view Documentation/spi/spi-summary @ 897:329ea0ccb344

balloon: try harder to balloon up under memory pressure.

Currently if the balloon driver is unable to increase the guest's
reservation it assumes the failure was due to reaching its full
allocation, gives up on the ballooning operation and records the limit
it reached as the "hard limit". The driver will not try again until
the target is set again (even to the same value).

However it is possible that ballooning has in fact failed due to
memory pressure in the host and therefore it is desirable to keep
attempting to reach the target in case memory becomes available. The
most likely scenario is that some guests are ballooning down while
others are ballooning up and therefore there is temporary memory
pressure while things stabilise. You would not expect a well behaved
toolstack to ask a domain to balloon to more than its allocation nor
would you expect it to deliberately over-commit memory by setting
balloon targets which exceed the total host memory.

This patch drops the concept of a hard limit and causes the balloon
driver to retry increasing the reservation on a timer in the same
manner as when decreasing the reservation.

Also if we partially succeed in increasing the reservation
(i.e. receive less pages than we asked for) then we may as well keep
those pages rather than returning them to Xen.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
author Keir Fraser <keir.fraser@citrix.com>
date Fri Jun 05 14:01:20 2009 +0100 (2009-06-05)
parents 831230e53067
line source
1 Overview of Linux kernel SPI support
2 ====================================
4 02-Dec-2005
6 What is SPI?
7 ------------
8 The "Serial Peripheral Interface" (SPI) is a synchronous four wire serial
9 link used to connect microcontrollers to sensors, memory, and peripherals.
11 The three signal wires hold a clock (SCLK, often on the order of 10 MHz),
12 and parallel data lines with "Master Out, Slave In" (MOSI) or "Master In,
13 Slave Out" (MISO) signals. (Other names are also used.) There are four
14 clocking modes through which data is exchanged; mode-0 and mode-3 are most
15 commonly used. Each clock cycle shifts data out and data in; the clock
16 doesn't cycle except when there is data to shift.
18 SPI masters may use a "chip select" line to activate a given SPI slave
19 device, so those three signal wires may be connected to several chips
20 in parallel. All SPI slaves support chipselects. Some devices have
21 other signals, often including an interrupt to the master.
23 Unlike serial busses like USB or SMBUS, even low level protocols for
24 SPI slave functions are usually not interoperable between vendors
25 (except for cases like SPI memory chips).
27 - SPI may be used for request/response style device protocols, as with
28 touchscreen sensors and memory chips.
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).
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.
36 In the same way, SPI slaves will only rarely support any kind of automatic
37 discovery/enumeration protocol. The tree of slave devices accessible from
38 a given SPI master will normally be set up manually, with configuration
39 tables.
41 SPI is only one of the names used by such four-wire protocols, and
42 most controllers have no problem handling "MicroWire" (think of it as
43 half-duplex SPI, for request/response protocols), SSP ("Synchronous
44 Serial Protocol"), PSP ("Programmable Serial Protocol"), and other
45 related protocols.
47 Microcontrollers often support both master and slave sides of the SPI
48 protocol. This document (and Linux) currently only supports the master
49 side of SPI interactions.
52 Who uses it? On what kinds of systems?
53 ---------------------------------------
54 Linux developers using SPI are probably writing device drivers for embedded
55 systems boards. SPI is used to control external chips, and it is also a
56 protocol supported by every MMC or SD memory card. (The older "DataFlash"
57 cards, predating MMC cards but using the same connectors and card shape,
58 support only SPI.) Some PC hardware uses SPI flash for BIOS code.
60 SPI slave chips range from digital/analog converters used for analog
61 sensors and codecs, to memory, to peripherals like USB controllers
62 or Ethernet adapters; and more.
64 Most systems using SPI will integrate a few devices on a mainboard.
65 Some provide SPI links on expansion connectors; in cases where no
66 dedicated SPI controller exists, GPIO pins can be used to create a
67 low speed "bitbanging" adapter. Very few systems will "hotplug" an SPI
68 controller; the reasons to use SPI focus on low cost and simple operation,
69 and if dynamic reconfiguration is important, USB will often be a more
70 appropriate low-pincount peripheral bus.
72 Many microcontrollers that can run Linux integrate one or more I/O
73 interfaces with SPI modes. Given SPI support, they could use MMC or SD
74 cards without needing a special purpose MMC/SD/SDIO controller.
77 How do these driver programming interfaces work?
78 ------------------------------------------------
79 The <linux/spi/spi.h> header file includes kerneldoc, as does the
80 main source code, and you should certainly read that. This is just
81 an overview, so you get the big picture before the details.
83 SPI requests always go into I/O queues. Requests for a given SPI device
84 are always executed in FIFO order, and complete asynchronously through
85 completion callbacks. There are also some simple synchronous wrappers
86 for those calls, including ones for common transaction types like writing
87 a command and then reading its response.
89 There are two types of SPI driver, here called:
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.
94 Or they can be PIO bitbangers, needing just GPIO pins.
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.
100 So for example one protocol driver might talk to the MTD layer to export
101 data to filesystems stored on SPI flash like DataFlash; and others might
102 control audio interfaces, present touchscreen sensors as input interfaces,
103 or monitor temperature and voltage levels during industrial processing.
104 And those might all be sharing the same controller driver.
106 A "struct spi_device" encapsulates the master-side interface between
107 those two types of driver. At this writing, Linux has no slave side
108 programming interface.
110 There is a minimal core of SPI programming interfaces, focussing on
111 using driver model to connect controller and protocol drivers using
112 device tables provided by board specific initialization code. SPI
113 shows up in sysfs in several locations:
115 /sys/devices/.../CTLR/spiB.C ... spi_device for on bus "B",
116 chipselect C, accessed through CTLR.
118 /sys/devices/.../CTLR/spiB.C/modalias ... identifies the driver
119 that should be used with this device (for hotplug/coldplug)
121 /sys/bus/spi/devices/spiB.C ... symlink to the physical
122 spiB-C device
124 /sys/bus/spi/drivers/D ... driver for one or more spi*.* devices
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.
131 How does board-specific init code declare SPI devices?
132 ------------------------------------------------------
133 Linux needs several kinds of information to properly configure SPI devices.
134 That information is normally provided by board-specific code, even for
135 chips that do support some of automated discovery/enumeration.
139 The first kind of information is a list of what SPI controllers exist.
140 For System-on-Chip (SOC) based boards, these will usually be platform
141 devices, and the controller may need some platform_data in order to
142 operate properly. The "struct platform_device" will include resources
143 like the physical address of the controller's first register and its IRQ.
145 Platforms will often abstract the "register SPI controller" operation,
146 maybe coupling it with code to initialize pin configurations, so that
147 the arch/.../mach-*/board-*.c files for several boards can all share the
148 same basic controller setup code. This is because most SOCs have several
149 SPI-capable controllers, and only the ones actually usable on a given
150 board should normally be set up and registered.
152 So for example arch/.../mach-*/board-*.c files might have code like:
154 #include <asm/arch/spi.h> /* for mysoc_spi_data */
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 = { ... };
161 static __init board_init(void)
162 {
163 ...
164 /* this board only uses SPI controller #2 */
165 mysoc_register_spi(2, &pdata);
166 ...
167 }
169 And SOC-specific utility code might look something like:
171 #include <asm/arch/spi.h>
173 static struct platform_device spi2 = { ... };
175 void mysoc_register_spi(unsigned n, struct mysoc_spi_data *pdata)
176 {
177 struct mysoc_spi_data *pdata2;
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);
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 }
195 Notice how the platform_data for boards may be different, even if the
196 same SOC controller is used. For example, on one board SPI might use
197 an external clock, where another derives the SPI clock from current
198 settings of some master clock.
203 The second kind of information is a list of what SPI slave devices exist
204 on the target board, often with some board-specific data needed for the
205 driver to work correctly.
207 Normally your arch/.../mach-*/board-*.c files would provide a small table
208 listing the SPI devices on each board. (This would typically be only a
209 small handful.) That might look like:
211 static struct ads7846_platform_data ads_info = {
212 .vref_delay_usecs = 100,
213 .x_plate_ohms = 580,
214 .y_plate_ohms = 410,
215 };
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 };
229 Again, notice how board-specific information is provided; each chip may need
230 several types. This example shows generic constraints like the fastest SPI
231 clock to allow (a function of board voltage in this case) or how an IRQ pin
232 is wired, plus chip-specific constraints like an important delay that's
233 changed by the capacitance at one pin.
235 (There's also "controller_data", information that may be useful to the
236 controller driver. An example would be peripheral-specific DMA tuning
237 data or chipselect callbacks. This is stored in spi_device later.)
239 The board_info should provide enough information to let the system work
240 without the chip's driver being loaded. The most troublesome aspect of
241 that is likely the SPI_CS_HIGH bit in the spi_device.mode field, since
242 sharing a bus with a device that interprets chipselect "backwards" is
243 not possible.
245 Then your board initialization code would register that table with the SPI
246 infrastructure, so that it's available later when the SPI master controller
247 driver is registered:
249 spi_register_board_info(spi_board_info, ARRAY_SIZE(spi_board_info));
251 Like with other static board-specific setup, you won't unregister those.
253 The widely used "card" style computers bundle memory, cpu, and little else
254 onto a card that's maybe just thirty square centimeters. On such systems,
255 your arch/.../mach-.../board-*.c file would primarily provide information
256 about the devices on the mainboard into which such a card is plugged. That
257 certainly includes SPI devices hooked up through the card connectors!
262 Developer boards often play by different rules than product boards, and one
263 example is the potential need to hotplug SPI devices and/or controllers.
265 For those cases you might need to use use spi_busnum_to_master() to look
266 up the spi bus master, and will likely need spi_new_device() to provide the
267 board info based on the board that was hotplugged. Of course, you'd later
268 call at least spi_unregister_device() when that board is removed.
270 When Linux includes support for MMC/SD/SDIO/DataFlash cards through SPI, those
271 configurations will also be dynamic. Fortunately, those devices all support
272 basic device identification probes, so that support should hotplug normally.
275 How do I write an "SPI Protocol Driver"?
276 ----------------------------------------
277 All SPI drivers are currently kernel drivers. A userspace driver API
278 would just be another kernel driver, probably offering some lowlevel
279 access through aio_read(), aio_write(), and ioctl() calls and using the
280 standard userspace sysfs mechanisms to bind to a given SPI device.
282 SPI protocol drivers somewhat resemble platform device drivers:
284 static struct spi_driver CHIP_driver = {
285 .driver = {
286 .name = "CHIP",
287 .bus = &spi_bus_type,
288 .owner = THIS_MODULE,
289 },
291 .probe = CHIP_probe,
292 .remove = __devexit_p(CHIP_remove),
293 .suspend = CHIP_suspend,
294 .resume = CHIP_resume,
295 };
297 The driver core will autmatically attempt to bind this driver to any SPI
298 device whose board_info gave a modalias of "CHIP". Your probe() code
299 might look like this unless you're creating a class_device:
301 static int __devinit CHIP_probe(struct spi_device *spi)
302 {
303 struct CHIP *chip;
304 struct CHIP_platform_data *pdata;
306 /* assuming the driver requires board-specific data: */
307 pdata = &spi->dev.platform_data;
308 if (!pdata)
309 return -ENODEV;
311 /* get memory for driver's per-chip state */
312 chip = kzalloc(sizeof *chip, GFP_KERNEL);
313 if (!chip)
314 return -ENOMEM;
315 dev_set_drvdata(&spi->dev, chip);
317 ... etc
318 return 0;
319 }
321 As soon as it enters probe(), the driver may issue I/O requests to
322 the SPI device using "struct spi_message". When remove() returns,
323 the driver guarantees that it won't submit any more such messages.
325 - An spi_message is a sequence of of protocol operations, executed
326 as one atomic sequence. SPI driver controls include:
328 + when bidirectional reads and writes start ... by how its
329 sequence of spi_transfer requests is arranged;
331 + optionally defining short delays after transfers ... using
332 the spi_transfer.delay_usecs setting;
334 + whether the chipselect becomes inactive after a transfer and
335 any delay ... by using the spi_transfer.cs_change flag;
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.
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).
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.
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.
354 After any detected error, the chip is deselected and processing
355 of that spi_message is aborted.
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().
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.
369 Some drivers may need to modify spi_device characteristics like the
370 transfer mode, wordsize, or clock rate. This is done with spi_setup(),
371 which would normally be called from probe() before the first I/O is
372 done to the device.
374 While "spi_device" would be the bottom boundary of the driver, the
375 upper boundaries might include sysfs (especially for sensor readings),
376 the input layer, ALSA, networking, MTD, the character device framework,
377 or other Linux subsystems.
379 Note that there are two types of memory your driver must manage as part
380 of interacting with SPI devices.
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".
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.
391 If you like, spi_message_alloc() and spi_message_free() convenience
392 routines are available to allocate and zero-initialize an spi_message
393 with several transfers.
396 How do I write an "SPI Master Controller Driver"?
397 -------------------------------------------------
398 An SPI controller will probably be registered on the platform_bus; write
399 a driver to bind to the device, whichever bus is involved.
401 The main task of this type of driver is to provide an "spi_master".
402 Use spi_alloc_master() to allocate the master, and class_get_devdata()
403 to get the driver-private data allocated for that device.
405 struct spi_master *master;
406 struct CONTROLLER *c;
408 master = spi_alloc_master(dev, sizeof *c);
409 if (!master)
410 return -ENODEV;
412 c = class_get_devdata(&master->cdev);
414 The driver will initialize the fields of that spi_master, including the
415 bus number (maybe the same as the platform device ID) and three methods
416 used to interact with the SPI core and SPI protocol drivers. It will
417 also initialize its own internal state. (See below about bus numbering
418 and those methods.)
420 After you initialize the spi_master, then use spi_register_master() to
421 publish it to the rest of the system. At that time, device nodes for
422 the controller and any predeclared spi devices will be made available,
423 and the driver model core will take care of binding them to drivers.
425 If you need to remove your SPI controller driver, spi_unregister_master()
426 will reverse the effect of spi_register_master().
431 Bus numbering is important, since that's how Linux identifies a given
432 SPI bus (shared SCK, MOSI, MISO). Valid bus numbers start at zero. On
433 SOC systems, the bus numbers should match the numbers defined by the chip
434 manufacturer. For example, hardware controller SPI2 would be bus number 2,
435 and spi_board_info for devices connected to it would use that number.
437 If you don't have such hardware-assigned bus number, and for some reason
438 you can't just assign them, then provide a negative bus number. That will
439 then be replaced by a dynamically assigned number. You'd then need to treat
440 this as a non-static configuration (see above).
445 master->setup(struct spi_device *spi)
446 This sets up the device clock rate, SPI mode, and word sizes.
447 Drivers may change the defaults provided by board_info, and then
448 call spi_setup(spi) to invoke this routine. It may sleep.
450 master->transfer(struct spi_device *spi, struct spi_message *message)
451 This must not sleep. Its responsibility is arrange that the
452 transfer happens and its complete() callback is issued; the two
453 will normally happen later, after other transfers complete.
455 master->cleanup(struct spi_device *spi)
456 Your controller driver may use spi_device.controller_state to hold
457 state it dynamically associates with that device. If you do that,
458 be sure to provide the cleanup() method to free that state.
463 The bulk of the driver will be managing the I/O queue fed by transfer().
465 That queue could be purely conceptual. For example, a driver used only
466 for low-frequency sensor acess might be fine using synchronous PIO.
468 But the queue will probably be very real, using message->queue, PIO,
469 often DMA (especially if the root filesystem is in SPI flash), and
470 execution contexts like IRQ handlers, tasklets, or workqueues (such
471 as keventd). Your driver can be as fancy, or as simple, as you need.
472 Such a transfer() method would normally just add the message to a
473 queue, and then start some asynchronous transfer engine (unless it's
474 already running).
478 ---------
479 Contributors to Linux-SPI discussions include (in alphabetical order,
480 by last name):
482 David Brownell
483 Russell King
484 Dmitry Pervushin
485 Stephen Street
486 Mark Underwood
487 Andrew Victor
488 Vitaly Wool