xref: /openbmc/u-boot/doc/driver-model/README.txt (revision 1005ccda)
1Driver Model
2============
3
4This README contains high-level information about driver model, a unified
5way of declaring and accessing drivers in U-Boot. The original work was done
6by:
7
8   Marek Vasut <marex@denx.de>
9   Pavel Herrmann <morpheus.ibis@gmail.com>
10   Viktor Křivák <viktor.krivak@gmail.com>
11   Tomas Hlavacek <tmshlvck@gmail.com>
12
13This has been both simplified and extended into the current implementation
14by:
15
16   Simon Glass <sjg@chromium.org>
17
18
19Terminology
20-----------
21
22Uclass - a group of devices which operate in the same way. A uclass provides
23	a way of accessing individual devices within the group, but always
24	using the same interface. For example a GPIO uclass provides
25	operations for get/set value. An I2C uclass may have 10 I2C ports,
26	4 with one driver, and 6 with another.
27
28Driver - some code which talks to a peripheral and presents a higher-level
29	interface to it.
30
31Device - an instance of a driver, tied to a particular port or peripheral.
32
33
34How to try it
35-------------
36
37Build U-Boot sandbox and run it:
38
39   make sandbox_defconfig
40   make
41   ./u-boot -d u-boot.dtb
42
43   (type 'reset' to exit U-Boot)
44
45
46There is a uclass called 'demo'. This uclass handles
47saying hello, and reporting its status. There are two drivers in this
48uclass:
49
50   - simple: Just prints a message for hello, doesn't implement status
51   - shape: Prints shapes and reports number of characters printed as status
52
53The demo class is pretty simple, but not trivial. The intention is that it
54can be used for testing, so it will implement all driver model features and
55provide good code coverage of them. It does have multiple drivers, it
56handles parameter data and platdata (data which tells the driver how
57to operate on a particular platform) and it uses private driver data.
58
59To try it, see the example session below:
60
61=>demo hello 1
62Hello '@' from 07981110: red 4
63=>demo status 2
64Status: 0
65=>demo hello 2
66g
67r@
68e@@
69e@@@
70n@@@@
71g@@@@@
72=>demo status 2
73Status: 21
74=>demo hello 4 ^
75  y^^^
76 e^^^^^
77l^^^^^^^
78l^^^^^^^
79 o^^^^^
80  w^^^
81=>demo status 4
82Status: 36
83=>
84
85
86Running the tests
87-----------------
88
89The intent with driver model is that the core portion has 100% test coverage
90in sandbox, and every uclass has its own test. As a move towards this, tests
91are provided in test/dm. To run them, try:
92
93   ./test/dm/test-dm.sh
94
95You should see something like this:
96
97    <...U-Boot banner...>
98    Running 53 driver model tests
99    Test: dm_test_autobind
100    Test: dm_test_autoprobe
101    Test: dm_test_bus_child_post_bind
102    Test: dm_test_bus_child_post_bind_uclass
103    Test: dm_test_bus_child_pre_probe_uclass
104    Test: dm_test_bus_children
105    Device 'c-test@0': seq 0 is in use by 'd-test'
106    Device 'c-test@1': seq 1 is in use by 'f-test'
107    Test: dm_test_bus_children_funcs
108    Test: dm_test_bus_children_iterators
109    Test: dm_test_bus_parent_data
110    Test: dm_test_bus_parent_data_uclass
111    Test: dm_test_bus_parent_ops
112    Test: dm_test_bus_parent_platdata
113    Test: dm_test_bus_parent_platdata_uclass
114    Test: dm_test_children
115    Test: dm_test_device_get_uclass_id
116    Test: dm_test_eth
117    Using eth@10002000 device
118    Using eth@10003000 device
119    Using eth@10004000 device
120    Test: dm_test_eth_alias
121    Using eth@10002000 device
122    Using eth@10004000 device
123    Using eth@10002000 device
124    Using eth@10003000 device
125    Test: dm_test_eth_prime
126    Using eth@10003000 device
127    Using eth@10002000 device
128    Test: dm_test_eth_rotate
129
130    Error: eth@10004000 address not set.
131
132    Error: eth@10004000 address not set.
133    Using eth@10002000 device
134
135    Error: eth@10004000 address not set.
136
137    Error: eth@10004000 address not set.
138    Using eth@10004000 device
139    Test: dm_test_fdt
140    Test: dm_test_fdt_offset
141    Test: dm_test_fdt_pre_reloc
142    Test: dm_test_fdt_uclass_seq
143    Test: dm_test_gpio
144    extra-gpios: get_value: error: gpio b5 not reserved
145    Test: dm_test_gpio_anon
146    Test: dm_test_gpio_copy
147    Test: dm_test_gpio_leak
148    extra-gpios: get_value: error: gpio b5 not reserved
149    Test: dm_test_gpio_phandles
150    Test: dm_test_gpio_requestf
151    Test: dm_test_i2c_bytewise
152    Test: dm_test_i2c_find
153    Test: dm_test_i2c_offset
154    Test: dm_test_i2c_offset_len
155    Test: dm_test_i2c_probe_empty
156    Test: dm_test_i2c_read_write
157    Test: dm_test_i2c_speed
158    Test: dm_test_leak
159    Test: dm_test_lifecycle
160    Test: dm_test_net_retry
161    Using eth@10004000 device
162    Using eth@10002000 device
163    Using eth@10004000 device
164    Test: dm_test_operations
165    Test: dm_test_ordering
166    Test: dm_test_pci_base
167    Test: dm_test_pci_swapcase
168    Test: dm_test_platdata
169    Test: dm_test_pre_reloc
170    Test: dm_test_remove
171    Test: dm_test_spi_find
172    Invalid chip select 0:0 (err=-19)
173    SF: Failed to get idcodes
174    SF: Detected M25P16 with page size 256 Bytes, erase size 64 KiB, total 2 MiB
175    Test: dm_test_spi_flash
176    2097152 bytes written in 0 ms
177    SF: Detected M25P16 with page size 256 Bytes, erase size 64 KiB, total 2 MiB
178    SPI flash test:
179    0 erase: 0 ticks, 65536000 KiB/s 524288.000 Mbps
180    1 check: 0 ticks, 65536000 KiB/s 524288.000 Mbps
181    2 write: 0 ticks, 65536000 KiB/s 524288.000 Mbps
182    3 read: 0 ticks, 65536000 KiB/s 524288.000 Mbps
183    Test passed
184    0 erase: 0 ticks, 65536000 KiB/s 524288.000 Mbps
185    1 check: 0 ticks, 65536000 KiB/s 524288.000 Mbps
186    2 write: 0 ticks, 65536000 KiB/s 524288.000 Mbps
187    3 read: 0 ticks, 65536000 KiB/s 524288.000 Mbps
188    Test: dm_test_spi_xfer
189    SF: Detected M25P16 with page size 256 Bytes, erase size 64 KiB, total 2 MiB
190    Test: dm_test_uclass
191    Test: dm_test_uclass_before_ready
192    Test: dm_test_usb_base
193    Test: dm_test_usb_flash
194    USB-1:   scanning bus 1 for devices... 2 USB Device(s) found
195    Failures: 0
196
197
198What is going on?
199-----------------
200
201Let's start at the top. The demo command is in common/cmd_demo.c. It does
202the usual command processing and then:
203
204	struct udevice *demo_dev;
205
206	ret = uclass_get_device(UCLASS_DEMO, devnum, &demo_dev);
207
208UCLASS_DEMO means the class of devices which implement 'demo'. Other
209classes might be MMC, or GPIO, hashing or serial. The idea is that the
210devices in the class all share a particular way of working. The class
211presents a unified view of all these devices to U-Boot.
212
213This function looks up a device for the demo uclass. Given a device
214number we can find the device because all devices have registered with
215the UCLASS_DEMO uclass.
216
217The device is automatically activated ready for use by uclass_get_device().
218
219Now that we have the device we can do things like:
220
221	return demo_hello(demo_dev, ch);
222
223This function is in the demo uclass. It takes care of calling the 'hello'
224method of the relevant driver. Bearing in mind that there are two drivers,
225this particular device may use one or other of them.
226
227The code for demo_hello() is in drivers/demo/demo-uclass.c:
228
229int demo_hello(struct udevice *dev, int ch)
230{
231	const struct demo_ops *ops = device_get_ops(dev);
232
233	if (!ops->hello)
234		return -ENOSYS;
235
236	return ops->hello(dev, ch);
237}
238
239As you can see it just calls the relevant driver method. One of these is
240in drivers/demo/demo-simple.c:
241
242static int simple_hello(struct udevice *dev, int ch)
243{
244	const struct dm_demo_pdata *pdata = dev_get_platdata(dev);
245
246	printf("Hello from %08x: %s %d\n", map_to_sysmem(dev),
247	       pdata->colour, pdata->sides);
248
249	return 0;
250}
251
252
253So that is a trip from top (command execution) to bottom (driver action)
254but it leaves a lot of topics to address.
255
256
257Declaring Drivers
258-----------------
259
260A driver declaration looks something like this (see
261drivers/demo/demo-shape.c):
262
263static const struct demo_ops shape_ops = {
264	.hello = shape_hello,
265	.status = shape_status,
266};
267
268U_BOOT_DRIVER(demo_shape_drv) = {
269	.name	= "demo_shape_drv",
270	.id	= UCLASS_DEMO,
271	.ops	= &shape_ops,
272	.priv_data_size = sizeof(struct shape_data),
273};
274
275
276This driver has two methods (hello and status) and requires a bit of
277private data (accessible through dev_get_priv(dev) once the driver has
278been probed). It is a member of UCLASS_DEMO so will register itself
279there.
280
281In U_BOOT_DRIVER it is also possible to specify special methods for bind
282and unbind, and these are called at appropriate times. For many drivers
283it is hoped that only 'probe' and 'remove' will be needed.
284
285The U_BOOT_DRIVER macro creates a data structure accessible from C,
286so driver model can find the drivers that are available.
287
288The methods a device can provide are documented in the device.h header.
289Briefly, they are:
290
291    bind - make the driver model aware of a device (bind it to its driver)
292    unbind - make the driver model forget the device
293    ofdata_to_platdata - convert device tree data to platdata - see later
294    probe - make a device ready for use
295    remove - remove a device so it cannot be used until probed again
296
297The sequence to get a device to work is bind, ofdata_to_platdata (if using
298device tree) and probe.
299
300
301Platform Data
302-------------
303
304Platform data is like Linux platform data, if you are familiar with that.
305It provides the board-specific information to start up a device.
306
307Why is this information not just stored in the device driver itself? The
308idea is that the device driver is generic, and can in principle operate on
309any board that has that type of device. For example, with modern
310highly-complex SoCs it is common for the IP to come from an IP vendor, and
311therefore (for example) the MMC controller may be the same on chips from
312different vendors. It makes no sense to write independent drivers for the
313MMC controller on each vendor's SoC, when they are all almost the same.
314Similarly, we may have 6 UARTs in an SoC, all of which are mostly the same,
315but lie at different addresses in the address space.
316
317Using the UART example, we have a single driver and it is instantiated 6
318times by supplying 6 lots of platform data. Each lot of platform data
319gives the driver name and a pointer to a structure containing information
320about this instance - e.g. the address of the register space. It may be that
321one of the UARTS supports RS-485 operation - this can be added as a flag in
322the platform data, which is set for this one port and clear for the rest.
323
324Think of your driver as a generic piece of code which knows how to talk to
325a device, but needs to know where it is, any variant/option information and
326so on. Platform data provides this link between the generic piece of code
327and the specific way it is bound on a particular board.
328
329Examples of platform data include:
330
331   - The base address of the IP block's register space
332   - Configuration options, like:
333         - the SPI polarity and maximum speed for a SPI controller
334         - the I2C speed to use for an I2C device
335         - the number of GPIOs available in a GPIO device
336
337Where does the platform data come from? It is either held in a structure
338which is compiled into U-Boot, or it can be parsed from the Device Tree
339(see 'Device Tree' below).
340
341For an example of how it can be compiled in, see demo-pdata.c which
342sets up a table of driver names and their associated platform data.
343The data can be interpreted by the drivers however they like - it is
344basically a communication scheme between the board-specific code and
345the generic drivers, which are intended to work on any board.
346
347Drivers can access their data via dev->info->platdata. Here is
348the declaration for the platform data, which would normally appear
349in the board file.
350
351	static const struct dm_demo_cdata red_square = {
352		.colour = "red",
353		.sides = 4.
354	};
355	static const struct driver_info info[] = {
356		{
357			.name = "demo_shape_drv",
358			.platdata = &red_square,
359		},
360	};
361
362	demo1 = driver_bind(root, &info[0]);
363
364
365Device Tree
366-----------
367
368While platdata is useful, a more flexible way of providing device data is
369by using device tree. With device tree we replace the above code with the
370following device tree fragment:
371
372	red-square {
373		compatible = "demo-shape";
374		colour = "red";
375		sides = <4>;
376	};
377
378This means that instead of having lots of U_BOOT_DEVICE() declarations in
379the board file, we put these in the device tree. This approach allows a lot
380more generality, since the same board file can support many types of boards
381(e,g. with the same SoC) just by using different device trees. An added
382benefit is that the Linux device tree can be used, thus further simplifying
383the task of board-bring up either for U-Boot or Linux devs (whoever gets to
384the board first!).
385
386The easiest way to make this work it to add a few members to the driver:
387
388	.platdata_auto_alloc_size = sizeof(struct dm_test_pdata),
389	.ofdata_to_platdata = testfdt_ofdata_to_platdata,
390
391The 'auto_alloc' feature allowed space for the platdata to be allocated
392and zeroed before the driver's ofdata_to_platdata() method is called. The
393ofdata_to_platdata() method, which the driver write supplies, should parse
394the device tree node for this device and place it in dev->platdata. Thus
395when the probe method is called later (to set up the device ready for use)
396the platform data will be present.
397
398Note that both methods are optional. If you provide an ofdata_to_platdata
399method then it will be called first (during activation). If you provide a
400probe method it will be called next. See Driver Lifecycle below for more
401details.
402
403If you don't want to have the platdata automatically allocated then you
404can leave out platdata_auto_alloc_size. In this case you can use malloc
405in your ofdata_to_platdata (or probe) method to allocate the required memory,
406and you should free it in the remove method.
407
408The driver model tree is intended to mirror that of the device tree. The
409root driver is at device tree offset 0 (the root node, '/'), and its
410children are the children of the root node.
411
412
413Declaring Uclasses
414------------------
415
416The demo uclass is declared like this:
417
418U_BOOT_CLASS(demo) = {
419	.id		= UCLASS_DEMO,
420};
421
422It is also possible to specify special methods for probe, etc. The uclass
423numbering comes from include/dm/uclass.h. To add a new uclass, add to the
424end of the enum there, then declare your uclass as above.
425
426
427Device Sequence Numbers
428-----------------------
429
430U-Boot numbers devices from 0 in many situations, such as in the command
431line for I2C and SPI buses, and the device names for serial ports (serial0,
432serial1, ...). Driver model supports this numbering and permits devices
433to be locating by their 'sequence'. This numbering uniquely identifies a
434device in its uclass, so no two devices within a particular uclass can have
435the same sequence number.
436
437Sequence numbers start from 0 but gaps are permitted. For example, a board
438may have I2C buses 1, 4, 5 but no 0, 2 or 3. The choice of how devices are
439numbered is up to a particular board, and may be set by the SoC in some
440cases. While it might be tempting to automatically renumber the devices
441where there are gaps in the sequence, this can lead to confusion and is
442not the way that U-Boot works.
443
444Each device can request a sequence number. If none is required then the
445device will be automatically allocated the next available sequence number.
446
447To specify the sequence number in the device tree an alias is typically
448used. Make sure that the uclass has the DM_UC_FLAG_SEQ_ALIAS flag set.
449
450aliases {
451	serial2 = "/serial@22230000";
452};
453
454This indicates that in the uclass called "serial", the named node
455("/serial@22230000") will be given sequence number 2. Any command or driver
456which requests serial device 2 will obtain this device.
457
458More commonly you can use node references, which expand to the full path:
459
460aliases {
461	serial2 = &serial_2;
462};
463...
464serial_2: serial@22230000 {
465...
466};
467
468The alias resolves to the same string in this case, but this version is
469easier to read.
470
471Device sequence numbers are resolved when a device is probed. Before then
472the sequence number is only a request which may or may not be honoured,
473depending on what other devices have been probed. However the numbering is
474entirely under the control of the board author so a conflict is generally
475an error.
476
477
478Bus Drivers
479-----------
480
481A common use of driver model is to implement a bus, a device which provides
482access to other devices. Example of buses include SPI and I2C. Typically
483the bus provides some sort of transport or translation that makes it
484possible to talk to the devices on the bus.
485
486Driver model provides some useful features to help with implementing buses.
487Firstly, a bus can request that its children store some 'parent data' which
488can be used to keep track of child state. Secondly, the bus can define
489methods which are called when a child is probed or removed. This is similar
490to the methods the uclass driver provides. Thirdly, per-child platform data
491can be provided to specify things like the child's address on the bus. This
492persists across child probe()/remove() cycles.
493
494For consistency and ease of implementation, the bus uclass can specify the
495per-child platform data, so that it can be the same for all children of buses
496in that uclass. There are also uclass methods which can be called when
497children are bound and probed.
498
499Here an explanation of how a bus fits with a uclass may be useful. Consider
500a USB bus with several devices attached to it, each from a different (made
501up) uclass:
502
503   xhci_usb (UCLASS_USB)
504      eth (UCLASS_ETHERNET)
505      camera (UCLASS_CAMERA)
506      flash (UCLASS_FLASH_STORAGE)
507
508Each of the devices is connected to a different address on the USB bus.
509The bus device wants to store this address and some other information such
510as the bus speed for each device.
511
512To achieve this, the bus device can use dev->parent_platdata in each of its
513three children. This can be auto-allocated if the bus driver (or bus uclass)
514has a non-zero value for per_child_platdata_auto_alloc_size. If not, then
515the bus device or uclass can allocate the space itself before the child
516device is probed.
517
518Also the bus driver can define the child_pre_probe() and child_post_remove()
519methods to allow it to do some processing before the child is activated or
520after it is deactivated.
521
522Similarly the bus uclass can define the child_post_bind() method to obtain
523the per-child platform data from the device tree and set it up for the child.
524The bus uclass can also provide a child_pre_probe() method. Very often it is
525the bus uclass that controls these features, since it avoids each driver
526having to do the same processing. Of course the driver can still tweak and
527override these activities.
528
529Note that the information that controls this behaviour is in the bus's
530driver, not the child's. In fact it is possible that child has no knowledge
531that it is connected to a bus. The same child device may even be used on two
532different bus types. As an example. the 'flash' device shown above may also
533be connected on a SATA bus or standalone with no bus:
534
535   xhci_usb (UCLASS_USB)
536      flash (UCLASS_FLASH_STORAGE)  - parent data/methods defined by USB bus
537
538   sata (UCLASS_SATA)
539      flash (UCLASS_FLASH_STORAGE)  - parent data/methods defined by SATA bus
540
541   flash (UCLASS_FLASH_STORAGE)  - no parent data/methods (not on a bus)
542
543Above you can see that the driver for xhci_usb/sata controls the child's
544bus methods. In the third example the device is not on a bus, and therefore
545will not have these methods at all. Consider the case where the flash
546device defines child methods. These would be used for *its* children, and
547would be quite separate from the methods defined by the driver for the bus
548that the flash device is connetced to. The act of attaching a device to a
549parent device which is a bus, causes the device to start behaving like a
550bus device, regardless of its own views on the matter.
551
552The uclass for the device can also contain data private to that uclass.
553But note that each device on the bus may be a memeber of a different
554uclass, and this data has nothing to do with the child data for each child
555on the bus. It is the bus' uclass that controls the child with respect to
556the bus.
557
558
559Driver Lifecycle
560----------------
561
562Here are the stages that a device goes through in driver model. Note that all
563methods mentioned here are optional - e.g. if there is no probe() method for
564a device then it will not be called. A simple device may have very few
565methods actually defined.
566
5671. Bind stage
568
569A device and its driver are bound using one of these two methods:
570
571   - Scan the U_BOOT_DEVICE() definitions. U-Boot It looks up the
572name specified by each, to find the appropriate driver. It then calls
573device_bind() to create a new device and bind' it to its driver. This will
574call the device's bind() method.
575
576   - Scan through the device tree definitions. U-Boot looks at top-level
577nodes in the the device tree. It looks at the compatible string in each node
578and uses the of_match part of the U_BOOT_DRIVER() structure to find the
579right driver for each node. It then calls device_bind() to bind the
580newly-created device to its driver (thereby creating a device structure).
581This will also call the device's bind() method.
582
583At this point all the devices are known, and bound to their drivers. There
584is a 'struct udevice' allocated for all devices. However, nothing has been
585activated (except for the root device). Each bound device that was created
586from a U_BOOT_DEVICE() declaration will hold the platdata pointer specified
587in that declaration. For a bound device created from the device tree,
588platdata will be NULL, but of_offset will be the offset of the device tree
589node that caused the device to be created. The uclass is set correctly for
590the device.
591
592The device's bind() method is permitted to perform simple actions, but
593should not scan the device tree node, not initialise hardware, nor set up
594structures or allocate memory. All of these tasks should be left for
595the probe() method.
596
597Note that compared to Linux, U-Boot's driver model has a separate step of
598probe/remove which is independent of bind/unbind. This is partly because in
599U-Boot it may be expensive to probe devices and we don't want to do it until
600they are needed, or perhaps until after relocation.
601
6022. Activation/probe
603
604When a device needs to be used, U-Boot activates it, by following these
605steps (see device_probe()):
606
607   a. If priv_auto_alloc_size is non-zero, then the device-private space
608   is allocated for the device and zeroed. It will be accessible as
609   dev->priv. The driver can put anything it likes in there, but should use
610   it for run-time information, not platform data (which should be static
611   and known before the device is probed).
612
613   b. If platdata_auto_alloc_size is non-zero, then the platform data space
614   is allocated. This is only useful for device tree operation, since
615   otherwise you would have to specific the platform data in the
616   U_BOOT_DEVICE() declaration. The space is allocated for the device and
617   zeroed. It will be accessible as dev->platdata.
618
619   c. If the device's uclass specifies a non-zero per_device_auto_alloc_size,
620   then this space is allocated and zeroed also. It is allocated for and
621   stored in the device, but it is uclass data. owned by the uclass driver.
622   It is possible for the device to access it.
623
624   d. If the device's immediate parent specifies a per_child_auto_alloc_size
625   then this space is allocated. This is intended for use by the parent
626   device to keep track of things related to the child. For example a USB
627   flash stick attached to a USB host controller would likely use this
628   space. The controller can hold information about the USB state of each
629   of its children.
630
631   e. All parent devices are probed. It is not possible to activate a device
632   unless its predecessors (all the way up to the root device) are activated.
633   This means (for example) that an I2C driver will require that its bus
634   be activated.
635
636   f. The device's sequence number is assigned, either the requested one
637   (assuming no conflicts) or the next available one if there is a conflict
638   or nothing particular is requested.
639
640   g. If the driver provides an ofdata_to_platdata() method, then this is
641   called to convert the device tree data into platform data. This should
642   do various calls like fdtdec_get_int(gd->fdt_blob, dev->of_offset, ...)
643   to access the node and store the resulting information into dev->platdata.
644   After this point, the device works the same way whether it was bound
645   using a device tree node or U_BOOT_DEVICE() structure. In either case,
646   the platform data is now stored in the platdata structure. Typically you
647   will use the platdata_auto_alloc_size feature to specify the size of the
648   platform data structure, and U-Boot will automatically allocate and zero
649   it for you before entry to ofdata_to_platdata(). But if not, you can
650   allocate it yourself in ofdata_to_platdata(). Note that it is preferable
651   to do all the device tree decoding in ofdata_to_platdata() rather than
652   in probe(). (Apart from the ugliness of mixing configuration and run-time
653   data, one day it is possible that U-Boot will cache platformat data for
654   devices which are regularly de/activated).
655
656   h. The device's probe() method is called. This should do anything that
657   is required by the device to get it going. This could include checking
658   that the hardware is actually present, setting up clocks for the
659   hardware and setting up hardware registers to initial values. The code
660   in probe() can access:
661
662      - platform data in dev->platdata (for configuration)
663      - private data in dev->priv (for run-time state)
664      - uclass data in dev->uclass_priv (for things the uclass stores
665        about this device)
666
667   Note: If you don't use priv_auto_alloc_size then you will need to
668   allocate the priv space here yourself. The same applies also to
669   platdata_auto_alloc_size. Remember to free them in the remove() method.
670
671   i. The device is marked 'activated'
672
673   j. The uclass's post_probe() method is called, if one exists. This may
674   cause the uclass to do some housekeeping to record the device as
675   activated and 'known' by the uclass.
676
6773. Running stage
678
679The device is now activated and can be used. From now until it is removed
680all of the above structures are accessible. The device appears in the
681uclass's list of devices (so if the device is in UCLASS_GPIO it will appear
682as a device in the GPIO uclass). This is the 'running' state of the device.
683
6844. Removal stage
685
686When the device is no-longer required, you can call device_remove() to
687remove it. This performs the probe steps in reverse:
688
689   a. The uclass's pre_remove() method is called, if one exists. This may
690   cause the uclass to do some housekeeping to record the device as
691   deactivated and no-longer 'known' by the uclass.
692
693   b. All the device's children are removed. It is not permitted to have
694   an active child device with a non-active parent. This means that
695   device_remove() is called for all the children recursively at this point.
696
697   c. The device's remove() method is called. At this stage nothing has been
698   deallocated so platform data, private data and the uclass data will all
699   still be present. This is where the hardware can be shut down. It is
700   intended that the device be completely inactive at this point, For U-Boot
701   to be sure that no hardware is running, it should be enough to remove
702   all devices.
703
704   d. The device memory is freed (platform data, private data, uclass data,
705   parent data).
706
707   Note: Because the platform data for a U_BOOT_DEVICE() is defined with a
708   static pointer, it is not de-allocated during the remove() method. For
709   a device instantiated using the device tree data, the platform data will
710   be dynamically allocated, and thus needs to be deallocated during the
711   remove() method, either:
712
713      1. if the platdata_auto_alloc_size is non-zero, the deallocation
714      happens automatically within the driver model core; or
715
716      2. when platdata_auto_alloc_size is 0, both the allocation (in probe()
717      or preferably ofdata_to_platdata()) and the deallocation in remove()
718      are the responsibility of the driver author.
719
720   e. The device sequence number is set to -1, meaning that it no longer
721   has an allocated sequence. If the device is later reactivated and that
722   sequence number is still free, it may well receive the name sequence
723   number again. But from this point, the sequence number previously used
724   by this device will no longer exist (think of SPI bus 2 being removed
725   and bus 2 is no longer available for use).
726
727   f. The device is marked inactive. Note that it is still bound, so the
728   device structure itself is not freed at this point. Should the device be
729   activated again, then the cycle starts again at step 2 above.
730
7315. Unbind stage
732
733The device is unbound. This is the step that actually destroys the device.
734If a parent has children these will be destroyed first. After this point
735the device does not exist and its memory has be deallocated.
736
737
738Data Structures
739---------------
740
741Driver model uses a doubly-linked list as the basic data structure. Some
742nodes have several lists running through them. Creating a more efficient
743data structure might be worthwhile in some rare cases, once we understand
744what the bottlenecks are.
745
746
747Changes since v1
748----------------
749
750For the record, this implementation uses a very similar approach to the
751original patches, but makes at least the following changes:
752
753- Tried to aggressively remove boilerplate, so that for most drivers there
754is little or no 'driver model' code to write.
755- Moved some data from code into data structure - e.g. store a pointer to
756the driver operations structure in the driver, rather than passing it
757to the driver bind function.
758- Rename some structures to make them more similar to Linux (struct udevice
759instead of struct instance, struct platdata, etc.)
760- Change the name 'core' to 'uclass', meaning U-Boot class. It seems that
761this concept relates to a class of drivers (or a subsystem). We shouldn't
762use 'class' since it is a C++ reserved word, so U-Boot class (uclass) seems
763better than 'core'.
764- Remove 'struct driver_instance' and just use a single 'struct udevice'.
765This removes a level of indirection that doesn't seem necessary.
766- Built in device tree support, to avoid the need for platdata
767- Removed the concept of driver relocation, and just make it possible for
768the new driver (created after relocation) to access the old driver data.
769I feel that relocation is a very special case and will only apply to a few
770drivers, many of which can/will just re-init anyway. So the overhead of
771dealing with this might not be worth it.
772- Implemented a GPIO system, trying to keep it simple
773
774
775Pre-Relocation Support
776----------------------
777
778For pre-relocation we simply call the driver model init function. Only
779drivers marked with DM_FLAG_PRE_RELOC or the device tree
780'u-boot,dm-pre-reloc' flag are initialised prior to relocation. This helps
781to reduce the driver model overhead.
782
783Then post relocation we throw that away and re-init driver model again.
784For drivers which require some sort of continuity between pre- and
785post-relocation devices, we can provide access to the pre-relocation
786device pointers, but this is not currently implemented (the root device
787pointer is saved but not made available through the driver model API).
788
789
790SPL Support
791-----------
792
793Driver model can operate in SPL. Its efficient implementation and small code
794size provide for a small overhead which is acceptable for all but the most
795constrained systems.
796
797To enable driver model in SPL, define CONFIG_SPL_DM. You might want to
798consider the following option also. See the main README for more details.
799
800   - CONFIG_SYS_MALLOC_SIMPLE
801   - CONFIG_DM_WARN
802   - CONFIG_DM_DEVICE_REMOVE
803   - CONFIG_DM_STDIO
804
805
806Enabling Driver Model
807---------------------
808
809Driver model is being brought into U-Boot gradually. As each subsystems gets
810support, a uclass is created and a CONFIG to enable use of driver model for
811that subsystem.
812
813For example CONFIG_DM_SERIAL enables driver model for serial. With that
814defined, the old serial support is not enabled, and your serial driver must
815conform to driver model. With that undefined, the old serial support is
816enabled and driver model is not available for serial. This means that when
817you convert a driver, you must either convert all its boards, or provide for
818the driver to be compiled both with and without driver model (generally this
819is not very hard).
820
821See the main README for full details of the available driver model CONFIG
822options.
823
824
825Things to punt for later
826------------------------
827
828Uclasses are statically numbered at compile time. It would be possible to
829change this to dynamic numbering, but then we would require some sort of
830lookup service, perhaps searching by name. This is slightly less efficient
831so has been left out for now. One small advantage of dynamic numbering might
832be fewer merge conflicts in uclass-id.h.
833
834
835Simon Glass
836sjg@chromium.org
837April 2013
838Updated 7-May-13
839Updated 14-Jun-13
840Updated 18-Oct-13
841Updated 5-Nov-13
842