Use one line comments in fwat_create_attrs() - Check propagate errors in fwat_create_attrs() - Add `mode` to fwat_attr_config and related macros to let users configure the `current_value` attribute mode - Use defined structs in fwat_attr_ops instead of anonymous ones - Move fwat_attr_type from config to ops [Patch 5] - Just transition to new API without chaing ABI - Link to v1: https://lore.kernel.org/r/20250509-fw-attrs-api-v1-0-258a... --- Kurt Borja (5): platform/x86: firmware_attributes_class: Add high level API for the attributes interface platform/x86: firmware_attributes_class: Move header to include directory platform/x86: samsung-galaxybook: Transition new firmware_attributes API Documentation: ABI: Update sysfs-class-firmware-attributes documentation MAINTAINERS: Add FIRMWARE ATTRIBUTES CLASS entry Thomas Weißschuh (1): platform/x86: firmware_attributes_class: Add device initialization methods .../ABI/testing/sysfs-class-firmware-attributes | 1 + MAINTAINERS | 8 + drivers/platform/x86/dell/dell-wmi-sysman/sysman.c | 2 +- drivers/platform/x86/firmware_attributes_class.c | 659 ++++++++++++++++++++- drivers/platform/x86/firmware_attributes_class.h | 12 - drivers/platform/x86/hp/hp-bioscfg/bioscfg.c | 2 +- drivers/platform/x86/lenovo/think-lmi.c | 2 +- drivers/platform/x86/samsung-galaxybook.c | 246 ++------ include/linux/firmware_attributes_class.h | 375 ++++++++++++ 9 files changed, 1107 insertions(+), 200 deletions(-) --- base-commit: 73f0f2b52c5ea67b3140b23f58d8079d158839c8 change-id: 20250326-fw-attrs-api-0eea7c0225b6 -- ~ Kurt