Commit | Line | Data |
---|---|---|
4c20386c DB |
1 | GPIO Interfaces |
2 | ||
3 | This provides an overview of GPIO access conventions on Linux. | |
4 | ||
7560fa60 DB |
5 | These calls use the gpio_* naming prefix. No other calls should use that |
6 | prefix, or the related __gpio_* prefix. | |
7 | ||
4c20386c DB |
8 | |
9 | What is a GPIO? | |
10 | =============== | |
11 | A "General Purpose Input/Output" (GPIO) is a flexible software-controlled | |
12 | digital signal. They are provided from many kinds of chip, and are familiar | |
13 | to Linux developers working with embedded and custom hardware. Each GPIO | |
14 | represents a bit connected to a particular pin, or "ball" on Ball Grid Array | |
15 | (BGA) packages. Board schematics show which external hardware connects to | |
16 | which GPIOs. Drivers can be written generically, so that board setup code | |
17 | passes such pin configuration data to drivers. | |
18 | ||
19 | System-on-Chip (SOC) processors heavily rely on GPIOs. In some cases, every | |
20 | non-dedicated pin can be configured as a GPIO; and most chips have at least | |
21 | several dozen of them. Programmable logic devices (like FPGAs) can easily | |
22 | provide GPIOs; multifunction chips like power managers, and audio codecs | |
23 | often have a few such pins to help with pin scarcity on SOCs; and there are | |
24 | also "GPIO Expander" chips that connect using the I2C or SPI serial busses. | |
25 | Most PC southbridges have a few dozen GPIO-capable pins (with only the BIOS | |
26 | firmware knowing how they're used). | |
27 | ||
28 | The exact capabilities of GPIOs vary between systems. Common options: | |
29 | ||
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 | |
1668be71 | 33 | for the other value (notably, "open drain" signaling). |
4c20386c DB |
34 | |
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 | |
7c2db759 | 38 | input de-glitch/debounce logic, sometimes with software controls. |
4c20386c DB |
39 | |
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. | |
43 | ||
44 | - Usually a GPIO will be configurable as either input or output, as needed | |
45 | by different product boards; single direction ones exist too. | |
46 | ||
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. | |
49 | ||
50 | On a given board each GPIO is used for one specific purpose like monitoring | |
51 | MMC/SD card insertion/removal, detecting card writeprotect status, driving | |
52 | a LED, configuring a transceiver, bitbanging a serial bus, poking a hardware | |
53 | watchdog, sensing a switch, and so on. | |
54 | ||
55 | ||
56 | GPIO conventions | |
57 | ================ | |
58 | Note that this is called a "convention" because you don't need to do it this | |
59 | way, and it's no crime if you don't. There **are** cases where portability | |
60 | is not the main issue; GPIOs are often used for the kind of board-specific | |
61 | glue logic that may even change between board revisions, and can't ever be | |
62 | used on a board that's wired differently. Only least-common-denominator | |
63 | functionality can be very portable. Other features are platform-specific, | |
64 | and that can be critical for glue logic. | |
65 | ||
7c2db759 | 66 | Plus, this doesn't require any implementation framework, just an interface. |
4c20386c DB |
67 | One platform might implement it as simple inline functions accessing chip |
68 | registers; another might implement it by delegating through abstractions | |
7c2db759 DB |
69 | used for several very different kinds of GPIO controller. (There is some |
70 | optional code supporting such an implementation strategy, described later | |
71 | in this document, but drivers acting as clients to the GPIO interface must | |
72 | not care how it's implemented.) | |
4c20386c DB |
73 | |
74 | That said, if the convention is supported on their platform, drivers should | |
7560fa60 DB |
75 | use it when possible. Platforms must declare GENERIC_GPIO support in their |
76 | Kconfig (boolean true), and provide an <asm/gpio.h> file. Drivers that can't | |
77 | work without standard GPIO calls should have Kconfig entries which depend | |
78 | on GENERIC_GPIO. The GPIO calls are available, either as "real code" or as | |
79 | optimized-away stubs, when drivers use the include file: | |
4c20386c | 80 | |
7560fa60 | 81 | #include <linux/gpio.h> |
4c20386c DB |
82 | |
83 | If you stick to this convention then it'll be easier for other developers to | |
84 | see what your code is doing, and help maintain it. | |
85 | ||
a0a99835 DB |
86 | Note that these operations include I/O barriers on platforms which need to |
87 | use them; drivers don't need to add them explicitly. | |
88 | ||
4c20386c DB |
89 | |
90 | Identifying GPIOs | |
91 | ----------------- | |
92 | GPIOs are identified by unsigned integers in the range 0..MAX_INT. That | |
93 | reserves "negative" numbers for other purposes like marking signals as | |
f5de6111 DB |
94 | "not available on this board", or indicating faults. Code that doesn't |
95 | touch the underlying hardware treats these integers as opaque cookies. | |
4c20386c DB |
96 | |
97 | Platforms define how they use those integers, and usually #define symbols | |
98 | for the GPIO lines so that board-specific setup code directly corresponds | |
99 | to the relevant schematics. In contrast, drivers should only use GPIO | |
100 | numbers passed to them from that setup code, using platform_data to hold | |
101 | board-specific pin configuration data (along with other board specific | |
102 | data they need). That avoids portability problems. | |
103 | ||
104 | So for example one platform uses numbers 32-159 for GPIOs; while another | |
105 | uses numbers 0..63 with one set of GPIO controllers, 64-79 with another | |
106 | type of GPIO controller, and on one particular board 80-95 with an FPGA. | |
107 | The numbers need not be contiguous; either of those platforms could also | |
108 | use numbers 2000-2063 to identify GPIOs in a bank of I2C GPIO expanders. | |
109 | ||
e6de1808 GL |
110 | If you want to initialize a structure with an invalid GPIO number, use |
111 | some negative number (perhaps "-EINVAL"); that will never be valid. To | |
112 | test if a number could reference a GPIO, you may use this predicate: | |
113 | ||
114 | int gpio_is_valid(int number); | |
115 | ||
116 | A number that's not valid will be rejected by calls which may request | |
117 | or free GPIOs (see below). Other numbers may also be rejected; for | |
118 | example, a number might be valid but unused on a given board. | |
119 | ||
4c20386c DB |
120 | Whether a platform supports multiple GPIO controllers is currently a |
121 | platform-specific implementation issue. | |
122 | ||
123 | ||
124 | Using GPIOs | |
125 | ----------- | |
126 | One of the first things to do with a GPIO, often in board setup code when | |
127 | setting up a platform_device using the GPIO, is mark its direction: | |
128 | ||
129 | /* set as input or output, returning 0 or negative errno */ | |
130 | int gpio_direction_input(unsigned gpio); | |
28735a72 | 131 | int gpio_direction_output(unsigned gpio, int value); |
4c20386c DB |
132 | |
133 | The return value is zero for success, else a negative errno. It should | |
134 | be checked, since the get/set calls don't have error returns and since | |
83c6590c DB |
135 | misconfiguration is possible. You should normally issue these calls from |
136 | a task context. However, for spinlock-safe GPIOs it's OK to use them | |
137 | before tasking is enabled, as part of early board setup. | |
4c20386c | 138 | |
28735a72 DB |
139 | For output GPIOs, the value provided becomes the initial output value. |
140 | This helps avoid signal glitching during system startup. | |
141 | ||
7c2db759 DB |
142 | For compatibility with legacy interfaces to GPIOs, setting the direction |
143 | of a GPIO implicitly requests that GPIO (see below) if it has not been | |
144 | requested already. That compatibility may be removed in the future; | |
145 | explicitly requesting GPIOs is strongly preferred. | |
146 | ||
4c20386c DB |
147 | Setting the direction can fail if the GPIO number is invalid, or when |
148 | that particular GPIO can't be used in that mode. It's generally a bad | |
149 | idea to rely on boot firmware to have set the direction correctly, since | |
150 | it probably wasn't validated to do more than boot Linux. (Similarly, | |
151 | that board setup code probably needs to multiplex that pin as a GPIO, | |
152 | and configure pullups/pulldowns appropriately.) | |
153 | ||
154 | ||
155 | Spinlock-Safe GPIO access | |
156 | ------------------------- | |
157 | Most GPIO controllers can be accessed with memory read/write instructions. | |
158 | That doesn't need to sleep, and can safely be done from inside IRQ handlers. | |
7c2db759 | 159 | (That includes hardirq contexts on RT kernels.) |
4c20386c DB |
160 | |
161 | Use these calls to access such GPIOs: | |
162 | ||
163 | /* GPIO INPUT: return zero or nonzero */ | |
164 | int gpio_get_value(unsigned gpio); | |
165 | ||
166 | /* GPIO OUTPUT */ | |
167 | void gpio_set_value(unsigned gpio, int value); | |
168 | ||
169 | The values are boolean, zero for low, nonzero for high. When reading the | |
170 | value of an output pin, the value returned should be what's seen on the | |
171 | pin ... that won't always match the specified output value, because of | |
7c2db759 | 172 | issues including open-drain signaling and output latencies. |
4c20386c DB |
173 | |
174 | The get/set calls have no error returns because "invalid GPIO" should have | |
be1ff386 | 175 | been reported earlier from gpio_direction_*(). However, note that not all |
4c20386c | 176 | platforms can read the value of output pins; those that can't should always |
f5de6111 DB |
177 | return zero. Also, using these calls for GPIOs that can't safely be accessed |
178 | without sleeping (see below) is an error. | |
4c20386c | 179 | |
f5de6111 | 180 | Platform-specific implementations are encouraged to optimize the two |
4c20386c DB |
181 | calls to access the GPIO value in cases where the GPIO number (and for |
182 | output, value) are constant. It's normal for them to need only a couple | |
183 | of instructions in such cases (reading or writing a hardware register), | |
184 | and not to need spinlocks. Such optimized calls can make bitbanging | |
185 | applications a lot more efficient (in both space and time) than spending | |
186 | dozens of instructions on subroutine calls. | |
187 | ||
188 | ||
189 | GPIO access that may sleep | |
190 | -------------------------- | |
191 | Some GPIO controllers must be accessed using message based busses like I2C | |
192 | or SPI. Commands to read or write those GPIO values require waiting to | |
193 | get to the head of a queue to transmit a command and get its response. | |
194 | This requires sleeping, which can't be done from inside IRQ handlers. | |
195 | ||
196 | Platforms that support this type of GPIO distinguish them from other GPIOs | |
7c2db759 DB |
197 | by returning nonzero from this call (which requires a valid GPIO number, |
198 | either explicitly or implicitly requested): | |
4c20386c DB |
199 | |
200 | int gpio_cansleep(unsigned gpio); | |
201 | ||
202 | To access such GPIOs, a different set of accessors is defined: | |
203 | ||
204 | /* GPIO INPUT: return zero or nonzero, might sleep */ | |
205 | int gpio_get_value_cansleep(unsigned gpio); | |
206 | ||
207 | /* GPIO OUTPUT, might sleep */ | |
208 | void gpio_set_value_cansleep(unsigned gpio, int value); | |
209 | ||
210 | Other than the fact that these calls might sleep, and will not be ignored | |
211 | for GPIOs that can't be accessed from IRQ handlers, these calls act the | |
212 | same as the spinlock-safe calls. | |
213 | ||
214 | ||
215 | Claiming and Releasing GPIOs (OPTIONAL) | |
216 | --------------------------------------- | |
217 | To help catch system configuration errors, two calls are defined. | |
218 | However, many platforms don't currently support this mechanism. | |
219 | ||
220 | /* request GPIO, returning 0 or negative errno. | |
221 | * non-null labels may be useful for diagnostics. | |
222 | */ | |
223 | int gpio_request(unsigned gpio, const char *label); | |
224 | ||
225 | /* release previously-claimed GPIO */ | |
226 | void gpio_free(unsigned gpio); | |
227 | ||
228 | Passing invalid GPIO numbers to gpio_request() will fail, as will requesting | |
229 | GPIOs that have already been claimed with that call. The return value of | |
83c6590c DB |
230 | gpio_request() must be checked. You should normally issue these calls from |
231 | a task context. However, for spinlock-safe GPIOs it's OK to request GPIOs | |
232 | before tasking is enabled, as part of early board setup. | |
4c20386c DB |
233 | |
234 | These calls serve two basic purposes. One is marking the signals which | |
235 | are actually in use as GPIOs, for better diagnostics; systems may have | |
236 | several hundred potential GPIOs, but often only a dozen are used on any | |
7c2db759 DB |
237 | given board. Another is to catch conflicts, identifying errors when |
238 | (a) two or more drivers wrongly think they have exclusive use of that | |
239 | signal, or (b) something wrongly believes it's safe to remove drivers | |
240 | needed to manage a signal that's in active use. That is, requesting a | |
241 | GPIO can serve as a kind of lock. | |
4c20386c DB |
242 | |
243 | These two calls are optional because not not all current Linux platforms | |
244 | offer such functionality in their GPIO support; a valid implementation | |
245 | could return success for all gpio_request() calls. Unlike the other calls, | |
246 | the state they represent doesn't normally match anything from a hardware | |
247 | register; it's just a software bitmap which clearly is not necessary for | |
248 | correct operation of hardware or (bug free) drivers. | |
249 | ||
250 | Note that requesting a GPIO does NOT cause it to be configured in any | |
251 | way; it just marks that GPIO as in use. Separate code must handle any | |
252 | pin setup (e.g. controlling which pin the GPIO uses, pullup/pulldown). | |
253 | ||
7c2db759 DB |
254 | Also note that it's your responsibility to have stopped using a GPIO |
255 | before you free it. | |
256 | ||
4c20386c DB |
257 | |
258 | GPIOs mapped to IRQs | |
259 | -------------------- | |
260 | GPIO numbers are unsigned integers; so are IRQ numbers. These make up | |
261 | two logically distinct namespaces (GPIO 0 need not use IRQ 0). You can | |
262 | map between them using calls like: | |
263 | ||
264 | /* map GPIO numbers to IRQ numbers */ | |
265 | int gpio_to_irq(unsigned gpio); | |
266 | ||
267 | /* map IRQ numbers to GPIO numbers */ | |
268 | int irq_to_gpio(unsigned irq); | |
269 | ||
270 | Those return either the corresponding number in the other namespace, or | |
271 | else a negative errno code if the mapping can't be done. (For example, | |
7c2db759 | 272 | some GPIOs can't be used as IRQs.) It is an unchecked error to use a GPIO |
be1ff386 | 273 | number that wasn't set up as an input using gpio_direction_input(), or |
4c20386c DB |
274 | to use an IRQ number that didn't originally come from gpio_to_irq(). |
275 | ||
276 | These two mapping calls are expected to cost on the order of a single | |
277 | addition or subtraction. They're not allowed to sleep. | |
278 | ||
279 | Non-error values returned from gpio_to_irq() can be passed to request_irq() | |
280 | or free_irq(). They will often be stored into IRQ resources for platform | |
281 | devices, by the board-specific initialization code. Note that IRQ trigger | |
282 | options are part of the IRQ interface, e.g. IRQF_TRIGGER_FALLING, as are | |
283 | system wakeup capabilities. | |
284 | ||
285 | Non-error values returned from irq_to_gpio() would most commonly be used | |
f5de6111 DB |
286 | with gpio_get_value(), for example to initialize or update driver state |
287 | when the IRQ is edge-triggered. | |
4c20386c DB |
288 | |
289 | ||
1668be71 DB |
290 | Emulating Open Drain Signals |
291 | ---------------------------- | |
292 | Sometimes shared signals need to use "open drain" signaling, where only the | |
293 | low signal level is actually driven. (That term applies to CMOS transistors; | |
294 | "open collector" is used for TTL.) A pullup resistor causes the high signal | |
295 | level. This is sometimes called a "wire-AND"; or more practically, from the | |
296 | negative logic (low=true) perspective this is a "wire-OR". | |
297 | ||
298 | One common example of an open drain signal is a shared active-low IRQ line. | |
299 | Also, bidirectional data bus signals sometimes use open drain signals. | |
300 | ||
301 | Some GPIO controllers directly support open drain outputs; many don't. When | |
302 | you need open drain signaling but your hardware doesn't directly support it, | |
303 | there's a common idiom you can use to emulate it with any GPIO pin that can | |
304 | be used as either an input or an output: | |
305 | ||
306 | LOW: gpio_direction_output(gpio, 0) ... this drives the signal | |
307 | and overrides the pullup. | |
308 | ||
309 | HIGH: gpio_direction_input(gpio) ... this turns off the output, | |
310 | so the pullup (or some other device) controls the signal. | |
311 | ||
312 | If you are "driving" the signal high but gpio_get_value(gpio) reports a low | |
313 | value (after the appropriate rise time passes), you know some other component | |
314 | is driving the shared signal low. That's not necessarily an error. As one | |
315 | common example, that's how I2C clocks are stretched: a slave that needs a | |
316 | slower clock delays the rising edge of SCK, and the I2C master adjusts its | |
317 | signaling rate accordingly. | |
318 | ||
4c20386c DB |
319 | |
320 | What do these conventions omit? | |
321 | =============================== | |
322 | One of the biggest things these conventions omit is pin multiplexing, since | |
323 | this is highly chip-specific and nonportable. One platform might not need | |
324 | explicit multiplexing; another might have just two options for use of any | |
325 | given pin; another might have eight options per pin; another might be able | |
326 | to route a given GPIO to any one of several pins. (Yes, those examples all | |
327 | come from systems that run Linux today.) | |
328 | ||
329 | Related to multiplexing is configuration and enabling of the pullups or | |
330 | pulldowns integrated on some platforms. Not all platforms support them, | |
331 | or support them in the same way; and any given board might use external | |
332 | pullups (or pulldowns) so that the on-chip ones should not be used. | |
7c2db759 | 333 | (When a circuit needs 5 kOhm, on-chip 100 kOhm resistors won't do.) |
7560fa60 DB |
334 | Likewise drive strength (2 mA vs 20 mA) and voltage (1.8V vs 3.3V) is a |
335 | platform-specific issue, as are models like (not) having a one-to-one | |
336 | correspondence between configurable pins and GPIOs. | |
4c20386c DB |
337 | |
338 | There are other system-specific mechanisms that are not specified here, | |
339 | like the aforementioned options for input de-glitching and wire-OR output. | |
340 | Hardware may support reading or writing GPIOs in gangs, but that's usually | |
f5de6111 | 341 | configuration dependent: for GPIOs sharing the same bank. (GPIOs are |
4c20386c | 342 | commonly grouped in banks of 16 or 32, with a given SOC having several such |
7c2db759 DB |
343 | banks.) Some systems can trigger IRQs from output GPIOs, or read values |
344 | from pins not managed as GPIOs. Code relying on such mechanisms will | |
345 | necessarily be nonportable. | |
4c20386c | 346 | |
7c2db759 | 347 | Dynamic definition of GPIOs is not currently standard; for example, as |
4c20386c DB |
348 | a side effect of configuring an add-on board with some GPIO expanders. |
349 | ||
350 | These calls are purely for kernel space, but a userspace API could be built | |
7c2db759 DB |
351 | on top of them. |
352 | ||
353 | ||
354 | GPIO implementor's framework (OPTIONAL) | |
355 | ======================================= | |
356 | As noted earlier, there is an optional implementation framework making it | |
357 | easier for platforms to support different kinds of GPIO controller using | |
358 | the same programming interface. | |
359 | ||
360 | As a debugging aid, if debugfs is available a /sys/kernel/debug/gpio file | |
361 | will be found there. That will list all the controllers registered through | |
362 | this framework, and the state of the GPIOs currently in use. | |
363 | ||
364 | ||
365 | Controller Drivers: gpio_chip | |
366 | ----------------------------- | |
367 | In this framework each GPIO controller is packaged as a "struct gpio_chip" | |
368 | with information common to each controller of that type: | |
369 | ||
370 | - methods to establish GPIO direction | |
371 | - methods used to access GPIO values | |
372 | - flag saying whether calls to its methods may sleep | |
373 | - optional debugfs dump method (showing extra state like pullup config) | |
374 | - label for diagnostics | |
375 | ||
376 | There is also per-instance data, which may come from device.platform_data: | |
377 | the number of its first GPIO, and how many GPIOs it exposes. | |
378 | ||
379 | The code implementing a gpio_chip should support multiple instances of the | |
380 | controller, possibly using the driver model. That code will configure each | |
381 | gpio_chip and issue gpiochip_add(). Removing a GPIO controller should be | |
382 | rare; use gpiochip_remove() when it is unavoidable. | |
383 | ||
384 | Most often a gpio_chip is part of an instance-specific structure with state | |
385 | not exposed by the GPIO interfaces, such as addressing, power management, | |
386 | and more. Chips such as codecs will have complex non-GPIO state, | |
387 | ||
388 | Any debugfs dump method should normally ignore signals which haven't been | |
389 | requested as GPIOs. They can use gpiochip_is_requested(), which returns | |
390 | either NULL or the label associated with that GPIO when it was requested. | |
391 | ||
392 | ||
393 | Platform Support | |
394 | ---------------- | |
395 | To support this framework, a platform's Kconfig will "select HAVE_GPIO_LIB" | |
396 | and arrange that its <asm/gpio.h> includes <asm-generic/gpio.h> and defines | |
397 | three functions: gpio_get_value(), gpio_set_value(), and gpio_cansleep(). | |
398 | They may also want to provide a custom value for ARCH_NR_GPIOS. | |
399 | ||
400 | Trivial implementations of those functions can directly use framework | |
401 | code, which always dispatches through the gpio_chip: | |
402 | ||
403 | #define gpio_get_value __gpio_get_value | |
404 | #define gpio_set_value __gpio_set_value | |
405 | #define gpio_cansleep __gpio_cansleep | |
406 | ||
407 | Fancier implementations could instead define those as inline functions with | |
408 | logic optimizing access to specific SOC-based GPIOs. For example, if the | |
409 | referenced GPIO is the constant "12", getting or setting its value could | |
410 | cost as little as two or three instructions, never sleeping. When such an | |
411 | optimization is not possible those calls must delegate to the framework | |
412 | code, costing at least a few dozen instructions. For bitbanged I/O, such | |
413 | instruction savings can be significant. | |
414 | ||
415 | For SOCs, platform-specific code defines and registers gpio_chip instances | |
416 | for each bank of on-chip GPIOs. Those GPIOs should be numbered/labeled to | |
417 | match chip vendor documentation, and directly match board schematics. They | |
418 | may well start at zero and go up to a platform-specific limit. Such GPIOs | |
419 | are normally integrated into platform initialization to make them always be | |
420 | available, from arch_initcall() or earlier; they can often serve as IRQs. | |
421 | ||
422 | ||
423 | Board Support | |
424 | ------------- | |
425 | For external GPIO controllers -- such as I2C or SPI expanders, ASICs, multi | |
426 | function devices, FPGAs or CPLDs -- most often board-specific code handles | |
427 | registering controller devices and ensures that their drivers know what GPIO | |
428 | numbers to use with gpiochip_add(). Their numbers often start right after | |
429 | platform-specific GPIOs. | |
430 | ||
431 | For example, board setup code could create structures identifying the range | |
432 | of GPIOs that chip will expose, and passes them to each GPIO expander chip | |
433 | using platform_data. Then the chip driver's probe() routine could pass that | |
434 | data to gpiochip_add(). | |
435 | ||
436 | Initialization order can be important. For example, when a device relies on | |
437 | an I2C-based GPIO, its probe() routine should only be called after that GPIO | |
438 | becomes available. That may mean the device should not be registered until | |
439 | calls for that GPIO can work. One way to address such dependencies is for | |
440 | such gpio_chip controllers to provide setup() and teardown() callbacks to | |
441 | board specific code; those board specific callbacks would register devices | |
442 | once all the necessary resources are available. |