Re: [PATCH v4 1/2] platform/x86: asus-wmi: add TUF keyboard RGB readback support

From: Denis Benato

Date: Mon Nov 16 2026 - 21:24:35 EST



On 11/17/26 01:07, bartu alev wrote:
> On 11/17/26 2:53 AM, Denis Benato wrote:
>
>> Yes, it's absolutely reasonable. If the return either doesn't have the
>> presence bit set or is 0xFFF..... it is to be counted as "not implemented".
>>
>> And also yes, we should respect ABIs whenever possible...
>>
>> However... now that I think about it asus_wmi_evaluate_method_buf() can
>> be used on setters and those don't have the presence bit, do we have that
>> covered?
> yes that wont break setters:
>
> 1. asus_wmi_evaluate_method_buf() is strictly a buffer-getter helper (its
>    comment states it returns an error whenever the ACPI output is not a
>    buffer).  Integer-returning DEVS setters use
> asus_wmi_evaluate_method*() /
>    asus_wmi_set_devstate() instead; even today, passing an integer `1`
>    from DEVS through _buf() returns 1 as an error.

Yeah I checked: only one caller. Make it check for known -ENODEV:
0xff...fe.

The only reference (as far as I can tell) to ~0 being "absent" is this:
absent=(l==0xffffffff);
inside asus_rfkill_hotplug, maybe you can exclude checking for that,
albeit it's not going to harm.

Perhaps add proper kernel-doc to that function where possible
errors are documented as well as input parameters.

> 2. We wouldn't check ASUS_WMI_DSTS_PRESENCE_BIT in _buf() at all.  A
>    buffer-returning call (getter or setter) still hits `case
>    ACPI_TYPE_BUFFER` and returns 0 unchanged.  Only `case ACPI_TYPE_INTEGER`
>    changes to decode via `u32 retval`:
>    - 0xfffffffe or (u32)~0 (0xffffffff) -> -ENODEV
>    - 0 or any other integer               -> -ENODATA (matching `default:`)

Sorry I am a bit spoiled by the past work of Marco where he reorganized
the whole thing.

We should really revive that, but it's a long work that touches quite a bit of
very old code...

> Then only kbd_rgb_read_status() maps -ENODATA -> -ENODEV so FX505DT /
> 0-returning firmware matches the ABI without altering setter or legacy
> buffer semantics.

Good. Very good.