blob: a5ffba33a35188a0d858871c06c090361de81b88 [file] [log] [blame]
David Brownell8ae12a02006-01-08 13:34:19 -08001Overview of Linux kernel SPI support
2====================================
3
David Brownellb8852442006-01-08 13:34:23 -0800402-Dec-2005
David Brownell8ae12a02006-01-08 13:34:19 -08005
6What is SPI?
7------------
David Brownellb8852442006-01-08 13:34:23 -08008The "Serial Peripheral Interface" (SPI) is a synchronous four wire serial
9link used to connect microcontrollers to sensors, memory, and peripherals.
David Brownell8ae12a02006-01-08 13:34:19 -080010
11The three signal wires hold a clock (SCLK, often on the order of 10 MHz),
12and parallel data lines with "Master Out, Slave In" (MOSI) or "Master In,
13Slave Out" (MISO) signals. (Other names are also used.) There are four
14clocking modes through which data is exchanged; mode-0 and mode-3 are most
David Brownellb8852442006-01-08 13:34:23 -080015commonly used. Each clock cycle shifts data out and data in; the clock
16doesn't cycle except when there is data to shift.
David Brownell8ae12a02006-01-08 13:34:19 -080017
18SPI masters may use a "chip select" line to activate a given SPI slave
19device, so those three signal wires may be connected to several chips
20in parallel. All SPI slaves support chipselects. Some devices have
21other signals, often including an interrupt to the master.
22
23Unlike serial busses like USB or SMBUS, even low level protocols for
24SPI slave functions are usually not interoperable between vendors
25(except for cases like SPI memory chips).
26
27 - SPI may be used for request/response style device protocols, as with
28 touchscreen sensors and memory chips.
29
30 - It may also be used to stream data in either direction (half duplex),
31 or both of them at the same time (full duplex).
32
33 - Some devices may use eight bit words. Others may different word
34 lengths, such as streams of 12-bit or 20-bit digital samples.
35
36In the same way, SPI slaves will only rarely support any kind of automatic
37discovery/enumeration protocol. The tree of slave devices accessible from
38a given SPI master will normally be set up manually, with configuration
39tables.
40
41SPI is only one of the names used by such four-wire protocols, and
42most controllers have no problem handling "MicroWire" (think of it as
43half-duplex SPI, for request/response protocols), SSP ("Synchronous
44Serial Protocol"), PSP ("Programmable Serial Protocol"), and other
45related protocols.
46
47Microcontrollers often support both master and slave sides of the SPI
48protocol. This document (and Linux) currently only supports the master
49side of SPI interactions.
50
51
52Who uses it? On what kinds of systems?
53---------------------------------------
54Linux developers using SPI are probably writing device drivers for embedded
55systems boards. SPI is used to control external chips, and it is also a
56protocol supported by every MMC or SD memory card. (The older "DataFlash"
57cards, predating MMC cards but using the same connectors and card shape,
58support only SPI.) Some PC hardware uses SPI flash for BIOS code.
59
60SPI slave chips range from digital/analog converters used for analog
61sensors and codecs, to memory, to peripherals like USB controllers
62or Ethernet adapters; and more.
63
64Most systems using SPI will integrate a few devices on a mainboard.
65Some provide SPI links on expansion connectors; in cases where no
66dedicated SPI controller exists, GPIO pins can be used to create a
67low speed "bitbanging" adapter. Very few systems will "hotplug" an SPI
68controller; the reasons to use SPI focus on low cost and simple operation,
69and if dynamic reconfiguration is important, USB will often be a more
70appropriate low-pincount peripheral bus.
71
72Many microcontrollers that can run Linux integrate one or more I/O
73interfaces with SPI modes. Given SPI support, they could use MMC or SD
74cards without needing a special purpose MMC/SD/SDIO controller.
75
76
77How do these driver programming interfaces work?
78------------------------------------------------
79The <linux/spi/spi.h> header file includes kerneldoc, as does the
80main source code, and you should certainly read that. This is just
81an overview, so you get the big picture before the details.
82
David Brownellb8852442006-01-08 13:34:23 -080083SPI requests always go into I/O queues. Requests for a given SPI device
84are always executed in FIFO order, and complete asynchronously through
85completion callbacks. There are also some simple synchronous wrappers
86for those calls, including ones for common transaction types like writing
87a command and then reading its response.
88
David Brownell8ae12a02006-01-08 13:34:19 -080089There are two types of SPI driver, here called:
90
91 Controller drivers ... these are often built in to System-On-Chip
92 processors, and often support both Master and Slave roles.
93 These drivers touch hardware registers and may use DMA.
David Brownellb8852442006-01-08 13:34:23 -080094 Or they can be PIO bitbangers, needing just GPIO pins.
David Brownell8ae12a02006-01-08 13:34:19 -080095
96 Protocol drivers ... these pass messages through the controller
97 driver to communicate with a Slave or Master device on the
98 other side of an SPI link.
99
100So for example one protocol driver might talk to the MTD layer to export
101data to filesystems stored on SPI flash like DataFlash; and others might
102control audio interfaces, present touchscreen sensors as input interfaces,
103or monitor temperature and voltage levels during industrial processing.
104And those might all be sharing the same controller driver.
105
106A "struct spi_device" encapsulates the master-side interface between
107those two types of driver. At this writing, Linux has no slave side
108programming interface.
109
110There is a minimal core of SPI programming interfaces, focussing on
111using driver model to connect controller and protocol drivers using
112device tables provided by board specific initialization code. SPI
113shows up in sysfs in several locations:
114
115 /sys/devices/.../CTLR/spiB.C ... spi_device for on bus "B",
116 chipselect C, accessed through CTLR.
117
David Brownell71117632006-01-08 13:34:29 -0800118 /sys/devices/.../CTLR/spiB.C/modalias ... identifies the driver
119 that should be used with this device (for hotplug/coldplug)
120
David Brownell8ae12a02006-01-08 13:34:19 -0800121 /sys/bus/spi/devices/spiB.C ... symlink to the physical
122 spiB-C device
123
124 /sys/bus/spi/drivers/D ... driver for one or more spi*.* devices
125
126 /sys/class/spi_master/spiB ... class device for the controller
127 managing bus "B". All the spiB.* devices share the same
128 physical SPI bus segment, with SCLK, MOSI, and MISO.
129
David Brownell8ae12a02006-01-08 13:34:19 -0800130
131How does board-specific init code declare SPI devices?
132------------------------------------------------------
133Linux needs several kinds of information to properly configure SPI devices.
134That information is normally provided by board-specific code, even for
135chips that do support some of automated discovery/enumeration.
136
137DECLARE CONTROLLERS
138
139The first kind of information is a list of what SPI controllers exist.
140For System-on-Chip (SOC) based boards, these will usually be platform
141devices, and the controller may need some platform_data in order to
142operate properly. The "struct platform_device" will include resources
143like the physical address of the controller's first register and its IRQ.
144
145Platforms will often abstract the "register SPI controller" operation,
146maybe coupling it with code to initialize pin configurations, so that
147the arch/.../mach-*/board-*.c files for several boards can all share the
148same basic controller setup code. This is because most SOCs have several
149SPI-capable controllers, and only the ones actually usable on a given
150board should normally be set up and registered.
151
152So for example arch/.../mach-*/board-*.c files might have code like:
153
154 #include <asm/arch/spi.h> /* for mysoc_spi_data */
155
156 /* if your mach-* infrastructure doesn't support kernels that can
157 * run on multiple boards, pdata wouldn't benefit from "__init".
158 */
159 static struct mysoc_spi_data __init pdata = { ... };
160
161 static __init board_init(void)
162 {
163 ...
164 /* this board only uses SPI controller #2 */
165 mysoc_register_spi(2, &pdata);
166 ...
167 }
168
169And SOC-specific utility code might look something like:
170
171 #include <asm/arch/spi.h>
172
173 static struct platform_device spi2 = { ... };
174
175 void mysoc_register_spi(unsigned n, struct mysoc_spi_data *pdata)
176 {
177 struct mysoc_spi_data *pdata2;
178
179 pdata2 = kmalloc(sizeof *pdata2, GFP_KERNEL);
180 *pdata2 = pdata;
181 ...
182 if (n == 2) {
183 spi2->dev.platform_data = pdata2;
184 register_platform_device(&spi2);
185
186 /* also: set up pin modes so the spi2 signals are
187 * visible on the relevant pins ... bootloaders on
188 * production boards may already have done this, but
189 * developer boards will often need Linux to do it.
190 */
191 }
192 ...
193 }
194
195Notice how the platform_data for boards may be different, even if the
196same SOC controller is used. For example, on one board SPI might use
197an external clock, where another derives the SPI clock from current
198settings of some master clock.
199
200
201DECLARE SLAVE DEVICES
202
203The second kind of information is a list of what SPI slave devices exist
204on the target board, often with some board-specific data needed for the
205driver to work correctly.
206
207Normally your arch/.../mach-*/board-*.c files would provide a small table
208listing the SPI devices on each board. (This would typically be only a
209small handful.) That might look like:
210
211 static struct ads7846_platform_data ads_info = {
212 .vref_delay_usecs = 100,
213 .x_plate_ohms = 580,
214 .y_plate_ohms = 410,
215 };
216
217 static struct spi_board_info spi_board_info[] __initdata = {
218 {
219 .modalias = "ads7846",
220 .platform_data = &ads_info,
221 .mode = SPI_MODE_0,
222 .irq = GPIO_IRQ(31),
223 .max_speed_hz = 120000 /* max sample rate at 3V */ * 16,
224 .bus_num = 1,
225 .chip_select = 0,
226 },
227 };
228
229Again, notice how board-specific information is provided; each chip may need
230several types. This example shows generic constraints like the fastest SPI
231clock to allow (a function of board voltage in this case) or how an IRQ pin
232is wired, plus chip-specific constraints like an important delay that's
233changed by the capacitance at one pin.
234
235(There's also "controller_data", information that may be useful to the
236controller driver. An example would be peripheral-specific DMA tuning
237data or chipselect callbacks. This is stored in spi_device later.)
238
239The board_info should provide enough information to let the system work
240without the chip's driver being loaded. The most troublesome aspect of
241that is likely the SPI_CS_HIGH bit in the spi_device.mode field, since
242sharing a bus with a device that interprets chipselect "backwards" is
243not possible.
244
245Then your board initialization code would register that table with the SPI
246infrastructure, so that it's available later when the SPI master controller
247driver is registered:
248
249 spi_register_board_info(spi_board_info, ARRAY_SIZE(spi_board_info));
250
251Like with other static board-specific setup, you won't unregister those.
252
David Brownell71117632006-01-08 13:34:29 -0800253The widely used "card" style computers bundle memory, cpu, and little else
254onto a card that's maybe just thirty square centimeters. On such systems,
255your arch/.../mach-.../board-*.c file would primarily provide information
256about the devices on the mainboard into which such a card is plugged. That
257certainly includes SPI devices hooked up through the card connectors!
258
David Brownell8ae12a02006-01-08 13:34:19 -0800259
260NON-STATIC CONFIGURATIONS
261
262Developer boards often play by different rules than product boards, and one
263example is the potential need to hotplug SPI devices and/or controllers.
264
265For those cases you might need to use use spi_busnum_to_master() to look
266up the spi bus master, and will likely need spi_new_device() to provide the
267board info based on the board that was hotplugged. Of course, you'd later
268call at least spi_unregister_device() when that board is removed.
269
David Brownell71117632006-01-08 13:34:29 -0800270When Linux includes support for MMC/SD/SDIO/DataFlash cards through SPI, those
271configurations will also be dynamic. Fortunately, those devices all support
272basic device identification probes, so that support should hotplug normally.
273
David Brownell8ae12a02006-01-08 13:34:19 -0800274
275How do I write an "SPI Protocol Driver"?
276----------------------------------------
277All SPI drivers are currently kernel drivers. A userspace driver API
278would just be another kernel driver, probably offering some lowlevel
279access through aio_read(), aio_write(), and ioctl() calls and using the
280standard userspace sysfs mechanisms to bind to a given SPI device.
281
David Brownellb8852442006-01-08 13:34:23 -0800282SPI protocol drivers somewhat resemble platform device drivers:
David Brownell8ae12a02006-01-08 13:34:19 -0800283
David Brownellb8852442006-01-08 13:34:23 -0800284 static struct spi_driver CHIP_driver = {
285 .driver = {
286 .name = "CHIP",
287 .bus = &spi_bus_type,
288 .owner = THIS_MODULE,
289 },
290
David Brownell8ae12a02006-01-08 13:34:19 -0800291 .probe = CHIP_probe,
David Brownellb8852442006-01-08 13:34:23 -0800292 .remove = __devexit_p(CHIP_remove),
David Brownell8ae12a02006-01-08 13:34:19 -0800293 .suspend = CHIP_suspend,
294 .resume = CHIP_resume,
295 };
296
David Brownellb8852442006-01-08 13:34:23 -0800297The driver core will autmatically attempt to bind this driver to any SPI
David Brownell8ae12a02006-01-08 13:34:19 -0800298device whose board_info gave a modalias of "CHIP". Your probe() code
299might look like this unless you're creating a class_device:
300
David Brownellb8852442006-01-08 13:34:23 -0800301 static int __devinit CHIP_probe(struct spi_device *spi)
David Brownell8ae12a02006-01-08 13:34:19 -0800302 {
David Brownell8ae12a02006-01-08 13:34:19 -0800303 struct CHIP *chip;
David Brownellb8852442006-01-08 13:34:23 -0800304 struct CHIP_platform_data *pdata;
305
306 /* assuming the driver requires board-specific data: */
307 pdata = &spi->dev.platform_data;
308 if (!pdata)
309 return -ENODEV;
David Brownell8ae12a02006-01-08 13:34:19 -0800310
311 /* get memory for driver's per-chip state */
312 chip = kzalloc(sizeof *chip, GFP_KERNEL);
313 if (!chip)
314 return -ENOMEM;
David Brownellb8852442006-01-08 13:34:23 -0800315 dev_set_drvdata(&spi->dev, chip);
David Brownell8ae12a02006-01-08 13:34:19 -0800316
317 ... etc
318 return 0;
319 }
320
321As soon as it enters probe(), the driver may issue I/O requests to
322the SPI device using "struct spi_message". When remove() returns,
323the driver guarantees that it won't submit any more such messages.
324
325 - An spi_message is a sequence of of protocol operations, executed
326 as one atomic sequence. SPI driver controls include:
327
328 + when bidirectional reads and writes start ... by how its
329 sequence of spi_transfer requests is arranged;
330
331 + optionally defining short delays after transfers ... using
332 the spi_transfer.delay_usecs setting;
333
334 + whether the chipselect becomes inactive after a transfer and
335 any delay ... by using the spi_transfer.cs_change flag;
336
337 + hinting whether the next message is likely to go to this same
338 device ... using the spi_transfer.cs_change flag on the last
339 transfer in that atomic group, and potentially saving costs
340 for chip deselect and select operations.
341
342 - Follow standard kernel rules, and provide DMA-safe buffers in
343 your messages. That way controller drivers using DMA aren't forced
344 to make extra copies unless the hardware requires it (e.g. working
345 around hardware errata that force the use of bounce buffering).
346
347 If standard dma_map_single() handling of these buffers is inappropriate,
348 you can use spi_message.is_dma_mapped to tell the controller driver
349 that you've already provided the relevant DMA addresses.
350
351 - The basic I/O primitive is spi_async(). Async requests may be
352 issued in any context (irq handler, task, etc) and completion
353 is reported using a callback provided with the message.
David Brownellb8852442006-01-08 13:34:23 -0800354 After any detected error, the chip is deselected and processing
355 of that spi_message is aborted.
David Brownell8ae12a02006-01-08 13:34:19 -0800356
357 - There are also synchronous wrappers like spi_sync(), and wrappers
358 like spi_read(), spi_write(), and spi_write_then_read(). These
359 may be issued only in contexts that may sleep, and they're all
360 clean (and small, and "optional") layers over spi_async().
361
362 - The spi_write_then_read() call, and convenience wrappers around
363 it, should only be used with small amounts of data where the
364 cost of an extra copy may be ignored. It's designed to support
365 common RPC-style requests, such as writing an eight bit command
366 and reading a sixteen bit response -- spi_w8r16() being one its
367 wrappers, doing exactly that.
368
369Some drivers may need to modify spi_device characteristics like the
370transfer mode, wordsize, or clock rate. This is done with spi_setup(),
371which would normally be called from probe() before the first I/O is
372done to the device.
373
374While "spi_device" would be the bottom boundary of the driver, the
375upper boundaries might include sysfs (especially for sensor readings),
376the input layer, ALSA, networking, MTD, the character device framework,
377or other Linux subsystems.
378
David Brownell0c868462006-01-08 13:34:25 -0800379Note that there are two types of memory your driver must manage as part
380of interacting with SPI devices.
381
382 - I/O buffers use the usual Linux rules, and must be DMA-safe.
383 You'd normally allocate them from the heap or free page pool.
384 Don't use the stack, or anything that's declared "static".
385
386 - The spi_message and spi_transfer metadata used to glue those
387 I/O buffers into a group of protocol transactions. These can
388 be allocated anywhere it's convenient, including as part of
389 other allocate-once driver data structures. Zero-init these.
390
391If you like, spi_message_alloc() and spi_message_free() convenience
392routines are available to allocate and zero-initialize an spi_message
393with several transfers.
394
David Brownell8ae12a02006-01-08 13:34:19 -0800395
396How do I write an "SPI Master Controller Driver"?
397-------------------------------------------------
398An SPI controller will probably be registered on the platform_bus; write
399a driver to bind to the device, whichever bus is involved.
400
401The main task of this type of driver is to provide an "spi_master".
402Use spi_alloc_master() to allocate the master, and class_get_devdata()
403to get the driver-private data allocated for that device.
404
405 struct spi_master *master;
406 struct CONTROLLER *c;
407
408 master = spi_alloc_master(dev, sizeof *c);
409 if (!master)
410 return -ENODEV;
411
412 c = class_get_devdata(&master->cdev);
413
414The driver will initialize the fields of that spi_master, including the
415bus number (maybe the same as the platform device ID) and three methods
416used to interact with the SPI core and SPI protocol drivers. It will
417also initialize its own internal state.
418
419 master->setup(struct spi_device *spi)
420 This sets up the device clock rate, SPI mode, and word sizes.
421 Drivers may change the defaults provided by board_info, and then
422 call spi_setup(spi) to invoke this routine. It may sleep.
423
424 master->transfer(struct spi_device *spi, struct spi_message *message)
425 This must not sleep. Its responsibility is arrange that the
426 transfer happens and its complete() callback is issued; the two
427 will normally happen later, after other transfers complete.
428
429 master->cleanup(struct spi_device *spi)
430 Your controller driver may use spi_device.controller_state to hold
431 state it dynamically associates with that device. If you do that,
432 be sure to provide the cleanup() method to free that state.
433
434The bulk of the driver will be managing the I/O queue fed by transfer().
435
436That queue could be purely conceptual. For example, a driver used only
437for low-frequency sensor acess might be fine using synchronous PIO.
438
439But the queue will probably be very real, using message->queue, PIO,
440often DMA (especially if the root filesystem is in SPI flash), and
441execution contexts like IRQ handlers, tasklets, or workqueues (such
442as keventd). Your driver can be as fancy, or as simple, as you need.
443
444
445THANKS TO
446---------
447Contributors to Linux-SPI discussions include (in alphabetical order,
448by last name):
449
450David Brownell
451Russell King
452Dmitry Pervushin
453Stephen Street
454Mark Underwood
455Andrew Victor
456Vitaly Wool
457