devm_gpiod_put_array() is meant to undo the effects of
devm_gpiod_get_array() - in particular, it should release the GPIOs
contained in the array acquired with the latter. It is meant to be the
resource-managed version of gpiod_put_array(), and it should behave
similar to the non-array version devm_gpiod_put().

Since commit d1d52c6622a6 ("gpiolib: devres: Finish the conversion to
use devm_add_action()") it doesn't do that anymore, it just removes the
devres action and frees associated memory, but it doesn't actually
release the GPIOs.

Fix by switching from devm_remove_action() to devm_release_action(),
which will in addition invoke the action to release the GPIOs.

Fixes: d1d52c6622a6 ("gpiolib: devres: Finish the conversion to use devm_add_action()")
Signed-off-by: André Draszik <andre.draszik@linaro.org>
Reported-by: Wattson CI <wattson-external@google.com>
Reported-by: Samuel Wu <wusamuel@google.com>
Link: https://lore.kernel.org/r/20250715-gpiolib-devres-put-array-fix-v1-1-970d82a8c887@linaro.org
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>

Pull gpio updates from Bartosz Golaszewski:
 "We have three new drivers, some refactoring in the GPIO core, lots of
  various changes across many drivers, new configfs interface for the
  virtual gpio-aggregator module and DT-bindings updates.

  The treewide conversion of GPIO drivers to using the new value setter
  callbacks is ongoing with another round of GPIO drivers updated. You
  will also see these commits coming in from other subsystems as with
  the relevant changes merged into mainline last cycle, I've started
  converting GPIO providers located elsewhere than drivers/gpio/.

  GPIO core:
   - use more lock guards where applicable
   - refactor GPIO ACPI code and shrink it in the process by 8%
   - move GPIO ACPI quirks into a separate file
   - remove unneeded #ifdef
   - convert GPIO devres helpers to using devm_add_action() where
     applicable which shrinks and simplifies the code
   - refactor GPIO descriptor validation in GPIO consumer interfaces
   - don't allow setting values on input lines in the GPIO core which
     will take off the burden from GPIO drivers of checking this down
     the line
   - provide gpiod_is_equal() as a way of safely comparing two GPIO
     descriptors (the only current user is in regulator core)

  New drivers:
   - add the GPIO module for the max77759 multifunction device
   - add the GPIO driver for the VeriSilicon BLZP1600 GPIO controller
   - add the GPIO driver for the Spacemit K1 SoC

  Driver improvements:
   - convert more drivers to using the new GPIO line value setter
     callbacks
   - convert more drivers to making the irq_chip immutable as is
     recommended by the interrupt subsystem
   - extend build testing coverage by enabling more modules to be built
     with COMPILE_TEST=y
   - extend the gpio-aggregator module with a configfs interface that
     makes the setup easier for user-space than the existing
     driver-level sysfs attributes and also adds more advanced
     configuration features (such as referring to aggregated lines by
     their original names or modifying their names as exposed by the
     aggregated chip)
   - add a missing mutex_destroy() in gpio-imx-scu
   - add an OF polarity quirk for s5m8767
   - allow building gpio-vf610 as a loadable module
   - make gpio-mxc not hardcode its GPIO base number with GPIO SYSFS
     interface disabled (another small step towards getting rid of the
     global GPIO numberspace)
   - add support for level-triggered interrupts to gpio-pca953x
   - don't double-check the ngpios property in gpio-ds4520 as GPIO core
     already does it
   - don't double-check the number of GPIOs in gpio-imx-scu as GPIO core
     already does it
   - remove unused callbacks from gpio-max3191x

  DT bindings:
   - add device-tree bindings for max77759, spacemit,k1 and blzp1600
     (new drivers added this cycle)
   - document more properties for gpio-vf610 and gpio-tegra186
   - document a new pca95xx variant
   - fix style of examples in several GPIO DT-binding documents

  Misc:
   - TODO list updates"

* tag 'gpio-updates-for-v6.16-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/brgl/linux: (123 commits)
  gpio: timberdale: select GPIOLIB_IRQCHIP
  gpio: lpc18xx: select GPIOLIB_IRQCHIP
  gpio: grgpio: select GPIOLIB_IRQCHIP
  gpio: bcm-kona: select GPIOLIB_IRQCHIP
  dt-bindings: gpio: vf610: add ngpios and gpio-reserved-ranges
  gpio: davinci: select GPIOLIB_IRQCHIP
  gpiolib-acpi: Update file references in the Documentation and MAINTAINERS
  gpiolib: acpi: Move quirks to a separate file
  gpiolib: acpi: Add acpi_gpio_need_run_edge_events_on_boot() getter
  gpiolib: acpi: Handle deferred list via new API
  gpiolib: acpi: Make sure we fill struct acpi_gpio_info
  gpiolib: acpi: Switch to use enum in acpi_gpio_in_ignore_list()
  gpiolib: acpi: Use temporary variable for struct acpi_gpio_info
  gpiolib: remove unneeded #ifdef
  gpio: mpc8xxx: select GPIOLIB_IRQCHIP
  gpio: pxa: select GPIOLIB_IRQCHIP
  gpio: pxa: Make irq_chip immutable
  gpio: timberdale: Make irq_chip immutable
  gpio: xgene-sb: Make irq_chip immutable
  gpio: davinci: Make irq_chip immutable
  ...

With a recently added helper we can complete the conversion of
the GPIOLIB code to use devm_add_action() in all suitable cases.
So do this.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Link: https://lore.kernel.org/r/20250220162238.2738038-4-andriy.shevchenko@linux.intel.com
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>

This function was introduced as a workaround for an issue with resource
ownership in the regulator subsystem. Rather than passing the ownership
of a GPIO, we should make the regulator core be able to deal with
resources it didn't request. Deprecate this function so that we don't
get more users in the tree.

Link: https://lore.kernel.org/r/20250401-gpio-todo-remove-nonexclusive-v2-2-7c1380797b0d@linaro.org
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>

There is no reason to keep deprecated legacy API implementations
in the gpiolib-devres.c. Consolidate devm_gpio_*() with other legacy
APIs. While at it, clean up header inclusion block in gpiolib-devres.c.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Link: https://lore.kernel.org/r/20240828151357.2677340-1-andriy.shevchenko@linux.intel.com
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>

$ scripts/kernel-doc -v -none -Wall drivers/gpio/gpiolib* 2>&1 | grep -w warning | wc -l
67

Fix these by adding Return sections. While at it, make sure all of
Return sections use the same style.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Randy Dunlap <rdunlap@infradead.org>
Tested-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Link: https://lore.kernel.org/r/20240828164449.2777666-1-andriy.shevchenko@linux.intel.com
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>

This devm API takes a consumer device as an argument to setup the devm
action, but throws it away when calling further into gpiolib. This leads
to odd debug messages like this:

 (NULL device *): using DT '/gpio-keys/switch-pen-insert' for '(null)' GPIO lookup

Let's pass the consumer device down, by directly calling what
fwnode_gpiod_get_index() calls but pass the device used for devm. This
changes the message to look like this instead:

 gpio-keys gpio-keys: using DT '/gpio-keys/switch-pen-insert' for '(null)' GPIO lookup

Note that callers of fwnode_gpiod_get_index() will still see the NULL
device pointer debug message, but there's not much we can do about that
because the API doesn't take a struct device.

Cc: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Fixes: 8eb1f71e7acc ("gpiolib: consolidate GPIO lookups")
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>

Now that everyone is using [devm_]fwnode_gpiod_get[_index]() APIs,
remove OF-specific [devm_]gpiod_get_from_of_node().

Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>

The last user, which in fact was a dead code, has gone a year ago,
previous one 3 years ago. On top of that we want to drop away the
legacy GPIO APIs in the kernel, so take a chance to get rid of
unused devm_gpio_free() and accompanying stuff.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Bartosz Golaszewski <brgl@bgdev.pl>

Several gpiolib functions receive pointer to struct device_node which is
later passed to OF functions.  These OF functions accept already pointer
to const, so gpiolib can follow similar approach to indicate they are
not modifying the struct device_node.

Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@canonical.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com>