About Kernel Documentation Linux Kernel Contact Linux Resources Linux Blog

Documentation / power / runtime_pm.txt




Custom Search

Based on kernel version 3.2. Page generated on 2012-01-05 23:29 EST.

1	Runtime Power Management Framework for I/O Devices
2	
3	(C) 2009-2011 Rafael J. Wysocki <rjw@sisk.pl>, Novell Inc.
4	(C) 2010 Alan Stern <stern@rowland.harvard.edu>
5	
6	1. Introduction
7	
8	Support for runtime power management (runtime 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 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	
18	* A number of runtime 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 runtime PM operations with one another.
21	
22	* Three device runtime 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 runtime 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 runtime PM callbacks present in 'struct dev_pm_ops', the device runtime PM
31	fields of 'struct dev_pm_info' and the core helper functions provided for
32	runtime PM are described below.
33	
34	2. Device Runtime PM Callbacks
35	
36	There are three device runtime 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);
42		int (*runtime_idle)(struct device *dev);
43		...
44	};
45	
46	The ->runtime_suspend(), ->runtime_resume() and ->runtime_idle() callbacks
47	are executed by the PM core for the device's subsystem that may be either of
48	the following:
49	
50	  1. PM domain of the device, if the device's PM domain object, dev->pm_domain,
51	     is present.
52	
53	  2. Device type of the device, if both dev->type and dev->type->pm are present.
54	
55	  3. Device class of the device, if both dev->class and dev->class->pm are
56	     present.
57	
58	  4. Bus type of the device, if both dev->bus and dev->bus->pm are present.
59	
60	The PM core always checks which callback to use in the order given above, so the
61	priority order of callbacks from high to low is: PM domain, device type, class
62	and bus type.  Moreover, the high-priority one will always take precedence over
63	a low-priority one.  The PM domain, bus type, device type and class callbacks
64	are referred to as subsystem-level callbacks in what follows.
65	
66	By default, the callbacks are always invoked in process context with interrupts
67	enabled.  However, subsystems can use the pm_runtime_irq_safe() helper function
68	to tell the PM core that their ->runtime_suspend(), ->runtime_resume() and
69	->runtime_idle() callbacks may be invoked in atomic context with interrupts
70	disabled for a given device.  This implies that the callback routines in
71	question must not block or sleep, but it also means that the synchronous helper
72	functions listed at the end of Section 4 may be used for that device within an
73	interrupt handler or generally in an atomic context.
74	
75	The subsystem-level suspend callback is _entirely_ _responsible_ for handling
76	the suspend of the device as appropriate, which may, but need not include
77	executing the device driver's own ->runtime_suspend() callback (from the
78	PM core's point of view it is not necessary to implement a ->runtime_suspend()
79	callback in a device driver as long as the subsystem-level suspend callback
80	knows what to do to handle the device).
81	
82	  * Once the subsystem-level suspend callback has completed successfully
83	    for given device, the PM core regards the device as suspended, which need
84	    not mean that the device has been put into a low power state.  It is
85	    supposed to mean, however, that the device will not process data and will
86	    not communicate with the CPU(s) and RAM until the subsystem-level resume
87	    callback is executed for it.  The runtime PM status of a device after
88	    successful execution of the subsystem-level suspend callback is 'suspended'.
89	
90	  * If the subsystem-level suspend callback returns -EBUSY or -EAGAIN,
91	    the device's runtime PM status is 'active', which means that the device
92	    _must_ be fully operational afterwards.
93	
94	  * If the subsystem-level suspend callback returns an error code different
95	    from -EBUSY or -EAGAIN, the PM core regards this as a fatal error and will
96	    refuse to run the helper functions described in Section 4 for the device,
97	    until the status of it is directly set either to 'active', or to 'suspended'
98	    (the PM core provides special helper functions for this purpose).
99	
100	In particular, if the driver requires remote wake-up capability (i.e. hardware
101	mechanism allowing the device to request a change of its power state, such as
102	PCI PME) for proper functioning and device_run_wake() returns 'false' for the
103	device, then ->runtime_suspend() should return -EBUSY.  On the other hand, if
104	device_run_wake() returns 'true' for the device and the device is put into a low
105	power state during the execution of the subsystem-level suspend callback, it is
106	expected that remote wake-up will be enabled for the device.  Generally, remote
107	wake-up should be enabled for all input devices put into a low power state at
108	run time.
109	
110	The subsystem-level resume callback is _entirely_ _responsible_ for handling the
111	resume of the device as appropriate, which may, but need not include executing
112	the device driver's own ->runtime_resume() callback (from the PM core's point of
113	view it is not necessary to implement a ->runtime_resume() callback in a device
114	driver as long as the subsystem-level resume callback knows what to do to handle
115	the device).
116	
117	  * Once the subsystem-level resume callback has completed successfully, the PM
118	    core regards the device as fully operational, which means that the device
119	    _must_ be able to complete I/O operations as needed.  The runtime PM status
120	    of the device is then 'active'.
121	
122	  * If the subsystem-level resume callback returns an error code, the PM core
123	    regards this as a fatal error and will refuse to run the helper functions
124	    described in Section 4 for the device, until its status is directly set
125	    either to 'active' or to 'suspended' (the PM core provides special helper
126	    functions for this purpose).
127	
128	The subsystem-level idle callback is executed by the PM core whenever the device
129	appears to be idle, which is indicated to the PM core by two counters, the
130	device's usage counter and the counter of 'active' children of the device.
131	
132	  * If any of these counters is decreased using a helper function provided by
133	    the PM core and it turns out to be equal to zero, the other counter is
134	    checked.  If that counter also is equal to zero, the PM core executes the
135	    subsystem-level idle callback with the device as an argument.
136	
137	The action performed by a subsystem-level idle callback is totally dependent on
138	the subsystem in question, but the expected and recommended action is to check
139	if the device can be suspended (i.e. if all of the conditions necessary for
140	suspending the device are satisfied) and to queue up a suspend request for the
141	device in that case.  The value returned by this callback is ignored by the PM
142	core.
143	
144	The helper functions provided by the PM core, described in Section 4, guarantee
145	that the following constraints are met with respect to the bus type's runtime
146	PM callbacks:
147	
148	(1) The callbacks are mutually exclusive (e.g. it is forbidden to execute
149	    ->runtime_suspend() in parallel with ->runtime_resume() or with another
150	    instance of ->runtime_suspend() for the same device) with the exception that
151	    ->runtime_suspend() or ->runtime_resume() can be executed in parallel with
152	    ->runtime_idle() (although ->runtime_idle() will not be started while any
153	    of the other callbacks is being executed for the same device).
154	
155	(2) ->runtime_idle() and ->runtime_suspend() can only be executed for 'active'
156	    devices (i.e. the PM core will only execute ->runtime_idle() or
157	    ->runtime_suspend() for the devices the runtime PM status of which is
158	    'active').
159	
160	(3) ->runtime_idle() and ->runtime_suspend() can only be executed for a device
161	    the usage counter of which is equal to zero _and_ either the counter of
162	    'active' children of which is equal to zero, or the 'power.ignore_children'
163	    flag of which is set.
164	
165	(4) ->runtime_resume() can only be executed for 'suspended' devices  (i.e. the
166	    PM core will only execute ->runtime_resume() for the devices the runtime
167	    PM status of which is 'suspended').
168	
169	Additionally, the helper functions provided by the PM core obey the following
170	rules:
171	
172	  * If ->runtime_suspend() is about to be executed or there's a pending request
173	    to execute it, ->runtime_idle() will not be executed for the same device.
174	
175	  * A request to execute or to schedule the execution of ->runtime_suspend()
176	    will cancel any pending requests to execute ->runtime_idle() for the same
177	    device.
178	
179	  * If ->runtime_resume() is about to be executed or there's a pending request
180	    to execute it, the other callbacks will not be executed for the same device.
181	
182	  * A request to execute ->runtime_resume() will cancel any pending or
183	    scheduled requests to execute the other callbacks for the same device,
184	    except for scheduled autosuspends.
185	
186	3. Runtime PM Device Fields
187	
188	The following device runtime PM fields are present in 'struct dev_pm_info', as
189	defined in include/linux/pm.h:
190	
191	  struct timer_list suspend_timer;
192	    - timer used for scheduling (delayed) suspend and autosuspend requests
193	
194	  unsigned long timer_expires;
195	    - timer expiration time, in jiffies (if this is different from zero, the
196	      timer is running and will expire at that time, otherwise the timer is not
197	      running)
198	
199	  struct work_struct work;
200	    - work structure used for queuing up requests (i.e. work items in pm_wq)
201	
202	  wait_queue_head_t wait_queue;
203	    - wait queue used if any of the helper functions needs to wait for another
204	      one to complete
205	
206	  spinlock_t lock;
207	    - lock used for synchronisation
208	
209	  atomic_t usage_count;
210	    - the usage counter of the device
211	
212	  atomic_t child_count;
213	    - the count of 'active' children of the device
214	
215	  unsigned int ignore_children;
216	    - if set, the value of child_count is ignored (but still updated)
217	
218	  unsigned int disable_depth;
219	    - used for disabling the helper funcions (they work normally if this is
220	      equal to zero); the initial value of it is 1 (i.e. runtime PM is
221	      initially disabled for all devices)
222	
223	  unsigned int runtime_error;
224	    - if set, there was a fatal error (one of the callbacks returned error code
225	      as described in Section 2), so the helper funtions will not work until
226	      this flag is cleared; this is the error code returned by the failing
227	      callback
228	
229	  unsigned int idle_notification;
230	    - if set, ->runtime_idle() is being executed
231	
232	  unsigned int request_pending;
233	    - if set, there's a pending request (i.e. a work item queued up into pm_wq)
234	
235	  enum rpm_request request;
236	    - type of request that's pending (valid if request_pending is set)
237	
238	  unsigned int deferred_resume;
239	    - set if ->runtime_resume() is about to be run while ->runtime_suspend() is
240	      being executed for that device and it is not practical to wait for the
241	      suspend to complete; means "start a resume as soon as you've suspended"
242	
243	  unsigned int run_wake;
244	    - set if the device is capable of generating runtime wake-up events
245	
246	  enum rpm_status runtime_status;
247	    - the runtime PM status of the device; this field's initial value is
248	      RPM_SUSPENDED, which means that each device is initially regarded by the
249	      PM core as 'suspended', regardless of its real hardware status
250	
251	  unsigned int runtime_auto;
252	    - if set, indicates that the user space has allowed the device driver to
253	      power manage the device at run time via the /sys/devices/.../power/control
254	      interface; it may only be modified with the help of the pm_runtime_allow()
255	      and pm_runtime_forbid() helper functions
256	
257	  unsigned int no_callbacks;
258	    - indicates that the device does not use the runtime PM callbacks (see
259	      Section 8); it may be modified only by the pm_runtime_no_callbacks()
260	      helper function
261	
262	  unsigned int irq_safe;
263	    - indicates that the ->runtime_suspend() and ->runtime_resume() callbacks
264	      will be invoked with the spinlock held and interrupts disabled
265	
266	  unsigned int use_autosuspend;
267	    - indicates that the device's driver supports delayed autosuspend (see
268	      Section 9); it may be modified only by the
269	      pm_runtime{_dont}_use_autosuspend() helper functions
270	
271	  unsigned int timer_autosuspends;
272	    - indicates that the PM core should attempt to carry out an autosuspend
273	      when the timer expires rather than a normal suspend
274	
275	  int autosuspend_delay;
276	    - the delay time (in milliseconds) to be used for autosuspend
277	
278	  unsigned long last_busy;
279	    - the time (in jiffies) when the pm_runtime_mark_last_busy() helper
280	      function was last called for this device; used in calculating inactivity
281	      periods for autosuspend
282	
283	All of the above fields are members of the 'power' member of 'struct device'.
284	
285	4. Runtime PM Device Helper Functions
286	
287	The following runtime PM helper functions are defined in
288	drivers/base/power/runtime.c and include/linux/pm_runtime.h:
289	
290	  void pm_runtime_init(struct device *dev);
291	    - initialize the device runtime PM fields in 'struct dev_pm_info'
292	
293	  void pm_runtime_remove(struct device *dev);
294	    - make sure that the runtime PM of the device will be disabled after
295	      removing the device from device hierarchy
296	
297	  int pm_runtime_idle(struct device *dev);
298	    - execute the subsystem-level idle callback for the device; returns 0 on
299	      success or error code on failure, where -EINPROGRESS means that
300	      ->runtime_idle() is already being executed
301	
302	  int pm_runtime_suspend(struct device *dev);
303	    - execute the subsystem-level suspend callback for the device; returns 0 on
304	      success, 1 if the device's runtime PM status was already 'suspended', or
305	      error code on failure, where -EAGAIN or -EBUSY means it is safe to attempt
306	      to suspend the device again in future and -EACCES means that
307	      'power.disable_depth' is different from 0
308	
309	  int pm_runtime_autosuspend(struct device *dev);
310	    - same as pm_runtime_suspend() except that the autosuspend delay is taken
311	      into account; if pm_runtime_autosuspend_expiration() says the delay has
312	      not yet expired then an autosuspend is scheduled for the appropriate time
313	      and 0 is returned
314	
315	  int pm_runtime_resume(struct device *dev);
316	    - execute the subsystem-level resume callback for the device; returns 0 on
317	      success, 1 if the device's runtime PM status was already 'active' or
318	      error code on failure, where -EAGAIN means it may be safe to attempt to
319	      resume the device again in future, but 'power.runtime_error' should be
320	      checked additionally, and -EACCES means that 'power.disable_depth' is
321	      different from 0
322	
323	  int pm_request_idle(struct device *dev);
324	    - submit a request to execute the subsystem-level idle callback for the
325	      device (the request is represented by a work item in pm_wq); returns 0 on
326	      success or error code if the request has not been queued up
327	
328	  int pm_request_autosuspend(struct device *dev);
329	    - schedule the execution of the subsystem-level suspend callback for the
330	      device when the autosuspend delay has expired; if the delay has already
331	      expired then the work item is queued up immediately
332	
333	  int pm_schedule_suspend(struct device *dev, unsigned int delay);
334	    - schedule the execution of the subsystem-level suspend callback for the
335	      device in future, where 'delay' is the time to wait before queuing up a
336	      suspend work item in pm_wq, in milliseconds (if 'delay' is zero, the work
337	      item is queued up immediately); returns 0 on success, 1 if the device's PM
338	      runtime status was already 'suspended', or error code if the request
339	      hasn't been scheduled (or queued up if 'delay' is 0); if the execution of
340	      ->runtime_suspend() is already scheduled and not yet expired, the new
341	      value of 'delay' will be used as the time to wait
342	
343	  int pm_request_resume(struct device *dev);
344	    - submit a request to execute the subsystem-level resume callback for the
345	      device (the request is represented by a work item in pm_wq); returns 0 on
346	      success, 1 if the device's runtime PM status was already 'active', or
347	      error code if the request hasn't been queued up
348	
349	  void pm_runtime_get_noresume(struct device *dev);
350	    - increment the device's usage counter
351	
352	  int pm_runtime_get(struct device *dev);
353	    - increment the device's usage counter, run pm_request_resume(dev) and
354	      return its result
355	
356	  int pm_runtime_get_sync(struct device *dev);
357	    - increment the device's usage counter, run pm_runtime_resume(dev) and
358	      return its result
359	
360	  void pm_runtime_put_noidle(struct device *dev);
361	    - decrement the device's usage counter
362	
363	  int pm_runtime_put(struct device *dev);
364	    - decrement the device's usage counter; if the result is 0 then run
365	      pm_request_idle(dev) and return its result
366	
367	  int pm_runtime_put_autosuspend(struct device *dev);
368	    - decrement the device's usage counter; if the result is 0 then run
369	      pm_request_autosuspend(dev) and return its result
370	
371	  int pm_runtime_put_sync(struct device *dev);
372	    - decrement the device's usage counter; if the result is 0 then run
373	      pm_runtime_idle(dev) and return its result
374	
375	  int pm_runtime_put_sync_suspend(struct device *dev);
376	    - decrement the device's usage counter; if the result is 0 then run
377	      pm_runtime_suspend(dev) and return its result
378	
379	  int pm_runtime_put_sync_autosuspend(struct device *dev);
380	    - decrement the device's usage counter; if the result is 0 then run
381	      pm_runtime_autosuspend(dev) and return its result
382	
383	  void pm_runtime_enable(struct device *dev);
384	    - decrement the device's 'power.disable_depth' field; if that field is equal
385	      to zero, the runtime PM helper functions can execute subsystem-level
386	      callbacks described in Section 2 for the device
387	
388	  int pm_runtime_disable(struct device *dev);
389	    - increment the device's 'power.disable_depth' field (if the value of that
390	      field was previously zero, this prevents subsystem-level runtime PM
391	      callbacks from being run for the device), make sure that all of the pending
392	      runtime PM operations on the device are either completed or canceled;
393	      returns 1 if there was a resume request pending and it was necessary to
394	      execute the subsystem-level resume callback for the device to satisfy that
395	      request, otherwise 0 is returned
396	
397	  int pm_runtime_barrier(struct device *dev);
398	    - check if there's a resume request pending for the device and resume it
399	      (synchronously) in that case, cancel any other pending runtime PM requests
400	      regarding it and wait for all runtime PM operations on it in progress to
401	      complete; returns 1 if there was a resume request pending and it was
402	      necessary to execute the subsystem-level resume callback for the device to
403	      satisfy that request, otherwise 0 is returned
404	
405	  void pm_suspend_ignore_children(struct device *dev, bool enable);
406	    - set/unset the power.ignore_children flag of the device
407	
408	  int pm_runtime_set_active(struct device *dev);
409	    - clear the device's 'power.runtime_error' flag, set the device's runtime
410	      PM status to 'active' and update its parent's counter of 'active'
411	      children as appropriate (it is only valid to use this function if
412	      'power.runtime_error' is set or 'power.disable_depth' is greater than
413	      zero); it will fail and return error code if the device has a parent
414	      which is not active and the 'power.ignore_children' flag of which is unset
415	
416	  void pm_runtime_set_suspended(struct device *dev);
417	    - clear the device's 'power.runtime_error' flag, set the device's runtime
418	      PM status to 'suspended' and update its parent's counter of 'active'
419	      children as appropriate (it is only valid to use this function if
420	      'power.runtime_error' is set or 'power.disable_depth' is greater than
421	      zero)
422	
423	  bool pm_runtime_suspended(struct device *dev);
424	    - return true if the device's runtime PM status is 'suspended' and its
425	      'power.disable_depth' field is equal to zero, or false otherwise
426	
427	  bool pm_runtime_status_suspended(struct device *dev);
428	    - return true if the device's runtime PM status is 'suspended'
429	
430	  void pm_runtime_allow(struct device *dev);
431	    - set the power.runtime_auto flag for the device and decrease its usage
432	      counter (used by the /sys/devices/.../power/control interface to
433	      effectively allow the device to be power managed at run time)
434	
435	  void pm_runtime_forbid(struct device *dev);
436	    - unset the power.runtime_auto flag for the device and increase its usage
437	      counter (used by the /sys/devices/.../power/control interface to
438	      effectively prevent the device from being power managed at run time)
439	
440	  void pm_runtime_no_callbacks(struct device *dev);
441	    - set the power.no_callbacks flag for the device and remove the runtime
442	      PM attributes from /sys/devices/.../power (or prevent them from being
443	      added when the device is registered)
444	
445	  void pm_runtime_irq_safe(struct device *dev);
446	    - set the power.irq_safe flag for the device, causing the runtime-PM
447	      callbacks to be invoked with interrupts off
448	
449	  void pm_runtime_mark_last_busy(struct device *dev);
450	    - set the power.last_busy field to the current time
451	
452	  void pm_runtime_use_autosuspend(struct device *dev);
453	    - set the power.use_autosuspend flag, enabling autosuspend delays
454	
455	  void pm_runtime_dont_use_autosuspend(struct device *dev);
456	    - clear the power.use_autosuspend flag, disabling autosuspend delays
457	
458	  void pm_runtime_set_autosuspend_delay(struct device *dev, int delay);
459	    - set the power.autosuspend_delay value to 'delay' (expressed in
460	      milliseconds); if 'delay' is negative then runtime suspends are
461	      prevented
462	
463	  unsigned long pm_runtime_autosuspend_expiration(struct device *dev);
464	    - calculate the time when the current autosuspend delay period will expire,
465	      based on power.last_busy and power.autosuspend_delay; if the delay time
466	      is 1000 ms or larger then the expiration time is rounded up to the
467	      nearest second; returns 0 if the delay period has already expired or
468	      power.use_autosuspend isn't set, otherwise returns the expiration time
469	      in jiffies
470	
471	It is safe to execute the following helper functions from interrupt context:
472	
473	pm_request_idle()
474	pm_request_autosuspend()
475	pm_schedule_suspend()
476	pm_request_resume()
477	pm_runtime_get_noresume()
478	pm_runtime_get()
479	pm_runtime_put_noidle()
480	pm_runtime_put()
481	pm_runtime_put_autosuspend()
482	pm_runtime_enable()
483	pm_suspend_ignore_children()
484	pm_runtime_set_active()
485	pm_runtime_set_suspended()
486	pm_runtime_suspended()
487	pm_runtime_mark_last_busy()
488	pm_runtime_autosuspend_expiration()
489	
490	If pm_runtime_irq_safe() has been called for a device then the following helper
491	functions may also be used in interrupt context:
492	
493	pm_runtime_idle()
494	pm_runtime_suspend()
495	pm_runtime_autosuspend()
496	pm_runtime_resume()
497	pm_runtime_get_sync()
498	pm_runtime_put_sync()
499	pm_runtime_put_sync_suspend()
500	pm_runtime_put_sync_autosuspend()
501	
502	5. Runtime PM Initialization, Device Probing and Removal
503	
504	Initially, the runtime PM is disabled for all devices, which means that the
505	majority of the runtime PM helper funtions described in Section 4 will return
506	-EAGAIN until pm_runtime_enable() is called for the device.
507	
508	In addition to that, the initial runtime PM status of all devices is
509	'suspended', but it need not reflect the actual physical state of the device.
510	Thus, if the device is initially active (i.e. it is able to process I/O), its
511	runtime PM status must be changed to 'active', with the help of
512	pm_runtime_set_active(), before pm_runtime_enable() is called for the device.
513	
514	However, if the device has a parent and the parent's runtime PM is enabled,
515	calling pm_runtime_set_active() for the device will affect the parent, unless
516	the parent's 'power.ignore_children' flag is set.  Namely, in that case the
517	parent won't be able to suspend at run time, using the PM core's helper
518	functions, as long as the child's status is 'active', even if the child's
519	runtime PM is still disabled (i.e. pm_runtime_enable() hasn't been called for
520	the child yet or pm_runtime_disable() has been called for it).  For this reason,
521	once pm_runtime_set_active() has been called for the device, pm_runtime_enable()
522	should be called for it too as soon as reasonably possible or its runtime PM
523	status should be changed back to 'suspended' with the help of
524	pm_runtime_set_suspended().
525	
526	If the default initial runtime PM status of the device (i.e. 'suspended')
527	reflects the actual state of the device, its bus type's or its driver's
528	->probe() callback will likely need to wake it up using one of the PM core's
529	helper functions described in Section 4.  In that case, pm_runtime_resume()
530	should be used.  Of course, for this purpose the device's runtime PM has to be
531	enabled earlier by calling pm_runtime_enable().
532	
533	If the device bus type's or driver's ->probe() callback runs
534	pm_runtime_suspend() or pm_runtime_idle() or their asynchronous counterparts,
535	they will fail returning -EAGAIN, because the device's usage counter is
536	incremented by the driver core before executing ->probe().  Still, it may be
537	desirable to suspend the device as soon as ->probe() has finished, so the driver
538	core uses pm_runtime_put_sync() to invoke the subsystem-level idle callback for
539	the device at that time.
540	
541	Moreover, the driver core prevents runtime PM callbacks from racing with the bus
542	notifier callback in __device_release_driver(), which is necessary, because the
543	notifier is used by some subsystems to carry out operations affecting the
544	runtime PM functionality.  It does so by calling pm_runtime_get_sync() before
545	driver_sysfs_remove() and the BUS_NOTIFY_UNBIND_DRIVER notifications.  This
546	resumes the device if it's in the suspended state and prevents it from
547	being suspended again while those routines are being executed.
548	
549	To allow bus types and drivers to put devices into the suspended state by
550	calling pm_runtime_suspend() from their ->remove() routines, the driver core
551	executes pm_runtime_put_sync() after running the BUS_NOTIFY_UNBIND_DRIVER
552	notifications in __device_release_driver().  This requires bus types and
553	drivers to make their ->remove() callbacks avoid races with runtime PM directly,
554	but also it allows of more flexibility in the handling of devices during the
555	removal of their drivers.
556	
557	The user space can effectively disallow the driver of the device to power manage
558	it at run time by changing the value of its /sys/devices/.../power/control
559	attribute to "on", which causes pm_runtime_forbid() to be called.  In principle,
560	this mechanism may also be used by the driver to effectively turn off the
561	runtime power management of the device until the user space turns it on.
562	Namely, during the initialization the driver can make sure that the runtime PM
563	status of the device is 'active' and call pm_runtime_forbid().  It should be
564	noted, however, that if the user space has already intentionally changed the
565	value of /sys/devices/.../power/control to "auto" to allow the driver to power
566	manage the device at run time, the driver may confuse it by using
567	pm_runtime_forbid() this way.
568	
569	6. Runtime PM and System Sleep
570	
571	Runtime PM and system sleep (i.e., system suspend and hibernation, also known
572	as suspend-to-RAM and suspend-to-disk) interact with each other in a couple of
573	ways.  If a device is active when a system sleep starts, everything is
574	straightforward.  But what should happen if the device is already suspended?
575	
576	The device may have different wake-up settings for runtime PM and system sleep.
577	For example, remote wake-up may be enabled for runtime suspend but disallowed
578	for system sleep (device_may_wakeup(dev) returns 'false').  When this happens,
579	the subsystem-level system suspend callback is responsible for changing the
580	device's wake-up setting (it may leave that to the device driver's system
581	suspend routine).  It may be necessary to resume the device and suspend it again
582	in order to do so.  The same is true if the driver uses different power levels
583	or other settings for runtime suspend and system sleep.
584	
585	During system resume, the simplest approach is to bring all devices back to full
586	power, even if they had been suspended before the system suspend began.  There
587	are several reasons for this, including:
588	
589	  * The device might need to switch power levels, wake-up settings, etc.
590	
591	  * Remote wake-up events might have been lost by the firmware.
592	
593	  * The device's children may need the device to be at full power in order
594	    to resume themselves.
595	
596	  * The driver's idea of the device state may not agree with the device's
597	    physical state.  This can happen during resume from hibernation.
598	
599	  * The device might need to be reset.
600	
601	  * Even though the device was suspended, if its usage counter was > 0 then most
602	    likely it would need a runtime resume in the near future anyway.
603	
604	If the device had been suspended before the system suspend began and it's
605	brought back to full power during resume, then its runtime PM status will have
606	to be updated to reflect the actual post-system sleep status.  The way to do
607	this is:
608	
609		pm_runtime_disable(dev);
610		pm_runtime_set_active(dev);
611		pm_runtime_enable(dev);
612	
613	The PM core always increments the runtime usage counter before calling the
614	->suspend() callback and decrements it after calling the ->resume() callback.
615	Hence disabling runtime PM temporarily like this will not cause any runtime
616	suspend attempts to be permanently lost.  If the usage count goes to zero
617	following the return of the ->resume() callback, the ->runtime_idle() callback
618	will be invoked as usual.
619	
620	On some systems, however, system sleep is not entered through a global firmware
621	or hardware operation.  Instead, all hardware components are put into low-power
622	states directly by the kernel in a coordinated way.  Then, the system sleep
623	state effectively follows from the states the hardware components end up in
624	and the system is woken up from that state by a hardware interrupt or a similar
625	mechanism entirely under the kernel's control.  As a result, the kernel never
626	gives control away and the states of all devices during resume are precisely
627	known to it.  If that is the case and none of the situations listed above takes
628	place (in particular, if the system is not waking up from hibernation), it may
629	be more efficient to leave the devices that had been suspended before the system
630	suspend began in the suspended state.
631	
632	The PM core does its best to reduce the probability of race conditions between
633	the runtime PM and system suspend/resume (and hibernation) callbacks by carrying
634	out the following operations:
635	
636	  * During system suspend it calls pm_runtime_get_noresume() and
637	    pm_runtime_barrier() for every device right before executing the
638	    subsystem-level .suspend() callback for it.  In addition to that it calls
639	    pm_runtime_disable() for every device right after executing the
640	    subsystem-level .suspend() callback for it.
641	
642	  * During system resume it calls pm_runtime_enable() and pm_runtime_put_sync()
643	    for every device right before and right after executing the subsystem-level
644	    .resume() callback for it, respectively.
645	
646	7. Generic subsystem callbacks
647	
648	Subsystems may wish to conserve code space by using the set of generic power
649	management callbacks provided by the PM core, defined in
650	driver/base/power/generic_ops.c:
651	
652	  int pm_generic_runtime_idle(struct device *dev);
653	    - invoke the ->runtime_idle() callback provided by the driver of this
654	      device, if defined, and call pm_runtime_suspend() for this device if the
655	      return value is 0 or the callback is not defined
656	
657	  int pm_generic_runtime_suspend(struct device *dev);
658	    - invoke the ->runtime_suspend() callback provided by the driver of this
659	      device and return its result, or return -EINVAL if not defined
660	
661	  int pm_generic_runtime_resume(struct device *dev);
662	    - invoke the ->runtime_resume() callback provided by the driver of this
663	      device and return its result, or return -EINVAL if not defined
664	
665	  int pm_generic_suspend(struct device *dev);
666	    - if the device has not been suspended at run time, invoke the ->suspend()
667	      callback provided by its driver and return its result, or return 0 if not
668	      defined
669	
670	  int pm_generic_suspend_noirq(struct device *dev);
671	    - if pm_runtime_suspended(dev) returns "false", invoke the ->suspend_noirq()
672	      callback provided by the device's driver and return its result, or return
673	      0 if not defined
674	
675	  int pm_generic_resume(struct device *dev);
676	    - invoke the ->resume() callback provided by the driver of this device and,
677	      if successful, change the device's runtime PM status to 'active'
678	
679	  int pm_generic_resume_noirq(struct device *dev);
680	    - invoke the ->resume_noirq() callback provided by the driver of this device
681	
682	  int pm_generic_freeze(struct device *dev);
683	    - if the device has not been suspended at run time, invoke the ->freeze()
684	      callback provided by its driver and return its result, or return 0 if not
685	      defined
686	
687	  int pm_generic_freeze_noirq(struct device *dev);
688	    - if pm_runtime_suspended(dev) returns "false", invoke the ->freeze_noirq()
689	      callback provided by the device's driver and return its result, or return
690	      0 if not defined
691	
692	  int pm_generic_thaw(struct device *dev);
693	    - if the device has not been suspended at run time, invoke the ->thaw()
694	      callback provided by its driver and return its result, or return 0 if not
695	      defined
696	
697	  int pm_generic_thaw_noirq(struct device *dev);
698	    - if pm_runtime_suspended(dev) returns "false", invoke the ->thaw_noirq()
699	      callback provided by the device's driver and return its result, or return
700	      0 if not defined
701	
702	  int pm_generic_poweroff(struct device *dev);
703	    - if the device has not been suspended at run time, invoke the ->poweroff()
704	      callback provided by its driver and return its result, or return 0 if not
705	      defined
706	
707	  int pm_generic_poweroff_noirq(struct device *dev);
708	    - if pm_runtime_suspended(dev) returns "false", run the ->poweroff_noirq()
709	      callback provided by the device's driver and return its result, or return
710	      0 if not defined
711	
712	  int pm_generic_restore(struct device *dev);
713	    - invoke the ->restore() callback provided by the driver of this device and,
714	      if successful, change the device's runtime PM status to 'active'
715	
716	  int pm_generic_restore_noirq(struct device *dev);
717	    - invoke the ->restore_noirq() callback provided by the device's driver
718	
719	These functions can be assigned to the ->runtime_idle(), ->runtime_suspend(),
720	->runtime_resume(), ->suspend(), ->suspend_noirq(), ->resume(),
721	->resume_noirq(), ->freeze(), ->freeze_noirq(), ->thaw(), ->thaw_noirq(),
722	->poweroff(), ->poweroff_noirq(), ->restore(), ->restore_noirq() callback
723	pointers in the subsystem-level dev_pm_ops structures.
724	
725	If a subsystem wishes to use all of them at the same time, it can simply assign
726	the GENERIC_SUBSYS_PM_OPS macro, defined in include/linux/pm.h, to its
727	dev_pm_ops structure pointer.
728	
729	Device drivers that wish to use the same function as a system suspend, freeze,
730	poweroff and runtime suspend callback, and similarly for system resume, thaw,
731	restore, and runtime resume, can achieve this with the help of the
732	UNIVERSAL_DEV_PM_OPS macro defined in include/linux/pm.h (possibly setting its
733	last argument to NULL).
734	
735	8. "No-Callback" Devices
736	
737	Some "devices" are only logical sub-devices of their parent and cannot be
738	power-managed on their own.  (The prototype example is a USB interface.  Entire
739	USB devices can go into low-power mode or send wake-up requests, but neither is
740	possible for individual interfaces.)  The drivers for these devices have no
741	need of runtime PM callbacks; if the callbacks did exist, ->runtime_suspend()
742	and ->runtime_resume() would always return 0 without doing anything else and
743	->runtime_idle() would always call pm_runtime_suspend().
744	
745	Subsystems can tell the PM core about these devices by calling
746	pm_runtime_no_callbacks().  This should be done after the device structure is
747	initialized and before it is registered (although after device registration is
748	also okay).  The routine will set the device's power.no_callbacks flag and
749	prevent the non-debugging runtime PM sysfs attributes from being created.
750	
751	When power.no_callbacks is set, the PM core will not invoke the
752	->runtime_idle(), ->runtime_suspend(), or ->runtime_resume() callbacks.
753	Instead it will assume that suspends and resumes always succeed and that idle
754	devices should be suspended.
755	
756	As a consequence, the PM core will never directly inform the device's subsystem
757	or driver about runtime power changes.  Instead, the driver for the device's
758	parent must take responsibility for telling the device's driver when the
759	parent's power state changes.
760	
761	9. Autosuspend, or automatically-delayed suspends
762	
763	Changing a device's power state isn't free; it requires both time and energy.
764	A device should be put in a low-power state only when there's some reason to
765	think it will remain in that state for a substantial time.  A common heuristic
766	says that a device which hasn't been used for a while is liable to remain
767	unused; following this advice, drivers should not allow devices to be suspended
768	at runtime until they have been inactive for some minimum period.  Even when
769	the heuristic ends up being non-optimal, it will still prevent devices from
770	"bouncing" too rapidly between low-power and full-power states.
771	
772	The term "autosuspend" is an historical remnant.  It doesn't mean that the
773	device is automatically suspended (the subsystem or driver still has to call
774	the appropriate PM routines); rather it means that runtime suspends will
775	automatically be delayed until the desired period of inactivity has elapsed.
776	
777	Inactivity is determined based on the power.last_busy field.  Drivers should
778	call pm_runtime_mark_last_busy() to update this field after carrying out I/O,
779	typically just before calling pm_runtime_put_autosuspend().  The desired length
780	of the inactivity period is a matter of policy.  Subsystems can set this length
781	initially by calling pm_runtime_set_autosuspend_delay(), but after device
782	registration the length should be controlled by user space, using the
783	/sys/devices/.../power/autosuspend_delay_ms attribute.
784	
785	In order to use autosuspend, subsystems or drivers must call
786	pm_runtime_use_autosuspend() (preferably before registering the device), and
787	thereafter they should use the various *_autosuspend() helper functions instead
788	of the non-autosuspend counterparts:
789	
790		Instead of: pm_runtime_suspend    use: pm_runtime_autosuspend;
791		Instead of: pm_schedule_suspend   use: pm_request_autosuspend;
792		Instead of: pm_runtime_put        use: pm_runtime_put_autosuspend;
793		Instead of: pm_runtime_put_sync   use: pm_runtime_put_sync_autosuspend.
794	
795	Drivers may also continue to use the non-autosuspend helper functions; they
796	will behave normally, not taking the autosuspend delay into account.
797	Similarly, if the power.use_autosuspend field isn't set then the autosuspend
798	helper functions will behave just like the non-autosuspend counterparts.
799	
800	Under some circumstances a driver or subsystem may want to prevent a device
801	from autosuspending immediately, even though the usage counter is zero and the
802	autosuspend delay time has expired.  If the ->runtime_suspend() callback
803	returns -EAGAIN or -EBUSY, and if the next autosuspend delay expiration time is
804	in the future (as it normally would be if the callback invoked
805	pm_runtime_mark_last_busy()), the PM core will automatically reschedule the
806	autosuspend.  The ->runtime_suspend() callback can't do this rescheduling
807	itself because no suspend requests of any kind are accepted while the device is
808	suspending (i.e., while the callback is running).
809	
810	The implementation is well suited for asynchronous use in interrupt contexts.
811	However such use inevitably involves races, because the PM core can't
812	synchronize ->runtime_suspend() callbacks with the arrival of I/O requests.
813	This synchronization must be handled by the driver, using its private lock.
814	Here is a schematic pseudo-code example:
815	
816		foo_read_or_write(struct foo_priv *foo, void *data)
817		{
818			lock(&foo->private_lock);
819			add_request_to_io_queue(foo, data);
820			if (foo->num_pending_requests++ == 0)
821				pm_runtime_get(&foo->dev);
822			if (!foo->is_suspended)
823				foo_process_next_request(foo);
824			unlock(&foo->private_lock);
825		}
826	
827		foo_io_completion(struct foo_priv *foo, void *req)
828		{
829			lock(&foo->private_lock);
830			if (--foo->num_pending_requests == 0) {
831				pm_runtime_mark_last_busy(&foo->dev);
832				pm_runtime_put_autosuspend(&foo->dev);
833			} else {
834				foo_process_next_request(foo);
835			}
836			unlock(&foo->private_lock);
837			/* Send req result back to the user ... */
838		}
839	
840		int foo_runtime_suspend(struct device *dev)
841		{
842			struct foo_priv foo = container_of(dev, ...);
843			int ret = 0;
844	
845			lock(&foo->private_lock);
846			if (foo->num_pending_requests > 0) {
847				ret = -EBUSY;
848			} else {
849				/* ... suspend the device ... */
850				foo->is_suspended = 1;
851			}
852			unlock(&foo->private_lock);
853			return ret;
854		}
855	
856		int foo_runtime_resume(struct device *dev)
857		{
858			struct foo_priv foo = container_of(dev, ...);
859	
860			lock(&foo->private_lock);
861			/* ... resume the device ... */
862			foo->is_suspended = 0;
863			pm_runtime_mark_last_busy(&foo->dev);
864			if (foo->num_pending_requests > 0)
865				foo_process_requests(foo);
866			unlock(&foo->private_lock);
867			return 0;
868		}
869	
870	The important point is that after foo_io_completion() asks for an autosuspend,
871	the foo_runtime_suspend() callback may race with foo_read_or_write().
872	Therefore foo_runtime_suspend() has to check whether there are any pending I/O
873	requests (while holding the private lock) before allowing the suspend to
874	proceed.
875	
876	In addition, the power.autosuspend_delay field can be changed by user space at
877	any time.  If a driver cares about this, it can call
878	pm_runtime_autosuspend_expiration() from within the ->runtime_suspend()
879	callback while holding its private lock.  If the function returns a nonzero
880	value then the delay has not yet expired and the callback should return
881	-EAGAIN.
Hide Line Numbers
About Kernel Documentation Linux Kernel Contact Linux Resources Linux Blog

Information is copyright its respective author. All material is available from the Linux Kernel Source distributed under a GPL License. This page is provided as a free service by mjmwired.net.