src/kernel/linux/v4.19/Documentation/acpi/enumeration.txt

ACPI based device enumeration
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
ACPI 5 introduced a set of new resources (UartTSerialBus, I2cSerialBus,
SpiSerialBus, GpioIo and GpioInt) which can be used in enumerating slave
devices behind serial bus controllers.

In addition we are starting to see peripherals integrated in the
SoC/Chipset to appear only in ACPI namespace. These are typically devices
that are accessed through memory-mapped registers.

In order to support this and re-use the existing drivers as much as
possible we decided to do following:

	o Devices that have no bus connector resource are represented as
	  platform devices.

	o Devices behind real busses where there is a connector resource
	  are represented as struct spi_device or struct i2c_device
	  (standard UARTs are not busses so there is no struct uart_device).

As both ACPI and Device Tree represent a tree of devices (and their
resources) this implementation follows the Device Tree way as much as
possible.

The ACPI implementation enumerates devices behind busses (platform, SPI and
I2C), creates the physical devices and binds them to their ACPI handle in
the ACPI namespace.

This means that when ACPI_HANDLE(dev) returns non-NULL the device was
enumerated from ACPI namespace. This handle can be used to extract other
device-specific configuration. There is an example of this below.

Platform bus support
~~~~~~~~~~~~~~~~~~~~
Since we are using platform devices to represent devices that are not
connected to any physical bus we only need to implement a platform driver
for the device and add supported ACPI IDs. If this same IP-block is used on
some other non-ACPI platform, the driver might work out of the box or needs
some minor changes.

Adding ACPI support for an existing driver should be pretty
straightforward. Here is the simplest example:

	#ifdef CONFIG_ACPI
	static const struct acpi_device_id mydrv_acpi_match[] = {
		/* ACPI IDs here */
		{ }
	};
	MODULE_DEVICE_TABLE(acpi, mydrv_acpi_match);
	#endif

	static struct platform_driver my_driver = {
		...
		.driver = {
			.acpi_match_table = ACPI_PTR(mydrv_acpi_match),
		},
	};

If the driver needs to perform more complex initialization like getting and
configuring GPIOs it can get its ACPI handle and extract this information
from ACPI tables.

DMA support
~~~~~~~~~~~
DMA controllers enumerated via ACPI should be registered in the system to
provide generic access to their resources. For example, a driver that would
like to be accessible to slave devices via generic API call
dma_request_slave_channel() must register itself at the end of the probe
function like this:

	err = devm_acpi_dma_controller_register(dev, xlate_func, dw);
	/* Handle the error if it's not a case of !CONFIG_ACPI */

and implement custom xlate function if needed (usually acpi_dma_simple_xlate()
is enough) which converts the FixedDMA resource provided by struct
acpi_dma_spec into the corresponding DMA channel. A piece of code for that case
could look like:

	#ifdef CONFIG_ACPI
	struct filter_args {
		/* Provide necessary information for the filter_func */
		...
	};

	static bool filter_func(struct dma_chan *chan, void *param)
	{
		/* Choose the proper channel */
		...
	}

	static struct dma_chan *xlate_func(struct acpi_dma_spec *dma_spec,
			struct acpi_dma *adma)
	{
		dma_cap_mask_t cap;
		struct filter_args args;

		/* Prepare arguments for filter_func */
		...
		return dma_request_channel(cap, filter_func, &args);
	}
	#else
	static struct dma_chan *xlate_func(struct acpi_dma_spec *dma_spec,
			struct acpi_dma *adma)
	{
		return NULL;
	}
	#endif

dma_request_slave_channel() will call xlate_func() for each registered DMA
controller. In the xlate function the proper channel must be chosen based on
information in struct acpi_dma_spec and the properties of the controller
provided by struct acpi_dma.

Clients must call dma_request_slave_channel() with the string parameter that
corresponds to a specific FixedDMA resource. By default "tx" means the first
entry of the FixedDMA resource array, "rx" means the second entry. The table
below shows a layout:

	Device (I2C0)
	{
		...
		Method (_CRS, 0, NotSerialized)
		{
			Name (DBUF, ResourceTemplate ()
			{
				FixedDMA (0x0018, 0x0004, Width32bit, _Y48)
				FixedDMA (0x0019, 0x0005, Width32bit, )
			})
		...
		}
	}

So, the FixedDMA with request line 0x0018 is "tx" and next one is "rx" in
this example.

In robust cases the client unfortunately needs to call
acpi_dma_request_slave_chan_by_index() directly and therefore choose the
specific FixedDMA resource by its index.

SPI serial bus support
~~~~~~~~~~~~~~~~~~~~~~
Slave devices behind SPI bus have SpiSerialBus resource attached to them.
This is extracted automatically by the SPI core and the slave devices are
enumerated once spi_register_master() is called by the bus driver.

Here is what the ACPI namespace for a SPI slave might look like:

	Device (EEP0)
	{
		Name (_ADR, 1)
		Name (_CID, Package() {
			"ATML0025",
			"AT25",
		})
		...
		Method (_CRS, 0, NotSerialized)
		{
			SPISerialBus(1, PolarityLow, FourWireMode, 8,
				ControllerInitiated, 1000000, ClockPolarityLow,
				ClockPhaseFirst, "\\_SB.PCI0.SPI1",)
		}
		...

The SPI device drivers only need to add ACPI IDs in a similar way than with
the platform device drivers. Below is an example where we add ACPI support
to at25 SPI eeprom driver (this is meant for the above ACPI snippet):

	#ifdef CONFIG_ACPI
	static const struct acpi_device_id at25_acpi_match[] = {
		{ "AT25", 0 },
		{ },
	};
	MODULE_DEVICE_TABLE(acpi, at25_acpi_match);
	#endif

	static struct spi_driver at25_driver = {
		.driver = {
			...
			.acpi_match_table = ACPI_PTR(at25_acpi_match),
		},
	};

Note that this driver actually needs more information like page size of the
eeprom etc. but at the time writing this there is no standard way of
passing those. One idea is to return this in _DSM method like:

	Device (EEP0)
	{
		...
		Method (_DSM, 4, NotSerialized)
		{
			Store (Package (6)
			{
				"byte-len", 1024,
				"addr-mode", 2,
				"page-size, 32
			}, Local0)

			// Check UUIDs etc.

			Return (Local0)
		}

Then the at25 SPI driver can get this configuration by calling _DSM on its
ACPI handle like:

	struct acpi_buffer output = { ACPI_ALLOCATE_BUFFER, NULL };
	struct acpi_object_list input;
	acpi_status status;

	/* Fill in the input buffer */

	status = acpi_evaluate_object(ACPI_HANDLE(&spi->dev), "_DSM",
				      &input, &output);
	if (ACPI_FAILURE(status))
		/* Handle the error */

	/* Extract the data here */

	kfree(output.pointer);

I2C serial bus support
~~~~~~~~~~~~~~~~~~~~~~
The slaves behind I2C bus controller only need to add the ACPI IDs like
with the platform and SPI drivers. The I2C core automatically enumerates
any slave devices behind the controller device once the adapter is
registered.

Below is an example of how to add ACPI support to the existing mpu3050
input driver:

	#ifdef CONFIG_ACPI
	static const struct acpi_device_id mpu3050_acpi_match[] = {
		{ "MPU3050", 0 },
		{ },
	};
	MODULE_DEVICE_TABLE(acpi, mpu3050_acpi_match);
	#endif

	static struct i2c_driver mpu3050_i2c_driver = {
		.driver	= {
			.name	= "mpu3050",
			.owner	= THIS_MODULE,
			.pm	= &mpu3050_pm,
			.of_match_table = mpu3050_of_match,
			.acpi_match_table = ACPI_PTR(mpu3050_acpi_match),
		},
		.probe		= mpu3050_probe,
		.remove		= mpu3050_remove,
		.id_table	= mpu3050_ids,
	};

GPIO support
~~~~~~~~~~~~
ACPI 5 introduced two new resources to describe GPIO connections: GpioIo
and GpioInt. These resources can be used to pass GPIO numbers used by
the device to the driver. ACPI 5.1 extended this with _DSD (Device
Specific Data) which made it possible to name the GPIOs among other things.

For example:

Device (DEV)
{
	Method (_CRS, 0, NotSerialized)
	{
		Name (SBUF, ResourceTemplate()
		{
			...
			// Used to power on/off the device
			GpioIo (Exclusive, PullDefault, 0x0000, 0x0000,
				IoRestrictionOutputOnly, "\\_SB.PCI0.GPI0",
				0x00, ResourceConsumer,,)
			{
				// Pin List
				0x0055
			}

			// Interrupt for the device
			GpioInt (Edge, ActiveHigh, ExclusiveAndWake, PullNone,
				 0x0000, "\\_SB.PCI0.GPI0", 0x00, ResourceConsumer,,)
			{
				// Pin list
				0x0058
			}

			...

		}

		Return (SBUF)
	}

	// ACPI 5.1 _DSD used for naming the GPIOs
	Name (_DSD, Package ()
	{
		ToUUID("daffd814-6eba-4d8c-8a91-bc9bbf4aa301"),
		Package ()
		{
			Package () {"power-gpios", Package() {^DEV, 0, 0, 0 }},
			Package () {"irq-gpios", Package() {^DEV, 1, 0, 0 }},
		}
	})
	...

These GPIO numbers are controller relative and path "\\_SB.PCI0.GPI0"
specifies the path to the controller. In order to use these GPIOs in Linux
we need to translate them to the corresponding Linux GPIO descriptors.

There is a standard GPIO API for that and is documented in
Documentation/gpio/.

In the above example we can get the corresponding two GPIO descriptors with
a code like this:

	#include <linux/gpio/consumer.h>
	...

	struct gpio_desc *irq_desc, *power_desc;

	irq_desc = gpiod_get(dev, "irq");
	if (IS_ERR(irq_desc))
		/* handle error */

	power_desc = gpiod_get(dev, "power");
	if (IS_ERR(power_desc))
		/* handle error */

	/* Now we can use the GPIO descriptors */

There are also devm_* versions of these functions which release the
descriptors once the device is released.

See Documentation/acpi/gpio-properties.txt for more information about the
_DSD binding related to GPIOs.

MFD devices
~~~~~~~~~~~
The MFD devices register their children as platform devices. For the child
devices there needs to be an ACPI handle that they can use to reference
parts of the ACPI namespace that relate to them. In the Linux MFD subsystem
we provide two ways:

	o The children share the parent ACPI handle.
	o The MFD cell can specify the ACPI id of the device.

For the first case, the MFD drivers do not need to do anything. The
resulting child platform device will have its ACPI_COMPANION() set to point
to the parent device.

If the ACPI namespace has a device that we can match using an ACPI id or ACPI
adr, the cell should be set like:

	static struct mfd_cell_acpi_match my_subdevice_cell_acpi_match = {
		.pnpid = "XYZ0001",
		.adr = 0,
	};

	static struct mfd_cell my_subdevice_cell = {
		.name = "my_subdevice",
		/* set the resources relative to the parent */
		.acpi_match = &my_subdevice_cell_acpi_match,
	};

The ACPI id "XYZ0001" is then used to lookup an ACPI device directly under
the MFD device and if found, that ACPI companion device is bound to the
resulting child platform device.

Device Tree namespace link device ID
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The Device Tree protocol uses device identification based on the "compatible"
property whose value is a string or an array of strings recognized as device
identifiers by drivers and the driver core.  The set of all those strings may be
regarded as a device identification namespace analogous to the ACPI/PNP device
ID namespace.  Consequently, in principle it should not be necessary to allocate
a new (and arguably redundant) ACPI/PNP device ID for a devices with an existing
identification string in the Device Tree (DT) namespace, especially if that ID
is only needed to indicate that a given device is compatible with another one,
presumably having a matching driver in the kernel already.

In ACPI, the device identification object called _CID (Compatible ID) is used to
list the IDs of devices the given one is compatible with, but those IDs must
belong to one of the namespaces prescribed by the ACPI specification (see
Section 6.1.2 of ACPI 6.0 for details) and the DT namespace is not one of them.
Moreover, the specification mandates that either a _HID or an _ADR identification
object be present for all ACPI objects representing devices (Section 6.1 of ACPI
6.0).  For non-enumerable bus types that object must be _HID and its value must
be a device ID from one of the namespaces prescribed by the specification too.

The special DT namespace link device ID, PRP0001, provides a means to use the
existing DT-compatible device identification in ACPI and to satisfy the above
requirements following from the ACPI specification at the same time.  Namely,
if PRP0001 is returned by _HID, the ACPI subsystem will look for the
"compatible" property in the device object's _DSD and will use the value of that
property to identify the corresponding device in analogy with the original DT
device identification algorithm.  If the "compatible" property is not present
or its value is not valid, the device will not be enumerated by the ACPI
subsystem.  Otherwise, it will be enumerated automatically as a platform device
(except when an I2C or SPI link from the device to its parent is present, in
which case the ACPI core will leave the device enumeration to the parent's
driver) and the identification strings from the "compatible" property value will
be used to find a driver for the device along with the device IDs listed by _CID
(if present).

Analogously, if PRP0001 is present in the list of device IDs returned by _CID,
the identification strings listed by the "compatible" property value (if present
and valid) will be used to look for a driver matching the device, but in that
case their relative priority with respect to the other device IDs listed by
_HID and _CID depends on the position of PRP0001 in the _CID return package.
Specifically, the device IDs returned by _HID and preceding PRP0001 in the _CID
return package will be checked first.  Also in that case the bus type the device
will be enumerated to depends on the device ID returned by _HID.

It is valid to define device objects with a _HID returning PRP0001 and without
the "compatible" property in the _DSD or a _CID as long as one of their
ancestors provides a _DSD with a valid "compatible" property.  Such device
objects are then simply regarded as additional "blocks" providing hierarchical
configuration information to the driver of the composite ancestor device.

However, PRP0001 can only be returned from either _HID or _CID of a device
object if all of the properties returned by the _DSD associated with it (either
the _DSD of the device object itself or the _DSD of its ancestor in the
"composite device" case described above) can be used in the ACPI environment.
Otherwise, the _DSD itself is regarded as invalid and therefore the "compatible"
property returned by it is meaningless.

Refer to DSD-properties-rules.txt for more information.