[linux-sunxi] Re: [PATCH] arm: sunxi: input: RFC: Add sysfs voltage for sun4i-lradc driver

[Hans de Goede]

Hi,

On 27-01-15 10:49, Priit Laes wrote:
>
> On Tue, 2015-01-27 at 10:18 +0100, Maxime Ripard wrote:
>> Hi,
>>
>> On Mon, Jan 26, 2015 at 06:58:32PM +0200, Priit Laes wrote:
>>> ---
>>
>> Like Hans was pointing out, commit log and signed-off-by please
>>
>>>   .../ABI/testing/sysfs-driver-input-sun4i-lradc     |  4 ++
>>>   drivers/input/keyboard/sun4i-lradc-keys.c          | 49
>>> +++++++++++++++++-----
>>>   2 files changed, 43 insertions(+), 10 deletions(-)
>>>   create mode 100644 Documentation/ABI/testing/sysfs-driver-input-
>>> sun4i-lradc
>>>
>>> diff --git a/Documentation/ABI/testing/sysfs-driver-input-sun4i-
>>> lradc b/Documentation/ABI/testing/sysfs-driver-input-sun4i-lradc
>>> new file mode 100644
>>> index 0000000..e4e6448
>>> --- /dev/null
>>> +++ b/Documentation/ABI/testing/sysfs-driver-input-sun4i-lradc
>>> @@ -0,0 +1,4 @@
>>> +What:          /sys/class/input/input(x)/device/voltage
>>> +Date:          February 2015
>>> +Contact:       Priit Laes <plaes at plaes.org>
>>> +Description:   ADC output voltage in microvolts or 0 if device is
>>> not opened.
>>
>> Why is it returning 0 when "device is not opened" ? What does that
>> even mean? You can't read that file without opening it.
>
> It means that something has to open the /dev/input/inputX device which
> sets up the ADC before the voltage can be read from the sysfs file.
>
> [...]
>
>
>>
>> As I told you already, if you're going to expose this an ADC in the
>> end, the proper solution is to use the IIO framework, not adding a
>> custom sysfs file.
>
> My intention was to expose just a simple debug output, so one can
> press the buttons and read the voltages for devicetree keymap.
>
> If anyone can suggest a simpler approach than current sysfs based one,
> I would do it.

The android driver always uses 0.2V / 200mV steps, so what I do is
simply create a mapping with 200mV mapped to KEY_VOLUMEUP, 400mV mapped
to KEY_VOLUMEDOWN, etc. following the hardcoded android driver mapping:

https://github.com/linux-sunxi/linux-sunxi/blob/sunxi-3.4/drivers/input/keyboard/sun4i-keyboard.c#L136

Usually this will be correct in one go, after testing one can shuffle
key codes as needed (usually not needed) and/or remove unused entries.

With that said I do think that a sysfs file to see the actual voltages,
or a kernel parameter to printk them on keypress interrupt would be useful.

I guess the printk option would be better as it would show the actual
keypress value read, not some semi-random sample.

Regards,

Hans