drivers: create a pin control subsystem

This creates a subsystem for handling of pin control devices.
These are devices that control different aspects of package
pins.

Currently it handles pinmuxing, i.e. assigning electronic
functions to groups of pins on primarily PGA and BGA type of
chip packages which are common in embedded systems.

The plan is to also handle other I/O pin control aspects
such as biasing, driving, input properties such as
schmitt-triggering, load capacitance etc within this
subsystem, to remove a lot of ARM arch code as well as
feature-creepy GPIO drivers which are implementing the same
thing over and over again.

This is being done to depopulate the arch/arm/* directory
of such custom drivers and try to abstract the infrastructure
they all need. See the Documentation/pinctrl.txt file that is
part of this patch for more details.

ChangeLog v1->v2:

- Various minor fixes from Joe's and Stephens review comments
- Added a pinmux_config() that can invoke custom configuration
  with arbitrary data passed in or out to/from the pinmux driver

ChangeLog v2->v3:

- Renamed subsystem folder to "pinctrl" since we will likely
  want to keep other pin control such as biasing in this
  subsystem too, so let us keep to something generic even though
  we're mainly doing pinmux now.
- As a consequence, register pins as an abstract entity separate
  from the pinmux. The muxing functions will claim pins out of the
  pin pool and make sure they do not collide. Pins can now be
  named by the pinctrl core.
- Converted the pin lookup from a static array into a radix tree,
  I agreed with Grant Likely to try to avoid any static allocation
  (which is crap for device tree stuff) so I just rewrote this
  to be dynamic, just like irq number descriptors. The
  platform-wide definition of number of pins goes away - this is
  now just the sum total of the pins registered to the subsystem.
- Make sure mappings with only a function name and no device
  works properly.

ChangeLog v3->v4:

- Define a number space per controller instead of globally,
  Stephen and Grant requested the same thing so now maps need to
  define target controller, and the radix tree of pin descriptors
  is a property on each pin controller device.
- Add a compulsory pinctrl device entry to the pinctrl mapping
  table. This must match the pinctrl device, like "pinctrl.0"
- Split the file core.c in two: core.c and pinmux.c where the
  latter carry all pinmux stuff, the core is for generic pin
  control, and use local headers to access functionality between
  files. It is now possible to implement a "blank" pin controller
  without pinmux capabilities. This split will make new additions
  like pindrive.c, pinbias.c etc possible for combined drivers
  and chunks of functionality which is a GoodThing(TM).
- Rewrite the interaction with the GPIO subsystem - the pin
  controller descriptor now handles this by defining an offset
  into the GPIO numberspace for its handled pin range. This is
  used to look up the apropriate pin controller for a GPIO pin.
  Then that specific GPIO range is matched 1-1 for the target
  controller instance.
- Fixed a number of review comments from Joe Perches.
- Broke out a header file pinctrl.h for the core pin handling
  stuff that will be reused by other stuff than pinmux.
- Fixed some erroneous EXPORT() stuff.
- Remove mispatched U300 Kconfig and Makefile entries
- Fixed a number of review comments from Stephen Warren, not all
  of them - still WIP. But I think the new mapping that will
  specify which function goes to which pin mux controller address
  50% of your concerns (else beat me up).

ChangeLog v4->v5:

- Defined a "position" for each function, so the pin controller now
  tracks a function in a certain position, and the pinmux maps define
  what position you want the function in. (Feedback from Stephen
  Warren and Sascha Hauer).
- Since we now need to request a combined function+position from
  the machine mapping table that connect mux settings to drivers,
  it was extended with a position field and a name field. The
  name field is now used if you e.g. need to switch between two
  mux map settings at runtime.
- Switched from a class device to using struct bus_type for this
  subsystem. Verified sysfs functionality: seems to work fine.
  (Feedback from Arnd Bergmann and Greg Kroah-Hartman)
- Define a per pincontroller list of GPIO ranges from the GPIO
  pin space that can be handled by the pin controller. These can
  be added one by one at runtime. (Feedback from Barry Song)
- Expanded documentation of regulator_[get|enable|disable|put]
  semantics.
- Fixed a number of review comments from Barry Song. (Thanks!)

ChangeLog v5->v6:

- Create an abstract pin group concept that can sort pins into
  named and enumerated groups no matter what the use of these
  groups may be, one possible usecase is a group of pins being
  muxed in or so. The intention is however to also use these
  groups for other pin control activities.
- Make it compulsory for pinmux functions to associate with
  at least one group, so the abstract pin group concept is used
  to define the groups of pins affected by a pinmux function.
  The pinmux driver interface has been altered so as to enforce
  a function to list applicable groups per function.
- Provide an optional .group entry in the pinmux machine map
  so the map can select beteween different available groups
  to be used with a certain function.
- Consequent changes all over the place so that e.g. debugfs
  present reasonable information about the world.
- Drop the per-pin mux (*config) function in the pinmux_ops
  struct - I was afraid that this would start to be used for
  things totally unrelated to muxing, we can introduce that to
  the generic struct pinctrl_ops if needed. I want to keep
  muxing orthogonal to other pin control subjects and not mix
  these things up.

ChangeLog v6->v7:

- Make it possible to have several map entries matching the
  same device, pin controller and function, but using
  a different group, and alter the semantics so that
  pinmux_get() will pick all matching map entries, and
  store the associated groups in a list. The list will
  then be iterated over at pinmux_enable()/pinmux_disable()
  and corresponding driver functions called for each
  defined group. Notice that you're only allowed to map
  multiple *groups* to the same
  { device, pin controller, function } triplet, attempts
  to map the same device to multiple pin controllers will
  for example fail. This is hopefully the crucial feature
  requested by Stephen Warren.
- Add a pinmux hogging field to the pinmux mapping entries,
  and enable the pinmux core to hog pinmux map entries.
  This currently only works for pinmuxes without assigned
  devices as it looks now, but with device trees we can
  look up the corresponding struct device * entries when
  we register the pinmux driver, and have it hog each
  pinmux map in turn, for a simple approach to
  non-dynamic pin muxing. This addresses an issue from
  Grant Likely that the machine should take care of as
  much of the pinmux setup as possible, not the devices.
  By supplying a list of hogs, it can now instruct the
  core to take care of any static mappings.
- Switch pinmux group retrieveal function to grab an
  array of strings representing the groups rather than an
  array of unsigned and rewrite accordingly.
- Alter debugfs to show the grouplist handled by each
  pinmux. Also add a list of hogs.
- Dynamically allocate a struct pinmux at pinmux_get() and
  free it at pinmux_put(), then add these to the global
  list of pinmuxes active as we go along.
- Go over the list of pinmux maps at pinmux_get() time
  and repeatedly apply matches.
- Retrieve applicable groups per function from the driver
  as a string array rather than a unsigned array, then
  lookup the enumerators.
- Make the device to pinmux map a singleton - only allow the
  mapping table to be registered once and even tag the
  registration function with __init so it surely won't be
  abused.
- Create a separate debugfs file to view the pinmux map at
  runtime.
- Introduce a spin lock to the pin descriptor struct, lock it
  when modifying pin status entries. Reported by Stijn Devriendt.
- Fix up the documentation after review from Stephen Warren.
- Let the GPIO ranges give names as const char * instead of some
  fixed-length string.
- add a function to unregister GPIO ranges to mirror the
  registration function.
- Privatized the struct pinctrl_device and removed it from the
  <linux/pinctrl/pinctrl.h> API, the drivers do not need to know
  the members of this struct. It is now in the local header
  "core.h".
- Rename the concept of "anonymous" mux maps to "system" muxes
  and add convenience macros and documentation.

ChangeLog v7->v8:

- Delete the leftover pinmux_config() function from the
 <linux/pinctrl/pinmux.h> header.
- Fix a race condition found by Stijn Devriendt in pin_request()

ChangeLog v8->v9:

- Drop the bus_type and the sysfs attributes and all, we're not on
  the clear about how this should be used for e.g. userspace
  interfaces so let us save this for the future.
- Use the right name in MAINTAINERS, PIN CONTROL rather than
  PINMUX
- Don't kfree() the device state holder, let the .remove() callback
  handle this.
- Fix up numerous kerneldoc headers to have one line for the function
  description and more verbose documentation below the parameters

ChangeLog v9->v10:
- pinctrl: EXPORT_SYMBOL needs export.h, folded in a patch
  from Steven Rothwell
- fix pinctrl_register error handling, folded in a patch from
  Axel Lin
- Various fixes to documentation text so that it's consistent.
- Removed pointless comment from drivers/Kconfig
- Removed dependency on SYSFS since we removed the bus in
  v9.
- Renamed hopelessly abbreviated pctldev_* functions to the
  more verbose pinctrl_dev_*
- Drop mutex properly when looking up GPIO ranges
- Return NULL instead of ERR_PTR() errors on registration of
  pin controllers, using cast pointers is fragile. We can
  live without the detailed error codes for sure.

Cc: Stijn Devriendt <highguy@gmail.com>
Cc: Joe Perches <joe@perches.com>
Cc: Russell King <linux@arm.linux.org.uk>
Acked-by: Grant Likely <grant.likely@secretlab.ca>
Acked-by: Stephen Warren <swarren@nvidia.com>
Tested-by: Barry Song <21cnbao@gmail.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>

1 files changed, 950 insertions, 0 deletions

diff --git a/Documentation/pinctrl.txt b/Documentation/pinctrl.txt
new file mode 100644
index 000000000000..b04cb7d45a16
--- /dev/null
+++ b/Documentation/pinctrl.txt
@@ -0,0 +1,950 @@
+PINCTRL (PIN CONTROL) subsystem
+This document outlines the pin control subsystem in Linux
+
+This subsystem deals with:
+
+- Enumerating and naming controllable pins
+
+- Multiplexing of pins, pads, fingers (etc) see below for details
+
+The intention is to also deal with:
+
+- Software-controlled biasing and driving mode specific pins, such as
+  pull-up/down, open drain etc, load capacitance configuration when controlled
+  by software, etc.
+
+
+Top-level interface
+===================
+
+Definition of PIN CONTROLLER:
+
+- A pin controller is a piece of hardware, usually a set of registers, that
+  can control PINs. It may be able to multiplex, bias, set load capacitance,
+  set drive strength etc for individual pins or groups of pins.
+
+Definition of PIN:
+
+- PINS are equal to pads, fingers, balls or whatever packaging input or
+  output line you want to control and these are denoted by unsigned integers
+  in the range 0..maxpin. This numberspace is local to each PIN CONTROLLER, so
+  there may be several such number spaces in a system. This pin space may
+  be sparse - i.e. there may be gaps in the space with numbers where no
+  pin exists.
+
+When a PIN CONTROLLER is instatiated, it will register a descriptor to the
+pin control framework, and this descriptor contains an array of pin descriptors
+describing the pins handled by this specific pin controller.
+
+Here is an example of a PGA (Pin Grid Array) chip seen from underneath:
+
+        A   B   C   D   E   F   G   H
+
+   8    o   o   o   o   o   o   o   o
+
+   7    o   o   o   o   o   o   o   o
+
+   6    o   o   o   o   o   o   o   o
+
+   5    o   o   o   o   o   o   o   o
+
+   4    o   o   o   o   o   o   o   o
+
+   3    o   o   o   o   o   o   o   o
+
+   2    o   o   o   o   o   o   o   o
+
+   1    o   o   o   o   o   o   o   o
+
+To register a pin controller and name all the pins on this package we can do
+this in our driver:
+
+#include <linux/pinctrl/pinctrl.h>
+
+const struct pinctrl_pin_desc __refdata foo_pins[] = {
+      PINCTRL_PIN(0, "A1"),
+      PINCTRL_PIN(1, "A2"),
+      PINCTRL_PIN(2, "A3"),
+      ...
+      PINCTRL_PIN(61, "H6"),
+      PINCTRL_PIN(62, "H7"),
+      PINCTRL_PIN(63, "H8"),
+};
+
+static struct pinctrl_desc foo_desc = {
+        .name = "foo",
+        .pins = foo_pins,
+        .npins = ARRAY_SIZE(foo_pins),
+        .maxpin = 63,
+        .owner = THIS_MODULE,
+};
+
+int __init foo_probe(void)
+{
+        struct pinctrl_dev *pctl;
+
+        pctl = pinctrl_register(&foo_desc, <PARENT>, NULL);
+        if (IS_ERR(pctl))
+                pr_err("could not register foo pin driver\n");
+}
+
+Pins usually have fancier names than this. You can find these in the dataheet
+for your chip. Notice that the core pinctrl.h file provides a fancy macro
+called PINCTRL_PIN() to create the struct entries. As you can see I enumerated
+the pins from 0 in the upper left corner to 63 in the lower right corner,
+this enumeration was arbitrarily chosen, in practice you need to think
+through your numbering system so that it matches the layout of registers
+and such things in your driver, or the code may become complicated. You must
+also consider matching of offsets to the GPIO ranges that may be handled by
+the pin controller.
+
+For a padring with 467 pads, as opposed to actual pins, I used an enumeration
+like this, walking around the edge of the chip, which seems to be industry
+standard too (all these pads had names, too):
+
+
+     0 ..... 104
+   466        105
+     .        .
+     .        .
+   358        224
+    357 .... 225
+
+
+Pin groups
+==========
+
+Many controllers need to deal with groups of pins, so the pin controller
+subsystem has a mechanism for enumerating groups of pins and retrieving the
+actual enumerated pins that are part of a certain group.
+
+For example, say that we have a group of pins dealing with an SPI interface
+on { 0, 8, 16, 24 }, and a group of pins dealing with an I2C interface on pins
+on { 24, 25 }.
+
+These two groups are presented to the pin control subsystem by implementing
+some generic pinctrl_ops like this:
+
+#include <linux/pinctrl/pinctrl.h>
+
+struct foo_group {
+        const char *name;
+        const unsigned int *pins;
+        const unsigned num_pins;
+};
+
+static unsigned int spi0_pins[] = { 0, 8, 16, 24 };
+static unsigned int i2c0_pins[] = { 24, 25 };
+
+static const struct foo_group foo_groups[] = {
+        {
+                .name = "spi0_grp",
+                .pins = spi0_pins,
+                .num_pins = ARRAY_SIZE(spi0_pins),
+        },
+        {
+                .name = "i2c0_grp",
+                .pins = i2c0_pins,
+                .num_pins = ARRAY_SIZE(i2c0_pins),
+        },
+};
+
+
+static int foo_list_groups(struct pinctrl_dev *pctldev, unsigned selector)
+{
+        if (selector >= ARRAY_SIZE(foo_groups))
+                return -EINVAL;
+        return 0;
+}
+
+static const char *foo_get_group_name(struct pinctrl_dev *pctldev,
+                                       unsigned selector)
+{
+        return foo_groups[selector].name;
+}
+
+static int foo_get_group_pins(struct pinctrl_dev *pctldev, unsigned selector,
+                               unsigned ** const pins,
+                               unsigned * const num_pins)
+{
+        *pins = (unsigned *) foo_groups[selector].pins;
+        *num_pins = foo_groups[selector].num_pins;
+        return 0;
+}
+
+static struct pinctrl_ops foo_pctrl_ops = {
+        .list_groups = foo_list_groups,
+        .get_group_name = foo_get_group_name,
+        .get_group_pins = foo_get_group_pins,
+};
+
+
+static struct pinctrl_desc foo_desc = {
+       ...
+       .pctlops = &foo_pctrl_ops,
+};
+
+The pin control subsystem will call the .list_groups() function repeatedly
+beginning on 0 until it returns non-zero to determine legal selectors, then
+it will call the other functions to retrieve the name and pins of the group.
+Maintaining the data structure of the groups is up to the driver, this is
+just a simple example - in practice you may need more entries in your group
+structure, for example specific register ranges associated with each group
+and so on.
+
+
+Interaction with the GPIO subsystem
+===================================
+
+The GPIO drivers may want to perform operations of various types on the same
+physical pins that are also registered as pin controller pins.
+
+Since the pin controller subsystem have its pinspace local to the pin
+controller we need a mapping so that the pin control subsystem can figure out
+which pin controller handles control of a certain GPIO pin. Since a single
+pin controller may be muxing several GPIO ranges (typically SoCs that have
+one set of pins but internally several GPIO silicon blocks, each modeled as
+a struct gpio_chip) any number of GPIO ranges can be added to a pin controller
+instance like this:
+
+struct gpio_chip chip_a;
+struct gpio_chip chip_b;
+
+static struct pinctrl_gpio_range gpio_range_a = {
+        .name = "chip a",
+        .id = 0,
+        .base = 32,
+        .npins = 16,
+        .gc = &chip_a;
+};
+
+static struct pinctrl_gpio_range gpio_range_a = {
+        .name = "chip b",
+        .id = 0,
+        .base = 48,
+        .npins = 8,
+        .gc = &chip_b;
+};
+
+
+{
+        struct pinctrl_dev *pctl;
+        ...
+        pinctrl_add_gpio_range(pctl, &gpio_range_a);
+        pinctrl_add_gpio_range(pctl, &gpio_range_b);
+}
+
+So this complex system has one pin controller handling two different
+GPIO chips. Chip a has 16 pins and chip b has 8 pins. They are mapped in
+the global GPIO pin space at:
+
+chip a: [32 .. 47]
+chip b: [48 .. 55]
+
+When GPIO-specific functions in the pin control subsystem are called, these
+ranges will be used to look up the apropriate pin controller by inspecting
+and matching the pin to the pin ranges across all controllers. When a
+pin controller handling the matching range is found, GPIO-specific functions
+will be called on that specific pin controller.
+
+For all functionalities dealing with pin biasing, pin muxing etc, the pin
+controller subsystem will subtract the range's .base offset from the passed
+in gpio pin number, and pass that on to the pin control driver, so the driver
+will get an offset into its handled number range. Further it is also passed
+the range ID value, so that the pin controller knows which range it should
+deal with.
+
+For example: if a user issues pinctrl_gpio_set_foo(50), the pin control
+subsystem will find that the second range on this pin controller matches,
+subtract the base 48 and call the
+pinctrl_driver_gpio_set_foo(pinctrl, range, 2) where the latter function has
+this signature:
+
+int pinctrl_driver_gpio_set_foo(struct pinctrl_dev *pctldev,
+    struct pinctrl_gpio_range *rangeid,
+    unsigned offset);
+
+Now the driver knows that we want to do some GPIO-specific operation on the
+second GPIO range handled by "chip b", at offset 2 in that specific range.
+
+(If the GPIO subsystem is ever refactored to use a local per-GPIO controller
+pin space, this mapping will need to be augmented accordingly.)
+
+
+PINMUX interfaces
+=================
+
+These calls use the pinmux_* naming prefix.  No other calls should use that
+prefix.
+
+
+What is pinmuxing?
+==================
+
+PINMUX, also known as padmux, ballmux, alternate functions or mission modes
+is a way for chip vendors producing some kind of electrical packages to use
+a certain physical pin (ball, pad, finger, etc) for multiple mutually exclusive
+functions, depending on the application. By "application" in this context
+we usually mean a way of soldering or wiring the package into an electronic
+system, even though the framework makes it possible to also change the function
+at runtime.
+
+Here is an example of a PGA (Pin Grid Array) chip seen from underneath:
+
+        A   B   C   D   E   F   G   H
+      +---+
+   8  | o | o   o   o   o   o   o   o
+      |   |
+   7  | o | o   o   o   o   o   o   o
+      |   |
+   6  | o | o   o   o   o   o   o   o
+      +---+---+
+   5  | o | o | o   o   o   o   o   o
+      +---+---+               +---+
+   4    o   o   o   o   o   o | o | o
+                              |   |
+   3    o   o   o   o   o   o | o | o
+                              |   |
+   2    o   o   o   o   o   o | o | o
+      +-------+-------+-------+---+---+
+   1  | o   o | o   o | o   o | o | o |
+      +-------+-------+-------+---+---+
+
+This is not tetris. The game to think of is chess. Not all PGA/BGA packages
+are chessboard-like, big ones have "holes" in some arrangement according to
+different design patterns, but we're using this as a simple example. Of the
+pins you see some will be taken by things like a few VCC and GND to feed power
+to the chip, and quite a few will be taken by large ports like an external
+memory interface. The remaining pins will often be subject to pin multiplexing.
+
+The example 8x8 PGA package above will have pin numbers 0 thru 63 assigned to
+its physical pins. It will name the pins { A1, A2, A3 ... H6, H7, H8 } using
+pinctrl_register_pins() and a suitable data set as shown earlier.
+
+In this 8x8 BGA package the pins { A8, A7, A6, A5 } can be used as an SPI port
+(these are four pins: CLK, RXD, TXD, FRM). In that case, pin B5 can be used as
+some general-purpose GPIO pin. However, in another setting, pins { A5, B5 } can
+be used as an I2C port (these are just two pins: SCL, SDA). Needless to say,
+we cannot use the SPI port and I2C port at the same time. However in the inside
+of the package the silicon performing the SPI logic can alternatively be routed
+out on pins { G4, G3, G2, G1 }.
+
+On the botton row at { A1, B1, C1, D1, E1, F1, G1, H1 } we have something
+special - it's an external MMC bus that can be 2, 4 or 8 bits wide, and it will
+consume 2, 4 or 8 pins respectively, so either { A1, B1 } are taken or
+{ A1, B1, C1, D1 } or all of them. If we use all 8 bits, we cannot use the SPI
+port on pins { G4, G3, G2, G1 } of course.
+
+This way the silicon blocks present inside the chip can be multiplexed "muxed"
+out on different pin ranges. Often contemporary SoC (systems on chip) will
+contain several I2C, SPI, SDIO/MMC, etc silicon blocks that can be routed to
+different pins by pinmux settings.
+
+Since general-purpose I/O pins (GPIO) are typically always in shortage, it is
+common to be able to use almost any pin as a GPIO pin if it is not currently
+in use by some other I/O port.
+
+
+Pinmux conventions
+==================
+
+The purpose of the pinmux functionality in the pin controller subsystem is to
+abstract and provide pinmux settings to the devices you choose to instantiate
+in your machine configuration. It is inspired by the clk, GPIO and regulator
+subsystems, so devices will request their mux setting, but it's also possible
+to request a single pin for e.g. GPIO.
+
+Definitions:
+
+- FUNCTIONS can be switched in and out by a driver residing with the pin
+  control subsystem in the drivers/pinctrl/* directory of the kernel. The
+  pin control driver knows the possible functions. In the example above you can
+  identify three pinmux functions, one for spi, one for i2c and one for mmc.
+
+- FUNCTIONS are assumed to be enumerable from zero in a one-dimensional array.
+  In this case the array could be something like: { spi0, i2c0, mmc0 }
+  for the three available functions.
+
+- FUNCTIONS have PIN GROUPS as defined on the generic level - so a certain
+  function is *always* associated with a certain set of pin groups, could
+  be just a single one, but could also be many. In the example above the
+  function i2c is associated with the pins { A5, B5 }, enumerated as
+  { 24, 25 } in the controller pin space.
+
+  The Function spi is associated with pin groups { A8, A7, A6, A5 }
+  and { G4, G3, G2, G1 }, which are enumerated as { 0, 8, 16, 24 } and
+  { 38, 46, 54, 62 } respectively.
+
+  Group names must be unique per pin controller, no two groups on the same
+  controller may have the same name.
+
+- The combination of a FUNCTION and a PIN GROUP determine a certain function
+  for a certain set of pins. The knowledge of the functions and pin groups
+  and their machine-specific particulars are kept inside the pinmux driver,
+  from the outside only the enumerators are known, and the driver core can:
+
+  - Request the name of a function with a certain selector (>= 0)
+  - A list of groups associated with a certain function
+  - Request that a certain group in that list to be activated for a certain
+    function
+
+  As already described above, pin groups are in turn self-descriptive, so
+  the core will retrieve the actual pin range in a certain group from the
+  driver.
+
+- FUNCTIONS and GROUPS on a certain PIN CONTROLLER are MAPPED to a certain
+  device by the board file, device tree or similar machine setup configuration
+  mechanism, similar to how regulators are connected to devices, usually by
+  name. Defining a pin controller, function and group thus uniquely identify
+  the set of pins to be used by a certain device. (If only one possible group
+  of pins is available for the function, no group name need to be supplied -
+  the core will simply select the first and only group available.)
+
+  In the example case we can define that this particular machine shall
+  use device spi0 with pinmux function fspi0 group gspi0 and i2c0 on function
+  fi2c0 group gi2c0, on the primary pin controller, we get mappings
+  like these:
+
+  {
+    {"map-spi0", spi0, pinctrl0, fspi0, gspi0},
+    {"map-i2c0", i2c0, pinctrl0, fi2c0, gi2c0}
+  }
+
+  Every map must be assigned a symbolic name, pin controller and function.
+  The group is not compulsory - if it is omitted the first group presented by
+  the driver as applicable for the function will be selected, which is
+  useful for simple cases.
+
+  The device name is present in map entries tied to specific devices. Maps
+  without device names are referred to as SYSTEM pinmuxes, such as can be taken
+  by the machine implementation on boot and not tied to any specific device.
+
+  It is possible to map several groups to the same combination of device,
+  pin controller and function. This is for cases where a certain function on
+  a certain pin controller may use different sets of pins in different
+  configurations.
+
+- PINS for a certain FUNCTION using a certain PIN GROUP on a certain
+  PIN CONTROLLER are provided on a first-come first-serve basis, so if some
+  other device mux setting or GPIO pin request has already taken your physical
+  pin, you will be denied the use of it. To get (activate) a new setting, the
+  old one has to be put (deactivated) first.
+
+Sometimes the documentation and hardware registers will be oriented around
+pads (or "fingers") rather than pins - these are the soldering surfaces on the
+silicon inside the package, and may or may not match the actual number of
+pins/balls underneath the capsule. Pick some enumeration that makes sense to
+you. Define enumerators only for the pins you can control if that makes sense.
+
+Assumptions:
+
+We assume that the number possible function maps to pin groups is limited by
+the hardware. I.e. we assume that there is no system where any function can be
+mapped to any pin, like in a phone exchange. So the available pins groups for
+a certain function will be limited to a few choices (say up to eight or so),
+not hundreds or any amount of choices. This is the characteristic we have found
+by inspecting available pinmux hardware, and a necessary assumption since we
+expect pinmux drivers to present *all* possible function vs pin group mappings
+to the subsystem.
+
+
+Pinmux drivers
+==============
+
+The pinmux core takes care of preventing conflicts on pins and calling
+the pin controller driver to execute different settings.
+
+It is the responsibility of the pinmux driver to impose further restrictions
+(say for example infer electronic limitations due to load etc) to determine
+whether or not the requested function can actually be allowed, and in case it
+is possible to perform the requested mux setting, poke the hardware so that
+this happens.
+
+Pinmux drivers are required to supply a few callback functions, some are
+optional. Usually the enable() and disable() functions are implemented,
+writing values into some certain registers to activate a certain mux setting
+for a certain pin.
+
+A simple driver for the above example will work by setting bits 0, 1, 2, 3 or 4
+into some register named MUX to select a certain function with a certain
+group of pins would work something like this:
+
+#include <linux/pinctrl/pinctrl.h>
+#include <linux/pinctrl/pinmux.h>
+
+struct foo_group {
+        const char *name;
+        const unsigned int *pins;
+        const unsigned num_pins;
+};
+
+static const unsigned spi0_0_pins[] = { 0, 8, 16, 24 };
+static const unsigned spi0_1_pins[] = { 38, 46, 54, 62 };
+static const unsigned i2c0_pins[] = { 24, 25 };
+static const unsigned mmc0_1_pins[] = { 56, 57 };
+static const unsigned mmc0_2_pins[] = { 58, 59 };
+static const unsigned mmc0_3_pins[] = { 60, 61, 62, 63 };
+
+static const struct foo_group foo_groups[] = {
+        {
+                .name = "spi0_0_grp",
+                .pins = spi0_0_pins,
+                .num_pins = ARRAY_SIZE(spi0_0_pins),
+        },
+        {
+                .name = "spi0_1_grp",
+                .pins = spi0_1_pins,
+                .num_pins = ARRAY_SIZE(spi0_1_pins),
+        },
+        {
+                .name = "i2c0_grp",
+                .pins = i2c0_pins,
+                .num_pins = ARRAY_SIZE(i2c0_pins),
+        },
+        {
+                .name = "mmc0_1_grp",
+                .pins = mmc0_1_pins,
+                .num_pins = ARRAY_SIZE(mmc0_1_pins),
+        },
+        {
+                .name = "mmc0_2_grp",
+                .pins = mmc0_2_pins,
+                .num_pins = ARRAY_SIZE(mmc0_2_pins),
+        },
+        {
+                .name = "mmc0_3_grp",
+                .pins = mmc0_3_pins,
+                .num_pins = ARRAY_SIZE(mmc0_3_pins),
+        },
+};
+
+
+static int foo_list_groups(struct pinctrl_dev *pctldev, unsigned selector)
+{
+        if (selector >= ARRAY_SIZE(foo_groups))
+                return -EINVAL;
+        return 0;
+}
+
+static const char *foo_get_group_name(struct pinctrl_dev *pctldev,
+                                       unsigned selector)
+{
+        return foo_groups[selector].name;
+}
+
+static int foo_get_group_pins(struct pinctrl_dev *pctldev, unsigned selector,
+                               unsigned ** const pins,
+                               unsigned * const num_pins)
+{
+        *pins = (unsigned *) foo_groups[selector].pins;
+        *num_pins = foo_groups[selector].num_pins;
+        return 0;
+}
+
+static struct pinctrl_ops foo_pctrl_ops = {
+        .list_groups = foo_list_groups,
+        .get_group_name = foo_get_group_name,
+        .get_group_pins = foo_get_group_pins,
+};
+
+struct foo_pmx_func {
+        const char *name;
+        const char * const *groups;
+        const unsigned num_groups;
+};
+
+static const char * const spi0_groups[] = { "spi0_1_grp" };
+static const char * const i2c0_groups[] = { "i2c0_grp" };
+static const char * const mmc0_groups[] = { "mmc0_1_grp", "mmc0_2_grp",
+                                        "mmc0_3_grp" };
+
+static const struct foo_pmx_func foo_functions[] = {
+        {
+                .name = "spi0",
+                .groups = spi0_groups,
+                .num_groups = ARRAY_SIZE(spi0_groups),
+        },
+        {
+                .name = "i2c0",
+                .groups = i2c0_groups,
+                .num_groups = ARRAY_SIZE(i2c0_groups),
+        },
+        {
+                .name = "mmc0",
+                .groups = mmc0_groups,
+                .num_groups = ARRAY_SIZE(mmc0_groups),
+        },
+};
+
+int foo_list_funcs(struct pinctrl_dev *pctldev, unsigned selector)
+{
+        if (selector >= ARRAY_SIZE(foo_functions))
+                return -EINVAL;
+        return 0;
+}
+
+const char *foo_get_fname(struct pinctrl_dev *pctldev, unsigned selector)
+{
+        return myfuncs[selector].name;
+}
+
+static int foo_get_groups(struct pinctrl_dev *pctldev, unsigned selector,
+                          const char * const **groups,
+                          unsigned * const num_groups)
+{
+        *groups = foo_functions[selector].groups;
+        *num_groups = foo_functions[selector].num_groups;
+        return 0;
+}
+
+int foo_enable(struct pinctrl_dev *pctldev, unsigned selector,
+                unsigned group)
+{
+        u8 regbit = (1 << group);
+
+        writeb((readb(MUX)|regbit), MUX)
+        return 0;
+}
+
+int foo_disable(struct pinctrl_dev *pctldev, unsigned selector,
+                unsigned group)
+{
+        u8 regbit = (1 << group);
+
+        writeb((readb(MUX) & ~(regbit)), MUX)
+        return 0;
+}
+
+struct pinmux_ops foo_pmxops = {
+        .list_functions = foo_list_funcs,
+        .get_function_name = foo_get_fname,
+        .get_function_groups = foo_get_groups,
+        .enable = foo_enable,
+        .disable = foo_disable,
+};
+
+/* Pinmux operations are handled by some pin controller */
+static struct pinctrl_desc foo_desc = {
+        ...
+        .pctlops = &foo_pctrl_ops,
+        .pmxops = &foo_pmxops,
+};
+
+In the example activating muxing 0 and 1 at the same time setting bits
+0 and 1, uses one pin in common so they would collide.
+
+The beauty of the pinmux subsystem is that since it keeps track of all
+pins and who is using them, it will already have denied an impossible
+request like that, so the driver does not need to worry about such
+things - when it gets a selector passed in, the pinmux subsystem makes
+sure no other device or GPIO assignment is already using the selected
+pins. Thus bits 0 and 1 in the control register will never be set at the
+same time.
+
+All the above functions are mandatory to implement for a pinmux driver.
+
+
+Pinmux interaction with the GPIO subsystem
+==========================================
+
+The function list could become long, especially if you can convert every
+individual pin into a GPIO pin independent of any other pins, and then try
+the approach to define every pin as a function.
+
+In this case, the function array would become 64 entries for each GPIO
+setting and then the device functions.
+
+For this reason there is an additional function a pinmux driver can implement
+to enable only GPIO on an individual pin: .gpio_request_enable(). The same
+.free() function as for other functions is assumed to be usable also for
+GPIO pins.
+
+This function will pass in the affected GPIO range identified by the pin
+controller core, so you know which GPIO pins are being affected by the request
+operation.
+
+Alternatively it is fully allowed to use named functions for each GPIO
+pin, the pinmux_request_gpio() will attempt to obtain the function "gpioN"
+where "N" is the global GPIO pin number if no special GPIO-handler is
+registered.
+
+
+Pinmux board/machine configuration
+==================================
+
+Boards and machines define how a certain complete running system is put
+together, including how GPIOs and devices are muxed, how regulators are
+constrained and how the clock tree looks. Of course pinmux settings are also
+part of this.
+
+A pinmux config for a machine looks pretty much like a simple regulator
+configuration, so for the example array above we want to enable i2c and
+spi on the second function mapping:
+
+#include <linux/pinctrl/machine.h>
+
+static struct pinmux_map pmx_mapping[] = {
+        {
+                .ctrl_dev_name = "pinctrl.0",
+                .function = "spi0",
+                .dev_name = "foo-spi.0",
+        },
+        {
+                .ctrl_dev_name = "pinctrl.0",
+                .function = "i2c0",
+                .dev_name = "foo-i2c.0",
+        },
+        {
+                .ctrl_dev_name = "pinctrl.0",
+                .function = "mmc0",
+                .dev_name = "foo-mmc.0",
+        },
+};
+
+The dev_name here matches to the unique device name that can be used to look
+up the device struct (just like with clockdev or regulators). The function name
+must match a function provided by the pinmux driver handling this pin range.
+
+As you can see we may have several pin controllers on the system and thus
+we need to specify which one of them that contain the functions we wish
+to map. The map can also use struct device * directly, so there is no
+inherent need to use strings to specify .dev_name or .ctrl_dev_name, these
+are for the situation where you do not have a handle to the struct device *,
+for example if they are not yet instantiated or cumbersome to obtain.
+
+You register this pinmux mapping to the pinmux subsystem by simply:
+
+       ret = pinmux_register_mappings(&pmx_mapping, ARRAY_SIZE(pmx_mapping));
+
+Since the above construct is pretty common there is a helper macro to make
+it even more compact which assumes you want to use pinctrl.0 and position
+0 for mapping, for example:
+
+static struct pinmux_map pmx_mapping[] = {
+       PINMUX_MAP_PRIMARY("I2CMAP", "i2c0", "foo-i2c.0"),
+};
+
+
+Complex mappings
+================
+
+As it is possible to map a function to different groups of pins an optional
+.group can be specified like this:
+
+...
+{
+        .name = "spi0-pos-A",
+        .ctrl_dev_name = "pinctrl.0",
+        .function = "spi0",
+        .group = "spi0_0_grp",
+        .dev_name = "foo-spi.0",
+},
+{
+        .name = "spi0-pos-B",
+        .ctrl_dev_name = "pinctrl.0",
+        .function = "spi0",
+        .group = "spi0_1_grp",
+        .dev_name = "foo-spi.0",
+},
+...
+
+This example mapping is used to switch between two positions for spi0 at
+runtime, as described further below under the heading "Runtime pinmuxing".
+
+Further it is possible to match several groups of pins to the same function
+for a single device, say for example in the mmc0 example above, where you can
+additively expand the mmc0 bus from 2 to 4 to 8 pins. If we want to use all
+three groups for a total of 2+2+4 = 8 pins (for an 8-bit MMC bus as is the
+case), we define a mapping like this:
+
+...
+{
+        .name "2bit"
+        .ctrl_dev_name = "pinctrl.0",
+        .function = "mmc0",
+        .group = "mmc0_0_grp",
+        .dev_name = "foo-mmc.0",
+},
+{
+        .name "4bit"
+        .ctrl_dev_name = "pinctrl.0",
+        .function = "mmc0",
+        .group = "mmc0_0_grp",
+        .dev_name = "foo-mmc.0",
+},
+{
+        .name "4bit"
+        .ctrl_dev_name = "pinctrl.0",
+        .function = "mmc0",
+        .group = "mmc0_1_grp",
+        .dev_name = "foo-mmc.0",
+},
+{
+        .name "8bit"
+        .ctrl_dev_name = "pinctrl.0",
+        .function = "mmc0",
+        .group = "mmc0_0_grp",
+        .dev_name = "foo-mmc.0",
+},
+{
+        .name "8bit"
+        .ctrl_dev_name = "pinctrl.0",
+        .function = "mmc0",
+        .group = "mmc0_1_grp",
+        .dev_name = "foo-mmc.0",
+},
+{
+        .name "8bit"
+        .ctrl_dev_name = "pinctrl.0",
+        .function = "mmc0",
+        .group = "mmc0_2_grp",
+        .dev_name = "foo-mmc.0",
+},
+...
+
+The result of grabbing this mapping from the device with something like
+this (see next paragraph):
+
+        pmx = pinmux_get(&device, "8bit");
+
+Will be that you activate all the three bottom records in the mapping at
+once. Since they share the same name, pin controller device, funcion and
+device, and since we allow multiple groups to match to a single device, they
+all get selected, and they all get enabled and disable simultaneously by the
+pinmux core.
+
+
+Pinmux requests from drivers
+============================
+
+Generally it is discouraged to let individual drivers get and enable pinmuxes.
+So if possible, handle the pinmuxes in platform code or some other place where
+you have access to all the affected struct device * pointers. In some cases
+where a driver needs to switch between different mux mappings at runtime
+this is not possible.
+
+A driver may request a certain mux to be activated, usually just the default
+mux like this:
+
+#include <linux/pinctrl/pinmux.h>
+
+struct foo_state {
+       struct pinmux *pmx;
+       ...
+};
+
+foo_probe()
+{
+        /* Allocate a state holder named "state" etc */
+        struct pinmux pmx;
+
+        pmx = pinmux_get(&device, NULL);
+        if IS_ERR(pmx)
+                return PTR_ERR(pmx);
+        pinmux_enable(pmx);
+
+        state->pmx = pmx;
+}
+
+foo_remove()
+{
+        pinmux_disable(state->pmx);
+        pinmux_put(state->pmx);
+}
+
+If you want to grab a specific mux mapping and not just the first one found for
+this device you can specify a specific mapping name, for example in the above
+example the second i2c0 setting: pinmux_get(&device, "spi0-pos-B");
+
+This get/enable/disable/put sequence can just as well be handled by bus drivers
+if you don't want each and every driver to handle it and you know the
+arrangement on your bus.
+
+The semantics of the get/enable respective disable/put is as follows:
+
+- pinmux_get() is called in process context to reserve the pins affected with
+  a certain mapping and set up the pinmux core and the driver. It will allocate
+  a struct from the kernel memory to hold the pinmux state.
+
+- pinmux_enable()/pinmux_disable() is quick and can be called from fastpath
+  (irq context) when you quickly want to set up/tear down the hardware muxing
+  when running a device driver. Usually it will just poke some values into a
+  register.
+
+- pinmux_disable() is called in process context to tear down the pin requests
+  and release the state holder struct for the mux setting.
+
+Usually the pinmux core handled the get/put pair and call out to the device
+drivers bookkeeping operations, like checking available functions and the
+associated pins, whereas the enable/disable pass on to the pin controller
+driver which takes care of activating and/or deactivating the mux setting by
+quickly poking some registers.
+
+The pins are allocated for your device when you issue the pinmux_get() call,
+after this you should be able to see this in the debugfs listing of all pins.
+
+
+System pinmux hogging
+=====================
+
+A system pinmux map entry, i.e. a pinmux setting that does not have a device
+associated with it, can be hogged by the core when the pin controller is
+registered. This means that the core will attempt to call pinmux_get() and
+pinmux_enable() on it immediately after the pin control device has been
+registered.
+
+This is enabled by simply setting the .hog_on_boot field in the map to true,
+like this:
+
+{
+        .name "POWERMAP"
+        .ctrl_dev_name = "pinctrl.0",
+        .function = "power_func",
+        .hog_on_boot = true,
+},
+
+Since it may be common to request the core to hog a few always-applicable
+mux settings on the primary pin controller, there is a convenience macro for
+this:
+
+PINMUX_MAP_PRIMARY_SYS_HOG("POWERMAP", "power_func")
+
+This gives the exact same result as the above construction.
+
+
+Runtime pinmuxing
+=================
+
+It is possible to mux a certain function in and out at runtime, say to move
+an SPI port from one set of pins to another set of pins. Say for example for
+spi0 in the example above, we expose two different groups of pins for the same
+function, but with different named in the mapping as described under
+"Advanced mapping" above. So we have two mappings named "spi0-pos-A" and
+"spi0-pos-B".
+
+This snippet first muxes the function in the pins defined by group A, enables
+it, disables and releases it, and muxes it in on the pins defined by group B:
+
+foo_switch()
+{
+        struct pinmux pmx;
+
+        /* Enable on position A */
+        pmx = pinmux_get(&device, "spi0-pos-A");
+        if IS_ERR(pmx)
+                return PTR_ERR(pmx);
+        pinmux_enable(pmx);
+
+        /* This releases the pins again */
+        pinmux_disable(pmx);
+        pinmux_put(pmx);
+
+        /* Enable on position B */
+        pmx = pinmux_get(&device, "spi0-pos-B");
+        if IS_ERR(pmx)
+                return PTR_ERR(pmx);
+        pinmux_enable(pmx);
+        ...
+}
+
+The above has to be done from process context.