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_config 40 make 41 ./u-boot 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 17 driver model tests 99 Test: dm_test_autobind 100 Test: dm_test_autoprobe 101 Test: dm_test_children 102 Test: dm_test_fdt 103 Device 'd-test': seq 3 is in use by 'b-test' 104 Test: dm_test_fdt_offset 105 Test: dm_test_fdt_pre_reloc 106 Test: dm_test_fdt_uclass_seq 107 Device 'd-test': seq 3 is in use by 'b-test' 108 Device 'a-test': seq 0 is in use by 'd-test' 109 Test: dm_test_gpio 110 sandbox_gpio: sb_gpio_get_value: error: offset 4 not reserved 111 Test: dm_test_leak 112 Test: dm_test_lifecycle 113 Test: dm_test_operations 114 Test: dm_test_ordering 115 Test: dm_test_platdata 116 Test: dm_test_pre_reloc 117 Test: dm_test_remove 118 Test: dm_test_uclass 119 Test: dm_test_uclass_before_ready 120 Failures: 0 121 122 123What is going on? 124----------------- 125 126Let's start at the top. The demo command is in common/cmd_demo.c. It does 127the usual command processing and then: 128 129 struct udevice *demo_dev; 130 131 ret = uclass_get_device(UCLASS_DEMO, devnum, &demo_dev); 132 133UCLASS_DEMO means the class of devices which implement 'demo'. Other 134classes might be MMC, or GPIO, hashing or serial. The idea is that the 135devices in the class all share a particular way of working. The class 136presents a unified view of all these devices to U-Boot. 137 138This function looks up a device for the demo uclass. Given a device 139number we can find the device because all devices have registered with 140the UCLASS_DEMO uclass. 141 142The device is automatically activated ready for use by uclass_get_device(). 143 144Now that we have the device we can do things like: 145 146 return demo_hello(demo_dev, ch); 147 148This function is in the demo uclass. It takes care of calling the 'hello' 149method of the relevant driver. Bearing in mind that there are two drivers, 150this particular device may use one or other of them. 151 152The code for demo_hello() is in drivers/demo/demo-uclass.c: 153 154int demo_hello(struct udevice *dev, int ch) 155{ 156 const struct demo_ops *ops = device_get_ops(dev); 157 158 if (!ops->hello) 159 return -ENOSYS; 160 161 return ops->hello(dev, ch); 162} 163 164As you can see it just calls the relevant driver method. One of these is 165in drivers/demo/demo-simple.c: 166 167static int simple_hello(struct udevice *dev, int ch) 168{ 169 const struct dm_demo_pdata *pdata = dev_get_platdata(dev); 170 171 printf("Hello from %08x: %s %d\n", map_to_sysmem(dev), 172 pdata->colour, pdata->sides); 173 174 return 0; 175} 176 177 178So that is a trip from top (command execution) to bottom (driver action) 179but it leaves a lot of topics to address. 180 181 182Declaring Drivers 183----------------- 184 185A driver declaration looks something like this (see 186drivers/demo/demo-shape.c): 187 188static const struct demo_ops shape_ops = { 189 .hello = shape_hello, 190 .status = shape_status, 191}; 192 193U_BOOT_DRIVER(demo_shape_drv) = { 194 .name = "demo_shape_drv", 195 .id = UCLASS_DEMO, 196 .ops = &shape_ops, 197 .priv_data_size = sizeof(struct shape_data), 198}; 199 200 201This driver has two methods (hello and status) and requires a bit of 202private data (accessible through dev_get_priv(dev) once the driver has 203been probed). It is a member of UCLASS_DEMO so will register itself 204there. 205 206In U_BOOT_DRIVER it is also possible to specify special methods for bind 207and unbind, and these are called at appropriate times. For many drivers 208it is hoped that only 'probe' and 'remove' will be needed. 209 210The U_BOOT_DRIVER macro creates a data structure accessible from C, 211so driver model can find the drivers that are available. 212 213The methods a device can provide are documented in the device.h header. 214Briefly, they are: 215 216 bind - make the driver model aware of a device (bind it to its driver) 217 unbind - make the driver model forget the device 218 ofdata_to_platdata - convert device tree data to platdata - see later 219 probe - make a device ready for use 220 remove - remove a device so it cannot be used until probed again 221 222The sequence to get a device to work is bind, ofdata_to_platdata (if using 223device tree) and probe. 224 225 226Platform Data 227------------- 228 229Platform data is like Linux platform data, if you are familiar with that. 230It provides the board-specific information to start up a device. 231 232Why is this information not just stored in the device driver itself? The 233idea is that the device driver is generic, and can in principle operate on 234any board that has that type of device. For example, with modern 235highly-complex SoCs it is common for the IP to come from an IP vendor, and 236therefore (for example) the MMC controller may be the same on chips from 237different vendors. It makes no sense to write independent drivers for the 238MMC controller on each vendor's SoC, when they are all almost the same. 239Similarly, we may have 6 UARTs in an SoC, all of which are mostly the same, 240but lie at different addresses in the address space. 241 242Using the UART example, we have a single driver and it is instantiated 6 243times by supplying 6 lots of platform data. Each lot of platform data 244gives the driver name and a pointer to a structure containing information 245about this instance - e.g. the address of the register space. It may be that 246one of the UARTS supports RS-485 operation - this can be added as a flag in 247the platform data, which is set for this one port and clear for the rest. 248 249Think of your driver as a generic piece of code which knows how to talk to 250a device, but needs to know where it is, any variant/option information and 251so on. Platform data provides this link between the generic piece of code 252and the specific way it is bound on a particular board. 253 254Examples of platform data include: 255 256 - The base address of the IP block's register space 257 - Configuration options, like: 258 - the SPI polarity and maximum speed for a SPI controller 259 - the I2C speed to use for an I2C device 260 - the number of GPIOs available in a GPIO device 261 262Where does the platform data come from? It is either held in a structure 263which is compiled into U-Boot, or it can be parsed from the Device Tree 264(see 'Device Tree' below). 265 266For an example of how it can be compiled in, see demo-pdata.c which 267sets up a table of driver names and their associated platform data. 268The data can be interpreted by the drivers however they like - it is 269basically a communication scheme between the board-specific code and 270the generic drivers, which are intended to work on any board. 271 272Drivers can access their data via dev->info->platdata. Here is 273the declaration for the platform data, which would normally appear 274in the board file. 275 276 static const struct dm_demo_cdata red_square = { 277 .colour = "red", 278 .sides = 4. 279 }; 280 static const struct driver_info info[] = { 281 { 282 .name = "demo_shape_drv", 283 .platdata = &red_square, 284 }, 285 }; 286 287 demo1 = driver_bind(root, &info[0]); 288 289 290Device Tree 291----------- 292 293While platdata is useful, a more flexible way of providing device data is 294by using device tree. With device tree we replace the above code with the 295following device tree fragment: 296 297 red-square { 298 compatible = "demo-shape"; 299 colour = "red"; 300 sides = <4>; 301 }; 302 303This means that instead of having lots of U_BOOT_DEVICE() declarations in 304the board file, we put these in the device tree. This approach allows a lot 305more generality, since the same board file can support many types of boards 306(e,g. with the same SoC) just by using different device trees. An added 307benefit is that the Linux device tree can be used, thus further simplifying 308the task of board-bring up either for U-Boot or Linux devs (whoever gets to 309the board first!). 310 311The easiest way to make this work it to add a few members to the driver: 312 313 .platdata_auto_alloc_size = sizeof(struct dm_test_pdata), 314 .ofdata_to_platdata = testfdt_ofdata_to_platdata, 315 316The 'auto_alloc' feature allowed space for the platdata to be allocated 317and zeroed before the driver's ofdata_to_platdata() method is called. The 318ofdata_to_platdata() method, which the driver write supplies, should parse 319the device tree node for this device and place it in dev->platdata. Thus 320when the probe method is called later (to set up the device ready for use) 321the platform data will be present. 322 323Note that both methods are optional. If you provide an ofdata_to_platdata 324method then it will be called first (during activation). If you provide a 325probe method it will be called next. See Driver Lifecycle below for more 326details. 327 328If you don't want to have the platdata automatically allocated then you 329can leave out platdata_auto_alloc_size. In this case you can use malloc 330in your ofdata_to_platdata (or probe) method to allocate the required memory, 331and you should free it in the remove method. 332 333 334Declaring Uclasses 335------------------ 336 337The demo uclass is declared like this: 338 339U_BOOT_CLASS(demo) = { 340 .id = UCLASS_DEMO, 341}; 342 343It is also possible to specify special methods for probe, etc. The uclass 344numbering comes from include/dm/uclass.h. To add a new uclass, add to the 345end of the enum there, then declare your uclass as above. 346 347 348Device Sequence Numbers 349----------------------- 350 351U-Boot numbers devices from 0 in many situations, such as in the command 352line for I2C and SPI buses, and the device names for serial ports (serial0, 353serial1, ...). Driver model supports this numbering and permits devices 354to be locating by their 'sequence'. 355 356Sequence numbers start from 0 but gaps are permitted. For example, a board 357may have I2C buses 0, 1, 4, 5 but no 2 or 3. The choice of how devices are 358numbered is up to a particular board, and may be set by the SoC in some 359cases. While it might be tempting to automatically renumber the devices 360where there are gaps in the sequence, this can lead to confusion and is 361not the way that U-Boot works. 362 363Each device can request a sequence number. If none is required then the 364device will be automatically allocated the next available sequence number. 365 366To specify the sequence number in the device tree an alias is typically 367used. 368 369aliases { 370 serial2 = "/serial@22230000"; 371}; 372 373This indicates that in the uclass called "serial", the named node 374("/serial@22230000") will be given sequence number 2. Any command or driver 375which requests serial device 2 will obtain this device. 376 377Some devices represent buses where the devices on the bus are numbered or 378addressed. For example, SPI typically numbers its slaves from 0, and I2C 379uses a 7-bit address. In these cases the 'reg' property of the subnode is 380used, for example: 381 382{ 383 aliases { 384 spi2 = "/spi@22300000"; 385 }; 386 387 spi@22300000 { 388 #address-cells = <1>; 389 #size-cells = <1>; 390 spi-flash@0 { 391 reg = <0>; 392 ... 393 } 394 eeprom@1 { 395 reg = <1>; 396 }; 397 }; 398 399In this case we have a SPI bus with two slaves at 0 and 1. The SPI bus 400itself is numbered 2. So we might access the SPI flash with: 401 402 sf probe 2:0 403 404and the eeprom with 405 406 sspi 2:1 32 ef 407 408These commands simply need to look up the 2nd device in the SPI uclass to 409find the right SPI bus. Then, they look at the children of that bus for the 410right sequence number (0 or 1 in this case). 411 412Typically the alias method is used for top-level nodes and the 'reg' method 413is used only for buses. 414 415Device sequence numbers are resolved when a device is probed. Before then 416the sequence number is only a request which may or may not be honoured, 417depending on what other devices have been probed. However the numbering is 418entirely under the control of the board author so a conflict is generally 419an error. 420 421 422Driver Lifecycle 423---------------- 424 425Here are the stages that a device goes through in driver model. Note that all 426methods mentioned here are optional - e.g. if there is no probe() method for 427a device then it will not be called. A simple device may have very few 428methods actually defined. 429 4301. Bind stage 431 432A device and its driver are bound using one of these two methods: 433 434 - Scan the U_BOOT_DEVICE() definitions. U-Boot It looks up the 435name specified by each, to find the appropriate driver. It then calls 436device_bind() to create a new device and bind' it to its driver. This will 437call the device's bind() method. 438 439 - Scan through the device tree definitions. U-Boot looks at top-level 440nodes in the the device tree. It looks at the compatible string in each node 441and uses the of_match part of the U_BOOT_DRIVER() structure to find the 442right driver for each node. It then calls device_bind() to bind the 443newly-created device to its driver (thereby creating a device structure). 444This will also call the device's bind() method. 445 446At this point all the devices are known, and bound to their drivers. There 447is a 'struct udevice' allocated for all devices. However, nothing has been 448activated (except for the root device). Each bound device that was created 449from a U_BOOT_DEVICE() declaration will hold the platdata pointer specified 450in that declaration. For a bound device created from the device tree, 451platdata will be NULL, but of_offset will be the offset of the device tree 452node that caused the device to be created. The uclass is set correctly for 453the device. 454 455The device's bind() method is permitted to perform simple actions, but 456should not scan the device tree node, not initialise hardware, nor set up 457structures or allocate memory. All of these tasks should be left for 458the probe() method. 459 460Note that compared to Linux, U-Boot's driver model has a separate step of 461probe/remove which is independent of bind/unbind. This is partly because in 462U-Boot it may be expensive to probe devices and we don't want to do it until 463they are needed, or perhaps until after relocation. 464 4652. Activation/probe 466 467When a device needs to be used, U-Boot activates it, by following these 468steps (see device_probe()): 469 470 a. If priv_auto_alloc_size is non-zero, then the device-private space 471 is allocated for the device and zeroed. It will be accessible as 472 dev->priv. The driver can put anything it likes in there, but should use 473 it for run-time information, not platform data (which should be static 474 and known before the device is probed). 475 476 b. If platdata_auto_alloc_size is non-zero, then the platform data space 477 is allocated. This is only useful for device tree operation, since 478 otherwise you would have to specific the platform data in the 479 U_BOOT_DEVICE() declaration. The space is allocated for the device and 480 zeroed. It will be accessible as dev->platdata. 481 482 c. If the device's uclass specifies a non-zero per_device_auto_alloc_size, 483 then this space is allocated and zeroed also. It is allocated for and 484 stored in the device, but it is uclass data. owned by the uclass driver. 485 It is possible for the device to access it. 486 487 d. All parent devices are probed. It is not possible to activate a device 488 unless its predecessors (all the way up to the root device) are activated. 489 This means (for example) that an I2C driver will require that its bus 490 be activated. 491 492 e. The device's sequence number is assigned, either the requested one 493 (assuming no conflicts) or the next available one if there is a conflict 494 or nothing particular is requested. 495 496 f. If the driver provides an ofdata_to_platdata() method, then this is 497 called to convert the device tree data into platform data. This should 498 do various calls like fdtdec_get_int(gd->fdt_blob, dev->of_offset, ...) 499 to access the node and store the resulting information into dev->platdata. 500 After this point, the device works the same way whether it was bound 501 using a device tree node or U_BOOT_DEVICE() structure. In either case, 502 the platform data is now stored in the platdata structure. Typically you 503 will use the platdata_auto_alloc_size feature to specify the size of the 504 platform data structure, and U-Boot will automatically allocate and zero 505 it for you before entry to ofdata_to_platdata(). But if not, you can 506 allocate it yourself in ofdata_to_platdata(). Note that it is preferable 507 to do all the device tree decoding in ofdata_to_platdata() rather than 508 in probe(). (Apart from the ugliness of mixing configuration and run-time 509 data, one day it is possible that U-Boot will cache platformat data for 510 devices which are regularly de/activated). 511 512 g. The device's probe() method is called. This should do anything that 513 is required by the device to get it going. This could include checking 514 that the hardware is actually present, setting up clocks for the 515 hardware and setting up hardware registers to initial values. The code 516 in probe() can access: 517 518 - platform data in dev->platdata (for configuration) 519 - private data in dev->priv (for run-time state) 520 - uclass data in dev->uclass_priv (for things the uclass stores 521 about this device) 522 523 Note: If you don't use priv_auto_alloc_size then you will need to 524 allocate the priv space here yourself. The same applies also to 525 platdata_auto_alloc_size. Remember to free them in the remove() method. 526 527 h. The device is marked 'activated' 528 529 i. The uclass's post_probe() method is called, if one exists. This may 530 cause the uclass to do some housekeeping to record the device as 531 activated and 'known' by the uclass. 532 5333. Running stage 534 535The device is now activated and can be used. From now until it is removed 536all of the above structures are accessible. The device appears in the 537uclass's list of devices (so if the device is in UCLASS_GPIO it will appear 538as a device in the GPIO uclass). This is the 'running' state of the device. 539 5404. Removal stage 541 542When the device is no-longer required, you can call device_remove() to 543remove it. This performs the probe steps in reverse: 544 545 a. The uclass's pre_remove() method is called, if one exists. This may 546 cause the uclass to do some housekeeping to record the device as 547 deactivated and no-longer 'known' by the uclass. 548 549 b. All the device's children are removed. It is not permitted to have 550 an active child device with a non-active parent. This means that 551 device_remove() is called for all the children recursively at this point. 552 553 c. The device's remove() method is called. At this stage nothing has been 554 deallocated so platform data, private data and the uclass data will all 555 still be present. This is where the hardware can be shut down. It is 556 intended that the device be completely inactive at this point, For U-Boot 557 to be sure that no hardware is running, it should be enough to remove 558 all devices. 559 560 d. The device memory is freed (platform data, private data, uclass data). 561 562 Note: Because the platform data for a U_BOOT_DEVICE() is defined with a 563 static pointer, it is not de-allocated during the remove() method. For 564 a device instantiated using the device tree data, the platform data will 565 be dynamically allocated, and thus needs to be deallocated during the 566 remove() method, either: 567 568 1. if the platdata_auto_alloc_size is non-zero, the deallocation 569 happens automatically within the driver model core; or 570 571 2. when platdata_auto_alloc_size is 0, both the allocation (in probe() 572 or preferably ofdata_to_platdata()) and the deallocation in remove() 573 are the responsibility of the driver author. 574 575 e. The device sequence number is set to -1, meaning that it no longer 576 has an allocated sequence. If the device is later reactivated and that 577 sequence number is still free, it may well receive the name sequence 578 number again. But from this point, the sequence number previously used 579 by this device will no longer exist (think of SPI bus 2 being removed 580 and bus 2 is no longer available for use). 581 582 f. The device is marked inactive. Note that it is still bound, so the 583 device structure itself is not freed at this point. Should the device be 584 activated again, then the cycle starts again at step 2 above. 585 5865. Unbind stage 587 588The device is unbound. This is the step that actually destroys the device. 589If a parent has children these will be destroyed first. After this point 590the device does not exist and its memory has be deallocated. 591 592 593Data Structures 594--------------- 595 596Driver model uses a doubly-linked list as the basic data structure. Some 597nodes have several lists running through them. Creating a more efficient 598data structure might be worthwhile in some rare cases, once we understand 599what the bottlenecks are. 600 601 602Changes since v1 603---------------- 604 605For the record, this implementation uses a very similar approach to the 606original patches, but makes at least the following changes: 607 608- Tried to aggressively remove boilerplate, so that for most drivers there 609is little or no 'driver model' code to write. 610- Moved some data from code into data structure - e.g. store a pointer to 611the driver operations structure in the driver, rather than passing it 612to the driver bind function. 613- Rename some structures to make them more similar to Linux (struct udevice 614instead of struct instance, struct platdata, etc.) 615- Change the name 'core' to 'uclass', meaning U-Boot class. It seems that 616this concept relates to a class of drivers (or a subsystem). We shouldn't 617use 'class' since it is a C++ reserved word, so U-Boot class (uclass) seems 618better than 'core'. 619- Remove 'struct driver_instance' and just use a single 'struct udevice'. 620This removes a level of indirection that doesn't seem necessary. 621- Built in device tree support, to avoid the need for platdata 622- Removed the concept of driver relocation, and just make it possible for 623the new driver (created after relocation) to access the old driver data. 624I feel that relocation is a very special case and will only apply to a few 625drivers, many of which can/will just re-init anyway. So the overhead of 626dealing with this might not be worth it. 627- Implemented a GPIO system, trying to keep it simple 628 629 630Pre-Relocation Support 631---------------------- 632 633For pre-relocation we simply call the driver model init function. Only 634drivers marked with DM_FLAG_PRE_RELOC or the device tree 635'u-boot,dm-pre-reloc' flag are initialised prior to relocation. This helps 636to reduce the driver model overhead. 637 638Then post relocation we throw that away and re-init driver model again. 639For drivers which require some sort of continuity between pre- and 640post-relocation devices, we can provide access to the pre-relocation 641device pointers, but this is not currently implemented (the root device 642pointer is saved but not made available through the driver model API). 643 644 645Things to punt for later 646------------------------ 647 648- SPL support - this will have to be present before many drivers can be 649converted, but it seems like we can add it once we are happy with the 650core implementation. 651 652That is not to say that no thinking has gone into this - in fact there 653is quite a lot there. However, getting these right is non-trivial and 654there is a high cost associated with going down the wrong path. 655 656For SPL, it may be possible to fit in a simplified driver model with only 657bind and probe methods, to reduce size. 658 659Uclasses are statically numbered at compile time. It would be possible to 660change this to dynamic numbering, but then we would require some sort of 661lookup service, perhaps searching by name. This is slightly less efficient 662so has been left out for now. One small advantage of dynamic numbering might 663be fewer merge conflicts in uclass-id.h. 664 665 666Simon Glass 667sjg@chromium.org 668April 2013 669Updated 7-May-13 670Updated 14-Jun-13 671Updated 18-Oct-13 672Updated 5-Nov-13 673