1Device Power Management
   3Copyright (c) 2010-2011 Rafael J. Wysocki <>, Novell Inc.
   4Copyright (c) 2010 Alan Stern <>
   7Most of the code in Linux is device drivers, so most of the Linux power
   8management (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.
  12This 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.
  18Two Models for Device Power Management
  20Drivers will use one or both of these models to put devices into low-power
  23    System Sleep model:
  24        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").
  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.
  34        Some drivers can manage hardware wakeup events, which make the system
  35        leave the low-power state.  This feature may be enabled or disabled
  36        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
  39        system enter low-power states more often.
  41    Runtime Power Management model:
  42        Devices may also be put into low-power states while the system is
  43        running, independently of other power management activity in principle.
  44        However, devices are not generally independent of each other (for
  45        example, a parent device cannot be suspended unless all of its child
  46        devices have been suspended).  Moreover, depending on the bus type the
  47        device is on, it may be necessary to carry out some bus-specific
  48        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).
  52        For these reasons not only the device driver itself, but also the
  53        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.
  59There'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.
  66Most 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.
  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).
  76Interfaces for Entering System Sleep States
  78There 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.
  84Device Power Management Operations
  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:
  90struct 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_late)(struct device *dev);
 100        int (*resume_early)(struct device *dev);
 101        int (*freeze_late)(struct device *dev);
 102        int (*thaw_early)(struct device *dev);
 103        int (*poweroff_late)(struct device *dev);
 104        int (*restore_early)(struct device *dev);
 105        int (*suspend_noirq)(struct device *dev);
 106        int (*resume_noirq)(struct device *dev);
 107        int (*freeze_noirq)(struct device *dev);
 108        int (*thaw_noirq)(struct device *dev);
 109        int (*poweroff_noirq)(struct device *dev);
 110        int (*restore_noirq)(struct device *dev);
 111        int (*runtime_suspend)(struct device *dev);
 112        int (*runtime_resume)(struct device *dev);
 113        int (*runtime_idle)(struct device *dev);
 116This structure is defined in include/linux/pm.h and the methods included in it
 117are also described in that file.  Their roles will be explained in what follows.
 118For now, it should be sufficient to remember that the last three methods are
 119specific to runtime power management while the remaining ones are used during
 120system-wide power transitions.
 122There also is a deprecated "old" or "legacy" interface for power management
 123operations available at least for some subsystems.  This approach does not use
 124struct dev_pm_ops objects and it is suitable only for implementing system sleep
 125power management methods.  Therefore it is not described in this document, so
 126please refer directly to the source code for more information about it.
 129Subsystem-Level Methods
 131The core methods to suspend and resume devices reside in struct dev_pm_ops
 132pointed to by the ops member of struct dev_pm_domain, or by the pm member of
 133struct bus_type, struct device_type and struct class.  They are mostly of
 134interest to the people writing infrastructure for platforms and buses, like PCI
 135or USB, or device type and device class drivers.  They also are relevant to the
 136writers of device drivers whose subsystems (PM domains, device types, device
 137classes and bus types) don't provide all power management methods.
 139Bus drivers implement these methods as appropriate for the hardware and the
 140drivers using it; PCI works differently from USB, and so on.  Not many people
 141write subsystem-level drivers; most driver code is a "device driver" that builds
 142on top of bus-specific framework code.
 144For more information on these driver calls, see the description later;
 145they are called in phases for every device, respecting the parent-child
 146sequencing in the driver model tree.
 149/sys/devices/.../power/wakeup files
 151All device objects in the driver model contain fields that control the handling
 152of system wakeup events (hardware signals that can force the system out of a
 153sleep state).  These fields are initialized by bus or device driver code using
 154device_set_wakeup_capable() and device_set_wakeup_enable(), defined in
 157The "power.can_wakeup" flag just records whether the device (and its driver) can
 158physically support wakeup events.  The device_set_wakeup_capable() routine
 159affects this flag.  The "power.wakeup" field is a pointer to an object of type
 160struct wakeup_source used for controlling whether or not the device should use
 161its system wakeup mechanism and for notifying the PM core of system wakeup
 162events signaled by the device.  This object is only present for wakeup-capable
 163devices (i.e. devices whose "can_wakeup" flags are set) and is created (or
 164removed) by device_set_wakeup_capable().
 166Whether or not a device is capable of issuing wakeup events is a hardware
 167matter, and the kernel is responsible for keeping track of it.  By contrast,
 168whether or not a wakeup-capable device should issue wakeup events is a policy
 169decision, and it is managed by user space through a sysfs attribute: the
 170"power/wakeup" file.  User space can write the strings "enabled" or "disabled"
 171to it to indicate whether or not, respectively, the device is supposed to signal
 172system wakeup.  This file is only present if the "power.wakeup" object exists
 173for the given device and is created (or removed) along with that object, by
 174device_set_wakeup_capable().  Reads from the file will return the corresponding
 177The "power/wakeup" file is supposed to contain the "disabled" string initially
 178for the majority of devices; the major exceptions are power buttons, keyboards,
 179and Ethernet adapters whose WoL (wake-on-LAN) feature has been set up with
 180ethtool.  It should also default to "enabled" for devices that don't generate
 181wakeup requests on their own but merely forward wakeup requests from one bus to
 182another (like PCI Express ports).
 184The device_may_wakeup() routine returns true only if the "power.wakeup" object
 185exists and the corresponding "power/wakeup" file contains the string "enabled".
 186This information is used by subsystems, like the PCI bus type code, to see
 187whether or not to enable the devices' wakeup mechanisms.  If device wakeup
 188mechanisms are enabled or disabled directly by drivers, they also should use
 189device_may_wakeup() to decide what to do during a system sleep transition.
 190Device drivers, however, are not supposed to call device_set_wakeup_enable()
 191directly in any case.
 193It ought to be noted that system wakeup is conceptually different from "remote
 194wakeup" used by runtime power management, although it may be supported by the
 195same physical mechanism.  Remote wakeup is a feature allowing devices in
 196low-power states to trigger specific interrupts to signal conditions in which
 197they should be put into the full-power state.  Those interrupts may or may not
 198be used to signal system wakeup events, depending on the hardware design.  On
 199some systems it is impossible to trigger them from system sleep states.  In any
 200case, remote wakeup should always be enabled for runtime power management for
 201all devices and drivers that support it.
 203/sys/devices/.../power/control files
 205Each device in the driver model has a flag to control whether it is subject to
 206runtime power management.  This flag, called runtime_auto, is initialized by the
 207bus type (or generally subsystem) code using pm_runtime_allow() or
 208pm_runtime_forbid(); the default is to allow runtime power management.
 210The setting can be adjusted by user space by writing either "on" or "auto" to
 211the device's power/control sysfs file.  Writing "auto" calls pm_runtime_allow(),
 212setting the flag and allowing the device to be runtime power-managed by its
 213driver.  Writing "on" calls pm_runtime_forbid(), clearing the flag, returning
 214the device to full power if it was in a low-power state, and preventing the
 215device from being runtime power-managed.  User space can check the current value
 216of the runtime_auto flag by reading the file.
 218The device's runtime_auto flag has no effect on the handling of system-wide
 219power transitions.  In particular, the device can (and in the majority of cases
 220should and will) be put into a low-power state during a system-wide transition
 221to a sleep state even though its runtime_auto flag is clear.
 223For more information about the runtime power management framework, refer to
 227Calling Drivers to Enter and Leave System Sleep States
 229When the system goes into a sleep state, each device's driver is asked to
 230suspend the device by putting it into a state compatible with the target
 231system state.  That's usually some version of "off", but the details are
 232system-specific.  Also, wakeup-enabled devices will usually stay partly
 233functional in order to wake the system.
 235When the system leaves that low-power state, the device's driver is asked to
 236resume it by returning it to full power.  The suspend and resume operations
 237always go together, and both are multi-phase operations.
 239For simple drivers, suspend might quiesce the device using class code
 240and then turn its hardware as "off" as possible during suspend_noirq.  The
 241matching resume calls would then completely reinitialize the hardware
 242before reactivating its class I/O queues.
 244More power-aware drivers might prepare the devices for triggering system wakeup
 248Call Sequence Guarantees
 250To ensure that bridges and similar links needing to talk to a device are
 251available when the device is suspended or resumed, the device tree is
 252walked in a bottom-up order to suspend devices.  A top-down order is
 253used to resume those devices.
 255The ordering of the device tree is defined by the order in which devices
 256get registered:  a child can never be registered, probed or resumed before
 257its parent; and can't be removed or suspended after that parent.
 259The policy is that the device tree should match hardware bus topology.
 260(Or at least the control bus, for devices which use multiple busses.)
 261In particular, this means that a device registration may fail if the parent of
 262the device is suspending (i.e. has been chosen by the PM core as the next
 263device to suspend) or has already suspended, as well as after all of the other
 264devices have been suspended.  Device drivers must be prepared to cope with such
 268System Power Management Phases
 270Suspending or resuming the system is done in several phases.  Different phases
 271are used for standby or memory sleep states ("suspend-to-RAM") and the
 272hibernation state ("suspend-to-disk").  Each phase involves executing callbacks
 273for every device before the next phase begins.  Not all busses or classes
 274support all these callbacks and not all drivers use all the callbacks.  The
 275various phases always run after tasks have been frozen and before they are
 276unfrozen.  Furthermore, the *_noirq phases run at a time when IRQ handlers have
 277been disabled (except for those marked with the IRQF_NO_SUSPEND flag).
 279All phases use PM domain, bus, type, class or driver callbacks (that is, methods
 280defined in dev->pm_domain->ops, dev->bus->pm, dev->type->pm, dev->class->pm or
 281dev->driver->pm).  These callbacks are regarded by the PM core as mutually
 282exclusive.  Moreover, PM domain callbacks always take precedence over all of the
 283other callbacks and, for example, type callbacks take precedence over bus, class
 284and driver callbacks.  To be precise, the following rules are used to determine
 285which callback to execute in the given phase:
 287    1.  If dev->pm_domain is present, the PM core will choose the callback
 288        included in dev->pm_domain->ops for execution
 290    2.  Otherwise, if both dev->type and dev->type->pm are present, the callback
 291        included in dev->type->pm will be chosen for execution.
 293    3.  Otherwise, if both dev->class and dev->class->pm are present, the
 294        callback included in dev->class->pm will be chosen for execution.
 296    4.  Otherwise, if both dev->bus and dev->bus->pm are present, the callback
 297        included in dev->bus->pm will be chosen for execution.
 299This allows PM domains and device types to override callbacks provided by bus
 300types or device classes if necessary.
 302The PM domain, type, class and bus callbacks may in turn invoke device- or
 303driver-specific methods stored in dev->driver->pm, but they don't have to do
 306If the subsystem callback chosen for execution is not present, the PM core will
 307execute the corresponding method from dev->driver->pm instead if there is one.
 310Entering System Suspend
 312When the system goes into the standby or memory sleep state, the phases are:
 314                prepare, suspend, suspend_late, suspend_noirq.
 316    1.  The prepare phase is meant to prevent races by preventing new devices
 317        from being registered; the PM core would never know that all the
 318        children of a device had been suspended if new children could be
 319        registered at will.  (By contrast, devices may be unregistered at any
 320        time.)  Unlike the other suspend-related phases, during the prepare
 321        phase the device tree is traversed top-down.
 323        After the prepare callback method returns, no new children may be
 324        registered below the device.  The method may also prepare the device or
 325        driver in some way for the upcoming system power transition, but it
 326        should not put the device into a low-power state.
 328    2.  The suspend methods should quiesce the device to stop it from performing
 329        I/O.  They also may save the device registers and put it into the
 330        appropriate low-power state, depending on the bus type the device is on,
 331        and they may enable wakeup events.
 333    3   For a number of devices it is convenient to split suspend into the
 334        "quiesce device" and "save device state" phases, in which cases
 335        suspend_late is meant to do the latter.  It is always executed after
 336        runtime power management has been disabled for all devices.
 338    4.  The suspend_noirq phase occurs after IRQ handlers have been disabled,
 339        which means that the driver's interrupt handler will not be called while
 340        the callback method is running.  The methods should save the values of
 341        the device's registers that weren't saved previously and finally put the
 342        device into the appropriate low-power state.
 344        The majority of subsystems and device drivers need not implement this
 345        callback.  However, bus types allowing devices to share interrupt
 346        vectors, like PCI, generally need it; otherwise a driver might encounter
 347        an error during the suspend phase by fielding a shared interrupt
 348        generated by some other device after its own device had been set to low
 349        power.
 351At the end of these phases, drivers should have stopped all I/O transactions
 352(DMA, IRQs), saved enough state that they can re-initialize or restore previous
 353state (as needed by the hardware), and placed the device into a low-power state.
 354On many platforms they will gate off one or more clock sources; sometimes they
 355will also switch off power supplies or reduce voltages.  (Drivers supporting
 356runtime PM may already have performed some or all of these steps.)
 358If device_may_wakeup(dev) returns true, the device should be prepared for
 359generating hardware wakeup signals to trigger a system wakeup event when the
 360system is in the sleep state.  For example, enable_irq_wake() might identify
 361GPIO signals hooked up to a switch or other external hardware, and
 362pci_enable_wake() does something similar for the PCI PME signal.
 364If any of these callbacks returns an error, the system won't enter the desired
 365low-power state.  Instead the PM core will unwind its actions by resuming all
 366the devices that were suspended.
 369Leaving System Suspend
 371When resuming from standby or memory sleep, the phases are:
 373                resume_noirq, resume_early, resume, complete.
 375    1.  The resume_noirq callback methods should perform any actions needed
 376        before the driver's interrupt handlers are invoked.  This generally
 377        means undoing the actions of the suspend_noirq phase.  If the bus type
 378        permits devices to share interrupt vectors, like PCI, the method should
 379        bring the device and its driver into a state in which the driver can
 380        recognize if the device is the source of incoming interrupts, if any,
 381        and handle them correctly.
 383        For example, the PCI bus type's ->pm.resume_noirq() puts the device into
 384        the full-power state (D0 in the PCI terminology) and restores the
 385        standard configuration registers of the device.  Then it calls the
 386        device driver's ->pm.resume_noirq() method to perform device-specific
 387        actions.
 389    2.  The resume_early methods should prepare devices for the execution of
 390        the resume methods.  This generally involves undoing the actions of the
 391        preceding suspend_late phase.
 393    3   The resume methods should bring the the device back to its operating
 394        state, so that it can perform normal I/O.  This generally involves
 395        undoing the actions of the suspend phase.
 397    4.  The complete phase should undo the actions of the prepare phase.  Note,
 398        however, that new children may be registered below the device as soon as
 399        the resume callbacks occur; it's not necessary to wait until the
 400        complete phase.
 402At the end of these phases, drivers should be as functional as they were before
 403suspending: I/O can be performed using DMA and IRQs, and the relevant clocks are
 404gated on.  Even if the device was in a low-power state before the system sleep
 405because of runtime power management, afterwards it should be back in its
 406full-power state.  There are multiple reasons why it's best to do this; they are
 407discussed in more detail in Documentation/power/runtime_pm.txt.
 409However, the details here may again be platform-specific.  For example,
 410some systems support multiple "run" states, and the mode in effect at
 411the end of resume might not be the one which preceded suspension.
 412That means availability of certain clocks or power supplies changed,
 413which could easily affect how a driver works.
 415Drivers need to be able to handle hardware which has been reset since the
 416suspend methods were called, for example by complete reinitialization.
 417This may be the hardest part, and the one most protected by NDA'd documents
 418and chip errata.  It's simplest if the hardware state hasn't changed since
 419the suspend was carried out, but that can't be guaranteed (in fact, it usually
 420is not the case).
 422Drivers must also be prepared to notice that the device has been removed
 423while the system was powered down, whenever that's physically possible.
 424PCMCIA, MMC, USB, Firewire, SCSI, and even IDE are common examples of busses
 425where common Linux platforms will see such removal.  Details of how drivers
 426will notice and handle such removals are currently bus-specific, and often
 427involve a separate thread.
 429These callbacks may return an error value, but the PM core will ignore such
 430errors since there's nothing it can do about them other than printing them in
 431the system log.
 434Entering Hibernation
 436Hibernating the system is more complicated than putting it into the standby or
 437memory sleep state, because it involves creating and saving a system image.
 438Therefore there are more phases for hibernation, with a different set of
 439callbacks.  These phases always run after tasks have been frozen and memory has
 440been freed.
 442The general procedure for hibernation is to quiesce all devices (freeze), create
 443an image of the system memory while everything is stable, reactivate all
 444devices (thaw), write the image to permanent storage, and finally shut down the
 445system (poweroff).  The phases used to accomplish this are:
 447        prepare, freeze, freeze_late, freeze_noirq, thaw_noirq, thaw_early,
 448        thaw, complete, prepare, poweroff, poweroff_late, poweroff_noirq
 450    1.  The prepare phase is discussed in the "Entering System Suspend" section
 451        above.
 453    2.  The freeze methods should quiesce the device so that it doesn't generate
 454        IRQs or DMA, and they may need to save the values of device registers.
 455        However the device does not have to be put in a low-power state, and to
 456        save time it's best not to do so.  Also, the device should not be
 457        prepared to generate wakeup events.
 459    3.  The freeze_late phase is analogous to the suspend_late phase described
 460        above, except that the device should not be put in a low-power state and
 461        should not be allowed to generate wakeup events by it.
 463    4.  The freeze_noirq phase is analogous to the suspend_noirq phase discussed
 464        above, except again that the device should not be put in a low-power
 465        state and should not be allowed to generate wakeup events.
 467At this point the system image is created.  All devices should be inactive and
 468the contents of memory should remain undisturbed while this happens, so that the
 469image forms an atomic snapshot of the system state.
 471    5.  The thaw_noirq phase is analogous to the resume_noirq phase discussed
 472        above.  The main difference is that its methods can assume the device is
 473        in the same state as at the end of the freeze_noirq phase.
 475    6.  The thaw_early phase is analogous to the resume_early phase described
 476        above.  Its methods should undo the actions of the preceding
 477        freeze_late, if necessary.
 479    7.  The thaw phase is analogous to the resume phase discussed above.  Its
 480        methods should bring the device back to an operating state, so that it
 481        can be used for saving the image if necessary.
 483    8.  The complete phase is discussed in the "Leaving System Suspend" section
 484        above.
 486At this point the system image is saved, and the devices then need to be
 487prepared for the upcoming system shutdown.  This is much like suspending them
 488before putting the system into the standby or memory sleep state, and the phases
 489are similar.
 491    9.  The prepare phase is discussed above.
 493    10. The poweroff phase is analogous to the suspend phase.
 495    11. The poweroff_late phase is analogous to the suspend_late phase.
 497    12. The poweroff_noirq phase is analogous to the suspend_noirq phase.
 499The poweroff, poweroff_late and poweroff_noirq callbacks should do essentially
 500the same things as the suspend, suspend_late and suspend_noirq callbacks,
 501respectively.  The only notable difference is that they need not store the
 502device register values, because the registers should already have been stored
 503during the freeze, freeze_late or freeze_noirq phases.
 506Leaving Hibernation
 508Resuming from hibernation is, again, more complicated than resuming from a sleep
 509state in which the contents of main memory are preserved, because it requires
 510a system image to be loaded into memory and the pre-hibernation memory contents
 511to be restored before control can be passed back to the image kernel.
 513Although in principle, the image might be loaded into memory and the
 514pre-hibernation memory contents restored by the boot loader, in practice this
 515can't be done because boot loaders aren't smart enough and there is no
 516established protocol for passing the necessary information.  So instead, the
 517boot loader loads a fresh instance of the kernel, called the boot kernel, into
 518memory and passes control to it in the usual way.  Then the boot kernel reads
 519the system image, restores the pre-hibernation memory contents, and passes
 520control to the image kernel.  Thus two different kernels are involved in
 521resuming from hibernation.  In fact, the boot kernel may be completely different
 522from the image kernel: a different configuration and even a different version.
 523This has important consequences for device drivers and their subsystems.
 525To be able to load the system image into memory, the boot kernel needs to
 526include at least a subset of device drivers allowing it to access the storage
 527medium containing the image, although it doesn't need to include all of the
 528drivers present in the image kernel.  After the image has been loaded, the
 529devices managed by the boot kernel need to be prepared for passing control back
 530to the image kernel.  This is very similar to the initial steps involved in
 531creating a system image, and it is accomplished in the same way, using prepare,
 532freeze, and freeze_noirq phases.  However the devices affected by these phases
 533are only those having drivers in the boot kernel; other devices will still be in
 534whatever state the boot loader left them.
 536Should the restoration of the pre-hibernation memory contents fail, the boot
 537kernel would go through the "thawing" procedure described above, using the
 538thaw_noirq, thaw, and complete phases, and then continue running normally.  This
 539happens only rarely.  Most often the pre-hibernation memory contents are
 540restored successfully and control is passed to the image kernel, which then
 541becomes responsible for bringing the system back to the working state.
 543To achieve this, the image kernel must restore the devices' pre-hibernation
 544functionality.  The operation is much like waking up from the memory sleep
 545state, although it involves different phases:
 547        restore_noirq, restore_early, restore, complete
 549    1.  The restore_noirq phase is analogous to the resume_noirq phase.
 551    2.  The restore_early phase is analogous to the resume_early phase.
 553    3.  The restore phase is analogous to the resume phase.
 555    4.  The complete phase is discussed above.
 557The main difference from resume[_early|_noirq] is that restore[_early|_noirq]
 558must assume the device has been accessed and reconfigured by the boot loader or
 559the boot kernel.  Consequently the state of the device may be different from the
 560state remembered from the freeze, freeze_late and freeze_noirq phases.  The
 561device may even need to be reset and completely re-initialized.  In many cases
 562this difference doesn't matter, so the resume[_early|_noirq] and
 563restore[_early|_norq] method pointers can be set to the same routines.
 564Nevertheless, different callback pointers are used in case there is a situation
 565where it actually does matter.
 568Device Power Management Domains
 570Sometimes devices share reference clocks or other power resources.  In those
 571cases it generally is not possible to put devices into low-power states
 572individually.  Instead, a set of devices sharing a power resource can be put
 573into a low-power state together at the same time by turning off the shared
 574power resource.  Of course, they also need to be put into the full-power state
 575together, by turning the shared power resource on.  A set of devices with this
 576property is often referred to as a power domain.
 578Support for power domains is provided through the pm_domain field of struct
 579device.  This field is a pointer to an object of type struct dev_pm_domain,
 580defined in include/linux/pm.h, providing a set of power management callbacks
 581analogous to the subsystem-level and device driver callbacks that are executed
 582for the given device during all power transitions, instead of the respective
 583subsystem-level callbacks.  Specifically, if a device's pm_domain pointer is
 584not NULL, the ->suspend() callback from the object pointed to by it will be
 585executed instead of its subsystem's (e.g. bus type's) ->suspend() callback and
 586analogously for all of the remaining callbacks.  In other words, power
 587management domain callbacks, if defined for the given device, always take
 588precedence over the callbacks provided by the device's subsystem (e.g. bus
 591The support for device power management domains is only relevant to platforms
 592needing to use the same device driver power management callbacks in many
 593different power domain configurations and wanting to avoid incorporating the
 594support for power domains into subsystem-level callbacks, for example by
 595modifying the platform bus type.  Other platforms need not implement it or take
 596it into account in any way.
 599Device Low Power (suspend) States
 601Device low-power states aren't standard.  One device might only handle
 602"on" and "off", while another might support a dozen different versions of
 603"on" (how many engines are active?), plus a state that gets back to "on"
 604faster than from a full "off".
 606Some busses define rules about what different suspend states mean.  PCI
 607gives one example:  after the suspend sequence completes, a non-legacy
 608PCI device may not perform DMA or issue IRQs, and any wakeup events it
 609issues would be issued through the PME# bus signal.  Plus, there are
 610several PCI-standard device states, some of which are optional.
 612In contrast, integrated system-on-chip processors often use IRQs as the
 613wakeup event sources (so drivers would call enable_irq_wake) and might
 614be able to treat DMA completion as a wakeup event (sometimes DMA can stay
 615active too, it'd only be the CPU and some peripherals that sleep).
 617Some details here may be platform-specific.  Systems may have devices that
 618can be fully active in certain sleep states, such as an LCD display that's
 619refreshed using DMA while most of the system is sleeping lightly ... and
 620its frame buffer might even be updated by a DSP or other non-Linux CPU while
 621the Linux control processor stays idle.
 623Moreover, the specific actions taken may depend on the target system state.
 624One target system state might allow a given device to be very operational;
 625another might require a hard shut down with re-initialization on resume.
 626And two different target systems might use the same device in different
 627ways; the aforementioned LCD might be active in one product's "standby",
 628but a different product using the same SOC might work differently.
 631Power Management Notifiers
 633There are some operations that cannot be carried out by the power management
 634callbacks discussed above, because the callbacks occur too late or too early.
 635To handle these cases, subsystems and device drivers may register power
 636management notifiers that are called before tasks are frozen and after they have
 637been thawed.  Generally speaking, the PM notifiers are suitable for performing
 638actions that either require user space to be available, or at least won't
 639interfere with user space.
 641For details refer to Documentation/power/notifiers.txt.
 644Runtime Power Management
 646Many devices are able to dynamically power down while the system is still
 647running. This feature is useful for devices that are not being used, and
 648can offer significant power savings on a running system.  These devices
 649often support a range of runtime power states, which might use names such
 650as "off", "sleep", "idle", "active", and so on.  Those states will in some
 651cases (like PCI) be partially constrained by the bus the device uses, and will
 652usually include hardware states that are also used in system sleep states.
 654A system-wide power transition can be started while some devices are in low
 655power states due to runtime power management.  The system sleep PM callbacks
 656should recognize such situations and react to them appropriately, but the
 657necessary actions are subsystem-specific.
 659In some cases the decision may be made at the subsystem level while in other
 660cases the device driver may be left to decide.  In some cases it may be
 661desirable to leave a suspended device in that state during a system-wide power
 662transition, but in other cases the device must be put back into the full-power
 663state temporarily, for example so that its system wakeup capability can be
 664disabled.  This all depends on the hardware and the design of the subsystem and
 665device driver in question.
 667During system-wide resume from a sleep state it's easiest to put devices into
 668the full-power state, as explained in Documentation/power/runtime_pm.txt.  Refer
 669to that document for more information regarding this particular issue as well as
 670for information on the device runtime power management framework in general.
 671 kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.