1================================================== 2Runtime Power Management Framework for I/O Devices 3================================================== 4 5(C) 2009-2011 Rafael J. Wysocki <rjw@sisk.pl>, Novell Inc. 6 7(C) 2010 Alan Stern <stern@rowland.harvard.edu> 8 9(C) 2014 Intel Corp., Rafael J. Wysocki <rafael.j.wysocki@intel.com> 10 111. Introduction 12=============== 13 14Support for runtime power management (runtime PM) of I/O devices is provided 15at the power management core (PM core) level by means of: 16 17* The power management workqueue pm_wq in which bus types and device drivers can 18 put their PM-related work items. It is strongly recommended that pm_wq be 19 used for queuing all work items related to runtime PM, because this allows 20 them to be synchronized with system-wide power transitions (suspend to RAM, 21 hibernation and resume from system sleep states). pm_wq is declared in 22 include/linux/pm_runtime.h and defined in kernel/power/main.c. 23 24* A number of runtime PM fields in the 'power' member of 'struct device' (which 25 is of the type 'struct dev_pm_info', defined in include/linux/pm.h) that can 26 be used for synchronizing runtime PM operations with one another. 27 28* Three device runtime PM callbacks in 'struct dev_pm_ops' (defined in 29 include/linux/pm.h). 30 31* A set of helper functions defined in drivers/base/power/runtime.c that can be 32 used for carrying out runtime PM operations in such a way that the 33 synchronization between them is taken care of by the PM core. Bus types and 34 device drivers are encouraged to use these functions. 35 36The runtime PM callbacks present in 'struct dev_pm_ops', the device runtime PM 37fields of 'struct dev_pm_info' and the core helper functions provided for 38runtime PM are described below. 39 402. Device Runtime PM Callbacks 41============================== 42 43There are three device runtime PM callbacks defined in 'struct dev_pm_ops':: 44 45 struct dev_pm_ops { 46 ... 47 int (*runtime_suspend)(struct device *dev); 48 int (*runtime_resume)(struct device *dev); 49 int (*runtime_idle)(struct device *dev); 50 ... 51 }; 52 53The ->runtime_suspend(), ->runtime_resume() and ->runtime_idle() callbacks 54are executed by the PM core for the device's subsystem that may be either of 55the following: 56 57 1. PM domain of the device, if the device's PM domain object, dev->pm_domain, 58 is present. 59 60 2. Device type of the device, if both dev->type and dev->type->pm are present. 61 62 3. Device class of the device, if both dev->class and dev->class->pm are 63 present. 64 65 4. Bus type of the device, if both dev->bus and dev->bus->pm are present. 66 67If the subsystem chosen by applying the above rules doesn't provide the relevant 68callback, the PM core will invoke the corresponding driver callback stored in 69dev->driver->pm directly (if present). 70 71The PM core always checks which callback to use in the order given above, so the 72priority order of callbacks from high to low is: PM domain, device type, class 73and bus type. Moreover, the high-priority one will always take precedence over 74a low-priority one. The PM domain, bus type, device type and class callbacks 75are referred to as subsystem-level callbacks in what follows. 76 77By default, the callbacks are always invoked in process context with interrupts 78enabled. However, the pm_runtime_irq_safe() helper function can be used to tell 79the PM core that it is safe to run the ->runtime_suspend(), ->runtime_resume() 80and ->runtime_idle() callbacks for the given device in atomic context with 81interrupts disabled. This implies that the callback routines in question must 82not block or sleep, but it also means that the synchronous helper functions 83listed at the end of Section 4 may be used for that device within an interrupt 84handler or generally in an atomic context. 85 86The subsystem-level suspend callback, if present, is _entirely_ _responsible_ 87for handling the suspend of the device as appropriate, which may, but need not 88include executing the device driver's own ->runtime_suspend() callback (from the 89PM core's point of view it is not necessary to implement a ->runtime_suspend() 90callback in a device driver as long as the subsystem-level suspend callback 91knows what to do to handle the device). 92 93 * Once the subsystem-level suspend callback (or the driver suspend callback, 94 if invoked directly) has completed successfully for the given device, the PM 95 core regards the device as suspended, which need not mean that it has been 96 put into a low power state. It is supposed to mean, however, that the 97 device will not process data and will not communicate with the CPU(s) and 98 RAM until the appropriate resume callback is executed for it. The runtime 99 PM status of a device after successful execution of the suspend callback is 100 'suspended'. 101 102 * If the suspend callback returns -EBUSY or -EAGAIN, the device's runtime PM 103 status remains 'active', which means that the device _must_ be fully 104 operational afterwards. 105 106 * If the suspend callback returns an error code different from -EBUSY and 107 -EAGAIN, the PM core regards this as a fatal error and will refuse to run 108 the helper functions described in Section 4 for the device until its status 109 is directly set to either 'active', or 'suspended' (the PM core provides 110 special helper functions for this purpose). 111 112In particular, if the driver requires remote wakeup capability (i.e. hardware 113mechanism allowing the device to request a change of its power state, such as 114PCI PME) for proper functioning and device_can_wakeup() returns 'false' for the 115device, then ->runtime_suspend() should return -EBUSY. On the other hand, if 116device_can_wakeup() returns 'true' for the device and the device is put into a 117low-power state during the execution of the suspend callback, it is expected 118that remote wakeup will be enabled for the device. Generally, remote wakeup 119should be enabled for all input devices put into low-power states at run time. 120 121The subsystem-level resume callback, if present, is **entirely responsible** for 122handling the resume of the device as appropriate, which may, but need not 123include executing the device driver's own ->runtime_resume() callback (from the 124PM core's point of view it is not necessary to implement a ->runtime_resume() 125callback in a device driver as long as the subsystem-level resume callback knows 126what to do to handle the device). 127 128 * Once the subsystem-level resume callback (or the driver resume callback, if 129 invoked directly) has completed successfully, the PM core regards the device 130 as fully operational, which means that the device _must_ be able to complete 131 I/O operations as needed. The runtime PM status of the device is then 132 'active'. 133 134 * If the resume callback returns an error code, the PM core regards this as a 135 fatal error and will refuse to run the helper functions described in Section 136 4 for the device, until its status is directly set to either 'active', or 137 'suspended' (by means of special helper functions provided by the PM core 138 for this purpose). 139 140The idle callback (a subsystem-level one, if present, or the driver one) is 141executed by the PM core whenever the device appears to be idle, which is 142indicated to the PM core by two counters, the device's usage counter and the 143counter of 'active' children of the device. 144 145 * If any of these counters is decreased using a helper function provided by 146 the PM core and it turns out to be equal to zero, the other counter is 147 checked. If that counter also is equal to zero, the PM core executes the 148 idle callback with the device as its argument. 149 150The action performed by the idle callback is totally dependent on the subsystem 151(or driver) in question, but the expected and recommended action is to check 152if the device can be suspended (i.e. if all of the conditions necessary for 153suspending the device are satisfied) and to queue up a suspend request for the 154device in that case. If there is no idle callback, or if the callback returns 1550, then the PM core will attempt to carry out a runtime suspend of the device, 156also respecting devices configured for autosuspend. In essence this means a 157call to pm_runtime_autosuspend(). To prevent this (for example, if the callback 158routine has started a delayed suspend), the routine must return a non-zero 159value. Negative error return codes are ignored by the PM core. 160 161The helper functions provided by the PM core, described in Section 4, guarantee 162that the following constraints are met with respect to runtime PM callbacks for 163one device: 164 165(1) The callbacks are mutually exclusive (e.g. it is forbidden to execute 166 ->runtime_suspend() in parallel with ->runtime_resume() or with another 167 instance of ->runtime_suspend() for the same device) with the exception that 168 ->runtime_suspend() or ->runtime_resume() can be executed in parallel with 169 ->runtime_idle() (although ->runtime_idle() will not be started while any 170 of the other callbacks is being executed for the same device). 171 172(2) ->runtime_idle() and ->runtime_suspend() can only be executed for 'active' 173 devices (i.e. the PM core will only execute ->runtime_idle() or 174 ->runtime_suspend() for the devices the runtime PM status of which is 175 'active'). 176 177(3) ->runtime_idle() and ->runtime_suspend() can only be executed for a device 178 the usage counter of which is equal to zero _and_ either the counter of 179 'active' children of which is equal to zero, or the 'power.ignore_children' 180 flag of which is set. 181 182(4) ->runtime_resume() can only be executed for 'suspended' devices (i.e. the 183 PM core will only execute ->runtime_resume() for the devices the runtime 184 PM status of which is 'suspended'). 185 186Additionally, the helper functions provided by the PM core obey the following 187rules: 188 189 * If ->runtime_suspend() is about to be executed or there's a pending request 190 to execute it, ->runtime_idle() will not be executed for the same device. 191 192 * A request to execute or to schedule the execution of ->runtime_suspend() 193 will cancel any pending requests to execute ->runtime_idle() for the same 194 device. 195 196 * If ->runtime_resume() is about to be executed or there's a pending request 197 to execute it, the other callbacks will not be executed for the same device. 198 199 * A request to execute ->runtime_resume() will cancel any pending or 200 scheduled requests to execute the other callbacks for the same device, 201 except for scheduled autosuspends. 202 2033. Runtime PM Device Fields 204=========================== 205 206The following device runtime PM fields are present in 'struct dev_pm_info', as 207defined in include/linux/pm.h: 208 209 `struct timer_list suspend_timer;` 210 - timer used for scheduling (delayed) suspend and autosuspend requests 211 212 `unsigned long timer_expires;` 213 - timer expiration time, in jiffies (if this is different from zero, the 214 timer is running and will expire at that time, otherwise the timer is not 215 running) 216 217 `struct work_struct work;` 218 - work structure used for queuing up requests (i.e. work items in pm_wq) 219 220 `wait_queue_head_t wait_queue;` 221 - wait queue used if any of the helper functions needs to wait for another 222 one to complete 223 224 `spinlock_t lock;` 225 - lock used for synchronization 226 227 `atomic_t usage_count;` 228 - the usage counter of the device 229 230 `atomic_t child_count;` 231 - the count of 'active' children of the device 232 233 `unsigned int ignore_children;` 234 - if set, the value of child_count is ignored (but still updated) 235 236 `unsigned int disable_depth;` 237 - used for disabling the helper functions (they work normally if this is 238 equal to zero); the initial value of it is 1 (i.e. runtime PM is 239 initially disabled for all devices) 240 241 `int runtime_error;` 242 - if set, there was a fatal error (one of the callbacks returned error code 243 as described in Section 2), so the helper functions will not work until 244 this flag is cleared; this is the error code returned by the failing 245 callback 246 247 `unsigned int idle_notification;` 248 - if set, ->runtime_idle() is being executed 249 250 `unsigned int request_pending;` 251 - if set, there's a pending request (i.e. a work item queued up into pm_wq) 252 253 `enum rpm_request request;` 254 - type of request that's pending (valid if request_pending is set) 255 256 `unsigned int deferred_resume;` 257 - set if ->runtime_resume() is about to be run while ->runtime_suspend() is 258 being executed for that device and it is not practical to wait for the 259 suspend to complete; means "start a resume as soon as you've suspended" 260 261 `enum rpm_status runtime_status;` 262 - the runtime PM status of the device; this field's initial value is 263 RPM_SUSPENDED, which means that each device is initially regarded by the 264 PM core as 'suspended', regardless of its real hardware status 265 266 `enum rpm_status last_status;` 267 - the last runtime PM status of the device captured before disabling runtime 268 PM for it (invalid initially and when disable_depth is 0) 269 270 `unsigned int runtime_auto;` 271 - if set, indicates that the user space has allowed the device driver to 272 power manage the device at run time via the /sys/devices/.../power/control 273 `interface;` it may only be modified with the help of the 274 pm_runtime_allow() and pm_runtime_forbid() helper functions 275 276 `unsigned int no_callbacks;` 277 - indicates that the device does not use the runtime PM callbacks (see 278 Section 8); it may be modified only by the pm_runtime_no_callbacks() 279 helper function 280 281 `unsigned int irq_safe;` 282 - indicates that the ->runtime_suspend() and ->runtime_resume() callbacks 283 will be invoked with the spinlock held and interrupts disabled 284 285 `unsigned int use_autosuspend;` 286 - indicates that the device's driver supports delayed autosuspend (see 287 Section 9); it may be modified only by the 288 pm_runtime{_dont}_use_autosuspend() helper functions 289 290 `unsigned int timer_autosuspends;` 291 - indicates that the PM core should attempt to carry out an autosuspend 292 when the timer expires rather than a normal suspend 293 294 `int autosuspend_delay;` 295 - the delay time (in milliseconds) to be used for autosuspend 296 297 `unsigned long last_busy;` 298 - the time (in jiffies) when the pm_runtime_mark_last_busy() helper 299 function was last called for this device; used in calculating inactivity 300 periods for autosuspend 301 302All of the above fields are members of the 'power' member of 'struct device'. 303 3044. Runtime PM Device Helper Functions 305===================================== 306 307The following runtime PM helper functions are defined in 308drivers/base/power/runtime.c and include/linux/pm_runtime.h: 309 310 `void pm_runtime_init(struct device *dev);` 311 - initialize the device runtime PM fields in 'struct dev_pm_info' 312 313 `void pm_runtime_remove(struct device *dev);` 314 - make sure that the runtime PM of the device will be disabled after 315 removing the device from device hierarchy 316 317 `int pm_runtime_idle(struct device *dev);` 318 - execute the subsystem-level idle callback for the device; returns an 319 error code on failure, where -EINPROGRESS means that ->runtime_idle() is 320 already being executed; if there is no callback or the callback returns 0 321 then run pm_runtime_autosuspend(dev) and return its result 322 323 `int pm_runtime_suspend(struct device *dev);` 324 - execute the subsystem-level suspend callback for the device; returns 0 on 325 success, 1 if the device's runtime PM status was already 'suspended', or 326 error code on failure, where -EAGAIN or -EBUSY means it is safe to attempt 327 to suspend the device again in future and -EACCES means that 328 'power.disable_depth' is different from 0 329 330 `int pm_runtime_autosuspend(struct device *dev);` 331 - same as pm_runtime_suspend() except that a call to 332 pm_runtime_mark_last_busy() is made and an autosuspend is scheduled for 333 the appropriate time and 0 is returned 334 335 `int pm_runtime_resume(struct device *dev);` 336 - execute the subsystem-level resume callback for the device; returns 0 on 337 success, 1 if the device's runtime PM status is already 'active' (also if 338 'power.disable_depth' is nonzero, but the status was 'active' when it was 339 changing from 0 to 1) or error code on failure, where -EAGAIN means it may 340 be safe to attempt to resume the device again in future, but 341 'power.runtime_error' should be checked additionally, and -EACCES means 342 that the callback could not be run, because 'power.disable_depth' was 343 different from 0 344 345 `int pm_runtime_resume_and_get(struct device *dev);` 346 - run pm_runtime_resume(dev) and if successful, increment the device's 347 usage counter; returns 0 on success (whether or not the device's 348 runtime PM status was already 'active') or the error code from 349 pm_runtime_resume() on failure. 350 351 `int pm_request_idle(struct device *dev);` 352 - submit a request to execute the subsystem-level idle callback for the 353 device (the request is represented by a work item in pm_wq); returns 0 on 354 success or error code if the request has not been queued up 355 356 `int pm_request_autosuspend(struct device *dev);` 357 - Call pm_runtime_mark_last_busy() and schedule the execution of the 358 subsystem-level suspend callback for the device when the autosuspend delay 359 expires 360 361 `int pm_schedule_suspend(struct device *dev, unsigned int delay);` 362 - schedule the execution of the subsystem-level suspend callback for the 363 device in future, where 'delay' is the time to wait before queuing up a 364 suspend work item in pm_wq, in milliseconds (if 'delay' is zero, the work 365 item is queued up immediately); returns 0 on success, 1 if the device's PM 366 runtime status was already 'suspended', or error code if the request 367 hasn't been scheduled (or queued up if 'delay' is 0); if the execution of 368 ->runtime_suspend() is already scheduled and not yet expired, the new 369 value of 'delay' will be used as the time to wait 370 371 `int pm_request_resume(struct device *dev);` 372 - submit a request to execute the subsystem-level resume callback for the 373 device (the request is represented by a work item in pm_wq); returns 0 on 374 success, 1 if the device's runtime PM status was already 'active', or 375 error code if the request hasn't been queued up 376 377 `void pm_runtime_get_noresume(struct device *dev);` 378 - increment the device's usage counter 379 380 `int pm_runtime_get(struct device *dev);` 381 - increment the device's usage counter, run pm_request_resume(dev) and 382 return its result 383 384 `int pm_runtime_get_sync(struct device *dev);` 385 - increment the device's usage counter, run pm_runtime_resume(dev) and 386 return its result; 387 note that it does not drop the device's usage counter on errors, so 388 consider using pm_runtime_resume_and_get() instead of it, especially 389 if its return value is checked by the caller, as this is likely to 390 result in cleaner code. 391 392 `int pm_runtime_get_if_in_use(struct device *dev);` 393 - return -EINVAL if 'power.disable_depth' is nonzero; otherwise, if the 394 runtime PM status is RPM_ACTIVE and the runtime PM usage counter is 395 nonzero, increment the counter and return 1; otherwise return 0 without 396 changing the counter 397 398 `int pm_runtime_get_if_active(struct device *dev);` 399 - return -EINVAL if 'power.disable_depth' is nonzero; otherwise, if the 400 runtime PM status is RPM_ACTIVE, increment the counter and 401 return 1; otherwise return 0 without changing the counter 402 403 `void pm_runtime_put_noidle(struct device *dev);` 404 - decrement the device's usage counter 405 406 `int pm_runtime_put(struct device *dev);` 407 - decrement the device's usage counter; if the result is 0 then run 408 pm_request_idle(dev) and return its result 409 410 `int pm_runtime_put_autosuspend(struct device *dev);` 411 - set the power.last_busy field to the current time and decrement the 412 device's usage counter; if the result is 0 then run 413 pm_request_autosuspend(dev) and return its result 414 415 `int __pm_runtime_put_autosuspend(struct device *dev);` 416 - decrement the device's usage counter; if the result is 0 then run 417 pm_request_autosuspend(dev) and return its result 418 419 `int pm_runtime_put_sync(struct device *dev);` 420 - decrement the device's usage counter; if the result is 0 then run 421 pm_runtime_idle(dev) and return its result 422 423 `int pm_runtime_put_sync_suspend(struct device *dev);` 424 - decrement the device's usage counter; if the result is 0 then run 425 pm_runtime_suspend(dev) and return its result 426 427 `int pm_runtime_put_sync_autosuspend(struct device *dev);` 428 - set the power.last_busy field to the current time and decrement the 429 device's usage counter; if the result is 0 then run 430 pm_runtime_autosuspend(dev) and return its result 431 432 `void pm_runtime_enable(struct device *dev);` 433 - decrement the device's 'power.disable_depth' field; if that field is equal 434 to zero, the runtime PM helper functions can execute subsystem-level 435 callbacks described in Section 2 for the device 436 437 `int pm_runtime_disable(struct device *dev);` 438 - increment the device's 'power.disable_depth' field (if the value of that 439 field was previously zero, this prevents subsystem-level runtime PM 440 callbacks from being run for the device), make sure that all of the 441 pending runtime PM operations on the device are either completed or 442 canceled; returns 1 if there was a resume request pending and it was 443 necessary to execute the subsystem-level resume callback for the device 444 to satisfy that request, otherwise 0 is returned 445 446 `int pm_runtime_barrier(struct device *dev);` 447 - check if there's a resume request pending for the device and resume it 448 (synchronously) in that case, cancel any other pending runtime PM requests 449 regarding it and wait for all runtime PM operations on it in progress to 450 complete; returns 1 if there was a resume request pending and it was 451 necessary to execute the subsystem-level resume callback for the device to 452 satisfy that request, otherwise 0 is returned 453 454 `void pm_suspend_ignore_children(struct device *dev, bool enable);` 455 - set/unset the power.ignore_children flag of the device 456 457 `int pm_runtime_set_active(struct device *dev);` 458 - clear the device's 'power.runtime_error' flag, set the device's runtime 459 PM status to 'active' and update its parent's counter of 'active' 460 children as appropriate (it is only valid to use this function if 461 'power.runtime_error' is set or 'power.disable_depth' is greater than 462 zero); it will fail and return error code if the device has a parent 463 which is not active and the 'power.ignore_children' flag of which is unset 464 465 `void pm_runtime_set_suspended(struct device *dev);` 466 - clear the device's 'power.runtime_error' flag, set the device's runtime 467 PM status to 'suspended' and update its parent's counter of 'active' 468 children as appropriate (it is only valid to use this function if 469 'power.runtime_error' is set or 'power.disable_depth' is greater than 470 zero) 471 472 `bool pm_runtime_active(struct device *dev);` 473 - return true if the device's runtime PM status is 'active' or its 474 'power.disable_depth' field is not equal to zero, or false otherwise 475 476 `bool pm_runtime_suspended(struct device *dev);` 477 - return true if the device's runtime PM status is 'suspended' and its 478 'power.disable_depth' field is equal to zero, or false otherwise 479 480 `bool pm_runtime_status_suspended(struct device *dev);` 481 - return true if the device's runtime PM status is 'suspended' 482 483 `void pm_runtime_allow(struct device *dev);` 484 - set the power.runtime_auto flag for the device and decrease its usage 485 counter (used by the /sys/devices/.../power/control interface to 486 effectively allow the device to be power managed at run time) 487 488 `void pm_runtime_forbid(struct device *dev);` 489 - unset the power.runtime_auto flag for the device and increase its usage 490 counter (used by the /sys/devices/.../power/control interface to 491 effectively prevent the device from being power managed at run time) 492 493 `void pm_runtime_no_callbacks(struct device *dev);` 494 - set the power.no_callbacks flag for the device and remove the runtime 495 PM attributes from /sys/devices/.../power (or prevent them from being 496 added when the device is registered) 497 498 `void pm_runtime_irq_safe(struct device *dev);` 499 - set the power.irq_safe flag for the device, causing the runtime-PM 500 callbacks to be invoked with interrupts off 501 502 `bool pm_runtime_is_irq_safe(struct device *dev);` 503 - return true if power.irq_safe flag was set for the device, causing 504 the runtime-PM callbacks to be invoked with interrupts off 505 506 `void pm_runtime_mark_last_busy(struct device *dev);` 507 - set the power.last_busy field to the current time 508 509 `void pm_runtime_use_autosuspend(struct device *dev);` 510 - set the power.use_autosuspend flag, enabling autosuspend delays; call 511 pm_runtime_get_sync if the flag was previously cleared and 512 power.autosuspend_delay is negative 513 514 `void pm_runtime_dont_use_autosuspend(struct device *dev);` 515 - clear the power.use_autosuspend flag, disabling autosuspend delays; 516 decrement the device's usage counter if the flag was previously set and 517 power.autosuspend_delay is negative; call pm_runtime_idle 518 519 `void pm_runtime_set_autosuspend_delay(struct device *dev, int delay);` 520 - set the power.autosuspend_delay value to 'delay' (expressed in 521 milliseconds); if 'delay' is negative then runtime suspends are 522 prevented; if power.use_autosuspend is set, pm_runtime_get_sync may be 523 called or the device's usage counter may be decremented and 524 pm_runtime_idle called depending on if power.autosuspend_delay is 525 changed to or from a negative value; if power.use_autosuspend is clear, 526 pm_runtime_idle is called 527 528 `unsigned long pm_runtime_autosuspend_expiration(struct device *dev);` 529 - calculate the time when the current autosuspend delay period will expire, 530 based on power.last_busy and power.autosuspend_delay; if the delay time 531 is 1000 ms or larger then the expiration time is rounded up to the 532 nearest second; returns 0 if the delay period has already expired or 533 power.use_autosuspend isn't set, otherwise returns the expiration time 534 in jiffies 535 536It is safe to execute the following helper functions from interrupt context: 537 538- pm_request_idle() 539- pm_request_autosuspend() 540- pm_schedule_suspend() 541- pm_request_resume() 542- pm_runtime_get_noresume() 543- pm_runtime_get() 544- pm_runtime_put_noidle() 545- pm_runtime_put() 546- pm_runtime_put_autosuspend() 547- __pm_runtime_put_autosuspend() 548- pm_runtime_enable() 549- pm_suspend_ignore_children() 550- pm_runtime_set_active() 551- pm_runtime_set_suspended() 552- pm_runtime_suspended() 553- pm_runtime_mark_last_busy() 554- pm_runtime_autosuspend_expiration() 555 556If pm_runtime_irq_safe() has been called for a device then the following helper 557functions may also be used in interrupt context: 558 559- pm_runtime_idle() 560- pm_runtime_suspend() 561- pm_runtime_autosuspend() 562- pm_runtime_resume() 563- pm_runtime_get_sync() 564- pm_runtime_put_sync() 565- pm_runtime_put_sync_suspend() 566- pm_runtime_put_sync_autosuspend() 567 5685. Runtime PM Initialization, Device Probing and Removal 569======================================================== 570 571Initially, the runtime PM is disabled for all devices, which means that the 572majority of the runtime PM helper functions described in Section 4 will return 573-EAGAIN until pm_runtime_enable() is called for the device. 574 575In addition to that, the initial runtime PM status of all devices is 576'suspended', but it need not reflect the actual physical state of the device. 577Thus, if the device is initially active (i.e. it is able to process I/O), its 578runtime PM status must be changed to 'active', with the help of 579pm_runtime_set_active(), before pm_runtime_enable() is called for the device. 580 581However, if the device has a parent and the parent's runtime PM is enabled, 582calling pm_runtime_set_active() for the device will affect the parent, unless 583the parent's 'power.ignore_children' flag is set. Namely, in that case the 584parent won't be able to suspend at run time, using the PM core's helper 585functions, as long as the child's status is 'active', even if the child's 586runtime PM is still disabled (i.e. pm_runtime_enable() hasn't been called for 587the child yet or pm_runtime_disable() has been called for it). For this reason, 588once pm_runtime_set_active() has been called for the device, pm_runtime_enable() 589should be called for it too as soon as reasonably possible or its runtime PM 590status should be changed back to 'suspended' with the help of 591pm_runtime_set_suspended(). 592 593If the default initial runtime PM status of the device (i.e. 'suspended') 594reflects the actual state of the device, its bus type's or its driver's 595->probe() callback will likely need to wake it up using one of the PM core's 596helper functions described in Section 4. In that case, pm_runtime_resume() 597should be used. Of course, for this purpose the device's runtime PM has to be 598enabled earlier by calling pm_runtime_enable(). 599 600Note, if the device may execute pm_runtime calls during the probe (such as 601if it is registered with a subsystem that may call back in) then the 602pm_runtime_get_sync() call paired with a pm_runtime_put() call will be 603appropriate to ensure that the device is not put back to sleep during the 604probe. This can happen with systems such as the network device layer. 605 606It may be desirable to suspend the device once ->probe() has finished. 607Therefore the driver core uses the asynchronous pm_request_idle() to submit a 608request to execute the subsystem-level idle callback for the device at that 609time. A driver that makes use of the runtime autosuspend feature may want to 610update the last busy mark before returning from ->probe(). 611 612Moreover, the driver core prevents runtime PM callbacks from racing with the bus 613notifier callback in __device_release_driver(), which is necessary because the 614notifier is used by some subsystems to carry out operations affecting the 615runtime PM functionality. It does so by calling pm_runtime_get_sync() before 616driver_sysfs_remove() and the BUS_NOTIFY_UNBIND_DRIVER notifications. This 617resumes the device if it's in the suspended state and prevents it from 618being suspended again while those routines are being executed. 619 620To allow bus types and drivers to put devices into the suspended state by 621calling pm_runtime_suspend() from their ->remove() routines, the driver core 622executes pm_runtime_put_sync() after running the BUS_NOTIFY_UNBIND_DRIVER 623notifications in __device_release_driver(). This requires bus types and 624drivers to make their ->remove() callbacks avoid races with runtime PM directly, 625but it also allows more flexibility in the handling of devices during the 626removal of their drivers. 627 628Drivers in ->remove() callback should undo the runtime PM changes done 629in ->probe(). Usually this means calling pm_runtime_disable(), 630pm_runtime_dont_use_autosuspend() etc. 631 632The user space can effectively disallow the driver of the device to power manage 633it at run time by changing the value of its /sys/devices/.../power/control 634attribute to "on", which causes pm_runtime_forbid() to be called. In principle, 635this mechanism may also be used by the driver to effectively turn off the 636runtime power management of the device until the user space turns it on. 637Namely, during the initialization the driver can make sure that the runtime PM 638status of the device is 'active' and call pm_runtime_forbid(). It should be 639noted, however, that if the user space has already intentionally changed the 640value of /sys/devices/.../power/control to "auto" to allow the driver to power 641manage the device at run time, the driver may confuse it by using 642pm_runtime_forbid() this way. 643 6446. Runtime PM and System Sleep 645============================== 646 647Runtime PM and system sleep (i.e., system suspend and hibernation, also known 648as suspend-to-RAM and suspend-to-disk) interact with each other in a couple of 649ways. If a device is active when a system sleep starts, everything is 650straightforward. But what should happen if the device is already suspended? 651 652The device may have different wake-up settings for runtime PM and system sleep. 653For example, remote wake-up may be enabled for runtime suspend but disallowed 654for system sleep (device_may_wakeup(dev) returns 'false'). When this happens, 655the subsystem-level system suspend callback is responsible for changing the 656device's wake-up setting (it may leave that to the device driver's system 657suspend routine). It may be necessary to resume the device and suspend it again 658in order to do so. The same is true if the driver uses different power levels 659or other settings for runtime suspend and system sleep. 660 661During system resume, the simplest approach is to bring all devices back to full 662power, even if they had been suspended before the system suspend began. There 663are several reasons for this, including: 664 665 * The device might need to switch power levels, wake-up settings, etc. 666 667 * Remote wake-up events might have been lost by the firmware. 668 669 * The device's children may need the device to be at full power in order 670 to resume themselves. 671 672 * The driver's idea of the device state may not agree with the device's 673 physical state. This can happen during resume from hibernation. 674 675 * The device might need to be reset. 676 677 * Even though the device was suspended, if its usage counter was > 0 then most 678 likely it would need a runtime resume in the near future anyway. 679 680If the device had been suspended before the system suspend began and it's 681brought back to full power during resume, then its runtime PM status will have 682to be updated to reflect the actual post-system sleep status. The way to do 683this is: 684 685 - pm_runtime_disable(dev); 686 - pm_runtime_set_active(dev); 687 - pm_runtime_enable(dev); 688 689The PM core always increments the runtime usage counter before calling the 690->suspend() callback and decrements it after calling the ->resume() callback. 691Hence disabling runtime PM temporarily like this will not cause any runtime 692suspend attempts to be permanently lost. If the usage count goes to zero 693following the return of the ->resume() callback, the ->runtime_idle() callback 694will be invoked as usual. 695 696On some systems, however, system sleep is not entered through a global firmware 697or hardware operation. Instead, all hardware components are put into low-power 698states directly by the kernel in a coordinated way. Then, the system sleep 699state effectively follows from the states the hardware components end up in 700and the system is woken up from that state by a hardware interrupt or a similar 701mechanism entirely under the kernel's control. As a result, the kernel never 702gives control away and the states of all devices during resume are precisely 703known to it. If that is the case and none of the situations listed above takes 704place (in particular, if the system is not waking up from hibernation), it may 705be more efficient to leave the devices that had been suspended before the system 706suspend began in the suspended state. 707 708To this end, the PM core provides a mechanism allowing some coordination between 709different levels of device hierarchy. Namely, if a system suspend .prepare() 710callback returns a positive number for a device, that indicates to the PM core 711that the device appears to be runtime-suspended and its state is fine, so it 712may be left in runtime suspend provided that all of its descendants are also 713left in runtime suspend. If that happens, the PM core will not execute any 714system suspend and resume callbacks for all of those devices, except for the 715.complete() callback, which is then entirely responsible for handling the device 716as appropriate. This only applies to system suspend transitions that are not 717related to hibernation (see Documentation/driver-api/pm/devices.rst for more 718information). 719 720The PM core does its best to reduce the probability of race conditions between 721the runtime PM and system suspend/resume (and hibernation) callbacks by carrying 722out the following operations: 723 724 * During system suspend pm_runtime_get_noresume() is called for every device 725 right before executing the subsystem-level .prepare() callback for it and 726 pm_runtime_barrier() is called for every device right before executing the 727 subsystem-level .suspend() callback for it. In addition to that the PM core 728 calls __pm_runtime_disable() with 'false' as the second argument for every 729 device right before executing the subsystem-level .suspend_late() callback 730 for it. 731 732 * During system resume pm_runtime_enable() and pm_runtime_put() are called for 733 every device right after executing the subsystem-level .resume_early() 734 callback and right after executing the subsystem-level .complete() callback 735 for it, respectively. 736 7377. Generic subsystem callbacks 738============================== 739 740Subsystems may wish to conserve code space by using the set of generic power 741management callbacks provided by the PM core, defined in 742driver/base/power/generic_ops.c: 743 744 `int pm_generic_runtime_suspend(struct device *dev);` 745 - invoke the ->runtime_suspend() callback provided by the driver of this 746 device and return its result, or return 0 if not defined 747 748 `int pm_generic_runtime_resume(struct device *dev);` 749 - invoke the ->runtime_resume() callback provided by the driver of this 750 device and return its result, or return 0 if not defined 751 752 `int pm_generic_suspend(struct device *dev);` 753 - if the device has not been suspended at run time, invoke the ->suspend() 754 callback provided by its driver and return its result, or return 0 if not 755 defined 756 757 `int pm_generic_suspend_noirq(struct device *dev);` 758 - if pm_runtime_suspended(dev) returns "false", invoke the ->suspend_noirq() 759 callback provided by the device's driver and return its result, or return 760 0 if not defined 761 762 `int pm_generic_resume(struct device *dev);` 763 - invoke the ->resume() callback provided by the driver of this device and, 764 if successful, change the device's runtime PM status to 'active' 765 766 `int pm_generic_resume_noirq(struct device *dev);` 767 - invoke the ->resume_noirq() callback provided by the driver of this device 768 769 `int pm_generic_freeze(struct device *dev);` 770 - if the device has not been suspended at run time, invoke the ->freeze() 771 callback provided by its driver and return its result, or return 0 if not 772 defined 773 774 `int pm_generic_freeze_noirq(struct device *dev);` 775 - if pm_runtime_suspended(dev) returns "false", invoke the ->freeze_noirq() 776 callback provided by the device's driver and return its result, or return 777 0 if not defined 778 779 `int pm_generic_thaw(struct device *dev);` 780 - if the device has not been suspended at run time, invoke the ->thaw() 781 callback provided by its driver and return its result, or return 0 if not 782 defined 783 784 `int pm_generic_thaw_noirq(struct device *dev);` 785 - if pm_runtime_suspended(dev) returns "false", invoke the ->thaw_noirq() 786 callback provided by the device's driver and return its result, or return 787 0 if not defined 788 789 `int pm_generic_poweroff(struct device *dev);` 790 - if the device has not been suspended at run time, invoke the ->poweroff() 791 callback provided by its driver and return its result, or return 0 if not 792 defined 793 794 `int pm_generic_poweroff_noirq(struct device *dev);` 795 - if pm_runtime_suspended(dev) returns "false", run the ->poweroff_noirq() 796 callback provided by the device's driver and return its result, or return 797 0 if not defined 798 799 `int pm_generic_restore(struct device *dev);` 800 - invoke the ->restore() callback provided by the driver of this device and, 801 if successful, change the device's runtime PM status to 'active' 802 803 `int pm_generic_restore_noirq(struct device *dev);` 804 - invoke the ->restore_noirq() callback provided by the device's driver 805 806These functions are the defaults used by the PM core if a subsystem doesn't 807provide its own callbacks for ->runtime_idle(), ->runtime_suspend(), 808->runtime_resume(), ->suspend(), ->suspend_noirq(), ->resume(), 809->resume_noirq(), ->freeze(), ->freeze_noirq(), ->thaw(), ->thaw_noirq(), 810->poweroff(), ->poweroff_noirq(), ->restore(), ->restore_noirq() in the 811subsystem-level dev_pm_ops structure. 812 813Device drivers that wish to use the same function as a system suspend, freeze, 814poweroff and runtime suspend callback, and similarly for system resume, thaw, 815restore, and runtime resume, can achieve similar behaviour with the help of the 816DEFINE_RUNTIME_DEV_PM_OPS() defined in include/linux/pm_runtime.h (possibly setting its 817last argument to NULL). 818 8198. "No-Callback" Devices 820======================== 821 822Some "devices" are only logical sub-devices of their parent and cannot be 823power-managed on their own. (The prototype example is a USB interface. Entire 824USB devices can go into low-power mode or send wake-up requests, but neither is 825possible for individual interfaces.) The drivers for these devices have no 826need of runtime PM callbacks; if the callbacks did exist, ->runtime_suspend() 827and ->runtime_resume() would always return 0 without doing anything else and 828->runtime_idle() would always call pm_runtime_suspend(). 829 830Subsystems can tell the PM core about these devices by calling 831pm_runtime_no_callbacks(). This should be done after the device structure is 832initialized and before it is registered (although after device registration is 833also okay). The routine will set the device's power.no_callbacks flag and 834prevent the non-debugging runtime PM sysfs attributes from being created. 835 836When power.no_callbacks is set, the PM core will not invoke the 837->runtime_idle(), ->runtime_suspend(), or ->runtime_resume() callbacks. 838Instead it will assume that suspends and resumes always succeed and that idle 839devices should be suspended. 840 841As a consequence, the PM core will never directly inform the device's subsystem 842or driver about runtime power changes. Instead, the driver for the device's 843parent must take responsibility for telling the device's driver when the 844parent's power state changes. 845 846Note that, in some cases it may not be desirable for subsystems/drivers to call 847pm_runtime_no_callbacks() for their devices. This could be because a subset of 848the runtime PM callbacks needs to be implemented, a platform dependent PM 849domain could get attached to the device or that the device is power managed 850through a supplier device link. For these reasons and to avoid boilerplate code 851in subsystems/drivers, the PM core allows runtime PM callbacks to be 852unassigned. More precisely, if a callback pointer is NULL, the PM core will act 853as though there was a callback and it returned 0. 854 8559. Autosuspend, or automatically-delayed suspends 856================================================= 857 858Changing a device's power state isn't free; it requires both time and energy. 859A device should be put in a low-power state only when there's some reason to 860think it will remain in that state for a substantial time. A common heuristic 861says that a device which hasn't been used for a while is liable to remain 862unused; following this advice, drivers should not allow devices to be suspended 863at runtime until they have been inactive for some minimum period. Even when 864the heuristic ends up being non-optimal, it will still prevent devices from 865"bouncing" too rapidly between low-power and full-power states. 866 867The term "autosuspend" is an historical remnant. It doesn't mean that the 868device is automatically suspended (the subsystem or driver still has to call 869the appropriate PM routines); rather it means that runtime suspends will 870automatically be delayed until the desired period of inactivity has elapsed. 871 872Inactivity is determined based on the power.last_busy field. The desired length 873of the inactivity period is a matter of policy. Subsystems can set this length 874initially by calling pm_runtime_set_autosuspend_delay(), but after device 875registration the length should be controlled by user space, using the 876/sys/devices/.../power/autosuspend_delay_ms attribute. 877 878In order to use autosuspend, subsystems or drivers must call 879pm_runtime_use_autosuspend() (preferably before registering the device), and 880thereafter they should use the various `*_autosuspend()` helper functions 881instead of the non-autosuspend counterparts:: 882 883 Instead of: pm_runtime_suspend use: pm_runtime_autosuspend; 884 Instead of: pm_schedule_suspend use: pm_request_autosuspend; 885 Instead of: pm_runtime_put use: pm_runtime_put_autosuspend; 886 Instead of: pm_runtime_put_sync use: pm_runtime_put_sync_autosuspend. 887 888Drivers may also continue to use the non-autosuspend helper functions; they 889will behave normally, which means sometimes taking the autosuspend delay into 890account (see pm_runtime_idle). The autosuspend variants of the functions also 891call pm_runtime_mark_last_busy(). 892 893Under some circumstances a driver or subsystem may want to prevent a device 894from autosuspending immediately, even though the usage counter is zero and the 895autosuspend delay time has expired. If the ->runtime_suspend() callback 896returns -EAGAIN or -EBUSY, and if the next autosuspend delay expiration time is 897in the future (as it normally would be if the callback invoked 898pm_runtime_mark_last_busy()), the PM core will automatically reschedule the 899autosuspend. The ->runtime_suspend() callback can't do this rescheduling 900itself because no suspend requests of any kind are accepted while the device is 901suspending (i.e., while the callback is running). 902 903The implementation is well suited for asynchronous use in interrupt contexts. 904However such use inevitably involves races, because the PM core can't 905synchronize ->runtime_suspend() callbacks with the arrival of I/O requests. 906This synchronization must be handled by the driver, using its private lock. 907Here is a schematic pseudo-code example:: 908 909 foo_read_or_write(struct foo_priv *foo, void *data) 910 { 911 lock(&foo->private_lock); 912 add_request_to_io_queue(foo, data); 913 if (foo->num_pending_requests++ == 0) 914 pm_runtime_get(&foo->dev); 915 if (!foo->is_suspended) 916 foo_process_next_request(foo); 917 unlock(&foo->private_lock); 918 } 919 920 foo_io_completion(struct foo_priv *foo, void *req) 921 { 922 lock(&foo->private_lock); 923 if (--foo->num_pending_requests == 0) 924 pm_runtime_put_autosuspend(&foo->dev); 925 else 926 foo_process_next_request(foo); 927 unlock(&foo->private_lock); 928 /* Send req result back to the user ... */ 929 } 930 931 int foo_runtime_suspend(struct device *dev) 932 { 933 struct foo_priv foo = container_of(dev, ...); 934 int ret = 0; 935 936 lock(&foo->private_lock); 937 if (foo->num_pending_requests > 0) { 938 ret = -EBUSY; 939 } else { 940 /* ... suspend the device ... */ 941 foo->is_suspended = 1; 942 } 943 unlock(&foo->private_lock); 944 return ret; 945 } 946 947 int foo_runtime_resume(struct device *dev) 948 { 949 struct foo_priv foo = container_of(dev, ...); 950 951 lock(&foo->private_lock); 952 /* ... resume the device ... */ 953 foo->is_suspended = 0; 954 pm_runtime_mark_last_busy(&foo->dev); 955 if (foo->num_pending_requests > 0) 956 foo_process_next_request(foo); 957 unlock(&foo->private_lock); 958 return 0; 959 } 960 961The important point is that after foo_io_completion() asks for an autosuspend, 962the foo_runtime_suspend() callback may race with foo_read_or_write(). 963Therefore foo_runtime_suspend() has to check whether there are any pending I/O 964requests (while holding the private lock) before allowing the suspend to 965proceed. 966 967In addition, the power.autosuspend_delay field can be changed by user space at 968any time. If a driver cares about this, it can call 969pm_runtime_autosuspend_expiration() from within the ->runtime_suspend() 970callback while holding its private lock. If the function returns a nonzero 971value then the delay has not yet expired and the callback should return 972-EAGAIN. 973