Re: [PATCH v2] pinctrl: qcom: support gpio_chip .set_config call

[Bjorn Andersson <bjorn.andersson@xxxxxxxxxx>]

On Wed 10 Mar 17:03 CST 2021, Linus Walleij wrote:

> On Wed, Mar 3, 2021 at 2:19 PM Shawn Guo <shawn.guo@xxxxxxxxxx> wrote:
> 
> > In case of ACPI boot, GPIO core does the right thing to parse GPIO pin
> > configs from ACPI table, and call into gpio_chip's .set_config hook for
> > setting them up.  It enables such support on qcom platform by using
> > generic config function, which in turn calls into .pin_config_set of
> > pinconf for setting up hardware.  For qcom platform, it's possible to
> > reuse pin group config functions for pin config hooks, because every pin
> > is maintained as a single group.
> >
> > This change fixes the problem that Touchpad of Lenovo Flex 5G laptop
> > doesn't work with ACPI boot, because PullUp config of Touchpad GpioInt
> > pin is not set up by the kernel.
> >
> > Signed-off-by: Shawn Guo <shawn.guo@xxxxxxxxxx>
> > ---
> > Changes for v2:
> > - Add pin config functions that simply call into group config ones.
> 
> Patch applied!
> 

As discussed in [1]; several key Qualcomm platforms have a gpio-ranges
representing the number of real GPIOs, but we then expose the UFS reset
GPO (no input) pin as a GPIO as well - making the two numbers differ.

I've moved ahead and merged the update to gpio-ranges, to make the two
match, but Shawn reports that with the introduction of .set_config() all
existing DTBs fail to probe storage because of the UFS code getting
EPROBE_DEFER back on its gpiod_get_optional().

I don't know how to make the transition, but can you please revert this
patch, to avoid breaking compatibility with DTBs out there?

[1] https://lore.kernel.org/linux-arm-msm/20210311230924.GX17424@dragon/#t

Regards,
Bjorn