Commit | Line | Data |
---|---|---|
5e928f77 RW |
1 | Run-time Power Management Framework for I/O Devices |
2 | ||
3 | (C) 2009 Rafael J. Wysocki <rjw@sisk.pl>, Novell Inc. | |
7490e442 | 4 | (C) 2010 Alan Stern <stern@rowland.harvard.edu> |
5e928f77 RW |
5 | |
6 | 1. Introduction | |
7 | ||
8 | Support for run-time power management (run-time PM) of I/O devices is provided | |
9 | at the power management core (PM core) level by means of: | |
10 | ||
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 run-time 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 | ||
18 | * A number of run-time PM fields in the 'power' member of 'struct device' (which | |
19 | is of the type 'struct dev_pm_info', defined in include/linux/pm.h) that can | |
20 | be used for synchronizing run-time PM operations with one another. | |
21 | ||
22 | * Three device run-time PM callbacks in 'struct dev_pm_ops' (defined in | |
23 | include/linux/pm.h). | |
24 | ||
25 | * A set of helper functions defined in drivers/base/power/runtime.c that can be | |
26 | used for carrying out run-time PM operations in such a way that the | |
27 | synchronization between them is taken care of by the PM core. Bus types and | |
28 | device drivers are encouraged to use these functions. | |
29 | ||
30 | The run-time PM callbacks present in 'struct dev_pm_ops', the device run-time PM | |
31 | fields of 'struct dev_pm_info' and the core helper functions provided for | |
32 | run-time PM are described below. | |
33 | ||
34 | 2. Device Run-time PM Callbacks | |
35 | ||
36 | There are three device run-time PM callbacks defined in 'struct dev_pm_ops': | |
37 | ||
38 | struct dev_pm_ops { | |
39 | ... | |
40 | int (*runtime_suspend)(struct device *dev); | |
41 | int (*runtime_resume)(struct device *dev); | |
e1b1903e | 42 | int (*runtime_idle)(struct device *dev); |
5e928f77 RW |
43 | ... |
44 | }; | |
45 | ||
a6ab7aa9 RW |
46 | The ->runtime_suspend(), ->runtime_resume() and ->runtime_idle() callbacks are |
47 | executed by the PM core for either the bus type, or device type (if the bus | |
48 | type's callback is not defined), or device class (if the bus type's and device | |
49 | type's callbacks are not defined) of given device. The bus type, device type | |
50 | and device class callbacks are referred to as subsystem-level callbacks in what | |
51 | follows. | |
52 | ||
53 | The subsystem-level suspend callback is _entirely_ _responsible_ for handling | |
54 | the suspend of the device as appropriate, which may, but need not include | |
55 | executing the device driver's own ->runtime_suspend() callback (from the | |
5e928f77 | 56 | PM core's point of view it is not necessary to implement a ->runtime_suspend() |
a6ab7aa9 RW |
57 | callback in a device driver as long as the subsystem-level suspend callback |
58 | knows what to do to handle the device). | |
5e928f77 | 59 | |
a6ab7aa9 | 60 | * Once the subsystem-level suspend callback has completed successfully |
5e928f77 RW |
61 | for given device, the PM core regards the device as suspended, which need |
62 | not mean that the device has been put into a low power state. It is | |
63 | supposed to mean, however, that the device will not process data and will | |
a6ab7aa9 RW |
64 | not communicate with the CPU(s) and RAM until the subsystem-level resume |
65 | callback is executed for it. The run-time PM status of a device after | |
66 | successful execution of the subsystem-level suspend callback is 'suspended'. | |
67 | ||
68 | * If the subsystem-level suspend callback returns -EBUSY or -EAGAIN, | |
69 | the device's run-time PM status is 'active', which means that the device | |
70 | _must_ be fully operational afterwards. | |
71 | ||
72 | * If the subsystem-level suspend callback returns an error code different | |
73 | from -EBUSY or -EAGAIN, the PM core regards this as a fatal error and will | |
74 | refuse to run the helper functions described in Section 4 for the device, | |
75 | until the status of it is directly set either to 'active', or to 'suspended' | |
76 | (the PM core provides special helper functions for this purpose). | |
77 | ||
78 | In particular, if the driver requires remote wake-up capability (i.e. hardware | |
79 | mechanism allowing the device to request a change of its power state, such as | |
80 | PCI PME) for proper functioning and device_run_wake() returns 'false' for the | |
81 | device, then ->runtime_suspend() should return -EBUSY. On the other hand, if | |
82 | device_run_wake() returns 'true' for the device and the device is put into a low | |
83 | power state during the execution of the subsystem-level suspend callback, it is | |
84 | expected that remote wake-up will be enabled for the device. Generally, remote | |
85 | wake-up should be enabled for all input devices put into a low power state at | |
86 | run time. | |
87 | ||
88 | The subsystem-level resume callback is _entirely_ _responsible_ for handling the | |
89 | resume of the device as appropriate, which may, but need not include executing | |
90 | the device driver's own ->runtime_resume() callback (from the PM core's point of | |
91 | view it is not necessary to implement a ->runtime_resume() callback in a device | |
92 | driver as long as the subsystem-level resume callback knows what to do to handle | |
93 | the device). | |
94 | ||
95 | * Once the subsystem-level resume callback has completed successfully, the PM | |
96 | core regards the device as fully operational, which means that the device | |
97 | _must_ be able to complete I/O operations as needed. The run-time PM status | |
98 | of the device is then 'active'. | |
99 | ||
100 | * If the subsystem-level resume callback returns an error code, the PM core | |
101 | regards this as a fatal error and will refuse to run the helper functions | |
102 | described in Section 4 for the device, until its status is directly set | |
103 | either to 'active' or to 'suspended' (the PM core provides special helper | |
104 | functions for this purpose). | |
105 | ||
106 | The subsystem-level idle callback is executed by the PM core whenever the device | |
107 | appears to be idle, which is indicated to the PM core by two counters, the | |
108 | device's usage counter and the counter of 'active' children of the device. | |
5e928f77 RW |
109 | |
110 | * If any of these counters is decreased using a helper function provided by | |
111 | the PM core and it turns out to be equal to zero, the other counter is | |
112 | checked. If that counter also is equal to zero, the PM core executes the | |
a6ab7aa9 | 113 | subsystem-level idle callback with the device as an argument. |
5e928f77 | 114 | |
a6ab7aa9 RW |
115 | The action performed by a subsystem-level idle callback is totally dependent on |
116 | the subsystem in question, but the expected and recommended action is to check | |
117 | if the device can be suspended (i.e. if all of the conditions necessary for | |
118 | suspending the device are satisfied) and to queue up a suspend request for the | |
119 | device in that case. The value returned by this callback is ignored by the PM | |
120 | core. | |
5e928f77 RW |
121 | |
122 | The helper functions provided by the PM core, described in Section 4, guarantee | |
123 | that the following constraints are met with respect to the bus type's run-time | |
124 | PM callbacks: | |
125 | ||
126 | (1) The callbacks are mutually exclusive (e.g. it is forbidden to execute | |
127 | ->runtime_suspend() in parallel with ->runtime_resume() or with another | |
128 | instance of ->runtime_suspend() for the same device) with the exception that | |
129 | ->runtime_suspend() or ->runtime_resume() can be executed in parallel with | |
130 | ->runtime_idle() (although ->runtime_idle() will not be started while any | |
131 | of the other callbacks is being executed for the same device). | |
132 | ||
133 | (2) ->runtime_idle() and ->runtime_suspend() can only be executed for 'active' | |
134 | devices (i.e. the PM core will only execute ->runtime_idle() or | |
135 | ->runtime_suspend() for the devices the run-time PM status of which is | |
136 | 'active'). | |
137 | ||
138 | (3) ->runtime_idle() and ->runtime_suspend() can only be executed for a device | |
139 | the usage counter of which is equal to zero _and_ either the counter of | |
140 | 'active' children of which is equal to zero, or the 'power.ignore_children' | |
141 | flag of which is set. | |
142 | ||
143 | (4) ->runtime_resume() can only be executed for 'suspended' devices (i.e. the | |
144 | PM core will only execute ->runtime_resume() for the devices the run-time | |
145 | PM status of which is 'suspended'). | |
146 | ||
147 | Additionally, the helper functions provided by the PM core obey the following | |
148 | rules: | |
149 | ||
150 | * If ->runtime_suspend() is about to be executed or there's a pending request | |
151 | to execute it, ->runtime_idle() will not be executed for the same device. | |
152 | ||
153 | * A request to execute or to schedule the execution of ->runtime_suspend() | |
154 | will cancel any pending requests to execute ->runtime_idle() for the same | |
155 | device. | |
156 | ||
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 | ||
160 | * A request to execute ->runtime_resume() will cancel any pending or | |
15bcb91d AS |
161 | scheduled requests to execute the other callbacks for the same device, |
162 | except for scheduled autosuspends. | |
5e928f77 RW |
163 | |
164 | 3. Run-time PM Device Fields | |
165 | ||
166 | The following device run-time PM fields are present in 'struct dev_pm_info', as | |
167 | defined in include/linux/pm.h: | |
168 | ||
169 | struct timer_list suspend_timer; | |
15bcb91d | 170 | - timer used for scheduling (delayed) suspend and autosuspend requests |
5e928f77 RW |
171 | |
172 | unsigned long timer_expires; | |
173 | - timer expiration time, in jiffies (if this is different from zero, the | |
174 | timer is running and will expire at that time, otherwise the timer is not | |
175 | running) | |
176 | ||
177 | struct work_struct work; | |
178 | - work structure used for queuing up requests (i.e. work items in pm_wq) | |
179 | ||
180 | wait_queue_head_t wait_queue; | |
181 | - wait queue used if any of the helper functions needs to wait for another | |
182 | one to complete | |
183 | ||
184 | spinlock_t lock; | |
185 | - lock used for synchronisation | |
186 | ||
187 | atomic_t usage_count; | |
188 | - the usage counter of the device | |
189 | ||
190 | atomic_t child_count; | |
191 | - the count of 'active' children of the device | |
192 | ||
193 | unsigned int ignore_children; | |
194 | - if set, the value of child_count is ignored (but still updated) | |
195 | ||
196 | unsigned int disable_depth; | |
197 | - used for disabling the helper funcions (they work normally if this is | |
198 | equal to zero); the initial value of it is 1 (i.e. run-time PM is | |
199 | initially disabled for all devices) | |
200 | ||
201 | unsigned int runtime_error; | |
202 | - if set, there was a fatal error (one of the callbacks returned error code | |
203 | as described in Section 2), so the helper funtions will not work until | |
204 | this flag is cleared; this is the error code returned by the failing | |
205 | callback | |
206 | ||
207 | unsigned int idle_notification; | |
208 | - if set, ->runtime_idle() is being executed | |
209 | ||
210 | unsigned int request_pending; | |
211 | - if set, there's a pending request (i.e. a work item queued up into pm_wq) | |
212 | ||
213 | enum rpm_request request; | |
214 | - type of request that's pending (valid if request_pending is set) | |
215 | ||
216 | unsigned int deferred_resume; | |
217 | - set if ->runtime_resume() is about to be run while ->runtime_suspend() is | |
218 | being executed for that device and it is not practical to wait for the | |
219 | suspend to complete; means "start a resume as soon as you've suspended" | |
220 | ||
7a1a8eb5 RW |
221 | unsigned int run_wake; |
222 | - set if the device is capable of generating run-time wake-up events | |
223 | ||
5e928f77 RW |
224 | enum rpm_status runtime_status; |
225 | - the run-time PM status of the device; this field's initial value is | |
226 | RPM_SUSPENDED, which means that each device is initially regarded by the | |
227 | PM core as 'suspended', regardless of its real hardware status | |
228 | ||
87d1b3e6 RW |
229 | unsigned int runtime_auto; |
230 | - if set, indicates that the user space has allowed the device driver to | |
231 | power manage the device at run time via the /sys/devices/.../power/control | |
232 | interface; it may only be modified with the help of the pm_runtime_allow() | |
233 | and pm_runtime_forbid() helper functions | |
234 | ||
7490e442 AS |
235 | unsigned int no_callbacks; |
236 | - indicates that the device does not use the run-time PM callbacks (see | |
237 | Section 8); it may be modified only by the pm_runtime_no_callbacks() | |
238 | helper function | |
239 | ||
15bcb91d AS |
240 | unsigned int use_autosuspend; |
241 | - indicates that the device's driver supports delayed autosuspend (see | |
242 | Section 9); it may be modified only by the | |
243 | pm_runtime{_dont}_use_autosuspend() helper functions | |
244 | ||
245 | unsigned int timer_autosuspends; | |
246 | - indicates that the PM core should attempt to carry out an autosuspend | |
247 | when the timer expires rather than a normal suspend | |
248 | ||
249 | int autosuspend_delay; | |
250 | - the delay time (in milliseconds) to be used for autosuspend | |
251 | ||
252 | unsigned long last_busy; | |
253 | - the time (in jiffies) when the pm_runtime_mark_last_busy() helper | |
254 | function was last called for this device; used in calculating inactivity | |
255 | periods for autosuspend | |
256 | ||
5e928f77 RW |
257 | All of the above fields are members of the 'power' member of 'struct device'. |
258 | ||
259 | 4. Run-time PM Device Helper Functions | |
260 | ||
261 | The following run-time PM helper functions are defined in | |
262 | drivers/base/power/runtime.c and include/linux/pm_runtime.h: | |
263 | ||
264 | void pm_runtime_init(struct device *dev); | |
265 | - initialize the device run-time PM fields in 'struct dev_pm_info' | |
266 | ||
267 | void pm_runtime_remove(struct device *dev); | |
268 | - make sure that the run-time PM of the device will be disabled after | |
269 | removing the device from device hierarchy | |
270 | ||
271 | int pm_runtime_idle(struct device *dev); | |
a6ab7aa9 RW |
272 | - execute the subsystem-level idle callback for the device; returns 0 on |
273 | success or error code on failure, where -EINPROGRESS means that | |
274 | ->runtime_idle() is already being executed | |
5e928f77 RW |
275 | |
276 | int pm_runtime_suspend(struct device *dev); | |
a6ab7aa9 | 277 | - execute the subsystem-level suspend callback for the device; returns 0 on |
5e928f77 RW |
278 | success, 1 if the device's run-time PM status was already 'suspended', or |
279 | error code on failure, where -EAGAIN or -EBUSY means it is safe to attempt | |
280 | to suspend the device again in future | |
281 | ||
15bcb91d AS |
282 | int pm_runtime_autosuspend(struct device *dev); |
283 | - same as pm_runtime_suspend() except that the autosuspend delay is taken | |
284 | into account; if pm_runtime_autosuspend_expiration() says the delay has | |
285 | not yet expired then an autosuspend is scheduled for the appropriate time | |
286 | and 0 is returned | |
287 | ||
5e928f77 | 288 | int pm_runtime_resume(struct device *dev); |
de8164fb | 289 | - execute the subsystem-level resume callback for the device; returns 0 on |
5e928f77 RW |
290 | success, 1 if the device's run-time PM status was already 'active' or |
291 | error code on failure, where -EAGAIN means it may be safe to attempt to | |
292 | resume the device again in future, but 'power.runtime_error' should be | |
293 | checked additionally | |
294 | ||
295 | int pm_request_idle(struct device *dev); | |
a6ab7aa9 RW |
296 | - submit a request to execute the subsystem-level idle callback for the |
297 | device (the request is represented by a work item in pm_wq); returns 0 on | |
298 | success or error code if the request has not been queued up | |
5e928f77 | 299 | |
15bcb91d AS |
300 | int pm_request_autosuspend(struct device *dev); |
301 | - schedule the execution of the subsystem-level suspend callback for the | |
302 | device when the autosuspend delay has expired; if the delay has already | |
303 | expired then the work item is queued up immediately | |
304 | ||
5e928f77 | 305 | int pm_schedule_suspend(struct device *dev, unsigned int delay); |
a6ab7aa9 RW |
306 | - schedule the execution of the subsystem-level suspend callback for the |
307 | device in future, where 'delay' is the time to wait before queuing up a | |
308 | suspend work item in pm_wq, in milliseconds (if 'delay' is zero, the work | |
309 | item is queued up immediately); returns 0 on success, 1 if the device's PM | |
5e928f77 RW |
310 | run-time status was already 'suspended', or error code if the request |
311 | hasn't been scheduled (or queued up if 'delay' is 0); if the execution of | |
312 | ->runtime_suspend() is already scheduled and not yet expired, the new | |
313 | value of 'delay' will be used as the time to wait | |
314 | ||
315 | int pm_request_resume(struct device *dev); | |
a6ab7aa9 RW |
316 | - submit a request to execute the subsystem-level resume callback for the |
317 | device (the request is represented by a work item in pm_wq); returns 0 on | |
5e928f77 RW |
318 | success, 1 if the device's run-time PM status was already 'active', or |
319 | error code if the request hasn't been queued up | |
320 | ||
321 | void pm_runtime_get_noresume(struct device *dev); | |
322 | - increment the device's usage counter | |
323 | ||
324 | int pm_runtime_get(struct device *dev); | |
325 | - increment the device's usage counter, run pm_request_resume(dev) and | |
326 | return its result | |
327 | ||
328 | int pm_runtime_get_sync(struct device *dev); | |
329 | - increment the device's usage counter, run pm_runtime_resume(dev) and | |
330 | return its result | |
331 | ||
332 | void pm_runtime_put_noidle(struct device *dev); | |
333 | - decrement the device's usage counter | |
334 | ||
335 | int pm_runtime_put(struct device *dev); | |
15bcb91d AS |
336 | - decrement the device's usage counter; if the result is 0 then run |
337 | pm_request_idle(dev) and return its result | |
338 | ||
339 | int pm_runtime_put_autosuspend(struct device *dev); | |
340 | - decrement the device's usage counter; if the result is 0 then run | |
341 | pm_request_autosuspend(dev) and return its result | |
5e928f77 RW |
342 | |
343 | int pm_runtime_put_sync(struct device *dev); | |
15bcb91d AS |
344 | - decrement the device's usage counter; if the result is 0 then run |
345 | pm_runtime_idle(dev) and return its result | |
346 | ||
347 | int pm_runtime_put_sync_autosuspend(struct device *dev); | |
348 | - decrement the device's usage counter; if the result is 0 then run | |
349 | pm_runtime_autosuspend(dev) and return its result | |
5e928f77 RW |
350 | |
351 | void pm_runtime_enable(struct device *dev); | |
352 | - enable the run-time PM helper functions to run the device bus type's | |
353 | run-time PM callbacks described in Section 2 | |
354 | ||
355 | int pm_runtime_disable(struct device *dev); | |
a6ab7aa9 RW |
356 | - prevent the run-time PM helper functions from running subsystem-level |
357 | run-time PM callbacks for the device, make sure that all of the pending | |
358 | run-time PM operations on the device are either completed or canceled; | |
359 | returns 1 if there was a resume request pending and it was necessary to | |
360 | execute the subsystem-level resume callback for the device to satisfy that | |
361 | request, otherwise 0 is returned | |
5e928f77 RW |
362 | |
363 | void pm_suspend_ignore_children(struct device *dev, bool enable); | |
364 | - set/unset the power.ignore_children flag of the device | |
365 | ||
366 | int pm_runtime_set_active(struct device *dev); | |
367 | - clear the device's 'power.runtime_error' flag, set the device's run-time | |
368 | PM status to 'active' and update its parent's counter of 'active' | |
369 | children as appropriate (it is only valid to use this function if | |
370 | 'power.runtime_error' is set or 'power.disable_depth' is greater than | |
371 | zero); it will fail and return error code if the device has a parent | |
372 | which is not active and the 'power.ignore_children' flag of which is unset | |
373 | ||
374 | void pm_runtime_set_suspended(struct device *dev); | |
375 | - clear the device's 'power.runtime_error' flag, set the device's run-time | |
376 | PM status to 'suspended' and update its parent's counter of 'active' | |
377 | children as appropriate (it is only valid to use this function if | |
378 | 'power.runtime_error' is set or 'power.disable_depth' is greater than | |
379 | zero) | |
380 | ||
d690b2cd | 381 | bool pm_runtime_suspended(struct device *dev); |
f08f5a0a RW |
382 | - return true if the device's runtime PM status is 'suspended' and its |
383 | 'power.disable_depth' field is equal to zero, or false otherwise | |
d690b2cd | 384 | |
87d1b3e6 RW |
385 | void pm_runtime_allow(struct device *dev); |
386 | - set the power.runtime_auto flag for the device and decrease its usage | |
387 | counter (used by the /sys/devices/.../power/control interface to | |
388 | effectively allow the device to be power managed at run time) | |
389 | ||
390 | void pm_runtime_forbid(struct device *dev); | |
391 | - unset the power.runtime_auto flag for the device and increase its usage | |
392 | counter (used by the /sys/devices/.../power/control interface to | |
393 | effectively prevent the device from being power managed at run time) | |
394 | ||
7490e442 AS |
395 | void pm_runtime_no_callbacks(struct device *dev); |
396 | - set the power.no_callbacks flag for the device and remove the run-time | |
397 | PM attributes from /sys/devices/.../power (or prevent them from being | |
398 | added when the device is registered) | |
399 | ||
15bcb91d AS |
400 | void pm_runtime_mark_last_busy(struct device *dev); |
401 | - set the power.last_busy field to the current time | |
402 | ||
403 | void pm_runtime_use_autosuspend(struct device *dev); | |
404 | - set the power.use_autosuspend flag, enabling autosuspend delays | |
405 | ||
406 | void pm_runtime_dont_use_autosuspend(struct device *dev); | |
407 | - clear the power.use_autosuspend flag, disabling autosuspend delays | |
408 | ||
409 | void pm_runtime_set_autosuspend_delay(struct device *dev, int delay); | |
410 | - set the power.autosuspend_delay value to 'delay' (expressed in | |
411 | milliseconds); if 'delay' is negative then run-time suspends are | |
412 | prevented | |
413 | ||
414 | unsigned long pm_runtime_autosuspend_expiration(struct device *dev); | |
415 | - calculate the time when the current autosuspend delay period will expire, | |
416 | based on power.last_busy and power.autosuspend_delay; if the delay time | |
417 | is 1000 ms or larger then the expiration time is rounded up to the | |
418 | nearest second; returns 0 if the delay period has already expired or | |
419 | power.use_autosuspend isn't set, otherwise returns the expiration time | |
420 | in jiffies | |
421 | ||
5e928f77 RW |
422 | It is safe to execute the following helper functions from interrupt context: |
423 | ||
424 | pm_request_idle() | |
15bcb91d | 425 | pm_request_autosuspend() |
5e928f77 RW |
426 | pm_schedule_suspend() |
427 | pm_request_resume() | |
428 | pm_runtime_get_noresume() | |
429 | pm_runtime_get() | |
430 | pm_runtime_put_noidle() | |
431 | pm_runtime_put() | |
15bcb91d AS |
432 | pm_runtime_put_autosuspend() |
433 | pm_runtime_enable() | |
5e928f77 RW |
434 | pm_suspend_ignore_children() |
435 | pm_runtime_set_active() | |
436 | pm_runtime_set_suspended() | |
15bcb91d AS |
437 | pm_runtime_suspended() |
438 | pm_runtime_mark_last_busy() | |
439 | pm_runtime_autosuspend_expiration() | |
5e928f77 RW |
440 | |
441 | 5. Run-time PM Initialization, Device Probing and Removal | |
442 | ||
443 | Initially, the run-time PM is disabled for all devices, which means that the | |
444 | majority of the run-time PM helper funtions described in Section 4 will return | |
445 | -EAGAIN until pm_runtime_enable() is called for the device. | |
446 | ||
447 | In addition to that, the initial run-time PM status of all devices is | |
448 | 'suspended', but it need not reflect the actual physical state of the device. | |
449 | Thus, if the device is initially active (i.e. it is able to process I/O), its | |
450 | run-time PM status must be changed to 'active', with the help of | |
451 | pm_runtime_set_active(), before pm_runtime_enable() is called for the device. | |
452 | ||
453 | However, if the device has a parent and the parent's run-time PM is enabled, | |
454 | calling pm_runtime_set_active() for the device will affect the parent, unless | |
455 | the parent's 'power.ignore_children' flag is set. Namely, in that case the | |
456 | parent won't be able to suspend at run time, using the PM core's helper | |
457 | functions, as long as the child's status is 'active', even if the child's | |
458 | run-time PM is still disabled (i.e. pm_runtime_enable() hasn't been called for | |
459 | the child yet or pm_runtime_disable() has been called for it). For this reason, | |
460 | once pm_runtime_set_active() has been called for the device, pm_runtime_enable() | |
461 | should be called for it too as soon as reasonably possible or its run-time PM | |
462 | status should be changed back to 'suspended' with the help of | |
463 | pm_runtime_set_suspended(). | |
464 | ||
465 | If the default initial run-time PM status of the device (i.e. 'suspended') | |
466 | reflects the actual state of the device, its bus type's or its driver's | |
467 | ->probe() callback will likely need to wake it up using one of the PM core's | |
468 | helper functions described in Section 4. In that case, pm_runtime_resume() | |
469 | should be used. Of course, for this purpose the device's run-time PM has to be | |
470 | enabled earlier by calling pm_runtime_enable(). | |
471 | ||
472 | If the device bus type's or driver's ->probe() or ->remove() callback runs | |
473 | pm_runtime_suspend() or pm_runtime_idle() or their asynchronous counterparts, | |
474 | they will fail returning -EAGAIN, because the device's usage counter is | |
475 | incremented by the core before executing ->probe() and ->remove(). Still, it | |
476 | may be desirable to suspend the device as soon as ->probe() or ->remove() has | |
a6ab7aa9 RW |
477 | finished, so the PM core uses pm_runtime_idle_sync() to invoke the |
478 | subsystem-level idle callback for the device at that time. | |
f1212ae1 | 479 | |
87d1b3e6 RW |
480 | The user space can effectively disallow the driver of the device to power manage |
481 | it at run time by changing the value of its /sys/devices/.../power/control | |
482 | attribute to "on", which causes pm_runtime_forbid() to be called. In principle, | |
483 | this mechanism may also be used by the driver to effectively turn off the | |
484 | run-time power management of the device until the user space turns it on. | |
485 | Namely, during the initialization the driver can make sure that the run-time PM | |
486 | status of the device is 'active' and call pm_runtime_forbid(). It should be | |
487 | noted, however, that if the user space has already intentionally changed the | |
488 | value of /sys/devices/.../power/control to "auto" to allow the driver to power | |
489 | manage the device at run time, the driver may confuse it by using | |
490 | pm_runtime_forbid() this way. | |
491 | ||
f1212ae1 AS |
492 | 6. Run-time PM and System Sleep |
493 | ||
494 | Run-time PM and system sleep (i.e., system suspend and hibernation, also known | |
495 | as suspend-to-RAM and suspend-to-disk) interact with each other in a couple of | |
496 | ways. If a device is active when a system sleep starts, everything is | |
497 | straightforward. But what should happen if the device is already suspended? | |
498 | ||
499 | The device may have different wake-up settings for run-time PM and system sleep. | |
500 | For example, remote wake-up may be enabled for run-time suspend but disallowed | |
501 | for system sleep (device_may_wakeup(dev) returns 'false'). When this happens, | |
502 | the subsystem-level system suspend callback is responsible for changing the | |
503 | device's wake-up setting (it may leave that to the device driver's system | |
504 | suspend routine). It may be necessary to resume the device and suspend it again | |
505 | in order to do so. The same is true if the driver uses different power levels | |
506 | or other settings for run-time suspend and system sleep. | |
507 | ||
508 | During system resume, devices generally should be brought back to full power, | |
509 | even if they were suspended before the system sleep began. There are several | |
510 | reasons for this, including: | |
511 | ||
512 | * The device might need to switch power levels, wake-up settings, etc. | |
513 | ||
514 | * Remote wake-up events might have been lost by the firmware. | |
515 | ||
516 | * The device's children may need the device to be at full power in order | |
517 | to resume themselves. | |
518 | ||
519 | * The driver's idea of the device state may not agree with the device's | |
520 | physical state. This can happen during resume from hibernation. | |
521 | ||
522 | * The device might need to be reset. | |
523 | ||
524 | * Even though the device was suspended, if its usage counter was > 0 then most | |
525 | likely it would need a run-time resume in the near future anyway. | |
526 | ||
527 | * Always going back to full power is simplest. | |
528 | ||
529 | If the device was suspended before the sleep began, then its run-time PM status | |
530 | will have to be updated to reflect the actual post-system sleep status. The way | |
531 | to do this is: | |
532 | ||
533 | pm_runtime_disable(dev); | |
534 | pm_runtime_set_active(dev); | |
535 | pm_runtime_enable(dev); | |
536 | ||
537 | The PM core always increments the run-time usage counter before calling the | |
538 | ->prepare() callback and decrements it after calling the ->complete() callback. | |
539 | Hence disabling run-time PM temporarily like this will not cause any run-time | |
540 | suspend callbacks to be lost. | |
d690b2cd RW |
541 | |
542 | 7. Generic subsystem callbacks | |
543 | ||
544 | Subsystems may wish to conserve code space by using the set of generic power | |
545 | management callbacks provided by the PM core, defined in | |
546 | driver/base/power/generic_ops.c: | |
547 | ||
548 | int pm_generic_runtime_idle(struct device *dev); | |
549 | - invoke the ->runtime_idle() callback provided by the driver of this | |
550 | device, if defined, and call pm_runtime_suspend() for this device if the | |
551 | return value is 0 or the callback is not defined | |
552 | ||
553 | int pm_generic_runtime_suspend(struct device *dev); | |
554 | - invoke the ->runtime_suspend() callback provided by the driver of this | |
555 | device and return its result, or return -EINVAL if not defined | |
556 | ||
557 | int pm_generic_runtime_resume(struct device *dev); | |
558 | - invoke the ->runtime_resume() callback provided by the driver of this | |
559 | device and return its result, or return -EINVAL if not defined | |
560 | ||
561 | int pm_generic_suspend(struct device *dev); | |
562 | - if the device has not been suspended at run time, invoke the ->suspend() | |
563 | callback provided by its driver and return its result, or return 0 if not | |
564 | defined | |
565 | ||
566 | int pm_generic_resume(struct device *dev); | |
567 | - invoke the ->resume() callback provided by the driver of this device and, | |
568 | if successful, change the device's runtime PM status to 'active' | |
569 | ||
570 | int pm_generic_freeze(struct device *dev); | |
571 | - if the device has not been suspended at run time, invoke the ->freeze() | |
572 | callback provided by its driver and return its result, or return 0 if not | |
573 | defined | |
574 | ||
575 | int pm_generic_thaw(struct device *dev); | |
576 | - if the device has not been suspended at run time, invoke the ->thaw() | |
577 | callback provided by its driver and return its result, or return 0 if not | |
578 | defined | |
579 | ||
580 | int pm_generic_poweroff(struct device *dev); | |
581 | - if the device has not been suspended at run time, invoke the ->poweroff() | |
582 | callback provided by its driver and return its result, or return 0 if not | |
583 | defined | |
584 | ||
585 | int pm_generic_restore(struct device *dev); | |
586 | - invoke the ->restore() callback provided by the driver of this device and, | |
587 | if successful, change the device's runtime PM status to 'active' | |
588 | ||
589 | These functions can be assigned to the ->runtime_idle(), ->runtime_suspend(), | |
590 | ->runtime_resume(), ->suspend(), ->resume(), ->freeze(), ->thaw(), ->poweroff(), | |
591 | or ->restore() callback pointers in the subsystem-level dev_pm_ops structures. | |
592 | ||
593 | If a subsystem wishes to use all of them at the same time, it can simply assign | |
594 | the GENERIC_SUBSYS_PM_OPS macro, defined in include/linux/pm.h, to its | |
595 | dev_pm_ops structure pointer. | |
596 | ||
597 | Device drivers that wish to use the same function as a system suspend, freeze, | |
598 | poweroff and run-time suspend callback, and similarly for system resume, thaw, | |
599 | restore, and run-time resume, can achieve this with the help of the | |
600 | UNIVERSAL_DEV_PM_OPS macro defined in include/linux/pm.h (possibly setting its | |
601 | last argument to NULL). | |
7490e442 AS |
602 | |
603 | 8. "No-Callback" Devices | |
604 | ||
605 | Some "devices" are only logical sub-devices of their parent and cannot be | |
606 | power-managed on their own. (The prototype example is a USB interface. Entire | |
607 | USB devices can go into low-power mode or send wake-up requests, but neither is | |
608 | possible for individual interfaces.) The drivers for these devices have no | |
609 | need of run-time PM callbacks; if the callbacks did exist, ->runtime_suspend() | |
610 | and ->runtime_resume() would always return 0 without doing anything else and | |
611 | ->runtime_idle() would always call pm_runtime_suspend(). | |
612 | ||
613 | Subsystems can tell the PM core about these devices by calling | |
614 | pm_runtime_no_callbacks(). This should be done after the device structure is | |
615 | initialized and before it is registered (although after device registration is | |
616 | also okay). The routine will set the device's power.no_callbacks flag and | |
617 | prevent the non-debugging run-time PM sysfs attributes from being created. | |
618 | ||
619 | When power.no_callbacks is set, the PM core will not invoke the | |
620 | ->runtime_idle(), ->runtime_suspend(), or ->runtime_resume() callbacks. | |
621 | Instead it will assume that suspends and resumes always succeed and that idle | |
622 | devices should be suspended. | |
623 | ||
624 | As a consequence, the PM core will never directly inform the device's subsystem | |
625 | or driver about run-time power changes. Instead, the driver for the device's | |
626 | parent must take responsibility for telling the device's driver when the | |
627 | parent's power state changes. | |
15bcb91d AS |
628 | |
629 | 9. Autosuspend, or automatically-delayed suspends | |
630 | ||
631 | Changing a device's power state isn't free; it requires both time and energy. | |
632 | A device should be put in a low-power state only when there's some reason to | |
633 | think it will remain in that state for a substantial time. A common heuristic | |
634 | says that a device which hasn't been used for a while is liable to remain | |
635 | unused; following this advice, drivers should not allow devices to be suspended | |
636 | at run-time until they have been inactive for some minimum period. Even when | |
637 | the heuristic ends up being non-optimal, it will still prevent devices from | |
638 | "bouncing" too rapidly between low-power and full-power states. | |
639 | ||
640 | The term "autosuspend" is an historical remnant. It doesn't mean that the | |
641 | device is automatically suspended (the subsystem or driver still has to call | |
642 | the appropriate PM routines); rather it means that run-time suspends will | |
643 | automatically be delayed until the desired period of inactivity has elapsed. | |
644 | ||
645 | Inactivity is determined based on the power.last_busy field. Drivers should | |
646 | call pm_runtime_mark_last_busy() to update this field after carrying out I/O, | |
647 | typically just before calling pm_runtime_put_autosuspend(). The desired length | |
648 | of the inactivity period is a matter of policy. Subsystems can set this length | |
649 | initially by calling pm_runtime_set_autosuspend_delay(), but after device | |
650 | registration the length should be controlled by user space, using the | |
651 | /sys/devices/.../power/autosuspend_delay_ms attribute. | |
652 | ||
653 | In order to use autosuspend, subsystems or drivers must call | |
654 | pm_runtime_use_autosuspend() (preferably before registering the device), and | |
655 | thereafter they should use the various *_autosuspend() helper functions instead | |
656 | of the non-autosuspend counterparts: | |
657 | ||
658 | Instead of: pm_runtime_suspend use: pm_runtime_autosuspend; | |
659 | Instead of: pm_schedule_suspend use: pm_request_autosuspend; | |
660 | Instead of: pm_runtime_put use: pm_runtime_put_autosuspend; | |
661 | Instead of: pm_runtime_put_sync use: pm_runtime_put_sync_autosuspend. | |
662 | ||
663 | Drivers may also continue to use the non-autosuspend helper functions; they | |
664 | will behave normally, not taking the autosuspend delay into account. | |
665 | Similarly, if the power.use_autosuspend field isn't set then the autosuspend | |
666 | helper functions will behave just like the non-autosuspend counterparts. | |
667 | ||
668 | The implementation is well suited for asynchronous use in interrupt contexts. | |
669 | However such use inevitably involves races, because the PM core can't | |
670 | synchronize ->runtime_suspend() callbacks with the arrival of I/O requests. | |
671 | This synchronization must be handled by the driver, using its private lock. | |
672 | Here is a schematic pseudo-code example: | |
673 | ||
674 | foo_read_or_write(struct foo_priv *foo, void *data) | |
675 | { | |
676 | lock(&foo->private_lock); | |
677 | add_request_to_io_queue(foo, data); | |
678 | if (foo->num_pending_requests++ == 0) | |
679 | pm_runtime_get(&foo->dev); | |
680 | if (!foo->is_suspended) | |
681 | foo_process_next_request(foo); | |
682 | unlock(&foo->private_lock); | |
683 | } | |
684 | ||
685 | foo_io_completion(struct foo_priv *foo, void *req) | |
686 | { | |
687 | lock(&foo->private_lock); | |
688 | if (--foo->num_pending_requests == 0) { | |
689 | pm_runtime_mark_last_busy(&foo->dev); | |
690 | pm_runtime_put_autosuspend(&foo->dev); | |
691 | } else { | |
692 | foo_process_next_request(foo); | |
693 | } | |
694 | unlock(&foo->private_lock); | |
695 | /* Send req result back to the user ... */ | |
696 | } | |
697 | ||
698 | int foo_runtime_suspend(struct device *dev) | |
699 | { | |
700 | struct foo_priv foo = container_of(dev, ...); | |
701 | int ret = 0; | |
702 | ||
703 | lock(&foo->private_lock); | |
704 | if (foo->num_pending_requests > 0) { | |
705 | ret = -EBUSY; | |
706 | } else { | |
707 | /* ... suspend the device ... */ | |
708 | foo->is_suspended = 1; | |
709 | } | |
710 | unlock(&foo->private_lock); | |
711 | return ret; | |
712 | } | |
713 | ||
714 | int foo_runtime_resume(struct device *dev) | |
715 | { | |
716 | struct foo_priv foo = container_of(dev, ...); | |
717 | ||
718 | lock(&foo->private_lock); | |
719 | /* ... resume the device ... */ | |
720 | foo->is_suspended = 0; | |
721 | pm_runtime_mark_last_busy(&foo->dev); | |
722 | if (foo->num_pending_requests > 0) | |
723 | foo_process_requests(foo); | |
724 | unlock(&foo->private_lock); | |
725 | return 0; | |
726 | } | |
727 | ||
728 | The important point is that after foo_io_completion() asks for an autosuspend, | |
729 | the foo_runtime_suspend() callback may race with foo_read_or_write(). | |
730 | Therefore foo_runtime_suspend() has to check whether there are any pending I/O | |
731 | requests (while holding the private lock) before allowing the suspend to | |
732 | proceed. | |
733 | ||
734 | In addition, the power.autosuspend_delay field can be changed by user space at | |
735 | any time. If a driver cares about this, it can call | |
736 | pm_runtime_autosuspend_expiration() from within the ->runtime_suspend() | |
737 | callback while holding its private lock. If the function returns a nonzero | |
738 | value then the delay has not yet expired and the callback should return | |
739 | -EAGAIN. |