runtime_pm.rst 45 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946
  1. ==================================================
  2. Runtime Power Management Framework for I/O Devices
  3. ==================================================
  4. (C) 2009-2011 Rafael J. Wysocki <rjw@sisk.pl>, Novell Inc.
  5. (C) 2010 Alan Stern <stern@rowland.harvard.edu>
  6. (C) 2014 Intel Corp., Rafael J. Wysocki <rafael.j.wysocki@intel.com>
  7. 1. Introduction
  8. ===============
  9. Support for runtime power management (runtime PM) of I/O devices is provided
  10. at the power management core (PM core) level by means of:
  11. * The power management workqueue pm_wq in which bus types and device drivers can
  12. put their PM-related work items. It is strongly recommended that pm_wq be
  13. used for queuing all work items related to runtime PM, because this allows
  14. them to be synchronized with system-wide power transitions (suspend to RAM,
  15. hibernation and resume from system sleep states). pm_wq is declared in
  16. include/linux/pm_runtime.h and defined in kernel/power/main.c.
  17. * A number of runtime PM fields in the 'power' member of 'struct device' (which
  18. is of the type 'struct dev_pm_info', defined in include/linux/pm.h) that can
  19. be used for synchronizing runtime PM operations with one another.
  20. * Three device runtime PM callbacks in 'struct dev_pm_ops' (defined in
  21. include/linux/pm.h).
  22. * A set of helper functions defined in drivers/base/power/runtime.c that can be
  23. used for carrying out runtime PM operations in such a way that the
  24. synchronization between them is taken care of by the PM core. Bus types and
  25. device drivers are encouraged to use these functions.
  26. The runtime PM callbacks present in 'struct dev_pm_ops', the device runtime PM
  27. fields of 'struct dev_pm_info' and the core helper functions provided for
  28. runtime PM are described below.
  29. 2. Device Runtime PM Callbacks
  30. ==============================
  31. There are three device runtime PM callbacks defined in 'struct dev_pm_ops'::
  32. struct dev_pm_ops {
  33. ...
  34. int (*runtime_suspend)(struct device *dev);
  35. int (*runtime_resume)(struct device *dev);
  36. int (*runtime_idle)(struct device *dev);
  37. ...
  38. };
  39. The ->runtime_suspend(), ->runtime_resume() and ->runtime_idle() callbacks
  40. are executed by the PM core for the device's subsystem that may be either of
  41. the following:
  42. 1. PM domain of the device, if the device's PM domain object, dev->pm_domain,
  43. is present.
  44. 2. Device type of the device, if both dev->type and dev->type->pm are present.
  45. 3. Device class of the device, if both dev->class and dev->class->pm are
  46. present.
  47. 4. Bus type of the device, if both dev->bus and dev->bus->pm are present.
  48. If the subsystem chosen by applying the above rules doesn't provide the relevant
  49. callback, the PM core will invoke the corresponding driver callback stored in
  50. dev->driver->pm directly (if present).
  51. The PM core always checks which callback to use in the order given above, so the
  52. priority order of callbacks from high to low is: PM domain, device type, class
  53. and bus type. Moreover, the high-priority one will always take precedence over
  54. a low-priority one. The PM domain, bus type, device type and class callbacks
  55. are referred to as subsystem-level callbacks in what follows.
  56. By default, the callbacks are always invoked in process context with interrupts
  57. enabled. However, the pm_runtime_irq_safe() helper function can be used to tell
  58. the PM core that it is safe to run the ->runtime_suspend(), ->runtime_resume()
  59. and ->runtime_idle() callbacks for the given device in atomic context with
  60. interrupts disabled. This implies that the callback routines in question must
  61. not block or sleep, but it also means that the synchronous helper functions
  62. listed at the end of Section 4 may be used for that device within an interrupt
  63. handler or generally in an atomic context.
  64. The subsystem-level suspend callback, if present, is _entirely_ _responsible_
  65. for handling the suspend of the device as appropriate, which may, but need not
  66. include executing the device driver's own ->runtime_suspend() callback (from the
  67. PM core's point of view it is not necessary to implement a ->runtime_suspend()
  68. callback in a device driver as long as the subsystem-level suspend callback
  69. knows what to do to handle the device).
  70. * Once the subsystem-level suspend callback (or the driver suspend callback,
  71. if invoked directly) has completed successfully for the given device, the PM
  72. core regards the device as suspended, which need not mean that it has been
  73. put into a low power state. It is supposed to mean, however, that the
  74. device will not process data and will not communicate with the CPU(s) and
  75. RAM until the appropriate resume callback is executed for it. The runtime
  76. PM status of a device after successful execution of the suspend callback is
  77. 'suspended'.
  78. * If the suspend callback returns -EBUSY or -EAGAIN, the device's runtime PM
  79. status remains 'active', which means that the device _must_ be fully
  80. operational afterwards.
  81. * If the suspend callback returns an error code different from -EBUSY and
  82. -EAGAIN, the PM core regards this as a fatal error and will refuse to run
  83. the helper functions described in Section 4 for the device until its status
  84. is directly set to either 'active', or 'suspended' (the PM core provides
  85. special helper functions for this purpose).
  86. In particular, if the driver requires remote wakeup capability (i.e. hardware
  87. mechanism allowing the device to request a change of its power state, such as
  88. PCI PME) for proper functioning and device_can_wakeup() returns 'false' for the
  89. device, then ->runtime_suspend() should return -EBUSY. On the other hand, if
  90. device_can_wakeup() returns 'true' for the device and the device is put into a
  91. low-power state during the execution of the suspend callback, it is expected
  92. that remote wakeup will be enabled for the device. Generally, remote wakeup
  93. should be enabled for all input devices put into low-power states at run time.
  94. The subsystem-level resume callback, if present, is **entirely responsible** for
  95. handling the resume of the device as appropriate, which may, but need not
  96. include executing the device driver's own ->runtime_resume() callback (from the
  97. PM core's point of view it is not necessary to implement a ->runtime_resume()
  98. callback in a device driver as long as the subsystem-level resume callback knows
  99. what to do to handle the device).
  100. * Once the subsystem-level resume callback (or the driver resume callback, if
  101. invoked directly) has completed successfully, the PM core regards the device
  102. as fully operational, which means that the device _must_ be able to complete
  103. I/O operations as needed. The runtime PM status of the device is then
  104. 'active'.
  105. * If the resume callback returns an error code, the PM core regards this as a
  106. fatal error and will refuse to run the helper functions described in Section
  107. 4 for the device, until its status is directly set to either 'active', or
  108. 'suspended' (by means of special helper functions provided by the PM core
  109. for this purpose).
  110. The idle callback (a subsystem-level one, if present, or the driver one) is
  111. executed by the PM core whenever the device appears to be idle, which is
  112. indicated to the PM core by two counters, the device's usage counter and the
  113. counter of 'active' children of the device.
  114. * If any of these counters is decreased using a helper function provided by
  115. the PM core and it turns out to be equal to zero, the other counter is
  116. checked. If that counter also is equal to zero, the PM core executes the
  117. idle callback with the device as its argument.
  118. The action performed by the idle callback is totally dependent on the subsystem
  119. (or driver) in question, but the expected and recommended action is to check
  120. if the device can be suspended (i.e. if all of the conditions necessary for
  121. suspending the device are satisfied) and to queue up a suspend request for the
  122. device in that case. If there is no idle callback, or if the callback returns
  123. 0, then the PM core will attempt to carry out a runtime suspend of the device,
  124. also respecting devices configured for autosuspend. In essence this means a
  125. call to pm_runtime_autosuspend() (do note that drivers needs to update the
  126. device last busy mark, pm_runtime_mark_last_busy(), to control the delay under
  127. this circumstance). To prevent this (for example, if the callback routine has
  128. started a delayed suspend), the routine must return a non-zero value. Negative
  129. error return codes are ignored by the PM core.
  130. The helper functions provided by the PM core, described in Section 4, guarantee
  131. that the following constraints are met with respect to runtime PM callbacks for
  132. one device:
  133. (1) The callbacks are mutually exclusive (e.g. it is forbidden to execute
  134. ->runtime_suspend() in parallel with ->runtime_resume() or with another
  135. instance of ->runtime_suspend() for the same device) with the exception that
  136. ->runtime_suspend() or ->runtime_resume() can be executed in parallel with
  137. ->runtime_idle() (although ->runtime_idle() will not be started while any
  138. of the other callbacks is being executed for the same device).
  139. (2) ->runtime_idle() and ->runtime_suspend() can only be executed for 'active'
  140. devices (i.e. the PM core will only execute ->runtime_idle() or
  141. ->runtime_suspend() for the devices the runtime PM status of which is
  142. 'active').
  143. (3) ->runtime_idle() and ->runtime_suspend() can only be executed for a device
  144. the usage counter of which is equal to zero _and_ either the counter of
  145. 'active' children of which is equal to zero, or the 'power.ignore_children'
  146. flag of which is set.
  147. (4) ->runtime_resume() can only be executed for 'suspended' devices (i.e. the
  148. PM core will only execute ->runtime_resume() for the devices the runtime
  149. PM status of which is 'suspended').
  150. Additionally, the helper functions provided by the PM core obey the following
  151. rules:
  152. * If ->runtime_suspend() is about to be executed or there's a pending request
  153. to execute it, ->runtime_idle() will not be executed for the same device.
  154. * A request to execute or to schedule the execution of ->runtime_suspend()
  155. will cancel any pending requests to execute ->runtime_idle() for the same
  156. device.
  157. * If ->runtime_resume() is about to be executed or there's a pending request
  158. to execute it, the other callbacks will not be executed for the same device.
  159. * A request to execute ->runtime_resume() will cancel any pending or
  160. scheduled requests to execute the other callbacks for the same device,
  161. except for scheduled autosuspends.
  162. 3. Runtime PM Device Fields
  163. ===========================
  164. The following device runtime PM fields are present in 'struct dev_pm_info', as
  165. defined in include/linux/pm.h:
  166. `struct timer_list suspend_timer;`
  167. - timer used for scheduling (delayed) suspend and autosuspend requests
  168. `unsigned long timer_expires;`
  169. - timer expiration time, in jiffies (if this is different from zero, the
  170. timer is running and will expire at that time, otherwise the timer is not
  171. running)
  172. `struct work_struct work;`
  173. - work structure used for queuing up requests (i.e. work items in pm_wq)
  174. `wait_queue_head_t wait_queue;`
  175. - wait queue used if any of the helper functions needs to wait for another
  176. one to complete
  177. `spinlock_t lock;`
  178. - lock used for synchronization
  179. `atomic_t usage_count;`
  180. - the usage counter of the device
  181. `atomic_t child_count;`
  182. - the count of 'active' children of the device
  183. `unsigned int ignore_children;`
  184. - if set, the value of child_count is ignored (but still updated)
  185. `unsigned int disable_depth;`
  186. - used for disabling the helper functions (they work normally if this is
  187. equal to zero); the initial value of it is 1 (i.e. runtime PM is
  188. initially disabled for all devices)
  189. `int runtime_error;`
  190. - if set, there was a fatal error (one of the callbacks returned error code
  191. as described in Section 2), so the helper functions will not work until
  192. this flag is cleared; this is the error code returned by the failing
  193. callback
  194. `unsigned int idle_notification;`
  195. - if set, ->runtime_idle() is being executed
  196. `unsigned int request_pending;`
  197. - if set, there's a pending request (i.e. a work item queued up into pm_wq)
  198. `enum rpm_request request;`
  199. - type of request that's pending (valid if request_pending is set)
  200. `unsigned int deferred_resume;`
  201. - set if ->runtime_resume() is about to be run while ->runtime_suspend() is
  202. being executed for that device and it is not practical to wait for the
  203. suspend to complete; means "start a resume as soon as you've suspended"
  204. `enum rpm_status runtime_status;`
  205. - the runtime PM status of the device; this field's initial value is
  206. RPM_SUSPENDED, which means that each device is initially regarded by the
  207. PM core as 'suspended', regardless of its real hardware status
  208. `unsigned int runtime_auto;`
  209. - if set, indicates that the user space has allowed the device driver to
  210. power manage the device at run time via the /sys/devices/.../power/control
  211. `interface;` it may only be modified with the help of the
  212. pm_runtime_allow() and pm_runtime_forbid() helper functions
  213. `unsigned int no_callbacks;`
  214. - indicates that the device does not use the runtime PM callbacks (see
  215. Section 8); it may be modified only by the pm_runtime_no_callbacks()
  216. helper function
  217. `unsigned int irq_safe;`
  218. - indicates that the ->runtime_suspend() and ->runtime_resume() callbacks
  219. will be invoked with the spinlock held and interrupts disabled
  220. `unsigned int use_autosuspend;`
  221. - indicates that the device's driver supports delayed autosuspend (see
  222. Section 9); it may be modified only by the
  223. pm_runtime{_dont}_use_autosuspend() helper functions
  224. `unsigned int timer_autosuspends;`
  225. - indicates that the PM core should attempt to carry out an autosuspend
  226. when the timer expires rather than a normal suspend
  227. `int autosuspend_delay;`
  228. - the delay time (in milliseconds) to be used for autosuspend
  229. `unsigned long last_busy;`
  230. - the time (in jiffies) when the pm_runtime_mark_last_busy() helper
  231. function was last called for this device; used in calculating inactivity
  232. periods for autosuspend
  233. All of the above fields are members of the 'power' member of 'struct device'.
  234. 4. Runtime PM Device Helper Functions
  235. =====================================
  236. The following runtime PM helper functions are defined in
  237. drivers/base/power/runtime.c and include/linux/pm_runtime.h:
  238. `void pm_runtime_init(struct device *dev);`
  239. - initialize the device runtime PM fields in 'struct dev_pm_info'
  240. `void pm_runtime_remove(struct device *dev);`
  241. - make sure that the runtime PM of the device will be disabled after
  242. removing the device from device hierarchy
  243. `int pm_runtime_idle(struct device *dev);`
  244. - execute the subsystem-level idle callback for the device; returns an
  245. error code on failure, where -EINPROGRESS means that ->runtime_idle() is
  246. already being executed; if there is no callback or the callback returns 0
  247. then run pm_runtime_autosuspend(dev) and return its result
  248. `int pm_runtime_suspend(struct device *dev);`
  249. - execute the subsystem-level suspend callback for the device; returns 0 on
  250. success, 1 if the device's runtime PM status was already 'suspended', or
  251. error code on failure, where -EAGAIN or -EBUSY means it is safe to attempt
  252. to suspend the device again in future and -EACCES means that
  253. 'power.disable_depth' is different from 0
  254. `int pm_runtime_autosuspend(struct device *dev);`
  255. - same as pm_runtime_suspend() except that the autosuspend delay is taken
  256. `into account;` if pm_runtime_autosuspend_expiration() says the delay has
  257. not yet expired then an autosuspend is scheduled for the appropriate time
  258. and 0 is returned
  259. `int pm_runtime_resume(struct device *dev);`
  260. - execute the subsystem-level resume callback for the device; returns 0 on
  261. success, 1 if the device's runtime PM status was already 'active' or
  262. error code on failure, where -EAGAIN means it may be safe to attempt to
  263. resume the device again in future, but 'power.runtime_error' should be
  264. checked additionally, and -EACCES means that 'power.disable_depth' is
  265. different from 0
  266. `int pm_request_idle(struct device *dev);`
  267. - submit a request to execute the subsystem-level idle callback for the
  268. device (the request is represented by a work item in pm_wq); returns 0 on
  269. success or error code if the request has not been queued up
  270. `int pm_request_autosuspend(struct device *dev);`
  271. - schedule the execution of the subsystem-level suspend callback for the
  272. device when the autosuspend delay has expired; if the delay has already
  273. expired then the work item is queued up immediately
  274. `int pm_schedule_suspend(struct device *dev, unsigned int delay);`
  275. - schedule the execution of the subsystem-level suspend callback for the
  276. device in future, where 'delay' is the time to wait before queuing up a
  277. suspend work item in pm_wq, in milliseconds (if 'delay' is zero, the work
  278. item is queued up immediately); returns 0 on success, 1 if the device's PM
  279. runtime status was already 'suspended', or error code if the request
  280. hasn't been scheduled (or queued up if 'delay' is 0); if the execution of
  281. ->runtime_suspend() is already scheduled and not yet expired, the new
  282. value of 'delay' will be used as the time to wait
  283. `int pm_request_resume(struct device *dev);`
  284. - submit a request to execute the subsystem-level resume callback for the
  285. device (the request is represented by a work item in pm_wq); returns 0 on
  286. success, 1 if the device's runtime PM status was already 'active', or
  287. error code if the request hasn't been queued up
  288. `void pm_runtime_get_noresume(struct device *dev);`
  289. - increment the device's usage counter
  290. `int pm_runtime_get(struct device *dev);`
  291. - increment the device's usage counter, run pm_request_resume(dev) and
  292. return its result
  293. `int pm_runtime_get_sync(struct device *dev);`
  294. - increment the device's usage counter, run pm_runtime_resume(dev) and
  295. return its result
  296. `int pm_runtime_get_if_in_use(struct device *dev);`
  297. - return -EINVAL if 'power.disable_depth' is nonzero; otherwise, if the
  298. runtime PM status is RPM_ACTIVE and the runtime PM usage counter is
  299. nonzero, increment the counter and return 1; otherwise return 0 without
  300. changing the counter
  301. `int pm_runtime_get_if_active(struct device *dev, bool ign_usage_count);`
  302. - return -EINVAL if 'power.disable_depth' is nonzero; otherwise, if the
  303. runtime PM status is RPM_ACTIVE, and either ign_usage_count is true
  304. or the device's usage_count is non-zero, increment the counter and
  305. return 1; otherwise return 0 without changing the counter
  306. `void pm_runtime_put_noidle(struct device *dev);`
  307. - decrement the device's usage counter
  308. `int pm_runtime_put(struct device *dev);`
  309. - decrement the device's usage counter; if the result is 0 then run
  310. pm_request_idle(dev) and return its result
  311. `int pm_runtime_put_autosuspend(struct device *dev);`
  312. - decrement the device's usage counter; if the result is 0 then run
  313. pm_request_autosuspend(dev) and return its result
  314. `int pm_runtime_put_sync(struct device *dev);`
  315. - decrement the device's usage counter; if the result is 0 then run
  316. pm_runtime_idle(dev) and return its result
  317. `int pm_runtime_put_sync_suspend(struct device *dev);`
  318. - decrement the device's usage counter; if the result is 0 then run
  319. pm_runtime_suspend(dev) and return its result
  320. `int pm_runtime_put_sync_autosuspend(struct device *dev);`
  321. - decrement the device's usage counter; if the result is 0 then run
  322. pm_runtime_autosuspend(dev) and return its result
  323. `void pm_runtime_enable(struct device *dev);`
  324. - decrement the device's 'power.disable_depth' field; if that field is equal
  325. to zero, the runtime PM helper functions can execute subsystem-level
  326. callbacks described in Section 2 for the device
  327. `int pm_runtime_disable(struct device *dev);`
  328. - increment the device's 'power.disable_depth' field (if the value of that
  329. field was previously zero, this prevents subsystem-level runtime PM
  330. callbacks from being run for the device), make sure that all of the
  331. pending runtime PM operations on the device are either completed or
  332. canceled; returns 1 if there was a resume request pending and it was
  333. necessary to execute the subsystem-level resume callback for the device
  334. to satisfy that request, otherwise 0 is returned
  335. `int pm_runtime_barrier(struct device *dev);`
  336. - check if there's a resume request pending for the device and resume it
  337. (synchronously) in that case, cancel any other pending runtime PM requests
  338. regarding it and wait for all runtime PM operations on it in progress to
  339. complete; returns 1 if there was a resume request pending and it was
  340. necessary to execute the subsystem-level resume callback for the device to
  341. satisfy that request, otherwise 0 is returned
  342. `void pm_suspend_ignore_children(struct device *dev, bool enable);`
  343. - set/unset the power.ignore_children flag of the device
  344. `int pm_runtime_set_active(struct device *dev);`
  345. - clear the device's 'power.runtime_error' flag, set the device's runtime
  346. PM status to 'active' and update its parent's counter of 'active'
  347. children as appropriate (it is only valid to use this function if
  348. 'power.runtime_error' is set or 'power.disable_depth' is greater than
  349. zero); it will fail and return error code if the device has a parent
  350. which is not active and the 'power.ignore_children' flag of which is unset
  351. `void pm_runtime_set_suspended(struct device *dev);`
  352. - clear the device's 'power.runtime_error' flag, set the device's runtime
  353. PM status to 'suspended' and update its parent's counter of 'active'
  354. children as appropriate (it is only valid to use this function if
  355. 'power.runtime_error' is set or 'power.disable_depth' is greater than
  356. zero)
  357. `bool pm_runtime_active(struct device *dev);`
  358. - return true if the device's runtime PM status is 'active' or its
  359. 'power.disable_depth' field is not equal to zero, or false otherwise
  360. `bool pm_runtime_suspended(struct device *dev);`
  361. - return true if the device's runtime PM status is 'suspended' and its
  362. 'power.disable_depth' field is equal to zero, or false otherwise
  363. `bool pm_runtime_status_suspended(struct device *dev);`
  364. - return true if the device's runtime PM status is 'suspended'
  365. `void pm_runtime_allow(struct device *dev);`
  366. - set the power.runtime_auto flag for the device and decrease its usage
  367. counter (used by the /sys/devices/.../power/control interface to
  368. effectively allow the device to be power managed at run time)
  369. `void pm_runtime_forbid(struct device *dev);`
  370. - unset the power.runtime_auto flag for the device and increase its usage
  371. counter (used by the /sys/devices/.../power/control interface to
  372. effectively prevent the device from being power managed at run time)
  373. `void pm_runtime_no_callbacks(struct device *dev);`
  374. - set the power.no_callbacks flag for the device and remove the runtime
  375. PM attributes from /sys/devices/.../power (or prevent them from being
  376. added when the device is registered)
  377. `void pm_runtime_irq_safe(struct device *dev);`
  378. - set the power.irq_safe flag for the device, causing the runtime-PM
  379. callbacks to be invoked with interrupts off
  380. `bool pm_runtime_is_irq_safe(struct device *dev);`
  381. - return true if power.irq_safe flag was set for the device, causing
  382. the runtime-PM callbacks to be invoked with interrupts off
  383. `void pm_runtime_mark_last_busy(struct device *dev);`
  384. - set the power.last_busy field to the current time
  385. `void pm_runtime_use_autosuspend(struct device *dev);`
  386. - set the power.use_autosuspend flag, enabling autosuspend delays; call
  387. pm_runtime_get_sync if the flag was previously cleared and
  388. power.autosuspend_delay is negative
  389. `void pm_runtime_dont_use_autosuspend(struct device *dev);`
  390. - clear the power.use_autosuspend flag, disabling autosuspend delays;
  391. decrement the device's usage counter if the flag was previously set and
  392. power.autosuspend_delay is negative; call pm_runtime_idle
  393. `void pm_runtime_set_autosuspend_delay(struct device *dev, int delay);`
  394. - set the power.autosuspend_delay value to 'delay' (expressed in
  395. milliseconds); if 'delay' is negative then runtime suspends are
  396. prevented; if power.use_autosuspend is set, pm_runtime_get_sync may be
  397. called or the device's usage counter may be decremented and
  398. pm_runtime_idle called depending on if power.autosuspend_delay is
  399. changed to or from a negative value; if power.use_autosuspend is clear,
  400. pm_runtime_idle is called
  401. `unsigned long pm_runtime_autosuspend_expiration(struct device *dev);`
  402. - calculate the time when the current autosuspend delay period will expire,
  403. based on power.last_busy and power.autosuspend_delay; if the delay time
  404. is 1000 ms or larger then the expiration time is rounded up to the
  405. nearest second; returns 0 if the delay period has already expired or
  406. power.use_autosuspend isn't set, otherwise returns the expiration time
  407. in jiffies
  408. It is safe to execute the following helper functions from interrupt context:
  409. - pm_request_idle()
  410. - pm_request_autosuspend()
  411. - pm_schedule_suspend()
  412. - pm_request_resume()
  413. - pm_runtime_get_noresume()
  414. - pm_runtime_get()
  415. - pm_runtime_put_noidle()
  416. - pm_runtime_put()
  417. - pm_runtime_put_autosuspend()
  418. - pm_runtime_enable()
  419. - pm_suspend_ignore_children()
  420. - pm_runtime_set_active()
  421. - pm_runtime_set_suspended()
  422. - pm_runtime_suspended()
  423. - pm_runtime_mark_last_busy()
  424. - pm_runtime_autosuspend_expiration()
  425. If pm_runtime_irq_safe() has been called for a device then the following helper
  426. functions may also be used in interrupt context:
  427. - pm_runtime_idle()
  428. - pm_runtime_suspend()
  429. - pm_runtime_autosuspend()
  430. - pm_runtime_resume()
  431. - pm_runtime_get_sync()
  432. - pm_runtime_put_sync()
  433. - pm_runtime_put_sync_suspend()
  434. - pm_runtime_put_sync_autosuspend()
  435. 5. Runtime PM Initialization, Device Probing and Removal
  436. ========================================================
  437. Initially, the runtime PM is disabled for all devices, which means that the
  438. majority of the runtime PM helper functions described in Section 4 will return
  439. -EAGAIN until pm_runtime_enable() is called for the device.
  440. In addition to that, the initial runtime PM status of all devices is
  441. 'suspended', but it need not reflect the actual physical state of the device.
  442. Thus, if the device is initially active (i.e. it is able to process I/O), its
  443. runtime PM status must be changed to 'active', with the help of
  444. pm_runtime_set_active(), before pm_runtime_enable() is called for the device.
  445. However, if the device has a parent and the parent's runtime PM is enabled,
  446. calling pm_runtime_set_active() for the device will affect the parent, unless
  447. the parent's 'power.ignore_children' flag is set. Namely, in that case the
  448. parent won't be able to suspend at run time, using the PM core's helper
  449. functions, as long as the child's status is 'active', even if the child's
  450. runtime PM is still disabled (i.e. pm_runtime_enable() hasn't been called for
  451. the child yet or pm_runtime_disable() has been called for it). For this reason,
  452. once pm_runtime_set_active() has been called for the device, pm_runtime_enable()
  453. should be called for it too as soon as reasonably possible or its runtime PM
  454. status should be changed back to 'suspended' with the help of
  455. pm_runtime_set_suspended().
  456. If the default initial runtime PM status of the device (i.e. 'suspended')
  457. reflects the actual state of the device, its bus type's or its driver's
  458. ->probe() callback will likely need to wake it up using one of the PM core's
  459. helper functions described in Section 4. In that case, pm_runtime_resume()
  460. should be used. Of course, for this purpose the device's runtime PM has to be
  461. enabled earlier by calling pm_runtime_enable().
  462. Note, if the device may execute pm_runtime calls during the probe (such as
  463. if it is registers with a subsystem that may call back in) then the
  464. pm_runtime_get_sync() call paired with a pm_runtime_put() call will be
  465. appropriate to ensure that the device is not put back to sleep during the
  466. probe. This can happen with systems such as the network device layer.
  467. It may be desirable to suspend the device once ->probe() has finished.
  468. Therefore the driver core uses the asynchronous pm_request_idle() to submit a
  469. request to execute the subsystem-level idle callback for the device at that
  470. time. A driver that makes use of the runtime autosuspend feature, may want to
  471. update the last busy mark before returning from ->probe().
  472. Moreover, the driver core prevents runtime PM callbacks from racing with the bus
  473. notifier callback in __device_release_driver(), which is necessary, because the
  474. notifier is used by some subsystems to carry out operations affecting the
  475. runtime PM functionality. It does so by calling pm_runtime_get_sync() before
  476. driver_sysfs_remove() and the BUS_NOTIFY_UNBIND_DRIVER notifications. This
  477. resumes the device if it's in the suspended state and prevents it from
  478. being suspended again while those routines are being executed.
  479. To allow bus types and drivers to put devices into the suspended state by
  480. calling pm_runtime_suspend() from their ->remove() routines, the driver core
  481. executes pm_runtime_put_sync() after running the BUS_NOTIFY_UNBIND_DRIVER
  482. notifications in __device_release_driver(). This requires bus types and
  483. drivers to make their ->remove() callbacks avoid races with runtime PM directly,
  484. but also it allows of more flexibility in the handling of devices during the
  485. removal of their drivers.
  486. Drivers in ->remove() callback should undo the runtime PM changes done
  487. in ->probe(). Usually this means calling pm_runtime_disable(),
  488. pm_runtime_dont_use_autosuspend() etc.
  489. The user space can effectively disallow the driver of the device to power manage
  490. it at run time by changing the value of its /sys/devices/.../power/control
  491. attribute to "on", which causes pm_runtime_forbid() to be called. In principle,
  492. this mechanism may also be used by the driver to effectively turn off the
  493. runtime power management of the device until the user space turns it on.
  494. Namely, during the initialization the driver can make sure that the runtime PM
  495. status of the device is 'active' and call pm_runtime_forbid(). It should be
  496. noted, however, that if the user space has already intentionally changed the
  497. value of /sys/devices/.../power/control to "auto" to allow the driver to power
  498. manage the device at run time, the driver may confuse it by using
  499. pm_runtime_forbid() this way.
  500. 6. Runtime PM and System Sleep
  501. ==============================
  502. Runtime PM and system sleep (i.e., system suspend and hibernation, also known
  503. as suspend-to-RAM and suspend-to-disk) interact with each other in a couple of
  504. ways. If a device is active when a system sleep starts, everything is
  505. straightforward. But what should happen if the device is already suspended?
  506. The device may have different wake-up settings for runtime PM and system sleep.
  507. For example, remote wake-up may be enabled for runtime suspend but disallowed
  508. for system sleep (device_may_wakeup(dev) returns 'false'). When this happens,
  509. the subsystem-level system suspend callback is responsible for changing the
  510. device's wake-up setting (it may leave that to the device driver's system
  511. suspend routine). It may be necessary to resume the device and suspend it again
  512. in order to do so. The same is true if the driver uses different power levels
  513. or other settings for runtime suspend and system sleep.
  514. During system resume, the simplest approach is to bring all devices back to full
  515. power, even if they had been suspended before the system suspend began. There
  516. are several reasons for this, including:
  517. * The device might need to switch power levels, wake-up settings, etc.
  518. * Remote wake-up events might have been lost by the firmware.
  519. * The device's children may need the device to be at full power in order
  520. to resume themselves.
  521. * The driver's idea of the device state may not agree with the device's
  522. physical state. This can happen during resume from hibernation.
  523. * The device might need to be reset.
  524. * Even though the device was suspended, if its usage counter was > 0 then most
  525. likely it would need a runtime resume in the near future anyway.
  526. If the device had been suspended before the system suspend began and it's
  527. brought back to full power during resume, then its runtime PM status will have
  528. to be updated to reflect the actual post-system sleep status. The way to do
  529. this is:
  530. - pm_runtime_disable(dev);
  531. - pm_runtime_set_active(dev);
  532. - pm_runtime_enable(dev);
  533. The PM core always increments the runtime usage counter before calling the
  534. ->suspend() callback and decrements it after calling the ->resume() callback.
  535. Hence disabling runtime PM temporarily like this will not cause any runtime
  536. suspend attempts to be permanently lost. If the usage count goes to zero
  537. following the return of the ->resume() callback, the ->runtime_idle() callback
  538. will be invoked as usual.
  539. On some systems, however, system sleep is not entered through a global firmware
  540. or hardware operation. Instead, all hardware components are put into low-power
  541. states directly by the kernel in a coordinated way. Then, the system sleep
  542. state effectively follows from the states the hardware components end up in
  543. and the system is woken up from that state by a hardware interrupt or a similar
  544. mechanism entirely under the kernel's control. As a result, the kernel never
  545. gives control away and the states of all devices during resume are precisely
  546. known to it. If that is the case and none of the situations listed above takes
  547. place (in particular, if the system is not waking up from hibernation), it may
  548. be more efficient to leave the devices that had been suspended before the system
  549. suspend began in the suspended state.
  550. To this end, the PM core provides a mechanism allowing some coordination between
  551. different levels of device hierarchy. Namely, if a system suspend .prepare()
  552. callback returns a positive number for a device, that indicates to the PM core
  553. that the device appears to be runtime-suspended and its state is fine, so it
  554. may be left in runtime suspend provided that all of its descendants are also
  555. left in runtime suspend. If that happens, the PM core will not execute any
  556. system suspend and resume callbacks for all of those devices, except for the
  557. complete callback, which is then entirely responsible for handling the device
  558. as appropriate. This only applies to system suspend transitions that are not
  559. related to hibernation (see Documentation/driver-api/pm/devices.rst for more
  560. information).
  561. The PM core does its best to reduce the probability of race conditions between
  562. the runtime PM and system suspend/resume (and hibernation) callbacks by carrying
  563. out the following operations:
  564. * During system suspend pm_runtime_get_noresume() is called for every device
  565. right before executing the subsystem-level .prepare() callback for it and
  566. pm_runtime_barrier() is called for every device right before executing the
  567. subsystem-level .suspend() callback for it. In addition to that the PM core
  568. calls __pm_runtime_disable() with 'false' as the second argument for every
  569. device right before executing the subsystem-level .suspend_late() callback
  570. for it.
  571. * During system resume pm_runtime_enable() and pm_runtime_put() are called for
  572. every device right after executing the subsystem-level .resume_early()
  573. callback and right after executing the subsystem-level .complete() callback
  574. for it, respectively.
  575. 7. Generic subsystem callbacks
  576. Subsystems may wish to conserve code space by using the set of generic power
  577. management callbacks provided by the PM core, defined in
  578. driver/base/power/generic_ops.c:
  579. `int pm_generic_runtime_suspend(struct device *dev);`
  580. - invoke the ->runtime_suspend() callback provided by the driver of this
  581. device and return its result, or return 0 if not defined
  582. `int pm_generic_runtime_resume(struct device *dev);`
  583. - invoke the ->runtime_resume() callback provided by the driver of this
  584. device and return its result, or return 0 if not defined
  585. `int pm_generic_suspend(struct device *dev);`
  586. - if the device has not been suspended at run time, invoke the ->suspend()
  587. callback provided by its driver and return its result, or return 0 if not
  588. defined
  589. `int pm_generic_suspend_noirq(struct device *dev);`
  590. - if pm_runtime_suspended(dev) returns "false", invoke the ->suspend_noirq()
  591. callback provided by the device's driver and return its result, or return
  592. 0 if not defined
  593. `int pm_generic_resume(struct device *dev);`
  594. - invoke the ->resume() callback provided by the driver of this device and,
  595. if successful, change the device's runtime PM status to 'active'
  596. `int pm_generic_resume_noirq(struct device *dev);`
  597. - invoke the ->resume_noirq() callback provided by the driver of this device
  598. `int pm_generic_freeze(struct device *dev);`
  599. - if the device has not been suspended at run time, invoke the ->freeze()
  600. callback provided by its driver and return its result, or return 0 if not
  601. defined
  602. `int pm_generic_freeze_noirq(struct device *dev);`
  603. - if pm_runtime_suspended(dev) returns "false", invoke the ->freeze_noirq()
  604. callback provided by the device's driver and return its result, or return
  605. 0 if not defined
  606. `int pm_generic_thaw(struct device *dev);`
  607. - if the device has not been suspended at run time, invoke the ->thaw()
  608. callback provided by its driver and return its result, or return 0 if not
  609. defined
  610. `int pm_generic_thaw_noirq(struct device *dev);`
  611. - if pm_runtime_suspended(dev) returns "false", invoke the ->thaw_noirq()
  612. callback provided by the device's driver and return its result, or return
  613. 0 if not defined
  614. `int pm_generic_poweroff(struct device *dev);`
  615. - if the device has not been suspended at run time, invoke the ->poweroff()
  616. callback provided by its driver and return its result, or return 0 if not
  617. defined
  618. `int pm_generic_poweroff_noirq(struct device *dev);`
  619. - if pm_runtime_suspended(dev) returns "false", run the ->poweroff_noirq()
  620. callback provided by the device's driver and return its result, or return
  621. 0 if not defined
  622. `int pm_generic_restore(struct device *dev);`
  623. - invoke the ->restore() callback provided by the driver of this device and,
  624. if successful, change the device's runtime PM status to 'active'
  625. `int pm_generic_restore_noirq(struct device *dev);`
  626. - invoke the ->restore_noirq() callback provided by the device's driver
  627. These functions are the defaults used by the PM core, if a subsystem doesn't
  628. provide its own callbacks for ->runtime_idle(), ->runtime_suspend(),
  629. ->runtime_resume(), ->suspend(), ->suspend_noirq(), ->resume(),
  630. ->resume_noirq(), ->freeze(), ->freeze_noirq(), ->thaw(), ->thaw_noirq(),
  631. ->poweroff(), ->poweroff_noirq(), ->restore(), ->restore_noirq() in the
  632. subsystem-level dev_pm_ops structure.
  633. Device drivers that wish to use the same function as a system suspend, freeze,
  634. poweroff and runtime suspend callback, and similarly for system resume, thaw,
  635. restore, and runtime resume, can achieve this with the help of the
  636. UNIVERSAL_DEV_PM_OPS macro defined in include/linux/pm.h (possibly setting its
  637. last argument to NULL).
  638. 8. "No-Callback" Devices
  639. ========================
  640. Some "devices" are only logical sub-devices of their parent and cannot be
  641. power-managed on their own. (The prototype example is a USB interface. Entire
  642. USB devices can go into low-power mode or send wake-up requests, but neither is
  643. possible for individual interfaces.) The drivers for these devices have no
  644. need of runtime PM callbacks; if the callbacks did exist, ->runtime_suspend()
  645. and ->runtime_resume() would always return 0 without doing anything else and
  646. ->runtime_idle() would always call pm_runtime_suspend().
  647. Subsystems can tell the PM core about these devices by calling
  648. pm_runtime_no_callbacks(). This should be done after the device structure is
  649. initialized and before it is registered (although after device registration is
  650. also okay). The routine will set the device's power.no_callbacks flag and
  651. prevent the non-debugging runtime PM sysfs attributes from being created.
  652. When power.no_callbacks is set, the PM core will not invoke the
  653. ->runtime_idle(), ->runtime_suspend(), or ->runtime_resume() callbacks.
  654. Instead it will assume that suspends and resumes always succeed and that idle
  655. devices should be suspended.
  656. As a consequence, the PM core will never directly inform the device's subsystem
  657. or driver about runtime power changes. Instead, the driver for the device's
  658. parent must take responsibility for telling the device's driver when the
  659. parent's power state changes.
  660. 9. Autosuspend, or automatically-delayed suspends
  661. =================================================
  662. Changing a device's power state isn't free; it requires both time and energy.
  663. A device should be put in a low-power state only when there's some reason to
  664. think it will remain in that state for a substantial time. A common heuristic
  665. says that a device which hasn't been used for a while is liable to remain
  666. unused; following this advice, drivers should not allow devices to be suspended
  667. at runtime until they have been inactive for some minimum period. Even when
  668. the heuristic ends up being non-optimal, it will still prevent devices from
  669. "bouncing" too rapidly between low-power and full-power states.
  670. The term "autosuspend" is an historical remnant. It doesn't mean that the
  671. device is automatically suspended (the subsystem or driver still has to call
  672. the appropriate PM routines); rather it means that runtime suspends will
  673. automatically be delayed until the desired period of inactivity has elapsed.
  674. Inactivity is determined based on the power.last_busy field. Drivers should
  675. call pm_runtime_mark_last_busy() to update this field after carrying out I/O,
  676. typically just before calling pm_runtime_put_autosuspend(). The desired length
  677. of the inactivity period is a matter of policy. Subsystems can set this length
  678. initially by calling pm_runtime_set_autosuspend_delay(), but after device
  679. registration the length should be controlled by user space, using the
  680. /sys/devices/.../power/autosuspend_delay_ms attribute.
  681. In order to use autosuspend, subsystems or drivers must call
  682. pm_runtime_use_autosuspend() (preferably before registering the device), and
  683. thereafter they should use the various `*_autosuspend()` helper functions
  684. instead of the non-autosuspend counterparts::
  685. Instead of: pm_runtime_suspend use: pm_runtime_autosuspend;
  686. Instead of: pm_schedule_suspend use: pm_request_autosuspend;
  687. Instead of: pm_runtime_put use: pm_runtime_put_autosuspend;
  688. Instead of: pm_runtime_put_sync use: pm_runtime_put_sync_autosuspend.
  689. Drivers may also continue to use the non-autosuspend helper functions; they
  690. will behave normally, which means sometimes taking the autosuspend delay into
  691. account (see pm_runtime_idle).
  692. Under some circumstances a driver or subsystem may want to prevent a device
  693. from autosuspending immediately, even though the usage counter is zero and the
  694. autosuspend delay time has expired. If the ->runtime_suspend() callback
  695. returns -EAGAIN or -EBUSY, and if the next autosuspend delay expiration time is
  696. in the future (as it normally would be if the callback invoked
  697. pm_runtime_mark_last_busy()), the PM core will automatically reschedule the
  698. autosuspend. The ->runtime_suspend() callback can't do this rescheduling
  699. itself because no suspend requests of any kind are accepted while the device is
  700. suspending (i.e., while the callback is running).
  701. The implementation is well suited for asynchronous use in interrupt contexts.
  702. However such use inevitably involves races, because the PM core can't
  703. synchronize ->runtime_suspend() callbacks with the arrival of I/O requests.
  704. This synchronization must be handled by the driver, using its private lock.
  705. Here is a schematic pseudo-code example::
  706. foo_read_or_write(struct foo_priv *foo, void *data)
  707. {
  708. lock(&foo->private_lock);
  709. add_request_to_io_queue(foo, data);
  710. if (foo->num_pending_requests++ == 0)
  711. pm_runtime_get(&foo->dev);
  712. if (!foo->is_suspended)
  713. foo_process_next_request(foo);
  714. unlock(&foo->private_lock);
  715. }
  716. foo_io_completion(struct foo_priv *foo, void *req)
  717. {
  718. lock(&foo->private_lock);
  719. if (--foo->num_pending_requests == 0) {
  720. pm_runtime_mark_last_busy(&foo->dev);
  721. pm_runtime_put_autosuspend(&foo->dev);
  722. } else {
  723. foo_process_next_request(foo);
  724. }
  725. unlock(&foo->private_lock);
  726. /* Send req result back to the user ... */
  727. }
  728. int foo_runtime_suspend(struct device *dev)
  729. {
  730. struct foo_priv foo = container_of(dev, ...);
  731. int ret = 0;
  732. lock(&foo->private_lock);
  733. if (foo->num_pending_requests > 0) {
  734. ret = -EBUSY;
  735. } else {
  736. /* ... suspend the device ... */
  737. foo->is_suspended = 1;
  738. }
  739. unlock(&foo->private_lock);
  740. return ret;
  741. }
  742. int foo_runtime_resume(struct device *dev)
  743. {
  744. struct foo_priv foo = container_of(dev, ...);
  745. lock(&foo->private_lock);
  746. /* ... resume the device ... */
  747. foo->is_suspended = 0;
  748. pm_runtime_mark_last_busy(&foo->dev);
  749. if (foo->num_pending_requests > 0)
  750. foo_process_next_request(foo);
  751. unlock(&foo->private_lock);
  752. return 0;
  753. }
  754. The important point is that after foo_io_completion() asks for an autosuspend,
  755. the foo_runtime_suspend() callback may race with foo_read_or_write().
  756. Therefore foo_runtime_suspend() has to check whether there are any pending I/O
  757. requests (while holding the private lock) before allowing the suspend to
  758. proceed.
  759. In addition, the power.autosuspend_delay field can be changed by user space at
  760. any time. If a driver cares about this, it can call
  761. pm_runtime_autosuspend_expiration() from within the ->runtime_suspend()
  762. callback while holding its private lock. If the function returns a nonzero
  763. value then the delay has not yet expired and the callback should return
  764. -EAGAIN.