Re: [PATCH v4 4/6 RESEND] platform/x86: Add Lenovo Capability Data 01 WMI Driver

[Armin Wolf <W_Armin@gmx.de>]

Am 02.04.25 um 22:47 schrieb Derek John Clark:

> On Wed, Mar 26, 2025 at 6:29 PM Armin Wolf <W_Armin@gmx.de> wrote:
>> Am 17.03.25 um 15:43 schrieb Derek J. Clark:
>>
>>> Adds lenovo-wmi-capdata01 driver which provides the
>>> LENOVO_CAPABILITY_DATA_01 WMI data block that comes on "Other Mode"
>>> enabled hardware. Provides an interface for querying if a given
>>> attribute is supported by the hardware, as well as its default_value,
>>> max_value, min_value, and step increment.
>>>
>>> Signed-off-by: Derek J. Clark <derekjohn.clark@gmail.com>
>>> ---
>>> v4:
>>>    - Make driver data a private struct, remove references from Other Mode
>>>      driver.
>>>    - Don't cache data at device initialization. Instead, on component bind,
>>>      cache the data on a member variable of the Other Mode driver data
>>>      passed as a void pointer.
>>>    - Add header file for capdata01 structs.
>>>    - Add new struct to pass capdata01 array data and array length to Other
>>>      Mode.
>>> v3:
>>> - Add as component to lenovo-wmi-other driver.
>>> v2:
>>> - Use devm_kmalloc to ensure driver can be instanced, remove global
>>>     reference.
>>> - Ensure reverse Christmas tree for all variable declarations.
>>> - Remove extra whitespace.
>>> - Use guard(mutex) in all mutex instances, global mutex.
>>> - Use pr_fmt instead of adding the driver name to each pr_err.
>>> - Remove noisy pr_info usage.
>>> - Rename capdata_wmi to lenovo_wmi_cd01_priv and cd01_wmi to priv.
>>> - Use list to get the lenovo_wmi_cd01_priv instance in
>>>     lenovo_wmi_capdata01_get as none of the data provided by the macros
>>>     that will use it can pass a member of the struct for use in
>>>     container_of.
>>> ---
>>>    MAINTAINERS                                 |   2 +
>>>    drivers/platform/x86/Kconfig                |   4 +
>>>    drivers/platform/x86/Makefile               |   1 +
>>>    drivers/platform/x86/lenovo-wmi-capdata01.c | 136 ++++++++++++++++++++
>>>    drivers/platform/x86/lenovo-wmi-capdata01.h |  29 +++++
>>>    5 files changed, 172 insertions(+)
>>>    create mode 100644 drivers/platform/x86/lenovo-wmi-capdata01.c
>>>    create mode 100644 drivers/platform/x86/lenovo-wmi-capdata01.h
>>>
>>> diff --git a/MAINTAINERS b/MAINTAINERS
>>> index 6dde75922aaf..56ead241a053 100644
>>> --- a/MAINTAINERS
>>> +++ b/MAINTAINERS
>>> @@ -13164,6 +13164,8 @@ L:    platform-driver-x86@vger.kernel.org
>>>    S:  Maintained
>>>    F:  Documentation/wmi/devices/lenovo-wmi-gamezone.rst
>>>    F:  Documentation/wmi/devices/lenovo-wmi-other.rst
>>> +F:   drivers/platform/x86/lenovo-wmi-capdata01.c
>>> +F:   drivers/platform/x86/lenovo-wmi-capdata01.h
>>>    F:  drivers/platform/x86/lenovo-wmi-events.c
>>>    F:  drivers/platform/x86/lenovo-wmi-events.h
>>>    F:  drivers/platform/x86/lenovo-wmi-helpers.c
>>> diff --git a/drivers/platform/x86/Kconfig b/drivers/platform/x86/Kconfig
>>> index 13b8f4ac5dc5..64663667f0cb 100644
>>> --- a/drivers/platform/x86/Kconfig
>>> +++ b/drivers/platform/x86/Kconfig
>>> @@ -467,6 +467,10 @@ config LENOVO_WMI_HELPERS
>>>        tristate
>>>        depends on ACPI_WMI
>>>
>>> +config LENOVO_WMI_DATA01
>>> +     tristate
>>> +     depends on ACPI_WMI
>>> +
>>>    config IDEAPAD_LAPTOP
>>>        tristate "Lenovo IdeaPad Laptop Extras"
>>>        depends on ACPI
>>> diff --git a/drivers/platform/x86/Makefile b/drivers/platform/x86/Makefile
>>> index fc039839286a..7a35c77221b7 100644
>>> --- a/drivers/platform/x86/Makefile
>>> +++ b/drivers/platform/x86/Makefile
>>> @@ -69,6 +69,7 @@ obj-$(CONFIG_THINKPAD_LMI)  += think-lmi.o
>>>    obj-$(CONFIG_YOGABOOK)              += lenovo-yogabook.o
>>>    obj-$(CONFIG_YT2_1380)              += lenovo-yoga-tab2-pro-1380-fastcharger.o
>>>    obj-$(CONFIG_LENOVO_WMI_CAMERA)     += lenovo-wmi-camera.o
>>> +obj-$(CONFIG_LENOVO_WMI_DATA01)      += lenovo-wmi-capdata01.o
>>>    obj-$(CONFIG_LENOVO_WMI_EVENTS)     += lenovo-wmi-events.o
>>>    obj-$(CONFIG_LENOVO_WMI_HELPERS)    += lenovo-wmi-helpers.o
>>>
>>> diff --git a/drivers/platform/x86/lenovo-wmi-capdata01.c b/drivers/platform/x86/lenovo-wmi-capdata01.c
>>> new file mode 100644
>>> index 000000000000..b6876611ffd9
>>> --- /dev/null
>>> +++ b/drivers/platform/x86/lenovo-wmi-capdata01.c
>>> @@ -0,0 +1,136 @@
>>> +// SPDX-License-Identifier: GPL-2.0-or-later
>>> +/*
>>> + * LENOVO_CAPABILITY_DATA_01 WMI data block driver. This interface provides
>>> + * information on tunable attributes used by the "Other Mode" WMI interface,
>>> + * including if it is supported by the hardware, the default_value, max_value,
>>> + * min_value, and step increment.
>>> + *
>>> + * Copyright(C) 2025 Derek J. Clark <derekjohn.clark@gmail.com>
>>> + */
>>> +
>>> +#include <linux/cleanup.h>
>>> +#include <linux/component.h>
>>> +#include <linux/container_of.h>
>>> +#include <linux/device.h>
>>> +#include <linux/gfp_types.h>
>>> +#include <linux/types.h>
>>> +#include <linux/wmi.h>
>>> +#include "lenovo-wmi-capdata01.h"
>> Hi,
>>
>> please also include linux/acpi.h, linux/export.h and linux/module.h.
>>
>>> +
>>> +/* Interface GUIDs */
>>> +#define LENOVO_CAPABILITY_DATA_01_GUID "7A8F5407-CB67-4D6E-B547-39B3BE018154"
>>> +
>>> +struct lwmi_cd01_priv {
>>> +     struct wmi_device *wdev;
>>> +};
>>> +
>>> +/*
>> /* -> /**
>>
>>> + * lenovo_cd01_component_bind() - On master bind, caches all capability data on
>>> + * the master device.
>>> + * @cd01_dev: Pointer to the capability data 01 parent device.
>>> + * @om_dev: Pointer to the other mode parent device.
>>> + * @data: capdata01_list object pointer to return the capability data with.
>>> + *
>>> + * Returns: 0, or an error.
>>> + */
>>> +static int lenovo_cd01_component_bind(struct device *cd01_dev,
>>> +                                   struct device *om_dev, void *data)
>>> +{
>>> +     struct lwmi_cd01_priv *priv = dev_get_drvdata(cd01_dev);
>>> +     int count, idx;
>>> +
>>> +     if (!priv)
>>> +             return -ENODEV;
>> This check is unnecessary, please drop.
>>
> Acked
>
>>> +
>>> +     count = wmidev_instance_count(priv->wdev);
>>> +
>>> +     if (count == 0)
>>> +             return -EINVAL;
>> The WMI driver core already ensures that WMI devices with 0 instances are
>> rejected. Please drop this check.
>>
> Good to know, thanks.
>
>>> +
>>> +     ((struct cd01_list *)data)->count = count;
>>> +     ((struct cd01_list *)data)->data = devm_kmalloc_array(om_dev, count,
>>> +                                                           sizeof(struct capdata01 *),
>>> +                                                           GFP_KERNEL);
>> Two things:
>>
>>    - using a local variable with a type of struct cd01_list * results in cleaner source code here
>>
>>    - using devres is not possible inside the component callbacks, since the lifetime of the component
>>      device is not necessarily tied to the lifetime of the underlying device.
>>
>> I suggest you move the whole WMI data querying into lwmi_cd01_probe(), because then you can keep using
>> devres.
>>
> Doing this in probe() puts the list on lwmi_cd01_priv. Should I copy
> that data on bind, or pass back a pointer to the cd01 device struct
> and use an exported function on cd01 and dev_get_drvdata to access
> priv->list->data[idx] when needed? I prefer the latter as this avoids
> needing to do devm memory allocation in component/master binds, then I
> can check for NULL when accessing and clear the pointer on
> master_unbind to avoid calling to a removed device driver.

Passing a pointer is OK, but please only pass a pointer to the struct cd01_list itself,
not the full cd01 device struct. Maybe you can also add a comment explaining that this
pointer will become invalid when unbinding from the component.

Explicitly NULL-ing this pointer upon unbinding seems unnecessary to me.

> (snip)
>
>>> +struct cd01_list {
>>> +     struct capdata01 **data;
>>> +     int count;
>>> +};
>> In order to save memory you could try something like this:
>>
>> struct cd01_list {
>>          size_t count;
>>          struct capdata01 data[];
>> };
>>
>> This way you
>>
>> 1. Avoid the memory fragmentation resulting from multiple memory allocations.
>>
>> 2. Omit two pointers when accessing the data.
>>
>> You can use struct_size() from linux/overflow.h to calculate the size of such
>> an array with a trailing flexible array.
>>
>> Thanks,
>> Armin Wolf
>>
> I think I have this part working in my branch. Using devm_kzalloc also
> allows me to omit manually setting NULL in a few cases, which is
> cleaner. Is it preferred to use struct_size() directly in the
> devm_kzalloc call, or create a separate `size` variable to set the
> result to and pass that into the function?
>
> - Derek
>
In most cases using struct_size() directly inside the devm_kzalloc() call is preferred.

Thanks,
Armin Wolf

>>> +
>>> +int lwmi_cd01_match(struct device *dev, void *data);
>>> +
>>> +#endif /* !_LENOVO_WMI_CAPDATA01_H_ */