Re: [PATCH v6 0/3] Support wakeup methods of Atmel maXTouch controllers

[Dmitry Osipenko <digetx@gmail.com>]

22.03.2021 01:44, Dmitry Torokhov пишет:
> Hi Dmitry,
> 
> On Sat, Mar 20, 2021 at 07:02:43PM +0300, Dmitry Osipenko wrote:
>> 02.03.2021 13:21, Dmitry Osipenko пишет:
>>> Some Atmel maXTouch controllers, like mXT1386 and mXT3432S1 for example,
>>> have a WAKE line that needs to be asserted in order to wake controller
>>> from a deep sleep, otherwise it will be unusable. This series implements
>>> support for the wakeup methods in accordance to the mXT1386 datasheet [1],
>>> see page 29 (chapter "5.8 WAKE Line").
>>>
>>> The mXT1386 is a widely used controller found on many older Android tablet
>>> devices. Touchscreen on Acer A500 tablet now works properly after this
>>> series.
>>>
>>> This patchset is a continuation of the work originally started by
>>> Jiada Wang [2].
>>>
>>> [1] https://ww1.microchip.com/downloads/en/DeviceDoc/mXT1386_1vx_Datasheet_LX.pdf
>>> [2] https://patchwork.kernel.org/project/linux-input/list/?series=357875
>>
>> Hi,
>>
>> This series is very wanted by Android tablet devices from Acer, Asus and
>> other vendors which use Maxtouch 1386 controller. Touchscreens don't
>> work without the wakeup support, i.e. without this series. The wakeup
>> support is implemented in accordance to the datasheet and touchscreens
>> are working excellent using these patches.
>>
>> Could you please take this series into v5.13?
>>
>> Or could you please let me know what exactly needs to be improved?
> 
> Sorry, I was still slightly unhappy that we still are not tracking the
> state of controller and opportunistically retrying failed I2C transfers,
> but as I am failing to find time to come up with another solution I have
> just applied your series.

Thank you! I don't have other solutions either, although /I think/
potentially it should be possible to differentiate the I2C error here.
On NVIDIA Tegra I see that I2C controller always gets a h/w NAK on TS
wake-up and it returns -EREMOTEIO in this case. IIRC, some other
non-NVIDIA I2C drivers always return -EIO on any error, so this method
isn't universal, but certainly it feels like there is some a room for
further improvements.