GPIO Interfaces

1This provides an overview of GPIO access conventions on Linux.
2
3These calls use the gpio_* naming prefix.  No other calls should use that
4prefix, or the related __gpio_* prefix.
5
6
7What is a GPIO?
8===============
9A "General Purpose Input/Output" (GPIO) is a flexible software-controlled
10digital signal.  They are provided from many kinds of chip, and are familiar
11to Linux developers working with embedded and custom hardware.  Each GPIO
12represents a bit connected to a particular pin, or "ball" on Ball Grid Array
13(BGA) packages.  Board schematics show which external hardware connects to
14which GPIOs.  Drivers can be written generically, so that board setup code
15passes such pin configuration data to drivers.
16
17System-on-Chip (SOC) processors heavily rely on GPIOs.  In some cases, every
18non-dedicated pin can be configured as a GPIO; and most chips have at least
19several dozen of them.  Programmable logic devices (like FPGAs) can easily
20provide GPIOs; multifunction chips like power managers, and audio codecs
21often have a few such pins to help with pin scarcity on SOCs; and there are
22also "GPIO Expander" chips that connect using the I2C or SPI serial busses.
23Most PC southbridges have a few dozen GPIO-capable pins (with only the BIOS
24firmware knowing how they're used).
25
26The exact capabilities of GPIOs vary between systems.  Common options:
27
28  - Output values are writable (high=1, low=0).  Some chips also have
29    options about how that value is driven, so that for example only one
30    value might be driven ... supporting "wire-OR" and similar schemes
31    for the other value (notably, "open drain" signaling).
32
33  - Input values are likewise readable (1, 0).  Some chips support readback
34    of pins configured as "output", which is very useful in such "wire-OR"
35    cases (to support bidirectional signaling).  GPIO controllers may have
36    input de-glitch/debounce logic, sometimes with software controls.
37
38  - Inputs can often be used as IRQ signals, often edge triggered but
39    sometimes level triggered.  Such IRQs may be configurable as system
40    wakeup events, to wake the system from a low power state.
41
42  - Usually a GPIO will be configurable as either input or output, as needed
43    by different product boards; single direction ones exist too.
44
45  - Most GPIOs can be accessed while holding spinlocks, but those accessed
46    through a serial bus normally can't.  Some systems support both types.
47
48On a given board each GPIO is used for one specific purpose like monitoring
49MMC/SD card insertion/removal, detecting card writeprotect status, driving
50a LED, configuring a transceiver, bitbanging a serial bus, poking a hardware
51watchdog, sensing a switch, and so on.
52
53
54GPIO conventions
55================
56Note that this is called a "convention" because you don't need to do it this
57way, and it's no crime if you don't.  There **are** cases where portability
58is not the main issue; GPIOs are often used for the kind of board-specific
59glue logic that may even change between board revisions, and can't ever be
60used on a board that's wired differently.  Only least-common-denominator
61functionality can be very portable.  Other features are platform-specific,
62and that can be critical for glue logic.
63
64Plus, this doesn't require any implementation framework, just an interface.
65One platform might implement it as simple inline functions accessing chip
66registers; another might implement it by delegating through abstractions
67used for several very different kinds of GPIO controller.  (There is some
68optional code supporting such an implementation strategy, described later
69in this document, but drivers acting as clients to the GPIO interface must
70not care how it's implemented.)
71
72That said, if the convention is supported on their platform, drivers should
73use it when possible.  Platforms must declare GENERIC_GPIO support in their
74Kconfig (boolean true), and provide an <asm/gpio.h> file.  Drivers that can't
75work without standard GPIO calls should have Kconfig entries which depend
76on GENERIC_GPIO.  The GPIO calls are available, either as "real code" or as
77optimized-away stubs, when drivers use the include file:
78
79	#include <linux/gpio.h>
80
81If you stick to this convention then it'll be easier for other developers to
82see what your code is doing, and help maintain it.
83
84Note that these operations include I/O barriers on platforms which need to
85use them; drivers don't need to add them explicitly.
86
87
88Identifying GPIOs
89-----------------
90GPIOs are identified by unsigned integers in the range 0..MAX_INT.  That
91reserves "negative" numbers for other purposes like marking signals as
92"not available on this board", or indicating faults.  Code that doesn't
93touch the underlying hardware treats these integers as opaque cookies.
94
95Platforms define how they use those integers, and usually #define symbols
96for the GPIO lines so that board-specific setup code directly corresponds
97to the relevant schematics.  In contrast, drivers should only use GPIO
98numbers passed to them from that setup code, using platform_data to hold
99board-specific pin configuration data (along with other board specific
100data they need).  That avoids portability problems.
101
102So for example one platform uses numbers 32-159 for GPIOs; while another
103uses numbers 0..63 with one set of GPIO controllers, 64-79 with another
104type of GPIO controller, and on one particular board 80-95 with an FPGA.
105The numbers need not be contiguous; either of those platforms could also
106use numbers 2000-2063 to identify GPIOs in a bank of I2C GPIO expanders.
107
108If you want to initialize a structure with an invalid GPIO number, use
109some negative number (perhaps "-EINVAL"); that will never be valid.  To
110test if such number from such a structure could reference a GPIO, you
111may use this predicate:
112
113	int gpio_is_valid(int number);
114
115A number that's not valid will be rejected by calls which may request
116or free GPIOs (see below).  Other numbers may also be rejected; for
117example, a number might be valid but temporarily unused on a given board.
118
119Whether a platform supports multiple GPIO controllers is a platform-specific
120implementation issue, as are whether that support can leave "holes" in the space
121of GPIO numbers, and whether new controllers can be added at runtime.  Such issues
122can affect things including whether adjacent GPIO numbers are both valid.
123
124Using GPIOs
125-----------
126The first thing a system should do with a GPIO is allocate it, using
127the gpio_request() call; see later.
128
129One of the next things to do with a GPIO, often in board setup code when
130setting up a platform_device using the GPIO, is mark its direction:
131
132	/* set as input or output, returning 0 or negative errno */
133	int gpio_direction_input(unsigned gpio);
134	int gpio_direction_output(unsigned gpio, int value);
135
136The return value is zero for success, else a negative errno.  It should
137be checked, since the get/set calls don't have error returns and since
138misconfiguration is possible.  You should normally issue these calls from
139a task context.  However, for spinlock-safe GPIOs it's OK to use them
140before tasking is enabled, as part of early board setup.
141
142For output GPIOs, the value provided becomes the initial output value.
143This helps avoid signal glitching during system startup.
144
145For compatibility with legacy interfaces to GPIOs, setting the direction
146of a GPIO implicitly requests that GPIO (see below) if it has not been
147requested already.  That compatibility is being removed from the optional
148gpiolib framework.
149
150Setting the direction can fail if the GPIO number is invalid, or when
151that particular GPIO can't be used in that mode.  It's generally a bad
152idea to rely on boot firmware to have set the direction correctly, since
153it probably wasn't validated to do more than boot Linux.  (Similarly,
154that board setup code probably needs to multiplex that pin as a GPIO,
155and configure pullups/pulldowns appropriately.)
156
157
158Spinlock-Safe GPIO access
159-------------------------
160Most GPIO controllers can be accessed with memory read/write instructions.
161Those don't need to sleep, and can safely be done from inside hard
162(nonthreaded) IRQ handlers and similar contexts.
163
164Use the following calls to access such GPIOs,
165for which gpio_cansleep() will always return false (see below):
166
167	/* GPIO INPUT:  return zero or nonzero */
168	int gpio_get_value(unsigned gpio);
169
170	/* GPIO OUTPUT */
171	void gpio_set_value(unsigned gpio, int value);
172
173The values are boolean, zero for low, nonzero for high.  When reading the
174value of an output pin, the value returned should be what's seen on the
175pin ... that won't always match the specified output value, because of
176issues including open-drain signaling and output latencies.
177
178The get/set calls have no error returns because "invalid GPIO" should have
179been reported earlier from gpio_direction_*().  However, note that not all
180platforms can read the value of output pins; those that can't should always
181return zero.  Also, using these calls for GPIOs that can't safely be accessed
182without sleeping (see below) is an error.
183
184Platform-specific implementations are encouraged to optimize the two
185calls to access the GPIO value in cases where the GPIO number (and for
186output, value) are constant.  It's normal for them to need only a couple
187of instructions in such cases (reading or writing a hardware register),
188and not to need spinlocks.  Such optimized calls can make bitbanging
189applications a lot more efficient (in both space and time) than spending
190dozens of instructions on subroutine calls.
191
192
193GPIO access that may sleep
194--------------------------
195Some GPIO controllers must be accessed using message based busses like I2C
196or SPI.  Commands to read or write those GPIO values require waiting to
197get to the head of a queue to transmit a command and get its response.
198This requires sleeping, which can't be done from inside IRQ handlers.
199
200Platforms that support this type of GPIO distinguish them from other GPIOs
201by returning nonzero from this call (which requires a valid GPIO number,
202which should have been previously allocated with gpio_request):
203
204	int gpio_cansleep(unsigned gpio);
205
206To access such GPIOs, a different set of accessors is defined:
207
208	/* GPIO INPUT:  return zero or nonzero, might sleep */
209	int gpio_get_value_cansleep(unsigned gpio);
210
211	/* GPIO OUTPUT, might sleep */
212	void gpio_set_value_cansleep(unsigned gpio, int value);
213
214
215Accessing such GPIOs requires a context which may sleep,  for example
216a threaded IRQ handler, and those accessors must be used instead of
217spinlock-safe accessors without the cansleep() name suffix.
218
219Other than the fact that these accessors might sleep, and will work
220on GPIOs that can't be accessed from hardIRQ handlers, these calls act
221the same as the spinlock-safe calls.
222
223  ** IN ADDITION ** calls to setup and configure such GPIOs must be made
224from contexts which may sleep, since they may need to access the GPIO
225controller chip too:  (These setup calls are usually made from board
226setup or driver probe/teardown code, so this is an easy constraint.)
227
228	gpio_direction_input()
229	gpio_direction_output()
230	gpio_request()
231
232## 	gpio_request_one()
233##	gpio_request_array()
234## 	gpio_free_array()
235
236	gpio_free()
237	gpio_set_debounce()
238
239
240
241Claiming and Releasing GPIOs
242----------------------------
243To help catch system configuration errors, two calls are defined.
244
245	/* request GPIO, returning 0 or negative errno.
246	 * non-null labels may be useful for diagnostics.
247	 */
248	int gpio_request(unsigned gpio, const char *label);
249
250	/* release previously-claimed GPIO */
251	void gpio_free(unsigned gpio);
252
253Passing invalid GPIO numbers to gpio_request() will fail, as will requesting
254GPIOs that have already been claimed with that call.  The return value of
255gpio_request() must be checked.  You should normally issue these calls from
256a task context.  However, for spinlock-safe GPIOs it's OK to request GPIOs
257before tasking is enabled, as part of early board setup.
258
259These calls serve two basic purposes.  One is marking the signals which
260are actually in use as GPIOs, for better diagnostics; systems may have
261several hundred potential GPIOs, but often only a dozen are used on any
262given board.  Another is to catch conflicts, identifying errors when
263(a) two or more drivers wrongly think they have exclusive use of that
264signal, or (b) something wrongly believes it's safe to remove drivers
265needed to manage a signal that's in active use.  That is, requesting a
266GPIO can serve as a kind of lock.
267
268Some platforms may also use knowledge about what GPIOs are active for
269power management, such as by powering down unused chip sectors and, more
270easily, gating off unused clocks.
271
272Note that requesting a GPIO does NOT cause it to be configured in any
273way; it just marks that GPIO as in use.  Separate code must handle any
274pin setup (e.g. controlling which pin the GPIO uses, pullup/pulldown).
275
276Also note that it's your responsibility to have stopped using a GPIO
277before you free it.
278
279Considering in most cases GPIOs are actually configured right after they
280are claimed, three additional calls are defined:
281
282	/* request a single GPIO, with initial configuration specified by
283	 * 'flags', identical to gpio_request() wrt other arguments and
284	 * return value
285	 */
286	int gpio_request_one(unsigned gpio, unsigned long flags, const char *label);
287
288	/* request multiple GPIOs in a single call
289	 */
290	int gpio_request_array(struct gpio *array, size_t num);
291
292	/* release multiple GPIOs in a single call
293	 */
294	void gpio_free_array(struct gpio *array, size_t num);
295
296where 'flags' is currently defined to specify the following properties:
297
298	* GPIOF_DIR_IN		- to configure direction as input
299	* GPIOF_DIR_OUT		- to configure direction as output
300
301	* GPIOF_INIT_LOW	- as output, set initial level to LOW
302	* GPIOF_INIT_HIGH	- as output, set initial level to HIGH
303
304since GPIOF_INIT_* are only valid when configured as output, so group valid
305combinations as:
306
307	* GPIOF_IN		- configure as input
308	* GPIOF_OUT_INIT_LOW	- configured as output, initial level LOW
309	* GPIOF_OUT_INIT_HIGH	- configured as output, initial level HIGH
310
311In the future, these flags can be extended to support more properties such
312as open-drain status.
313
314Further more, to ease the claim/release of multiple GPIOs, 'struct gpio' is
315introduced to encapsulate all three fields as:
316
317	struct gpio {
318		unsigned	gpio;
319		unsigned long	flags;
320		const char	*label;
321	};
322
323A typical example of usage:
324
325	static struct gpio leds_gpios[] = {
326		{ 32, GPIOF_OUT_INIT_HIGH, "Power LED" }, /* default to ON */
327		{ 33, GPIOF_OUT_INIT_LOW,  "Green LED" }, /* default to OFF */
328		{ 34, GPIOF_OUT_INIT_LOW,  "Red LED"   }, /* default to OFF */
329		{ 35, GPIOF_OUT_INIT_LOW,  "Blue LED"  }, /* default to OFF */
330		{ ... },
331	};
332
333	err = gpio_request_one(31, GPIOF_IN, "Reset Button");
334	if (err)
335		...
336
337	err = gpio_request_array(leds_gpios, ARRAY_SIZE(leds_gpios));
338	if (err)
339		...
340
341	gpio_free_array(leds_gpios, ARRAY_SIZE(leds_gpios));
342
343
344GPIOs mapped to IRQs
345--------------------
346GPIO numbers are unsigned integers; so are IRQ numbers.  These make up
347two logically distinct namespaces (GPIO 0 need not use IRQ 0).  You can
348map between them using calls like:
349
350	/* map GPIO numbers to IRQ numbers */
351	int gpio_to_irq(unsigned gpio);
352
353	/* map IRQ numbers to GPIO numbers (avoid using this) */
354	int irq_to_gpio(unsigned irq);
355
356Those return either the corresponding number in the other namespace, or
357else a negative errno code if the mapping can't be done.  (For example,
358some GPIOs can't be used as IRQs.)  It is an unchecked error to use a GPIO
359number that wasn't set up as an input using gpio_direction_input(), or
360to use an IRQ number that didn't originally come from gpio_to_irq().
361
362These two mapping calls are expected to cost on the order of a single
363addition or subtraction.  They're not allowed to sleep.
364
365Non-error values returned from gpio_to_irq() can be passed to request_irq()
366or free_irq().  They will often be stored into IRQ resources for platform
367devices, by the board-specific initialization code.  Note that IRQ trigger
368options are part of the IRQ interface, e.g. IRQF_TRIGGER_FALLING, as are
369system wakeup capabilities.
370
371Non-error values returned from irq_to_gpio() would most commonly be used
372with gpio_get_value(), for example to initialize or update driver state
373when the IRQ is edge-triggered.  Note that some platforms don't support
374this reverse mapping, so you should avoid using it.
375
376
377Emulating Open Drain Signals
378----------------------------
379Sometimes shared signals need to use "open drain" signaling, where only the
380low signal level is actually driven.  (That term applies to CMOS transistors;
381"open collector" is used for TTL.)  A pullup resistor causes the high signal
382level.  This is sometimes called a "wire-AND"; or more practically, from the
383negative logic (low=true) perspective this is a "wire-OR".
384
385One common example of an open drain signal is a shared active-low IRQ line.
386Also, bidirectional data bus signals sometimes use open drain signals.
387
388Some GPIO controllers directly support open drain outputs; many don't.  When
389you need open drain signaling but your hardware doesn't directly support it,
390there's a common idiom you can use to emulate it with any GPIO pin that can
391be used as either an input or an output:
392
393 LOW:	gpio_direction_output(gpio, 0) ... this drives the signal
394	and overrides the pullup.
395
396 HIGH:	gpio_direction_input(gpio) ... this turns off the output,
397	so the pullup (or some other device) controls the signal.
398
399If you are "driving" the signal high but gpio_get_value(gpio) reports a low
400value (after the appropriate rise time passes), you know some other component
401is driving the shared signal low.  That's not necessarily an error.  As one
402common example, that's how I2C clocks are stretched:  a slave that needs a
403slower clock delays the rising edge of SCK, and the I2C master adjusts its
404signaling rate accordingly.
405
406
407What do these conventions omit?
408===============================
409One of the biggest things these conventions omit is pin multiplexing, since
410this is highly chip-specific and nonportable.  One platform might not need
411explicit multiplexing; another might have just two options for use of any
412given pin; another might have eight options per pin; another might be able
413to route a given GPIO to any one of several pins.  (Yes, those examples all
414come from systems that run Linux today.)
415
416Related to multiplexing is configuration and enabling of the pullups or
417pulldowns integrated on some platforms.  Not all platforms support them,
418or support them in the same way; and any given board might use external
419pullups (or pulldowns) so that the on-chip ones should not be used.
420(When a circuit needs 5 kOhm, on-chip 100 kOhm resistors won't do.)
421Likewise drive strength (2 mA vs 20 mA) and voltage (1.8V vs 3.3V) is a
422platform-specific issue, as are models like (not) having a one-to-one
423correspondence between configurable pins and GPIOs.
424
425There are other system-specific mechanisms that are not specified here,
426like the aforementioned options for input de-glitching and wire-OR output.
427Hardware may support reading or writing GPIOs in gangs, but that's usually
428configuration dependent:  for GPIOs sharing the same bank.  (GPIOs are
429commonly grouped in banks of 16 or 32, with a given SOC having several such
430banks.)  Some systems can trigger IRQs from output GPIOs, or read values
431from pins not managed as GPIOs.  Code relying on such mechanisms will
432necessarily be nonportable.
433
434Dynamic definition of GPIOs is not currently standard; for example, as
435a side effect of configuring an add-on board with some GPIO expanders.
436
437
438GPIO implementor's framework (OPTIONAL)
439=======================================
440As noted earlier, there is an optional implementation framework making it
441easier for platforms to support different kinds of GPIO controller using
442the same programming interface.  This framework is called "gpiolib".
443
444As a debugging aid, if debugfs is available a /sys/kernel/debug/gpio file
445will be found there.  That will list all the controllers registered through
446this framework, and the state of the GPIOs currently in use.
447
448
449Controller Drivers: gpio_chip
450-----------------------------
451In this framework each GPIO controller is packaged as a "struct gpio_chip"
452with information common to each controller of that type:
453
454 - methods to establish GPIO direction
455 - methods used to access GPIO values
456 - flag saying whether calls to its methods may sleep
457 - optional debugfs dump method (showing extra state like pullup config)
458 - label for diagnostics
459
460There is also per-instance data, which may come from device.platform_data:
461the number of its first GPIO, and how many GPIOs it exposes.
462
463The code implementing a gpio_chip should support multiple instances of the
464controller, possibly using the driver model.  That code will configure each
465gpio_chip and issue gpiochip_add().  Removing a GPIO controller should be
466rare; use gpiochip_remove() when it is unavoidable.
467
468Most often a gpio_chip is part of an instance-specific structure with state
469not exposed by the GPIO interfaces, such as addressing, power management,
470and more.  Chips such as codecs will have complex non-GPIO state.
471
472Any debugfs dump method should normally ignore signals which haven't been
473requested as GPIOs.  They can use gpiochip_is_requested(), which returns
474either NULL or the label associated with that GPIO when it was requested.
475
476
477Platform Support
478----------------
479To support this framework, a platform's Kconfig will "select" either
480ARCH_REQUIRE_GPIOLIB or ARCH_WANT_OPTIONAL_GPIOLIB
481and arrange that its <asm/gpio.h> includes <asm-generic/gpio.h> and defines
482three functions: gpio_get_value(), gpio_set_value(), and gpio_cansleep().
483
484It may also provide a custom value for ARCH_NR_GPIOS, so that it better
485reflects the number of GPIOs in actual use on that platform, without
486wasting static table space.  (It should count both built-in/SoC GPIOs and
487also ones on GPIO expanders.
488
489ARCH_REQUIRE_GPIOLIB means that the gpiolib code will always get compiled
490into the kernel on that architecture.
491
492ARCH_WANT_OPTIONAL_GPIOLIB means the gpiolib code defaults to off and the user
493can enable it and build it into the kernel optionally.
494
495If neither of these options are selected, the platform does not support
496GPIOs through GPIO-lib and the code cannot be enabled by the user.
497
498Trivial implementations of those functions can directly use framework
499code, which always dispatches through the gpio_chip:
500
501  #define gpio_get_value	__gpio_get_value
502  #define gpio_set_value	__gpio_set_value
503  #define gpio_cansleep		__gpio_cansleep
504
505Fancier implementations could instead define those as inline functions with
506logic optimizing access to specific SOC-based GPIOs.  For example, if the
507referenced GPIO is the constant "12", getting or setting its value could
508cost as little as two or three instructions, never sleeping.  When such an
509optimization is not possible those calls must delegate to the framework
510code, costing at least a few dozen instructions.  For bitbanged I/O, such
511instruction savings can be significant.
512
513For SOCs, platform-specific code defines and registers gpio_chip instances
514for each bank of on-chip GPIOs.  Those GPIOs should be numbered/labeled to
515match chip vendor documentation, and directly match board schematics.  They
516may well start at zero and go up to a platform-specific limit.  Such GPIOs
517are normally integrated into platform initialization to make them always be
518available, from arch_initcall() or earlier; they can often serve as IRQs.
519
520
521Board Support
522-------------
523For external GPIO controllers -- such as I2C or SPI expanders, ASICs, multi
524function devices, FPGAs or CPLDs -- most often board-specific code handles
525registering controller devices and ensures that their drivers know what GPIO
526numbers to use with gpiochip_add().  Their numbers often start right after
527platform-specific GPIOs.
528
529For example, board setup code could create structures identifying the range
530of GPIOs that chip will expose, and passes them to each GPIO expander chip
531using platform_data.  Then the chip driver's probe() routine could pass that
532data to gpiochip_add().
533
534Initialization order can be important.  For example, when a device relies on
535an I2C-based GPIO, its probe() routine should only be called after that GPIO
536becomes available.  That may mean the device should not be registered until
537calls for that GPIO can work.  One way to address such dependencies is for
538such gpio_chip controllers to provide setup() and teardown() callbacks to
539board specific code; those board specific callbacks would register devices
540once all the necessary resources are available, and remove them later when
541the GPIO controller device becomes unavailable.
542
543
544Sysfs Interface for Userspace (OPTIONAL)
545========================================
546Platforms which use the "gpiolib" implementors framework may choose to
547configure a sysfs user interface to GPIOs.  This is different from the
548debugfs interface, since it provides control over GPIO direction and
549value instead of just showing a gpio state summary.  Plus, it could be
550present on production systems without debugging support.
551
552Given appropriate hardware documentation for the system, userspace could
553know for example that GPIO #23 controls the write protect line used to
554protect boot loader segments in flash memory.  System upgrade procedures
555may need to temporarily remove that protection, first importing a GPIO,
556then changing its output state, then updating the code before re-enabling
557the write protection.  In normal use, GPIO #23 would never be touched,
558and the kernel would have no need to know about it.
559
560Again depending on appropriate hardware documentation, on some systems
561userspace GPIO can be used to determine system configuration data that
562standard kernels won't know about.  And for some tasks, simple userspace
563GPIO drivers could be all that the system really needs.
564
565Note that standard kernel drivers exist for common "LEDs and Buttons"
566GPIO tasks:  "leds-gpio" and "gpio_keys", respectively.  Use those
567instead of talking directly to the GPIOs; they integrate with kernel
568frameworks better than your userspace code could.
569
570
571Paths in Sysfs
572--------------
573There are three kinds of entry in /sys/class/gpio:
574
575   -	Control interfaces used to get userspace control over GPIOs;
576
577   -	GPIOs themselves; and
578
579   -	GPIO controllers ("gpio_chip" instances).
580
581That's in addition to standard files including the "device" symlink.
582
583The control interfaces are write-only:
584
585    /sys/class/gpio/
586
587    	"export" ... Userspace may ask the kernel to export control of
588		a GPIO to userspace by writing its number to this file.
589
590		Example:  "echo 19 > export" will create a "gpio19" node
591		for GPIO #19, if that's not requested by kernel code.
592
593    	"unexport" ... Reverses the effect of exporting to userspace.
594
595		Example:  "echo 19 > unexport" will remove a "gpio19"
596		node exported using the "export" file.
597
598GPIO signals have paths like /sys/class/gpio/gpio42/ (for GPIO #42)
599and have the following read/write attributes:
600
601    /sys/class/gpio/gpioN/
602
603	"direction" ... reads as either "in" or "out".  This value may
604		normally be written.  Writing as "out" defaults to
605		initializing the value as low.  To ensure glitch free
606		operation, values "low" and "high" may be written to
607		configure the GPIO as an output with that initial value.
608
609		Note that this attribute *will not exist* if the kernel
610		doesn't support changing the direction of a GPIO, or
611		it was exported by kernel code that didn't explicitly
612		allow userspace to reconfigure this GPIO's direction.
613
614	"value" ... reads as either 0 (low) or 1 (high).  If the GPIO
615		is configured as an output, this value may be written;
616		any nonzero value is treated as high.
617
618		If the pin can be configured as interrupt-generating interrupt
619		and if it has been configured to generate interrupts (see the
620		description of "edge"), you can poll(2) on that file and
621		poll(2) will return whenever the interrupt was triggered. If
622		you use poll(2), set the events POLLPRI and POLLERR. If you
623		use select(2), set the file descriptor in exceptfds. After
624		poll(2) returns, either lseek(2) to the beginning of the sysfs
625		file and read the new value or close the file and re-open it
626		to read the value.
627
628	"edge" ... reads as either "none", "rising", "falling", or
629		"both". Write these strings to select the signal edge(s)
630		that will make poll(2) on the "value" file return.
631
632		This file exists only if the pin can be configured as an
633		interrupt generating input pin.
634
635	"active_low" ... reads as either 0 (false) or 1 (true).  Write
636		any nonzero value to invert the value attribute both
637		for reading and writing.  Existing and subsequent
638		poll(2) support configuration via the edge attribute
639		for "rising" and "falling" edges will follow this
640		setting.
641
642GPIO controllers have paths like /sys/class/gpio/gpiochip42/ (for the
643controller implementing GPIOs starting at #42) and have the following
644read-only attributes:
645
646    /sys/class/gpio/gpiochipN/
647
648    	"base" ... same as N, the first GPIO managed by this chip
649
650    	"label" ... provided for diagnostics (not always unique)
651
652    	"ngpio" ... how many GPIOs this manges (N to N + ngpio - 1)
653
654Board documentation should in most cases cover what GPIOs are used for
655what purposes.  However, those numbers are not always stable; GPIOs on
656a daughtercard might be different depending on the base board being used,
657or other cards in the stack.  In such cases, you may need to use the
658gpiochip nodes (possibly in conjunction with schematics) to determine
659the correct GPIO number to use for a given signal.
660
661
662Exporting from Kernel code
663--------------------------
664Kernel code can explicitly manage exports of GPIOs which have already been
665requested using gpio_request():
666
667	/* export the GPIO to userspace */
668	int gpio_export(unsigned gpio, bool direction_may_change);
669
670	/* reverse gpio_export() */
671	void gpio_unexport();
672
673	/* create a sysfs link to an exported GPIO node */
674	int gpio_export_link(struct device *dev, const char *name,
675		unsigned gpio)
676
677	/* change the polarity of a GPIO node in sysfs */
678	int gpio_sysfs_set_active_low(unsigned gpio, int value);
679
680After a kernel driver requests a GPIO, it may only be made available in
681the sysfs interface by gpio_export().  The driver can control whether the
682signal direction may change.  This helps drivers prevent userspace code
683from accidentally clobbering important system state.
684
685This explicit exporting can help with debugging (by making some kinds
686of experiments easier), or can provide an always-there interface that's
687suitable for documenting as part of a board support package.
688
689After the GPIO has been exported, gpio_export_link() allows creating
690symlinks from elsewhere in sysfs to the GPIO sysfs node.  Drivers can
691use this to provide the interface under their own device in sysfs with
692a descriptive name.
693
694Drivers can use gpio_sysfs_set_active_low() to hide GPIO line polarity
695differences between boards from user space.  This only affects the
696sysfs interface.  Polarity change can be done both before and after
697gpio_export(), and previously enabled poll(2) support for either
698rising or falling edge will be reconfigured to follow this setting.