blob: f023ba6bba62c2dcbd514691de267ab7229d0942 [file] [log] [blame]
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +01001Device Power Management
2
Rafael J. Wysocki7538e3d2011-02-16 21:53:17 +01003Copyright (c) 2010-2011 Rafael J. Wysocki <rjw@sisk.pl>, Novell Inc.
Alan Sternd6f9cda2010-03-26 23:53:55 +01004Copyright (c) 2010 Alan Stern <stern@rowland.harvard.edu>
5
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +01006
David Brownell4fc08402006-08-10 16:38:28 -07007Most of the code in Linux is device drivers, so most of the Linux power
Alan Sternd6f9cda2010-03-26 23:53:55 +01008management (PM) code is also driver-specific. Most drivers will do very
9little; others, especially for platforms with small batteries (like cell
10phones), will do a lot.
Linus Torvalds1da177e2005-04-16 15:20:36 -070011
David Brownell4fc08402006-08-10 16:38:28 -070012This writeup gives an overview of how drivers interact with system-wide
13power management goals, emphasizing the models and interfaces that are
14shared by everything that hooks up to the driver model core. Read it as
15background for the domain-specific work you'd do with any specific driver.
Linus Torvalds1da177e2005-04-16 15:20:36 -070016
17
David Brownell4fc08402006-08-10 16:38:28 -070018Two Models for Device Power Management
19======================================
20Drivers will use one or both of these models to put devices into low-power
21states:
22
23 System Sleep model:
Alan Sternd6f9cda2010-03-26 23:53:55 +010024 Drivers can enter low-power states as part of entering system-wide
25 low-power states like "suspend" (also known as "suspend-to-RAM"), or
26 (mostly for systems with disks) "hibernation" (also known as
27 "suspend-to-disk").
David Brownell4fc08402006-08-10 16:38:28 -070028
29 This is something that device, bus, and class drivers collaborate on
30 by implementing various role-specific suspend and resume methods to
31 cleanly power down hardware and software subsystems, then reactivate
32 them without loss of data.
33
34 Some drivers can manage hardware wakeup events, which make the system
Alan Sternd6f9cda2010-03-26 23:53:55 +010035 leave the low-power state. This feature may be enabled or disabled
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +010036 using the relevant /sys/devices/.../power/wakeup file (for Ethernet
37 drivers the ioctl interface used by ethtool may also be used for this
38 purpose); enabling it may cost some power usage, but let the whole
Alan Sternd6f9cda2010-03-26 23:53:55 +010039 system enter low-power states more often.
David Brownell4fc08402006-08-10 16:38:28 -070040
41 Runtime Power Management model:
Alan Sternd6f9cda2010-03-26 23:53:55 +010042 Devices may also be put into low-power states while the system is
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +010043 running, independently of other power management activity in principle.
44 However, devices are not generally independent of each other (for
Alan Sternd6f9cda2010-03-26 23:53:55 +010045 example, a parent device cannot be suspended unless all of its child
46 devices have been suspended). Moreover, depending on the bus type the
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +010047 device is on, it may be necessary to carry out some bus-specific
Alan Sternd6f9cda2010-03-26 23:53:55 +010048 operations on the device for this purpose. Devices put into low power
49 states at run time may require special handling during system-wide power
50 transitions (suspend or hibernation).
David Brownell4fc08402006-08-10 16:38:28 -070051
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +010052 For these reasons not only the device driver itself, but also the
Alan Sternd6f9cda2010-03-26 23:53:55 +010053 appropriate subsystem (bus type, device type or device class) driver and
54 the PM core are involved in runtime power management. As in the system
55 sleep power management case, they need to collaborate by implementing
56 various role-specific suspend and resume methods, so that the hardware
57 is cleanly powered down and reactivated without data or service loss.
David Brownell4fc08402006-08-10 16:38:28 -070058
Alan Sternd6f9cda2010-03-26 23:53:55 +010059There's not a lot to be said about those low-power states except that they are
60very system-specific, and often device-specific. Also, that if enough devices
61have been put into low-power states (at runtime), the effect may be very similar
62to entering some system-wide low-power state (system sleep) ... and that
63synergies exist, so that several drivers using runtime PM might put the system
64into a state where even deeper power saving options are available.
David Brownell4fc08402006-08-10 16:38:28 -070065
Alan Sternd6f9cda2010-03-26 23:53:55 +010066Most suspended devices will have quiesced all I/O: no more DMA or IRQs (except
67for wakeup events), no more data read or written, and requests from upstream
68drivers are no longer accepted. A given bus or platform may have different
69requirements though.
David Brownell4fc08402006-08-10 16:38:28 -070070
71Examples of hardware wakeup events include an alarm from a real time clock,
72network wake-on-LAN packets, keyboard or mouse activity, and media insertion
73or removal (for PCMCIA, MMC/SD, USB, and so on).
Linus Torvalds1da177e2005-04-16 15:20:36 -070074
75
David Brownell4fc08402006-08-10 16:38:28 -070076Interfaces for Entering System Sleep States
77===========================================
Alan Sternd6f9cda2010-03-26 23:53:55 +010078There are programming interfaces provided for subsystems (bus type, device type,
79device class) and device drivers to allow them to participate in the power
80management of devices they are concerned with. These interfaces cover both
81system sleep and runtime power management.
Linus Torvalds1da177e2005-04-16 15:20:36 -070082
David Brownell4fc08402006-08-10 16:38:28 -070083
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +010084Device Power Management Operations
85----------------------------------
86Device power management operations, at the subsystem level as well as at the
87device driver level, are implemented by defining and populating objects of type
88struct dev_pm_ops:
Linus Torvalds1da177e2005-04-16 15:20:36 -070089
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +010090struct dev_pm_ops {
91 int (*prepare)(struct device *dev);
92 void (*complete)(struct device *dev);
93 int (*suspend)(struct device *dev);
94 int (*resume)(struct device *dev);
95 int (*freeze)(struct device *dev);
96 int (*thaw)(struct device *dev);
97 int (*poweroff)(struct device *dev);
98 int (*restore)(struct device *dev);
99 int (*suspend_noirq)(struct device *dev);
100 int (*resume_noirq)(struct device *dev);
101 int (*freeze_noirq)(struct device *dev);
102 int (*thaw_noirq)(struct device *dev);
103 int (*poweroff_noirq)(struct device *dev);
104 int (*restore_noirq)(struct device *dev);
105 int (*runtime_suspend)(struct device *dev);
106 int (*runtime_resume)(struct device *dev);
107 int (*runtime_idle)(struct device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108};
109
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100110This structure is defined in include/linux/pm.h and the methods included in it
111are also described in that file. Their roles will be explained in what follows.
Alan Sternd6f9cda2010-03-26 23:53:55 +0100112For now, it should be sufficient to remember that the last three methods are
113specific to runtime power management while the remaining ones are used during
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100114system-wide power transitions.
115
Alan Sternd6f9cda2010-03-26 23:53:55 +0100116There also is a deprecated "old" or "legacy" interface for power management
117operations available at least for some subsystems. This approach does not use
118struct dev_pm_ops objects and it is suitable only for implementing system sleep
119power management methods. Therefore it is not described in this document, so
120please refer directly to the source code for more information about it.
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100121
122
123Subsystem-Level Methods
124-----------------------
125The core methods to suspend and resume devices reside in struct dev_pm_ops
126pointed to by the pm member of struct bus_type, struct device_type and
127struct class. They are mostly of interest to the people writing infrastructure
128for buses, like PCI or USB, or device type and device class drivers.
129
Alan Sternd6f9cda2010-03-26 23:53:55 +0100130Bus drivers implement these methods as appropriate for the hardware and the
131drivers using it; PCI works differently from USB, and so on. Not many people
132write subsystem-level drivers; most driver code is a "device driver" that builds
133on top of bus-specific framework code.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134
David Brownell4fc08402006-08-10 16:38:28 -0700135For more information on these driver calls, see the description later;
136they are called in phases for every device, respecting the parent-child
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100137sequencing in the driver model tree.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700138
139
David Brownell4fc08402006-08-10 16:38:28 -0700140/sys/devices/.../power/wakeup files
141-----------------------------------
Alan Sternd6f9cda2010-03-26 23:53:55 +0100142All devices in the driver model have two flags to control handling of wakeup
143events (hardware signals that can force the device and/or system out of a low
144power state). These flags are initialized by bus or device driver code using
145device_set_wakeup_capable() and device_set_wakeup_enable(), defined in
146include/linux/pm_wakeup.h.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147
David Brownell4fc08402006-08-10 16:38:28 -0700148The "can_wakeup" flag just records whether the device (and its driver) can
Alan Sternd6f9cda2010-03-26 23:53:55 +0100149physically support wakeup events. The device_set_wakeup_capable() routine
150affects this flag. The "should_wakeup" flag controls whether the device should
151try to use its wakeup mechanism. device_set_wakeup_enable() affects this flag;
152for the most part drivers should not change its value. The initial value of
153should_wakeup is supposed to be false for the majority of devices; the major
154exceptions are power buttons, keyboards, and Ethernet adapters whose WoL
155(wake-on-LAN) feature has been set up with ethtool.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156
Alan Sternd6f9cda2010-03-26 23:53:55 +0100157Whether or not a device is capable of issuing wakeup events is a hardware
158matter, and the kernel is responsible for keeping track of it. By contrast,
159whether or not a wakeup-capable device should issue wakeup events is a policy
160decision, and it is managed by user space through a sysfs attribute: the
161power/wakeup file. User space can write the strings "enabled" or "disabled" to
Rafael J. Wysockicb8f51b2011-02-08 23:26:02 +0100162set or clear the "should_wakeup" flag, respectively. This file is only present
163for wakeup-capable devices (i.e. devices whose "can_wakeup" flags are set)
164and is created (or removed) by device_set_wakeup_capable(). Reads from the
165file will return the corresponding string.
Alan Sternd6f9cda2010-03-26 23:53:55 +0100166
167The device_may_wakeup() routine returns true only if both flags are set.
Rafael J. Wysockicb8f51b2011-02-08 23:26:02 +0100168This information is used by subsystems, like the PCI bus type code, to see
169whether or not to enable the devices' wakeup mechanisms. If device wakeup
170mechanisms are enabled or disabled directly by drivers, they also should use
171device_may_wakeup() to decide what to do during a system sleep transition.
172However for runtime power management, wakeup events should be enabled whenever
173the device and driver both support them, regardless of the should_wakeup flag.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100176/sys/devices/.../power/control files
177------------------------------------
Alan Sternd6f9cda2010-03-26 23:53:55 +0100178Each device in the driver model has a flag to control whether it is subject to
179runtime power management. This flag, called runtime_auto, is initialized by the
180bus type (or generally subsystem) code using pm_runtime_allow() or
181pm_runtime_forbid(); the default is to allow runtime power management.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182
Alan Sternd6f9cda2010-03-26 23:53:55 +0100183The setting can be adjusted by user space by writing either "on" or "auto" to
184the device's power/control sysfs file. Writing "auto" calls pm_runtime_allow(),
185setting the flag and allowing the device to be runtime power-managed by its
186driver. Writing "on" calls pm_runtime_forbid(), clearing the flag, returning
187the device to full power if it was in a low-power state, and preventing the
188device from being runtime power-managed. User space can check the current value
189of the runtime_auto flag by reading the file.
David Brownell4fc08402006-08-10 16:38:28 -0700190
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100191The device's runtime_auto flag has no effect on the handling of system-wide
Alan Sternd6f9cda2010-03-26 23:53:55 +0100192power transitions. In particular, the device can (and in the majority of cases
193should and will) be put into a low-power state during a system-wide transition
194to a sleep state even though its runtime_auto flag is clear.
David Brownell4fc08402006-08-10 16:38:28 -0700195
Alan Sternd6f9cda2010-03-26 23:53:55 +0100196For more information about the runtime power management framework, refer to
197Documentation/power/runtime_pm.txt.
David Brownell4fc08402006-08-10 16:38:28 -0700198
199
Alan Sternd6f9cda2010-03-26 23:53:55 +0100200Calling Drivers to Enter and Leave System Sleep States
201======================================================
202When the system goes into a sleep state, each device's driver is asked to
203suspend the device by putting it into a state compatible with the target
David Brownell4fc08402006-08-10 16:38:28 -0700204system state. That's usually some version of "off", but the details are
205system-specific. Also, wakeup-enabled devices will usually stay partly
206functional in order to wake the system.
207
Alan Sternd6f9cda2010-03-26 23:53:55 +0100208When the system leaves that low-power state, the device's driver is asked to
209resume it by returning it to full power. The suspend and resume operations
210always go together, and both are multi-phase operations.
David Brownell4fc08402006-08-10 16:38:28 -0700211
Alan Sternd6f9cda2010-03-26 23:53:55 +0100212For simple drivers, suspend might quiesce the device using class code
213and then turn its hardware as "off" as possible during suspend_noirq. The
David Brownell4fc08402006-08-10 16:38:28 -0700214matching resume calls would then completely reinitialize the hardware
215before reactivating its class I/O queues.
216
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100217More power-aware drivers might prepare the devices for triggering system wakeup
218events.
David Brownell4fc08402006-08-10 16:38:28 -0700219
220
221Call Sequence Guarantees
222------------------------
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100223To ensure that bridges and similar links needing to talk to a device are
David Brownell4fc08402006-08-10 16:38:28 -0700224available when the device is suspended or resumed, the device tree is
225walked in a bottom-up order to suspend devices. A top-down order is
226used to resume those devices.
227
228The ordering of the device tree is defined by the order in which devices
229get registered: a child can never be registered, probed or resumed before
230its parent; and can't be removed or suspended after that parent.
231
232The policy is that the device tree should match hardware bus topology.
233(Or at least the control bus, for devices which use multiple busses.)
Rafael J. Wysocki58aca232008-03-12 00:57:22 +0100234In particular, this means that a device registration may fail if the parent of
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100235the device is suspending (i.e. has been chosen by the PM core as the next
Rafael J. Wysocki58aca232008-03-12 00:57:22 +0100236device to suspend) or has already suspended, as well as after all of the other
237devices have been suspended. Device drivers must be prepared to cope with such
238situations.
David Brownell4fc08402006-08-10 16:38:28 -0700239
240
Alan Sternd6f9cda2010-03-26 23:53:55 +0100241System Power Management Phases
242------------------------------
243Suspending or resuming the system is done in several phases. Different phases
244are used for standby or memory sleep states ("suspend-to-RAM") and the
245hibernation state ("suspend-to-disk"). Each phase involves executing callbacks
246for every device before the next phase begins. Not all busses or classes
247support all these callbacks and not all drivers use all the callbacks. The
248various phases always run after tasks have been frozen and before they are
249unfrozen. Furthermore, the *_noirq phases run at a time when IRQ handlers have
250been disabled (except for those marked with the IRQ_WAKEUP flag).
David Brownell4fc08402006-08-10 16:38:28 -0700251
Rafael J. Wysocki9659cc02011-02-18 23:20:21 +0100252All phases use bus, type, or class callbacks (that is, methods defined in
253dev->bus->pm, dev->type->pm, or dev->class->pm). These callbacks are mutually
254exclusive, so if the device type provides a struct dev_pm_ops object pointed to
255by its pm field (i.e. both dev->type and dev->type->pm are defined), the
256callbacks included in that object (i.e. dev->type->pm) will be used. Otherwise,
257if the class provides a struct dev_pm_ops object pointed to by its pm field
258(i.e. both dev->class and dev->class->pm are defined), the PM core will use the
259callbacks from that object (i.e. dev->class->pm). Finally, if the pm fields of
260both the device type and class objects are NULL (or those objects do not exist),
261the callbacks provided by the bus (that is, the callbacks from dev->bus->pm)
262will be used (this allows device types to override callbacks provided by bus
263types or classes if necessary).
David Brownell4fc08402006-08-10 16:38:28 -0700264
Alan Sternd6f9cda2010-03-26 23:53:55 +0100265These callbacks may in turn invoke device- or driver-specific methods stored in
266dev->driver->pm, but they don't have to.
David Brownell4fc08402006-08-10 16:38:28 -0700267
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100268
Alan Sternd6f9cda2010-03-26 23:53:55 +0100269Entering System Suspend
270-----------------------
271When the system goes into the standby or memory sleep state, the phases are:
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100272
Alan Sternd6f9cda2010-03-26 23:53:55 +0100273 prepare, suspend, suspend_noirq.
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100274
Alan Sternd6f9cda2010-03-26 23:53:55 +0100275 1. The prepare phase is meant to prevent races by preventing new devices
276 from being registered; the PM core would never know that all the
277 children of a device had been suspended if new children could be
278 registered at will. (By contrast, devices may be unregistered at any
279 time.) Unlike the other suspend-related phases, during the prepare
280 phase the device tree is traversed top-down.
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100281
Alan Sternd6f9cda2010-03-26 23:53:55 +0100282 The prepare phase uses only a bus callback. After the callback method
283 returns, no new children may be registered below the device. The method
284 may also prepare the device or driver in some way for the upcoming
285 system power transition, but it should not put the device into a
286 low-power state.
287
288 2. The suspend methods should quiesce the device to stop it from performing
289 I/O. They also may save the device registers and put it into the
290 appropriate low-power state, depending on the bus type the device is on,
291 and they may enable wakeup events.
292
293 3. The suspend_noirq phase occurs after IRQ handlers have been disabled,
294 which means that the driver's interrupt handler will not be called while
295 the callback method is running. The methods should save the values of
296 the device's registers that weren't saved previously and finally put the
297 device into the appropriate low-power state.
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100298
299 The majority of subsystems and device drivers need not implement this
Alan Sternd6f9cda2010-03-26 23:53:55 +0100300 callback. However, bus types allowing devices to share interrupt
301 vectors, like PCI, generally need it; otherwise a driver might encounter
302 an error during the suspend phase by fielding a shared interrupt
303 generated by some other device after its own device had been set to low
304 power.
David Brownell4fc08402006-08-10 16:38:28 -0700305
Alan Sternd6f9cda2010-03-26 23:53:55 +0100306At the end of these phases, drivers should have stopped all I/O transactions
307(DMA, IRQs), saved enough state that they can re-initialize or restore previous
308state (as needed by the hardware), and placed the device into a low-power state.
309On many platforms they will gate off one or more clock sources; sometimes they
310will also switch off power supplies or reduce voltages. (Drivers supporting
311runtime PM may already have performed some or all of these steps.)
David Brownell4fc08402006-08-10 16:38:28 -0700312
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100313If device_may_wakeup(dev) returns true, the device should be prepared for
Alan Sternd6f9cda2010-03-26 23:53:55 +0100314generating hardware wakeup signals to trigger a system wakeup event when the
315system is in the sleep state. For example, enable_irq_wake() might identify
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100316GPIO signals hooked up to a switch or other external hardware, and
317pci_enable_wake() does something similar for the PCI PME signal.
David Brownell4fc08402006-08-10 16:38:28 -0700318
Alan Sternd6f9cda2010-03-26 23:53:55 +0100319If any of these callbacks returns an error, the system won't enter the desired
320low-power state. Instead the PM core will unwind its actions by resuming all
321the devices that were suspended.
David Brownell4fc08402006-08-10 16:38:28 -0700322
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100323
Alan Sternd6f9cda2010-03-26 23:53:55 +0100324Leaving System Suspend
325----------------------
326When resuming from standby or memory sleep, the phases are:
327
328 resume_noirq, resume, complete.
329
330 1. The resume_noirq callback methods should perform any actions needed
331 before the driver's interrupt handlers are invoked. This generally
332 means undoing the actions of the suspend_noirq phase. If the bus type
333 permits devices to share interrupt vectors, like PCI, the method should
334 bring the device and its driver into a state in which the driver can
335 recognize if the device is the source of incoming interrupts, if any,
336 and handle them correctly.
337
338 For example, the PCI bus type's ->pm.resume_noirq() puts the device into
339 the full-power state (D0 in the PCI terminology) and restores the
340 standard configuration registers of the device. Then it calls the
341 device driver's ->pm.resume_noirq() method to perform device-specific
342 actions.
343
344 2. The resume methods should bring the the device back to its operating
345 state, so that it can perform normal I/O. This generally involves
346 undoing the actions of the suspend phase.
347
348 3. The complete phase uses only a bus callback. The method should undo the
349 actions of the prepare phase. Note, however, that new children may be
350 registered below the device as soon as the resume callbacks occur; it's
351 not necessary to wait until the complete phase.
352
353At the end of these phases, drivers should be as functional as they were before
354suspending: I/O can be performed using DMA and IRQs, and the relevant clocks are
355gated on. Even if the device was in a low-power state before the system sleep
356because of runtime power management, afterwards it should be back in its
357full-power state. There are multiple reasons why it's best to do this; they are
358discussed in more detail in Documentation/power/runtime_pm.txt.
359
360However, the details here may again be platform-specific. For example,
361some systems support multiple "run" states, and the mode in effect at
362the end of resume might not be the one which preceded suspension.
363That means availability of certain clocks or power supplies changed,
364which could easily affect how a driver works.
365
366Drivers need to be able to handle hardware which has been reset since the
367suspend methods were called, for example by complete reinitialization.
368This may be the hardest part, and the one most protected by NDA'd documents
369and chip errata. It's simplest if the hardware state hasn't changed since
370the suspend was carried out, but that can't be guaranteed (in fact, it ususally
371is not the case).
372
373Drivers must also be prepared to notice that the device has been removed
374while the system was powered down, whenever that's physically possible.
375PCMCIA, MMC, USB, Firewire, SCSI, and even IDE are common examples of busses
376where common Linux platforms will see such removal. Details of how drivers
377will notice and handle such removals are currently bus-specific, and often
378involve a separate thread.
379
380These callbacks may return an error value, but the PM core will ignore such
381errors since there's nothing it can do about them other than printing them in
382the system log.
383
384
385Entering Hibernation
386--------------------
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100387Hibernating the system is more complicated than putting it into the standby or
Alan Sternd6f9cda2010-03-26 23:53:55 +0100388memory sleep state, because it involves creating and saving a system image.
389Therefore there are more phases for hibernation, with a different set of
390callbacks. These phases always run after tasks have been frozen and memory has
391been freed.
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100392
Alan Sternd6f9cda2010-03-26 23:53:55 +0100393The general procedure for hibernation is to quiesce all devices (freeze), create
394an image of the system memory while everything is stable, reactivate all
395devices (thaw), write the image to permanent storage, and finally shut down the
396system (poweroff). The phases used to accomplish this are:
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100397
Alan Sternd6f9cda2010-03-26 23:53:55 +0100398 prepare, freeze, freeze_noirq, thaw_noirq, thaw, complete,
399 prepare, poweroff, poweroff_noirq
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100400
Alan Sternd6f9cda2010-03-26 23:53:55 +0100401 1. The prepare phase is discussed in the "Entering System Suspend" section
402 above.
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100403
Alan Sternd6f9cda2010-03-26 23:53:55 +0100404 2. The freeze methods should quiesce the device so that it doesn't generate
405 IRQs or DMA, and they may need to save the values of device registers.
406 However the device does not have to be put in a low-power state, and to
407 save time it's best not to do so. Also, the device should not be
408 prepared to generate wakeup events.
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100409
Alan Sternd6f9cda2010-03-26 23:53:55 +0100410 3. The freeze_noirq phase is analogous to the suspend_noirq phase discussed
411 above, except again that the device should not be put in a low-power
412 state and should not be allowed to generate wakeup events.
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100413
Alan Sternd6f9cda2010-03-26 23:53:55 +0100414At this point the system image is created. All devices should be inactive and
415the contents of memory should remain undisturbed while this happens, so that the
416image forms an atomic snapshot of the system state.
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100417
Alan Sternd6f9cda2010-03-26 23:53:55 +0100418 4. The thaw_noirq phase is analogous to the resume_noirq phase discussed
419 above. The main difference is that its methods can assume the device is
420 in the same state as at the end of the freeze_noirq phase.
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100421
Alan Sternd6f9cda2010-03-26 23:53:55 +0100422 5. The thaw phase is analogous to the resume phase discussed above. Its
423 methods should bring the device back to an operating state, so that it
424 can be used for saving the image if necessary.
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100425
Alan Sternd6f9cda2010-03-26 23:53:55 +0100426 6. The complete phase is discussed in the "Leaving System Suspend" section
427 above.
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100428
Alan Sternd6f9cda2010-03-26 23:53:55 +0100429At this point the system image is saved, and the devices then need to be
430prepared for the upcoming system shutdown. This is much like suspending them
431before putting the system into the standby or memory sleep state, and the phases
432are similar.
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100433
Alan Sternd6f9cda2010-03-26 23:53:55 +0100434 7. The prepare phase is discussed above.
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100435
Alan Sternd6f9cda2010-03-26 23:53:55 +0100436 8. The poweroff phase is analogous to the suspend phase.
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100437
Alan Sternd6f9cda2010-03-26 23:53:55 +0100438 9. The poweroff_noirq phase is analogous to the suspend_noirq phase.
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100439
Alan Sternd6f9cda2010-03-26 23:53:55 +0100440The poweroff and poweroff_noirq callbacks should do essentially the same things
441as the suspend and suspend_noirq callbacks. The only notable difference is that
442they need not store the device register values, because the registers should
443already have been stored during the freeze or freeze_noirq phases.
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100444
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100445
Alan Sternd6f9cda2010-03-26 23:53:55 +0100446Leaving Hibernation
447-------------------
448Resuming from hibernation is, again, more complicated than resuming from a sleep
449state in which the contents of main memory are preserved, because it requires
450a system image to be loaded into memory and the pre-hibernation memory contents
451to be restored before control can be passed back to the image kernel.
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100452
Alan Sternd6f9cda2010-03-26 23:53:55 +0100453Although in principle, the image might be loaded into memory and the
454pre-hibernation memory contents restored by the boot loader, in practice this
455can't be done because boot loaders aren't smart enough and there is no
456established protocol for passing the necessary information. So instead, the
457boot loader loads a fresh instance of the kernel, called the boot kernel, into
458memory and passes control to it in the usual way. Then the boot kernel reads
459the system image, restores the pre-hibernation memory contents, and passes
460control to the image kernel. Thus two different kernels are involved in
461resuming from hibernation. In fact, the boot kernel may be completely different
462from the image kernel: a different configuration and even a different version.
463This has important consequences for device drivers and their subsystems.
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100464
Alan Sternd6f9cda2010-03-26 23:53:55 +0100465To be able to load the system image into memory, the boot kernel needs to
466include at least a subset of device drivers allowing it to access the storage
467medium containing the image, although it doesn't need to include all of the
468drivers present in the image kernel. After the image has been loaded, the
469devices managed by the boot kernel need to be prepared for passing control back
470to the image kernel. This is very similar to the initial steps involved in
471creating a system image, and it is accomplished in the same way, using prepare,
472freeze, and freeze_noirq phases. However the devices affected by these phases
473are only those having drivers in the boot kernel; other devices will still be in
474whatever state the boot loader left them.
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100475
Alan Sternd6f9cda2010-03-26 23:53:55 +0100476Should the restoration of the pre-hibernation memory contents fail, the boot
477kernel would go through the "thawing" procedure described above, using the
478thaw_noirq, thaw, and complete phases, and then continue running normally. This
479happens only rarely. Most often the pre-hibernation memory contents are
480restored successfully and control is passed to the image kernel, which then
481becomes responsible for bringing the system back to the working state.
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100482
Alan Sternd6f9cda2010-03-26 23:53:55 +0100483To achieve this, the image kernel must restore the devices' pre-hibernation
484functionality. The operation is much like waking up from the memory sleep
485state, although it involves different phases:
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100486
Alan Sternd6f9cda2010-03-26 23:53:55 +0100487 restore_noirq, restore, complete
488
489 1. The restore_noirq phase is analogous to the resume_noirq phase.
490
491 2. The restore phase is analogous to the resume phase.
492
493 3. The complete phase is discussed above.
494
495The main difference from resume[_noirq] is that restore[_noirq] must assume the
496device has been accessed and reconfigured by the boot loader or the boot kernel.
497Consequently the state of the device may be different from the state remembered
498from the freeze and freeze_noirq phases. The device may even need to be reset
499and completely re-initialized. In many cases this difference doesn't matter, so
500the resume[_noirq] and restore[_norq] method pointers can be set to the same
501routines. Nevertheless, different callback pointers are used in case there is a
502situation where it actually matters.
503
504
Rafael J. Wysocki7538e3d2011-02-16 21:53:17 +0100505Device Power Domains
506--------------------
507Sometimes devices share reference clocks or other power resources. In those
508cases it generally is not possible to put devices into low-power states
509individually. Instead, a set of devices sharing a power resource can be put
510into a low-power state together at the same time by turning off the shared
511power resource. Of course, they also need to be put into the full-power state
512together, by turning the shared power resource on. A set of devices with this
513property is often referred to as a power domain.
514
515Support for power domains is provided through the pwr_domain field of struct
516device. This field is a pointer to an object of type struct dev_power_domain,
517defined in include/linux/pm.h, providing a set of power management callbacks
518analogous to the subsystem-level and device driver callbacks that are executed
519for the given device during all power transitions, in addition to the respective
520subsystem-level callbacks. Specifically, the power domain "suspend" callbacks
521(i.e. ->runtime_suspend(), ->suspend(), ->freeze(), ->poweroff(), etc.) are
522executed after the analogous subsystem-level callbacks, while the power domain
523"resume" callbacks (i.e. ->runtime_resume(), ->resume(), ->thaw(), ->restore,
524etc.) are executed before the analogous subsystem-level callbacks. Error codes
525returned by the "suspend" and "resume" power domain callbacks are ignored.
526
527Power domain ->runtime_idle() callback is executed before the subsystem-level
528->runtime_idle() callback and the result returned by it is not ignored. Namely,
529if it returns error code, the subsystem-level ->runtime_idle() callback will not
530be called and the helper function rpm_idle() executing it will return error
531code. This mechanism is intended to help platforms where saving device state
532is a time consuming operation and should only be carried out if all devices
533in the power domain are idle, before turning off the shared power resource(s).
534Namely, the power domain ->runtime_idle() callback may return error code until
535the pm_runtime_idle() helper (or its asychronous version) has been called for
536all devices in the power domain (it is recommended that the returned error code
537be -EBUSY in those cases), preventing the subsystem-level ->runtime_idle()
538callback from being run prematurely.
539
540The support for device power domains is only relevant to platforms needing to
541use the same subsystem-level (e.g. platform bus type) and device driver power
542management callbacks in many different power domain configurations and wanting
543to avoid incorporating the support for power domains into the subsystem-level
544callbacks. The other platforms need not implement it or take it into account
545in any way.
546
547
Alan Sternd6f9cda2010-03-26 23:53:55 +0100548System Devices
549--------------
550System devices (sysdevs) follow a slightly different API, which can be found in
551
552 include/linux/sysdev.h
553 drivers/base/sys.c
554
555System devices will be suspended with interrupts disabled, and after all other
556devices have been suspended. On resume, they will be resumed before any other
557devices, and also with interrupts disabled. These things occur in special
558"sysdev_driver" phases, which affect only system devices.
559
560Thus, after the suspend_noirq (or freeze_noirq or poweroff_noirq) phase, when
561the non-boot CPUs are all offline and IRQs are disabled on the remaining online
562CPU, then a sysdev_driver.suspend phase is carried out, and the system enters a
563sleep state (or a system image is created). During resume (or after the image
564has been created or loaded) a sysdev_driver.resume phase is carried out, IRQs
565are enabled on the only online CPU, the non-boot CPUs are enabled, and the
566resume_noirq (or thaw_noirq or restore_noirq) phase begins.
567
568Code to actually enter and exit the system-wide low power state sometimes
569involves hardware details that are only known to the boot firmware, and
570may leave a CPU running software (from SRAM or flash memory) that monitors
571the system and manages its wakeup sequence.
David Brownell4fc08402006-08-10 16:38:28 -0700572
573
574Device Low Power (suspend) States
575---------------------------------
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100576Device low-power states aren't standard. One device might only handle
David Brownell4fc08402006-08-10 16:38:28 -0700577"on" and "off, while another might support a dozen different versions of
578"on" (how many engines are active?), plus a state that gets back to "on"
579faster than from a full "off".
580
581Some busses define rules about what different suspend states mean. PCI
582gives one example: after the suspend sequence completes, a non-legacy
583PCI device may not perform DMA or issue IRQs, and any wakeup events it
584issues would be issued through the PME# bus signal. Plus, there are
585several PCI-standard device states, some of which are optional.
586
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100587In contrast, integrated system-on-chip processors often use IRQs as the
David Brownell4fc08402006-08-10 16:38:28 -0700588wakeup event sources (so drivers would call enable_irq_wake) and might
589be able to treat DMA completion as a wakeup event (sometimes DMA can stay
590active too, it'd only be the CPU and some peripherals that sleep).
591
592Some details here may be platform-specific. Systems may have devices that
593can be fully active in certain sleep states, such as an LCD display that's
594refreshed using DMA while most of the system is sleeping lightly ... and
595its frame buffer might even be updated by a DSP or other non-Linux CPU while
596the Linux control processor stays idle.
597
598Moreover, the specific actions taken may depend on the target system state.
599One target system state might allow a given device to be very operational;
600another might require a hard shut down with re-initialization on resume.
601And two different target systems might use the same device in different
602ways; the aforementioned LCD might be active in one product's "standby",
603but a different product using the same SOC might work differently.
604
605
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100606Power Management Notifiers
607--------------------------
Alan Sternd6f9cda2010-03-26 23:53:55 +0100608There are some operations that cannot be carried out by the power management
609callbacks discussed above, because the callbacks occur too late or too early.
610To handle these cases, subsystems and device drivers may register power
611management notifiers that are called before tasks are frozen and after they have
612been thawed. Generally speaking, the PM notifiers are suitable for performing
613actions that either require user space to be available, or at least won't
614interfere with user space.
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100615
616For details refer to Documentation/power/notifiers.txt.
617
618
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619Runtime Power Management
David Brownell4fc08402006-08-10 16:38:28 -0700620========================
621Many devices are able to dynamically power down while the system is still
622running. This feature is useful for devices that are not being used, and
623can offer significant power savings on a running system. These devices
624often support a range of runtime power states, which might use names such
625as "off", "sleep", "idle", "active", and so on. Those states will in some
Alan Sternd6f9cda2010-03-26 23:53:55 +0100626cases (like PCI) be partially constrained by the bus the device uses, and will
David Brownell4fc08402006-08-10 16:38:28 -0700627usually include hardware states that are also used in system sleep states.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628
Alan Sternd6f9cda2010-03-26 23:53:55 +0100629A system-wide power transition can be started while some devices are in low
630power states due to runtime power management. The system sleep PM callbacks
631should recognize such situations and react to them appropriately, but the
632necessary actions are subsystem-specific.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633
Alan Sternd6f9cda2010-03-26 23:53:55 +0100634In some cases the decision may be made at the subsystem level while in other
635cases the device driver may be left to decide. In some cases it may be
636desirable to leave a suspended device in that state during a system-wide power
637transition, but in other cases the device must be put back into the full-power
638state temporarily, for example so that its system wakeup capability can be
639disabled. This all depends on the hardware and the design of the subsystem and
640device driver in question.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641
Alan Sternd6f9cda2010-03-26 23:53:55 +0100642During system-wide resume from a sleep state it's best to put devices into the
643full-power state, as explained in Documentation/power/runtime_pm.txt. Refer to
644that document for more information regarding this particular issue as well as
Rafael J. Wysocki624f6ec2010-03-26 23:53:42 +0100645for information on the device runtime power management framework in general.