1GPIO Interfaces
   3This provides an overview of GPIO access conventions on Linux.
   5These calls use the gpio_* naming prefix.  No other calls should use that
   6prefix, or the related __gpio_* prefix.
   9What is a GPIO?
  11A "General Purpose Input/Output" (GPIO) is a flexible software-controlled
  12digital signal.  They are provided from many kinds of chip, and are familiar
  13to Linux developers working with embedded and custom hardware.  Each GPIO
  14represents a bit connected to a particular pin, or "ball" on Ball Grid Array
  15(BGA) packages.  Board schematics show which external hardware connects to
  16which GPIOs.  Drivers can be written generically, so that board setup code
  17passes such pin configuration data to drivers.
  19System-on-Chip (SOC) processors heavily rely on GPIOs.  In some cases, every
  20non-dedicated pin can be configured as a GPIO; and most chips have at least
  21several dozen of them.  Programmable logic devices (like FPGAs) can easily
  22provide GPIOs; multifunction chips like power managers, and audio codecs
  23often have a few such pins to help with pin scarcity on SOCs; and there are
  24also "GPIO Expander" chips that connect using the I2C or SPI serial busses.
  25Most PC southbridges have a few dozen GPIO-capable pins (with only the BIOS
  26firmware knowing how they're used).
  28The exact capabilities of GPIOs vary between systems.  Common options:
  30  - Output values are writable (high=1, low=0).  Some chips also have
  31    options about how that value is driven, so that for example only one
  32    value might be driven ... supporting "wire-OR" and similar schemes
  33    for the other value (notably, "open drain" signaling).
  35  - Input values are likewise readable (1, 0).  Some chips support readback
  36    of pins configured as "output", which is very useful in such "wire-OR"
  37    cases (to support bidirectional signaling).  GPIO controllers may have
  38    input de-glitch/debounce logic, sometimes with software controls.
  40  - Inputs can often be used as IRQ signals, often edge triggered but
  41    sometimes level triggered.  Such IRQs may be configurable as system
  42    wakeup events, to wake the system from a low power state.
  44  - Usually a GPIO will be configurable as either input or output, as needed
  45    by different product boards; single direction ones exist too.
  47  - Most GPIOs can be accessed while holding spinlocks, but those accessed
  48    through a serial bus normally can't.  Some systems support both types.
  50On a given board each GPIO is used for one specific purpose like monitoring
  51MMC/SD card insertion/removal, detecting card writeprotect status, driving
  52a LED, configuring a transceiver, bitbanging a serial bus, poking a hardware
  53watchdog, sensing a switch, and so on.
  56GPIO conventions
  58Note that this is called a "convention" because you don't need to do it this
  59way, and it's no crime if you don't.  There **are** cases where portability
  60is not the main issue; GPIOs are often used for the kind of board-specific
  61glue logic that may even change between board revisions, and can't ever be
  62used on a board that's wired differently.  Only least-common-denominator
  63functionality can be very portable.  Other features are platform-specific,
  64and that can be critical for glue logic.
  66Plus, this doesn't require any implementation framework, just an interface.
  67One platform might implement it as simple inline functions accessing chip
  68registers; another might implement it by delegating through abstractions
  69used for several very different kinds of GPIO controller.  (There is some
  70optional code supporting such an implementation strategy, described later
  71in this document, but drivers acting as clients to the GPIO interface must
  72not care how it's implemented.)
  74That said, if the convention is supported on their platform, drivers should
  75use it when possible.  Platforms must declare GENERIC_GPIO support in their
  76Kconfig (boolean true), and provide an <asm/gpio.h> file.  Drivers that can't
  77work without standard GPIO calls should have Kconfig entries which depend
  78on GENERIC_GPIO.  The GPIO calls are available, either as "real code" or as
  79optimized-away stubs, when drivers use the include file:
  81        #include <linux/gpio.h>
  83If you stick to this convention then it'll be easier for other developers to
  84see what your code is doing, and help maintain it.
  86Note that these operations include I/O barriers on platforms which need to
  87use them; drivers don't need to add them explicitly.
  90Identifying GPIOs
  92GPIOs are identified by unsigned integers in the range 0..MAX_INT.  That
  93reserves "negative" numbers for other purposes like marking signals as
  94"not available on this board", or indicating faults.  Code that doesn't
  95touch the underlying hardware treats these integers as opaque cookies.
  97Platforms define how they use those integers, and usually #define symbols
  98for the GPIO lines so that board-specific setup code directly corresponds
  99to the relevant schematics.  In contrast, drivers should only use GPIO
 100numbers passed to them from that setup code, using platform_data to hold
 101board-specific pin configuration data (along with other board specific
 102data they need).  That avoids portability problems.
 104So for example one platform uses numbers 32-159 for GPIOs; while another
 105uses numbers 0..63 with one set of GPIO controllers, 64-79 with another
 106type of GPIO controller, and on one particular board 80-95 with an FPGA.
 107The numbers need not be contiguous; either of those platforms could also
 108use numbers 2000-2063 to identify GPIOs in a bank of I2C GPIO expanders.
 110If you want to initialize a structure with an invalid GPIO number, use
 111some negative number (perhaps "-EINVAL"); that will never be valid.  To
 112test if such number from such a structure could reference a GPIO, you
 113may use this predicate:
 115        int gpio_is_valid(int number);
 117A number that's not valid will be rejected by calls which may request
 118or free GPIOs (see below).  Other numbers may also be rejected; for
 119example, a number might be valid but temporarily unused on a given board.
 121Whether a platform supports multiple GPIO controllers is a platform-specific
 122implementation issue, as are whether that support can leave "holes" in the space
 123of GPIO numbers, and whether new controllers can be added at runtime.  Such issues
 124can affect things including whether adjacent GPIO numbers are both valid.
 126Using GPIOs
 128The first thing a system should do with a GPIO is allocate it, using
 129the gpio_request() call; see later.
 131One of the next things to do with a GPIO, often in board setup code when
 132setting up a platform_device using the GPIO, is mark its direction:
 134        /* set as input or output, returning 0 or negative errno */
 135        int gpio_direction_input(unsigned gpio);
 136        int gpio_direction_output(unsigned gpio, int value);
 138The return value is zero for success, else a negative errno.  It should
 139be checked, since the get/set calls don't have error returns and since
 140misconfiguration is possible.  You should normally issue these calls from
 141a task context.  However, for spinlock-safe GPIOs it's OK to use them
 142before tasking is enabled, as part of early board setup.
 144For output GPIOs, the value provided becomes the initial output value.
 145This helps avoid signal glitching during system startup.
 147For compatibility with legacy interfaces to GPIOs, setting the direction
 148of a GPIO implicitly requests that GPIO (see below) if it has not been
 149requested already.  That compatibility is being removed from the optional
 150gpiolib framework.
 152Setting the direction can fail if the GPIO number is invalid, or when
 153that particular GPIO can't be used in that mode.  It's generally a bad
 154idea to rely on boot firmware to have set the direction correctly, since
 155it probably wasn't validated to do more than boot Linux.  (Similarly,
 156that board setup code probably needs to multiplex that pin as a GPIO,
 157and configure pullups/pulldowns appropriately.)
 160Spinlock-Safe GPIO access
 162Most GPIO controllers can be accessed with memory read/write instructions.
 163Those don't need to sleep, and can safely be done from inside hard
 164(nonthreaded) IRQ handlers and similar contexts.
 166Use the following calls to access such GPIOs,
 167for which gpio_cansleep() will always return false (see below):
 169        /* GPIO INPUT:  return zero or nonzero */
 170        int gpio_get_value(unsigned gpio);
 172        /* GPIO OUTPUT */
 173        void gpio_set_value(unsigned gpio, int value);
 175The values are boolean, zero for low, nonzero for high.  When reading the
 176value of an output pin, the value returned should be what's seen on the
 177pin ... that won't always match the specified output value, because of
 178issues including open-drain signaling and output latencies.
 180The get/set calls have no error returns because "invalid GPIO" should have
 181been reported earlier from gpio_direction_*().  However, note that not all
 182platforms can read the value of output pins; those that can't should always
 183return zero.  Also, using these calls for GPIOs that can't safely be accessed
 184without sleeping (see below) is an error.
 186Platform-specific implementations are encouraged to optimize the two
 187calls to access the GPIO value in cases where the GPIO number (and for
 188output, value) are constant.  It's normal for them to need only a couple
 189of instructions in such cases (reading or writing a hardware register),
 190and not to need spinlocks.  Such optimized calls can make bitbanging
 191applications a lot more efficient (in both space and time) than spending
 192dozens of instructions on subroutine calls.
 195GPIO access that may sleep
 197Some GPIO controllers must be accessed using message based busses like I2C
 198or SPI.  Commands to read or write those GPIO values require waiting to
 199get to the head of a queue to transmit a command and get its response.
 200This requires sleeping, which can't be done from inside IRQ handlers.
 202Platforms that support this type of GPIO distinguish them from other GPIOs
 203by returning nonzero from this call (which requires a valid GPIO number,
 204which should have been previously allocated with gpio_request):
 206        int gpio_cansleep(unsigned gpio);
 208To access such GPIOs, a different set of accessors is defined:
 210        /* GPIO INPUT:  return zero or nonzero, might sleep */
 211        int gpio_get_value_cansleep(unsigned gpio);
 213        /* GPIO OUTPUT, might sleep */
 214        void gpio_set_value_cansleep(unsigned gpio, int value);
 217Accessing such GPIOs requires a context which may sleep,  for example
 218a threaded IRQ handler, and those accessors must be used instead of
 219spinlock-safe accessors without the cansleep() name suffix.
 221Other than the fact that these accessors might sleep, and will work
 222on GPIOs that can't be accessed from hardIRQ handlers, these calls act
 223the same as the spinlock-safe calls.
 225  ** IN ADDITION ** calls to setup and configure such GPIOs must be made
 226from contexts which may sleep, since they may need to access the GPIO
 227controller chip too:  (These setup calls are usually made from board
 228setup or driver probe/teardown code, so this is an easy constraint.)
 230        gpio_direction_input()
 231        gpio_direction_output()
 232        gpio_request()
 234##      gpio_request_one()
 235##      gpio_request_array()
 236##      gpio_free_array()
 238        gpio_free()
 239        gpio_set_debounce()
 243Claiming and Releasing GPIOs
 245To help catch system configuration errors, two calls are defined.
 247        /* request GPIO, returning 0 or negative errno.
 248         * non-null labels may be useful for diagnostics.
 249         */
 250        int gpio_request(unsigned gpio, const char *label);
 252        /* release previously-claimed GPIO */
 253        void gpio_free(unsigned gpio);
 255Passing invalid GPIO numbers to gpio_request() will fail, as will requesting
 256GPIOs that have already been claimed with that call.  The return value of
 257gpio_request() must be checked.  You should normally issue these calls from
 258a task context.  However, for spinlock-safe GPIOs it's OK to request GPIOs
 259before tasking is enabled, as part of early board setup.
 261These calls serve two basic purposes.  One is marking the signals which
 262are actually in use as GPIOs, for better diagnostics; systems may have
 263several hundred potential GPIOs, but often only a dozen are used on any
 264given board.  Another is to catch conflicts, identifying errors when
 265(a) two or more drivers wrongly think they have exclusive use of that
 266signal, or (b) something wrongly believes it's safe to remove drivers
 267needed to manage a signal that's in active use.  That is, requesting a
 268GPIO can serve as a kind of lock.
 270Some platforms may also use knowledge about what GPIOs are active for
 271power management, such as by powering down unused chip sectors and, more
 272easily, gating off unused clocks.
 274For GPIOs that use pins known to the pinctrl subsystem, that subsystem should
 275be informed of their use; a gpiolib driver's .request() operation may call
 276pinctrl_request_gpio(), and a gpiolib driver's .free() operation may call
 277pinctrl_free_gpio(). The pinctrl subsystem allows a pinctrl_request_gpio()
 278to succeed concurrently with a pin or pingroup being "owned" by a device for
 279pin multiplexing.
 281Any programming of pin multiplexing hardware that is needed to route the
 282GPIO signal to the appropriate pin should occur within a GPIO driver's
 283.direction_input() or .direction_output() operations, and occur after any
 284setup of an output GPIO's value. This allows a glitch-free migration from a
 285pin's special function to GPIO. This is sometimes required when using a GPIO
 286to implement a workaround on signals typically driven by a non-GPIO HW block.
 288Some platforms allow some or all GPIO signals to be routed to different pins.
 289Similarly, other aspects of the GPIO or pin may need to be configured, such as
 290pullup/pulldown. Platform software should arrange that any such details are
 291configured prior to gpio_request() being called for those GPIOs, e.g. using
 292the pinctrl subsystem's mapping table, so that GPIO users need not be aware
 293of these details.
 295Also note that it's your responsibility to have stopped using a GPIO
 296before you free it.
 298Considering in most cases GPIOs are actually configured right after they
 299are claimed, three additional calls are defined:
 301        /* request a single GPIO, with initial configuration specified by
 302         * 'flags', identical to gpio_request() wrt other arguments and
 303         * return value
 304         */
 305        int gpio_request_one(unsigned gpio, unsigned long flags, const char *label);
 307        /* request multiple GPIOs in a single call
 308         */
 309        int gpio_request_array(struct gpio *array, size_t num);
 311        /* release multiple GPIOs in a single call
 312         */
 313        void gpio_free_array(struct gpio *array, size_t num);
 315where 'flags' is currently defined to specify the following properties:
 317        * GPIOF_DIR_IN          - to configure direction as input
 318        * GPIOF_DIR_OUT         - to configure direction as output
 320        * GPIOF_INIT_LOW        - as output, set initial level to LOW
 321        * GPIOF_INIT_HIGH       - as output, set initial level to HIGH
 322        * GPIOF_OPEN_DRAIN      - gpio pin is open drain type.
 323        * GPIOF_OPEN_SOURCE     - gpio pin is open source type.
 325        * GPIOF_EXPORT_DIR_FIXED        - export gpio to sysfs, keep direction
 326        * GPIOF_EXPORT_DIR_CHANGEABLE   - also export, allow changing direction
 328since GPIOF_INIT_* are only valid when configured as output, so group valid
 329combinations as:
 331        * GPIOF_IN              - configure as input
 332        * GPIOF_OUT_INIT_LOW    - configured as output, initial level LOW
 333        * GPIOF_OUT_INIT_HIGH   - configured as output, initial level HIGH
 335When setting the flag as GPIOF_OPEN_DRAIN then it will assume that pins is
 336open drain type. Such pins will not be driven to 1 in output mode. It is
 337require to connect pull-up on such pins. By enabling this flag, gpio lib will
 338make the direction to input when it is asked to set value of 1 in output mode
 339to make the pin HIGH. The pin is make to LOW by driving value 0 in output mode.
 341When setting the flag as GPIOF_OPEN_SOURCE then it will assume that pins is
 342open source type. Such pins will not be driven to 0 in output mode. It is
 343require to connect pull-down on such pin. By enabling this flag, gpio lib will
 344make the direction to input when it is asked to set value of 0 in output mode
 345to make the pin LOW. The pin is make to HIGH by driving value 1 in output mode.
 347In the future, these flags can be extended to support more properties.
 349Further more, to ease the claim/release of multiple GPIOs, 'struct gpio' is
 350introduced to encapsulate all three fields as:
 352        struct gpio {
 353                unsigned        gpio;
 354                unsigned long   flags;
 355                const char      *label;
 356        };
 358A typical example of usage:
 360        static struct gpio leds_gpios[] = {
 361                { 32, GPIOF_OUT_INIT_HIGH, "Power LED" }, /* default to ON */
 362                { 33, GPIOF_OUT_INIT_LOW,  "Green LED" }, /* default to OFF */
 363                { 34, GPIOF_OUT_INIT_LOW,  "Red LED"   }, /* default to OFF */
 364                { 35, GPIOF_OUT_INIT_LOW,  "Blue LED"  }, /* default to OFF */
 365                { ... },
 366        };
 368        err = gpio_request_one(31, GPIOF_IN, "Reset Button");
 369        if (err)
 370                ...
 372        err = gpio_request_array(leds_gpios, ARRAY_SIZE(leds_gpios));
 373        if (err)
 374                ...
 376        gpio_free_array(leds_gpios, ARRAY_SIZE(leds_gpios));
 379GPIOs mapped to IRQs
 381GPIO numbers are unsigned integers; so are IRQ numbers.  These make up
 382two logically distinct namespaces (GPIO 0 need not use IRQ 0).  You can
 383map between them using calls like:
 385        /* map GPIO numbers to IRQ numbers */
 386        int gpio_to_irq(unsigned gpio);
 388        /* map IRQ numbers to GPIO numbers (avoid using this) */
 389        int irq_to_gpio(unsigned irq);
 391Those return either the corresponding number in the other namespace, or
 392else a negative errno code if the mapping can't be done.  (For example,
 393some GPIOs can't be used as IRQs.)  It is an unchecked error to use a GPIO
 394number that wasn't set up as an input using gpio_direction_input(), or
 395to use an IRQ number that didn't originally come from gpio_to_irq().
 397These two mapping calls are expected to cost on the order of a single
 398addition or subtraction.  They're not allowed to sleep.
 400Non-error values returned from gpio_to_irq() can be passed to request_irq()
 401or free_irq().  They will often be stored into IRQ resources for platform
 402devices, by the board-specific initialization code.  Note that IRQ trigger
 403options are part of the IRQ interface, e.g. IRQF_TRIGGER_FALLING, as are
 404system wakeup capabilities.
 406Non-error values returned from irq_to_gpio() would most commonly be used
 407with gpio_get_value(), for example to initialize or update driver state
 408when the IRQ is edge-triggered.  Note that some platforms don't support
 409this reverse mapping, so you should avoid using it.
 412Emulating Open Drain Signals
 414Sometimes shared signals need to use "open drain" signaling, where only the
 415low signal level is actually driven.  (That term applies to CMOS transistors;
 416"open collector" is used for TTL.)  A pullup resistor causes the high signal
 417level.  This is sometimes called a "wire-AND"; or more practically, from the
 418negative logic (low=true) perspective this is a "wire-OR".
 420One common example of an open drain signal is a shared active-low IRQ line.
 421Also, bidirectional data bus signals sometimes use open drain signals.
 423Some GPIO controllers directly support open drain outputs; many don't.  When
 424you need open drain signaling but your hardware doesn't directly support it,
 425there's a common idiom you can use to emulate it with any GPIO pin that can
 426be used as either an input or an output:
 428 LOW:   gpio_direction_output(gpio, 0) ... this drives the signal
 429        and overrides the pullup.
 431 HIGH:  gpio_direction_input(gpio) ... this turns off the output,
 432        so the pullup (or some other device) controls the signal.
 434If you are "driving" the signal high but gpio_get_value(gpio) reports a low
 435value (after the appropriate rise time passes), you know some other component
 436is driving the shared signal low.  That's not necessarily an error.  As one
 437common example, that's how I2C clocks are stretched:  a slave that needs a
 438slower clock delays the rising edge of SCK, and the I2C master adjusts its
 439signaling rate accordingly.
 442GPIO controllers and the pinctrl subsystem
 445A GPIO controller on a SOC might be tightly coupled with the pinctrl
 446subsystem, in the sense that the pins can be used by other functions
 447together with an optional gpio feature. We have already covered the
 448case where e.g. a GPIO controller need to reserve a pin or set the
 449direction of a pin by calling any of:
 456But how does the pin control subsystem cross-correlate the GPIO
 457numbers (which are a global business) to a certain pin on a certain
 458pin controller?
 460This is done by registering "ranges" of pins, which are essentially
 461cross-reference tables. These are described in
 464While the pin allocation is totally managed by the pinctrl subsystem,
 465gpio (under gpiolib) is still maintained by gpio drivers. It may happen
 466that different pin ranges in a SoC is managed by different gpio drivers.
 468This makes it logical to let gpio drivers announce their pin ranges to
 469the pin ctrl subsystem before it will call 'pinctrl_request_gpio' in order
 470to request the corresponding pin to be prepared by the pinctrl subsystem
 471before any gpio usage.
 473For this, the gpio controller can register its pin range with pinctrl
 474subsystem. There are two ways of doing it currently: with or without DT.
 476For with DT support refer to Documentation/devicetree/bindings/gpio/gpio.txt.
 478For non-DT support, user can call gpiochip_add_pin_range() with appropriate
 479parameters to register a range of gpio pins with a pinctrl driver. For this
 480exact name string of pinctrl device has to be passed as one of the
 481argument to this routine.
 484What do these conventions omit?
 486One of the biggest things these conventions omit is pin multiplexing, since
 487this is highly chip-specific and nonportable.  One platform might not need
 488explicit multiplexing; another might have just two options for use of any
 489given pin; another might have eight options per pin; another might be able
 490to route a given GPIO to any one of several pins.  (Yes, those examples all
 491come from systems that run Linux today.)
 493Related to multiplexing is configuration and enabling of the pullups or
 494pulldowns integrated on some platforms.  Not all platforms support them,
 495or support them in the same way; and any given board might use external
 496pullups (or pulldowns) so that the on-chip ones should not be used.
 497(When a circuit needs 5 kOhm, on-chip 100 kOhm resistors won't do.)
 498Likewise drive strength (2 mA vs 20 mA) and voltage (1.8V vs 3.3V) is a
 499platform-specific issue, as are models like (not) having a one-to-one
 500correspondence between configurable pins and GPIOs.
 502There are other system-specific mechanisms that are not specified here,
 503like the aforementioned options for input de-glitching and wire-OR output.
 504Hardware may support reading or writing GPIOs in gangs, but that's usually
 505configuration dependent:  for GPIOs sharing the same bank.  (GPIOs are
 506commonly grouped in banks of 16 or 32, with a given SOC having several such
 507banks.)  Some systems can trigger IRQs from output GPIOs, or read values
 508from pins not managed as GPIOs.  Code relying on such mechanisms will
 509necessarily be nonportable.
 511Dynamic definition of GPIOs is not currently standard; for example, as
 512a side effect of configuring an add-on board with some GPIO expanders.
 515GPIO implementor's framework (OPTIONAL)
 517As noted earlier, there is an optional implementation framework making it
 518easier for platforms to support different kinds of GPIO controller using
 519the same programming interface.  This framework is called "gpiolib".
 521As a debugging aid, if debugfs is available a /sys/kernel/debug/gpio file
 522will be found there.  That will list all the controllers registered through
 523this framework, and the state of the GPIOs currently in use.
 526Controller Drivers: gpio_chip
 528In this framework each GPIO controller is packaged as a "struct gpio_chip"
 529with information common to each controller of that type:
 531 - methods to establish GPIO direction
 532 - methods used to access GPIO values
 533 - flag saying whether calls to its methods may sleep
 534 - optional debugfs dump method (showing extra state like pullup config)
 535 - label for diagnostics
 537There is also per-instance data, which may come from device.platform_data:
 538the number of its first GPIO, and how many GPIOs it exposes.
 540The code implementing a gpio_chip should support multiple instances of the
 541controller, possibly using the driver model.  That code will configure each
 542gpio_chip and issue gpiochip_add().  Removing a GPIO controller should be
 543rare; use gpiochip_remove() when it is unavoidable.
 545Most often a gpio_chip is part of an instance-specific structure with state
 546not exposed by the GPIO interfaces, such as addressing, power management,
 547and more.  Chips such as codecs will have complex non-GPIO state.
 549Any debugfs dump method should normally ignore signals which haven't been
 550requested as GPIOs.  They can use gpiochip_is_requested(), which returns
 551either NULL or the label associated with that GPIO when it was requested.
 554Platform Support
 556To support this framework, a platform's Kconfig will "select" either
 558and arrange that its <asm/gpio.h> includes <asm-generic/gpio.h> and defines
 559three functions: gpio_get_value(), gpio_set_value(), and gpio_cansleep().
 561It may also provide a custom value for ARCH_NR_GPIOS, so that it better
 562reflects the number of GPIOs in actual use on that platform, without
 563wasting static table space.  (It should count both built-in/SoC GPIOs and
 564also ones on GPIO expanders.
 566ARCH_REQUIRE_GPIOLIB means that the gpiolib code will always get compiled
 567into the kernel on that architecture.
 569ARCH_WANT_OPTIONAL_GPIOLIB means the gpiolib code defaults to off and the user
 570can enable it and build it into the kernel optionally.
 572If neither of these options are selected, the platform does not support
 573GPIOs through GPIO-lib and the code cannot be enabled by the user.
 575Trivial implementations of those functions can directly use framework
 576code, which always dispatches through the gpio_chip:
 578  #define gpio_get_value        __gpio_get_value
 579  #define gpio_set_value        __gpio_set_value
 580  #define gpio_cansleep         __gpio_cansleep
 582Fancier implementations could instead define those as inline functions with
 583logic optimizing access to specific SOC-based GPIOs.  For example, if the
 584referenced GPIO is the constant "12", getting or setting its value could
 585cost as little as two or three instructions, never sleeping.  When such an
 586optimization is not possible those calls must delegate to the framework
 587code, costing at least a few dozen instructions.  For bitbanged I/O, such
 588instruction savings can be significant.
 590For SOCs, platform-specific code defines and registers gpio_chip instances
 591for each bank of on-chip GPIOs.  Those GPIOs should be numbered/labeled to
 592match chip vendor documentation, and directly match board schematics.  They
 593may well start at zero and go up to a platform-specific limit.  Such GPIOs
 594are normally integrated into platform initialization to make them always be
 595available, from arch_initcall() or earlier; they can often serve as IRQs.
 598Board Support
 600For external GPIO controllers -- such as I2C or SPI expanders, ASICs, multi
 601function devices, FPGAs or CPLDs -- most often board-specific code handles
 602registering controller devices and ensures that their drivers know what GPIO
 603numbers to use with gpiochip_add().  Their numbers often start right after
 604platform-specific GPIOs.
 606For example, board setup code could create structures identifying the range
 607of GPIOs that chip will expose, and passes them to each GPIO expander chip
 608using platform_data.  Then the chip driver's probe() routine could pass that
 609data to gpiochip_add().
 611Initialization order can be important.  For example, when a device relies on
 612an I2C-based GPIO, its probe() routine should only be called after that GPIO
 613becomes available.  That may mean the device should not be registered until
 614calls for that GPIO can work.  One way to address such dependencies is for
 615such gpio_chip controllers to provide setup() and teardown() callbacks to
 616board specific code; those board specific callbacks would register devices
 617once all the necessary resources are available, and remove them later when
 618the GPIO controller device becomes unavailable.
 621Sysfs Interface for Userspace (OPTIONAL)
 623Platforms which use the "gpiolib" implementors framework may choose to
 624configure a sysfs user interface to GPIOs.  This is different from the
 625debugfs interface, since it provides control over GPIO direction and
 626value instead of just showing a gpio state summary.  Plus, it could be
 627present on production systems without debugging support.
 629Given appropriate hardware documentation for the system, userspace could
 630know for example that GPIO #23 controls the write protect line used to
 631protect boot loader segments in flash memory.  System upgrade procedures
 632may need to temporarily remove that protection, first importing a GPIO,
 633then changing its output state, then updating the code before re-enabling
 634the write protection.  In normal use, GPIO #23 would never be touched,
 635and the kernel would have no need to know about it.
 637Again depending on appropriate hardware documentation, on some systems
 638userspace GPIO can be used to determine system configuration data that
 639standard kernels won't know about.  And for some tasks, simple userspace
 640GPIO drivers could be all that the system really needs.
 642Note that standard kernel drivers exist for common "LEDs and Buttons"
 643GPIO tasks:  "leds-gpio" and "gpio_keys", respectively.  Use those
 644instead of talking directly to the GPIOs; they integrate with kernel
 645frameworks better than your userspace code could.
 648Paths in Sysfs
 650There are three kinds of entry in /sys/class/gpio:
 652   -    Control interfaces used to get userspace control over GPIOs;
 654   -    GPIOs themselves; and
 656   -    GPIO controllers ("gpio_chip" instances).
 658That's in addition to standard files including the "device" symlink.
 660The control interfaces are write-only:
 662    /sys/class/gpio/
 664        "export" ... Userspace may ask the kernel to export control of
 665                a GPIO to userspace by writing its number to this file.
 667                Example:  "echo 19 > export" will create a "gpio19" node
 668                for GPIO #19, if that's not requested by kernel code.
 670        "unexport" ... Reverses the effect of exporting to userspace.
 672                Example:  "echo 19 > unexport" will remove a "gpio19"
 673                node exported using the "export" file.
 675GPIO signals have paths like /sys/class/gpio/gpio42/ (for GPIO #42)
 676and have the following read/write attributes:
 678    /sys/class/gpio/gpioN/
 680        "direction" ... reads as either "in" or "out".  This value may
 681                normally be written.  Writing as "out" defaults to
 682                initializing the value as low.  To ensure glitch free
 683                operation, values "low" and "high" may be written to
 684                configure the GPIO as an output with that initial value.
 686                Note that this attribute *will not exist* if the kernel
 687                doesn't support changing the direction of a GPIO, or
 688                it was exported by kernel code that didn't explicitly
 689                allow userspace to reconfigure this GPIO's direction.
 691        "value" ... reads as either 0 (low) or 1 (high).  If the GPIO
 692                is configured as an output, this value may be written;
 693                any nonzero value is treated as high.
 695                If the pin can be configured as interrupt-generating interrupt
 696                and if it has been configured to generate interrupts (see the
 697                description of "edge"), you can poll(2) on that file and
 698                poll(2) will return whenever the interrupt was triggered. If
 699                you use poll(2), set the events POLLPRI and POLLERR. If you
 700                use select(2), set the file descriptor in exceptfds. After
 701                poll(2) returns, either lseek(2) to the beginning of the sysfs
 702                file and read the new value or close the file and re-open it
 703                to read the value.
 705        "edge" ... reads as either "none", "rising", "falling", or
 706                "both". Write these strings to select the signal edge(s)
 707                that will make poll(2) on the "value" file return.
 709                This file exists only if the pin can be configured as an
 710                interrupt generating input pin.
 712        "active_low" ... reads as either 0 (false) or 1 (true).  Write
 713                any nonzero value to invert the value attribute both
 714                for reading and writing.  Existing and subsequent
 715                poll(2) support configuration via the edge attribute
 716                for "rising" and "falling" edges will follow this
 717                setting.
 719GPIO controllers have paths like /sys/class/gpio/gpiochip42/ (for the
 720controller implementing GPIOs starting at #42) and have the following
 721read-only attributes:
 723    /sys/class/gpio/gpiochipN/
 725        "base" ... same as N, the first GPIO managed by this chip
 727        "label" ... provided for diagnostics (not always unique)
 729        "ngpio" ... how many GPIOs this manges (N to N + ngpio - 1)
 731Board documentation should in most cases cover what GPIOs are used for
 732what purposes.  However, those numbers are not always stable; GPIOs on
 733a daughtercard might be different depending on the base board being used,
 734or other cards in the stack.  In such cases, you may need to use the
 735gpiochip nodes (possibly in conjunction with schematics) to determine
 736the correct GPIO number to use for a given signal.
 739Exporting from Kernel code
 741Kernel code can explicitly manage exports of GPIOs which have already been
 742requested using gpio_request():
 744        /* export the GPIO to userspace */
 745        int gpio_export(unsigned gpio, bool direction_may_change);
 747        /* reverse gpio_export() */
 748        void gpio_unexport();
 750        /* create a sysfs link to an exported GPIO node */
 751        int gpio_export_link(struct device *dev, const char *name,
 752                unsigned gpio)
 754        /* change the polarity of a GPIO node in sysfs */
 755        int gpio_sysfs_set_active_low(unsigned gpio, int value);
 757After a kernel driver requests a GPIO, it may only be made available in
 758the sysfs interface by gpio_export().  The driver can control whether the
 759signal direction may change.  This helps drivers prevent userspace code
 760from accidentally clobbering important system state.
 762This explicit exporting can help with debugging (by making some kinds
 763of experiments easier), or can provide an always-there interface that's
 764suitable for documenting as part of a board support package.
 766After the GPIO has been exported, gpio_export_link() allows creating
 767symlinks from elsewhere in sysfs to the GPIO sysfs node.  Drivers can
 768use this to provide the interface under their own device in sysfs with
 769a descriptive name.
 771Drivers can use gpio_sysfs_set_active_low() to hide GPIO line polarity
 772differences between boards from user space.  This only affects the
 773sysfs interface.  Polarity change can be done both before and after
 774gpio_export(), and previously enabled poll(2) support for either
 775rising or falling edge will be reconfigured to follow this setting.