The helper function for adding a GPIO chip compiles in a lockdep
key for debugging, the same key is needed for nested chips as
well.

The macro construction is unreadable, replace this with two
static inlines instead.

The _gpiochip_irqchip_add prefixed function is not helpful,
rename it with gpiochip_irqchip_add_key() that tell us what the
function is actually doing.

Fixes: d245b3f9 ("gpio: simplify adding threaded interrupts")
Cc: Roger Quadros <rogerq@ti.com>
Reported-by: NClemens Gruber <clemens.gruber@pqgruber.com>
Reported-by: NRoger Quadros <rogerq@ti.com>
Reported-by: NGrygorii Strashko <grygorii.strashko@ti.com>
Tested-by: NClemens Gruber <clemens.gruber@pqgruber.com>
Tested-by: NGrygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

When removing a gpiochip that uses GPIO hogging (e.g. by unloading the
chip's DT overlay), a warning is printed:

    gpio gpiochip8: REMOVING GPIOCHIP WITH GPIOS STILL REQUESTED

This happens because gpiochip_free_hogs() is called after the gdev->chip
pointer is reset to NULL. Hence __gpiod_free() cannot determine the
chip in use, and cannot clear flags nor call the optional chip-specific
.free() callback.

Move the call to gpiochip_free_hogs() up to fix this.

Cc: stable@vger.kernel.org
Fixes: ff2b1359 ("gpio: make the gpiochip a real device")
Signed-off-by: NGeert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

The GPIO chardev is used for management tasks (allocating line and event
handles) and does neither support read() nor write() operations. Hence it
does not make much sense to allow seek operations.

Currently the chardev uses noop_llseek() for its seek implementation. This
function does not move the pointer and simply returns the current position
(always 0 for the GPIO chardev). noop_llseek() is primarily meant for
devices that can not support seek, but where there might be a user that
depends on the seek() operation succeeding. For newly added devices that
can not support seek operations it is recommended to use no_llseek(), which
will return an error. For more information see commit 6038f373
("llseek: automatically add .llseek fop").

Unfortunately this was overlooked when the GPIO chardev ABI was introduced.
But it is highly unlikely that since then userspace applications have
appeared that rely on being able to perform non-failing seek operations on
a GPIO chardev file descriptor. So it should be safe to change from
noop_llseel() to no_seek(). Also use nonseekable_open() in the chardev
open() callback to clear the FMODE_SEEK, FMODE_PREAD and FMODE_PWRITE flags
from the file. Neither of these should be set on a file that does not
support seek operations.

Cc: stable@vger.kernel.org
Fixes: 3c702e99 ("gpio: add a userspace chardev ABI for GPIOs")
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

This tries to simplify the use of CONFIG_GPIOLIB_IRQCHIP when
using threaded interrupts: add a new call
gpiochip_irqchip_add_nested() to indicate that we're dealing
with a nested rather than a chained irqchip, then create a
separate gpiochip_set_nested_irqchip() to mirror
the gpiochip_set_chained_irqchip() call to connect the
parent and child interrupts.

In the nested case gpiochip_set_nested_irqchip() does nothing
more than call irq_set_parent() on each valid child interrupt,
which has little semantic effect in the kernel, but this is
probably still formally correct.

Update all drivers using nested interrupts to use
gpiochip_irqchip_add_nested() so we can now see clearly
which these users are.

The DLN2 driver can drop its specific hack with
.irq_not_threaded as we now recognize whether a chip is
threaded or not from its use of gpiochip_irqchip_add_nested()
signature rather than from inspecting .can_sleep.

We rename the .irq_parent to .irq_chained_parent since this
parent IRQ is only really kept around for the chained
interrupt handlers.

Cc: Lars Poeschel <poeschel@lemonage.de>
Cc: Octavian Purdila <octavian.purdila@intel.com>
Cc: Daniel Baluta <daniel.baluta@intel.com>
Cc: Bin Gao <bin.gao@linux.intel.com>
Cc: Mika Westerberg <mika.westerberg@linux.intel.com>
Cc: Ajay Thomas <ajay.thomas.david.rajamanickam@intel.com>
Cc: Semen Protsenko <semen.protsenko@globallogic.com>
Cc: Alexander Stein <alexander.stein@systec-electronic.com>
Cc: Phil Reid <preid@electromag.com.au>
Cc: Bartosz Golaszewski <bgolaszewski@baylibre.com>
Cc: Patrice Chotard <patrice.chotard@st.com>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

When a GPIO line is marked as used for an interrupt, it is
helpful to set the label to "interrupt" so we know what is
going on when inspecting the lines.

If a GPIO is already properly named by gpiod_get*() we don't
need to do this. It only happens when a line is used from
the irqchip side of a GPIO driver without communicating
with the GPIO side, such as when gpiochip is used as interrupt
provider in the device tree.

If the line is still marked as used by "interrupt" when we
unmark it as used by an interrupt, also remove this label
from the descriptor.

Also shape up the code around unmarking IRQ lines.
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

I saw weird values != [0,1] being passed down to drivers
in their .set_direction_output() callbacks. Go over the
gpiolib and make sure to hammer it to [0,1] before hitting
the driver to avoid undesired side effects.
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

When locking a GPIO line as IRQ, we go to lengths to
double-check that the line is really set as input before
marking it as used for IRQ. This is not good on GPIO chips
that can sleep, because this function is called in IRQ-safe
context. Just skip this if it can't be checked quickly.

Currently this happens on sleeping expanders such as STMPE
or TC3589x:

BUG: scheduling while atomic: swapper/1/0x00000002
Modules linked in:
CPU: 0 PID: 1 Comm: swapper Not tainted 4.9.0-rc1+ #38
Hardware name: Nomadik STn8815
[<c000f2e0>] (unwind_backtrace) from [<c000d244>] (show_stack+0x10/0x14)
[<c000d244>] (show_stack) from [<c0037b78>] (__schedule_bug+0x54/0x80)
[<c0037b78>] (__schedule_bug) from [<c042df14>] (__schedule+0x3a0/0x460)
[<c042df14>] (__schedule) from [<c042e028>] (schedule+0x54/0xb8)
(...)

This patch fixes that problem and relies on the direction
read from the chip when it was added.

Cc: stable@vger.kernel.org
Fixes: 9c10280d ("gpio: flush direction status in gpiochip_lock_as_irq()")
Cc: Patrice Chotard <patrice.chotard@st.com>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

When allocating a new line handle or event a file is allocated that it is
associated to. The file is attached to a file descriptor of the current
process and the file descriptor is returned to userspace using
copy_to_user(). If this copy operation fails the line handle or event
allocation is aborted, all acquired resources are freed and an error is
returned.

But the file struct is not freed and left attached to the userspace
application and even though the file descriptor number was not copied it is
trivial to guess. If a userspace application performs a IOCTL on such a
left over file descriptor it will trigger a use-after-free and if the file
descriptor is closed (latest when the application exits) a double-free is
triggered.

anon_inode_getfd() performs 3 tasks, allocate a file struct, allocate a
file descriptor for the current process and install the file struct in the
file descriptor. As soon as the file struct is installed in the file
descriptor it is accessible by userspace (even if the IOCTL itself hasn't
completed yet), this means uninstalling the fd on the error path is not an
option, since userspace might already got a reference to the file.

Instead anon_inode_getfd() needs to be broken into its individual steps.
The allocation of the file struct and file descriptor is done first, then
the copy_to_user() is executed and only if it succeeds the file is
installed.

Since the file struct is reference counted it can not be just freed, but
its reference needs to be dropped, which will also call the release()
callback, which will free the state attached to the file. So in this case
the normal error cleanup path should not be taken.

Cc: stable@vger.kernel.org
Fixes: d932cd49 ("gpio: free handles in fringe cases")
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

The GPIO_GET_LINEEVENT_IOCTL currently ignores unknown or undefined
linehandle and lineevent flags. From a backwards and forwards compatibility
viewpoint it is highly desirable to reject unknown flags though.

On one hand an application that is using newer flags and is running on
an older kernel has no way to detect if the new flags were handled
correctly if they are silently discarded.

On the other hand an application that (accidentally) passes undefined flags
will run fine on an older kernel, but may break on a newer kernel when
these flags get defined.

Ensure that requests that have undefined flags set are rejected with an
error, rather than silently discarding the undefined flags.

Cc: stable@vger.kernel.org
Fixes: 61f922db ("gpio: userspace ABI for reading GPIO line events")
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

The GPIO_GET_LINEHANDLE_IOCTL currently ignores unknown or undefined
linehandle flags. From a backwards and forwards compatibility viewpoint it
is highly desirable to reject unknown flags though.

On one hand an application that is using newer flags and is running on
an older kernel has no way to detect if the new flags were handled
correctly if they are silently discarded.

On the other hand an application that (accidentally) passes undefined flags
will run fine on an older kernel, but may break on a newer kernel when
these flags get defined.

Ensure that requests that have undefined flags set are rejected with an
error, rather than silently discarding the undefined flags.

Cc: stable@vger.kernel.org
Fixes: d7c51b47 ("gpio: userspace ABI for reading/writing GPIO lines")
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

The GPIOHANDLE_GET_LINE_VALUES_IOCTL handler allocates a gpiohandle_data
struct on the stack and then passes it to copy_to_user(). But depending on
the number of requested line handles the struct is only partially
initialized.

This exposes the previous, potentially sensitive, stack content to the
issuing userspace application. To avoid this make sure that the struct is
fully initialized.

Cc: stable@vger.kernel.org
Fixes: d7c51b47 ("gpio: userspace ABI for reading/writing GPIO lines")
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

The line offset that is used as an index into the descs array is provided
by userspace and might go beyond the bounds of the array. If that happens
undefined behavior will occur.

Make sure that the offset is within the bounds of the desc array and reject
any requests that specify a value outside of it.

Cc: stable@vger.kernel.org
Fixes: 61f922db ("gpio: userspace ABI for reading GPIO line events")
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

The GPIOHANDLE_GET_LINE_VALUES_IOCTL handler allocates a gpiohandle_data
struct on the stack and then passes it to copy_to_user(). But only the
first element of the values array in the struct is set, which leaves the
struct partially initialized.

This exposes the previous, potentially sensitive, stack content to the
issuing userspace application. To avoid this make sure that the struct is
fully initialized.

Cc: stable@vger.kernel.org
Fixes: 61f922db ("gpio: userspace ABI for reading GPIO line events")
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

The line offset that is used as an index into the descs array is provided
by userspace and might go beyond the bounds of the array. If that happens
undefined behavior will occur.

Make sure that the offset is within the bounds of the desc array and reject
any requests that specify a value outside of it.

Cc: stable@vger.kernel.org
Fixes: d7c51b47 ("gpio: userspace ABI for reading/writing GPIO lines")
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

The GPIO_GET_CHIPINFO_IOCTL handler allocates a gpiochip_info struct on the
stack and then passes it to copy_to_user(). But depending on the length of
the GPIO chip name and label the struct is only partially initialized.

This exposes the previous, potentially sensitive, stack content to the
issuing userspace application. To avoid this make sure that the struct is
fully initialized.

Cc: stable@vger.kernel.org
Fixes: 521a2ad6 ("gpio: add userspace ABI for GPIO line information")
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

The current line offset validation is off by one. Depending on the data
stored behind the descs array this can either cause undefined behavior or
disclose arbitrary, potentially sensitive, memory to the issuing userspace
application.

Make sure that offset is within the bounds of the desc array.

Cc: stable@vger.kernel.org
Fixes: 521a2ad6 ("gpio: add userspace ABI for GPIO line information")
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

The generic GPIO library directly implement code for acpi_find_gpio()
which is only used with CONFIG_ACPI. This was probably done because
OF did the same thing, but I removed that so remove this too.

Rename the internal acpi_find_gpio() in gpiolib-acpi.c to
acpi_populate_gpio_lookup() which seems to be more appropriate anyway
so as to avoid a namespace clash with the same function.

Make the stub return -ENOENT rather than -ENOSYS (as that is for
syscalls!).

For some reason the sunxi pin control driver was including the private
gpiolib header, it works just fine without it so remove that oneliner.

Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: NMika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

The generic GPIO library directly implement code for of_find_gpio()
which is only used with CONFIG_OF and causes compilation problems
on archs that do not even have stubs for OF functions, especially
on UM that does not implement any IO remap functions.

Move the function to gpiolib-of.c, implement a static inline stub
in gpiolib.h returning PTR_ERR(-ENOENT) if CONFIG_OF_GPIO is not
set and be done with it.
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

When using GPIO irqchip helpers to setup irqchip for a gpiolib based
driver, it is not possible to select which GPIOs to add to the IRQ domain.
Instead it just adds all GPIOs which is not always desired. For example
there might be GPIOs that for some reason cannot generated normal
interrupts at all.

To support this we add a flag irq_need_valid_mask to struct gpio_chip. When
this flag is set the core allocates irq_valid_mask that holds one bit for
each GPIO the chip has. By default all bits are set but drivers can
manipulate this using set_bit() and clear_bit() accordingly.

Then when gpiochip_irqchip_add() is called, this mask is checked and all
GPIOs with bit is set are added to the IRQ domain created for the GPIO
chip.
Suggested-by: NLinus Walleij <linus.walleij@linaro.org>
Signed-off-by: NMika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

Follow DT and forbid default trigger if the GPIO irqchip device is
enumerated from ACPI. Triggering for these devices will be configured
automatically from ACPI interrupt resources provided by the BIOS.
Suggested-by: NLinus Walleij <linus.walleij@linaro.org>
Signed-off-by: NMika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

Using a default trigger is a bad idea if using DT to configure
interrupts, as the device's interrupt specifier will always contain
the trigger configuration.

Let's warn about that particular situation, and revert to not
having a default. Hopefully, the couple of drivers still using
this feature will quickly be fixed.
Signed-off-by: NMarc Zyngier <marc.zyngier@arm.com>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

The if...else... block after the loop can be dropped with
a slight refactoring.
Signed-off-by: NMasahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

Remove duplicated include.
Signed-off-by: NWei Yongjun <yongjun_wei@trendmicro.com.cn>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

This reverts commit 7e7c059c.

I was wrong about trying to do this, as it breaks the
orthogonality between gpiochips and irqchips.
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

Since commit dd34c37a ("gpio: of: Allow -gpio suffix for property
names") when requesting a GPIO from the devicetree gpiolib looks for
properties with both the '-gpio' and the '-gpios' suffix. This was
implemented by first searching for the property with the '-gpios' suffix
and if that yields an error try the '-gpio' suffix. This approach has the
issue that any error returned when looking for the '-gpios' suffix is
silently discarded.

Commit 06fc3b70 ("gpio: of: Fix handling for deferred probe for -gpio
suffix") partially addressed the issue by treating the EPROBE_DEFER error
as a special condition. This fixed the case when the property is specified,
but the GPIO provider is not ready yet. But there are other cases in which
of_get_named_gpiod_flags() returns an error even though the property is
specified, e.g. if the specification is incorrect.

of_find_gpio() should only try to look for the property with the '-gpio'
suffix if no property with the '-gpios' suffix was found. If the property
was not found of_get_named_gpiod_flags() will return -ENOENT, so update the
condition to abort and propagate the error to the caller in all other
cases.

This is important for gpiod_get_optinal() and friends to behave correctly
in case the specifier contains errors. Without this patch they'll return
NULL if the property uses the '-gpios' suffix and the specifier contains
errors, which falsely indicates to the caller that no GPIO was specified.
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

When registering a GPIO chip, drivers can override the device tree node
associated with the chip by setting the chip's ->of_node field. If set,
this field is supposed to take precedence over the ->parent->of_node
field, but the code doesn't actually do that.

Commit 762c2e46 ("gpio: of: remove of_gpiochip_and_xlate() and
struct gg_data") exposes this because it now no longer matches on the
GPIO chip's ->of_node field, but the GPIO device's ->of_node field that
is set using the procedure described above.
Signed-off-by: NThierry Reding <treding@nvidia.com>
Acked-by: NAlexandre Courbot <acourbot@nvidia.com>
Reviewed-by: NMasahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

If we fail when copying the ioctl() struct to userspace we still
need to clean up the cruft otherwise left behind or it will stay
around until the issuing process terminates the file handle.
Reported-by: NAlexander Stein <alexander.stein@systec-electronic.com>
Acked-by: NAlexander Stein <alexander.stein@systec-electronic.com>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

This reverts commit 923b93e4.

Make sure consumers do not overwrite gpio flags for pins that have
already been claimed.

While adding support for gpio drivers to refuse a request using
unsupported flags, the order of when the requested flag was checked and
the new flags were applied was reversed to that consumers could
overwrite flags for already requested gpios.

This not only affects device-tree setups where two drivers could request
the same gpio using conflicting configurations, but also allowed user
space to clear gpio flags for already claimed pins simply by attempting
to export them through the sysfs interface. By for example clearing the
FLAG_ACTIVE_LOW flag this way, user space could effectively change the
polarity of a signal.

Reverting this change obviously prevents gpio drivers from doing sanity
checks on the flags in their request callbacks. Fortunately only one
recently added driver (gpio-tps65218 in v4.6) appears to do this, and a
follow up patch could restore this functionality through a different
interface.

Cc: stable <stable@vger.kernel.org>	# 4.4
Signed-off-by: NJohan Hovold <johan@kernel.org>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

The generic IRQ helper library just checks if the IRQ line is
set as input before activating it for interrupts. As we
recently started to check things better with .get_dir() it
turns out that it's good to try to convince the line to become
an input before attempting to lock it as IRQ.
Reviewed-by: NBjörn Andersson <bjorn.andersson@linaro.org>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

Most functions that take a GPIO descriptor in need to check the
descriptor for IS_ERR(). We do this mostly in the VALIDATE_DESC()
macro except for the gpiod_to_irq() function which needs special
handling.

Cc: stable@vger.kernel.org
Reported-by: NGrygorii Strashko <grygorii.strashko@ti.com>
Reviewed-by: NGrygorii Strashko <grygorii.strashko@ti.com>
Acked-by: NAlexandre Courbot <acourbot@nvidia.com>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

commit 54d77198
("gpio: bail out silently on NULL descriptors")
doesn't work for gpiod_to_irq(): drivers assume that NULL
descriptors will give negative IRQ numbers in return.

It has been pointed out that returning 0 is NO_IRQ and that
drivers should be amended to treat this as an error, but that
is for the longer term: now let us repair the semantics.

Cc: Maxime Ripard <maxime.ripard@free-electrons.com>
Reported-by: NHans de Goede <hdegoede@redhat.com>
Tested-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

The lineevent_irq_thread is not exported, so make it static
to fix the following warning:

drivers/gpio/gpiolib.c:654:13: warning: symbol 'lineevent_irq_thread' was not declared. Should it be static?
Signed-off-by: NBen Dooks <ben.dooks@codethink.co.uk>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

When initializing the GPIO handles, we use the iterator (i)
to back off if something goes wrong. But since the iterator
is also used after we pass the loop, we must decrement by
one after exiting the loop so that we point at the last
element in the array.
Reported-by: NDan Carpenter <dan.carpenter@oracle.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Walter Harms <wharms@bfs.de>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

Most functions that take a GPIO descriptor in need to check the
descriptor for IS_ERR(). We do this mostly in the VALIDATE_DESC()
macro except for the gpiod_to_irq() function which needs special
handling.

Cc: stable@vger.kernel.org
Reported-by: NGrygorii Strashko <grygorii.strashko@ti.com>
Reviewed-by: NGrygorii Strashko <grygorii.strashko@ti.com>
Acked-by: NAlexandre Courbot <acourbot@nvidia.com>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

commit 54d77198
("gpio: bail out silently on NULL descriptors")
doesn't work for gpiod_to_irq(): drivers assume that NULL
descriptors will give negative IRQ numbers in return.

It has been pointed out that returning 0 is NO_IRQ and that
drivers should be amended to treat this as an error, but that
is for the longer term: now let us repair the semantics.

Cc: Maxime Ripard <maxime.ripard@free-electrons.com>
Reported-by: NHans de Goede <hdegoede@redhat.com>
Tested-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

gcc reports a theoretical case for returning uninitialized data in
the kfifo when a GPIO interrupt happens and neither
GPIOEVENT_REQUEST_RISING_EDGE nor GPIOEVENT_REQUEST_FALLING_EDGE
are set:

drivers/gpio/gpiolib.c: In function 'lineevent_irq_thread':
drivers/gpio/gpiolib.c:683:87: error: 'ge.id' may be used uninitialized in this function [-Werror=maybe-uninitialized]

This case should not happen, but to be on the safe side, let's
return from the irq handler without adding data to the FIFO
to ensure we can never leak stack data to user space.
Signed-off-by: NArnd Bergmann <arnd@arndb.de>
Fixes: 61f922db ("gpio: userspace ABI for reading GPIO line events")
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

This adds an ABI for listening to events on GPIO lines.
The mechanism returns an anonymous file handle to a request
to listen to a specific offset on a specific gpiochip.
To fetch the stream of events from the file handle, userspace
simply reads an event.

- Events can be requested with the same flags as ordinary
  handles, i.e. open drain or open source. An ioctl() call
  GPIO_GET_LINEEVENT_IOCTL is issued indicating the desired
  line.

- Events can be requested for falling edge events, rising
  edge events, or both.

- All events are timestamped using the kernel real time
  nanosecond timestamp (the same as is used by IIO).

- The supplied consumer label will appear in "lsgpio"
  listings of the lines, and in /proc/interrupts as the
  mechanism will request an interrupt from the gpio chip.

- Events are not supported on gpiochips that do not serve
  interrupts (no legal .to_irq() call). The event interrupt
  is threaded to avoid any realtime problems.

- It is possible to also directly read the current value
  of the registered GPIO line by issuing the same
  GPIOHANDLE_GET_LINE_VALUES_IOCTL as used by the
  line handles. Setting the value is not supported: we
  do not listen to events on output lines.

This ABI is strongly influenced by Industrial I/O and surpasses
the old sysfs ABI by providing proper precision timestamps,
making it possible to set flags like open drain, and put
consumer names on the GPIO lines.
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

This adds a userspace ABI for reading and writing GPIO lines.
The mechanism returns an anonymous file handle to a request
to read/write n offsets from a gpiochip. This file handle
in turn accepts two ioctl()s: one that reads and one that
writes values to the selected lines.

- Handles can be requested as input/output, active low,
  open drain, open source, however when you issue a request
  for n lines with GPIO_GET_LINEHANDLE_IOCTL, they must all
  have the same flags, i.e. all inputs or all outputs, all
  open drain etc. If a granular control of the flags for
  each line is desired, they need to be requested
  individually, not in a batch.

- The GPIOHANDLE_GET_LINE_VALUES_IOCTL read ioctl() can be
  issued also to output lines to verify that the hardware
  is in the expected state.

- It reads and writes up to GPIOHANDLES_MAX lines at once,
  utilizing the .set_multiple() call in the driver if
  possible, making the call efficient if several lines
  can be written with a single register update.

The limitation of GPIOHANDLES_MAX to 64 lines is done under
the assumption that we may expect hardware that can issue a
transaction updating 64 bits at an instant but unlikely
anything larger than that.

ChangeLog v2->v3:
- Use gpiod_get_value_cansleep() so we support also slowpath
  GPIO drivers.
- Fix up the UAPI docs kerneldoc.
- Allocate the anonymous fd last, so that the release
  function don't get called until that point of something
  fails. After this point, skip the errorpath.
ChangeLog v1->v2:
- Handle ioctl_compat() properly based on a similar patch
  to the other ioctl() handling code.
- Use _IOWR() as we pass pointers both in and out of the
  ioctl()
- Use kmalloc() and kfree() for the linehandled, do not
  try to be fancy with devm_* it doesn't work the way I
  thought.
- Fix const-correctness on the linehandle name field.
Acked-by: NMichael Welling <mwelling@ieee.org>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

gpiolib relies on the reference counters to clean up the gpio_device
structure.

Although the number of get/put is properly aligned on gpiolib.c
itself, it does not take into consideration how the referece counters
are affected by other external functions such as cdev_add and device_add.

Because of this, after the last call to put_device, the reference counter
has a value of +3, therefore never calling gpiodevice_release.

Due to the fact that some of the device  has already been cleaned on
gpiochip_remove, the library will end up OOPsing the kernel (e.g. a call
to of_gpiochip_find_and_xlate).

Cc: stable@vger.kernel.org
Signed-off-by: NRicardo Ribalda Delgado <ricardo.ribalda@gmail.com>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>

Under some circumstances, a gpiochip might be half cleaned from the
gpio_device list.

This patch makes sure that the chip pointer is still valid, before
calling the match function.

[  104.088296] BUG: unable to handle kernel NULL pointer dereference at
0000000000000090
[  104.089772] IP: [<ffffffff813d2045>] of_gpiochip_find_and_xlate+0x15/0x80
[  104.128273] Call Trace:
[  104.129802]  [<ffffffff813d2030>] ? of_parse_own_gpio+0x1f0/0x1f0
[  104.131353]  [<ffffffff813cd910>] gpiochip_find+0x60/0x90
[  104.132868]  [<ffffffff813d21ba>] of_get_named_gpiod_flags+0x9a/0x120
...
[  104.141586]  [<ffffffff8163d12b>] gpio_led_probe+0x11b/0x360

Cc: stable@vger.kernel.org
Signed-off-by: NRicardo Ribalda Delgado <ricardo.ribalda@gmail.com>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>