Device Power Management Basics¶
Copyright (c) 2010-2011 Rafael J. Wysocki <firstname.lastname@example.org>, Novell Inc. Copyright (c) 2010 Alan Stern <email@example.com> Copyright (c) 2016 Intel Corp., Rafael J. Wysocki <firstname.lastname@example.org>
Most of the code in Linux is device drivers, so most of the Linux power management (PM) code is also driver-specific. Most drivers will do very little; others, especially for platforms with small batteries (like cell phones), will do a lot.
This writeup gives an overview of how drivers interact with system-wide power management goals, emphasizing the models and interfaces that are shared by everything that hooks up to the driver model core. Read it as background for the domain-specific work you’d do with any specific driver.
Two Models for Device Power Management¶
Drivers will use one or both of these models to put devices into low-power states:
System Sleep model:
Drivers can enter low-power states as part of entering system-wide low-power states like “suspend” (also known as “suspend-to-RAM”), or (mostly for systems with disks) “hibernation” (also known as “suspend-to-disk”).
This is something that device, bus, and class drivers collaborate on by implementing various role-specific suspend and resume methods to cleanly power down hardware and software subsystems, then reactivate them without loss of data.
Some drivers can manage hardware wakeup events, which make the system leave the low-power state. This feature may be enabled or disabled using the relevant
/sys/devices/.../power/wakeupfile (for Ethernet drivers the ioctl interface used by ethtool may also be used for this purpose); enabling it may cost some power usage, but let the whole system enter low-power states more often.
Runtime Power Management model:
Devices may also be put into low-power states while the system is running, independently of other power management activity in principle. However, devices are not generally independent of each other (for example, a parent device cannot be suspended unless all of its child devices have been suspended). Moreover, depending on the bus type the device is on, it may be necessary to carry out some bus-specific operations on the device for this purpose. Devices put into low power states at run time may require special handling during system-wide power transitions (suspend or hibernation).
For these reasons not only the device driver itself, but also the appropriate subsystem (bus type, device type or device class) driver and the PM core are involved in runtime power management. As in the system sleep power management case, they need to collaborate by implementing various role-specific suspend and resume methods, so that the hardware is cleanly powered down and reactivated without data or service loss.
There’s not a lot to be said about those low-power states except that they are very system-specific, and often device-specific. Also, that if enough devices have been put into low-power states (at runtime), the effect may be very similar to entering some system-wide low-power state (system sleep) ... and that synergies exist, so that several drivers using runtime PM might put the system into a state where even deeper power saving options are available.
Most suspended devices will have quiesced all I/O: no more DMA or IRQs (except for wakeup events), no more data read or written, and requests from upstream drivers are no longer accepted. A given bus or platform may have different requirements though.
Examples of hardware wakeup events include an alarm from a real time clock, network wake-on-LAN packets, keyboard or mouse activity, and media insertion or removal (for PCMCIA, MMC/SD, USB, and so on).
Interfaces for Entering System Sleep States¶
There are programming interfaces provided for subsystems (bus type, device type, device class) and device drivers to allow them to participate in the power management of devices they are concerned with. These interfaces cover both system sleep and runtime power management.
Device Power Management Operations¶
Device power management operations, at the subsystem level as well as at the
device driver level, are implemented by defining and populating objects of type
struct dev_pm_ops defined in
include/linux/pm.h. The roles of the
methods included in it will be explained in what follows. For now, it should be
sufficient to remember that the last three methods are specific to runtime power
management while the remaining ones are used during system-wide power
There also is a deprecated “old” or “legacy” interface for power management
operations available at least for some subsystems. This approach does not use
struct dev_pm_ops objects and it is suitable only for implementing system
sleep power management methods in a limited way. Therefore it is not described
in this document, so please refer directly to the source code for more
information about it.
The core methods to suspend and resume devices reside in
struct dev_pm_ops pointed to by the
ops member of
struct dev_pm_domain, or by the
pm member of
struct device_type and
struct class. They are mostly of interest to the
people writing infrastructure for platforms and buses, like PCI or USB, or
device type and device class drivers. They also are relevant to the writers of
device drivers whose subsystems (PM domains, device types, device classes and
bus types) don’t provide all power management methods.
Bus drivers implement these methods as appropriate for the hardware and the drivers using it; PCI works differently from USB, and so on. Not many people write subsystem-level drivers; most driver code is a “device driver” that builds on top of bus-specific framework code.
For more information on these driver calls, see the description later; they are called in phases for every device, respecting the parent-child sequencing in the driver model tree.
All device objects in the driver model contain fields that control the handling
of system wakeup events (hardware signals that can force the system out of a
sleep state). These fields are initialized by bus or device driver code using
power.can_wakeup flag just records whether the device (and its
driver) can physically support wakeup events. The
device_set_wakeup_capable() routine affects this flag. The
power.wakeup field is a pointer to an object of type
struct wakeup_source used for controlling whether or not the device should use
its system wakeup mechanism and for notifying the PM core of system wakeup
events signaled by the device. This object is only present for wakeup-capable
devices (i.e. devices whose
can_wakeup flags are set) and is created
(or removed) by
Whether or not a device is capable of issuing wakeup events is a hardware
matter, and the kernel is responsible for keeping track of it. By contrast,
whether or not a wakeup-capable device should issue wakeup events is a policy
decision, and it is managed by user space through a sysfs attribute: the
power/wakeup file. User space can write the “enabled” or “disabled”
strings to it to indicate whether or not, respectively, the device is supposed
to signal system wakeup. This file is only present if the
power.wakeup object exists for the given device and is created (or
removed) along with that object, by
Reads from the file will return the corresponding string.
The initial value in the
power/wakeup file is “disabled” for the
majority of devices; the major exceptions are power buttons, keyboards, and
Ethernet adapters whose WoL (wake-on-LAN) feature has been set up with ethtool.
It should also default to “enabled” for devices that don’t generate wakeup
requests on their own but merely forward wakeup requests from one bus to another
(like PCI Express ports).
device_may_wakeup() routine returns true only if the
power.wakeup object exists and the corresponding
file contains the “enabled” string. This information is used by subsystems,
like the PCI bus type code, to see whether or not to enable the devices’ wakeup
mechanisms. If device wakeup mechanisms are enabled or disabled directly by
drivers, they also should use
device_may_wakeup() to decide what to do
during a system sleep transition. Device drivers, however, are not expected to
device_set_wakeup_enable() directly in any case.
It ought to be noted that system wakeup is conceptually different from “remote wakeup” used by runtime power management, although it may be supported by the same physical mechanism. Remote wakeup is a feature allowing devices in low-power states to trigger specific interrupts to signal conditions in which they should be put into the full-power state. Those interrupts may or may not be used to signal system wakeup events, depending on the hardware design. On some systems it is impossible to trigger them from system sleep states. In any case, remote wakeup should always be enabled for runtime power management for all devices and drivers that support it.
Each device in the driver model has a flag to control whether it is subject to
runtime power management. This flag,
runtime_auto, is initialized
by the bus type (or generally subsystem) code using
pm_runtime_forbid(); the default is to allow runtime power
The setting can be adjusted by user space by writing either “on” or “auto” to
power/control sysfs file. Writing “auto” calls
pm_runtime_allow(), setting the flag and allowing the device to be
runtime power-managed by its driver. Writing “on” calls
pm_runtime_forbid(), clearing the flag, returning the device to full
power if it was in a low-power state, and preventing the
device from being runtime power-managed. User space can check the current value
runtime_auto flag by reading that file.
runtime_auto flag has no effect on the handling of
system-wide power transitions. In particular, the device can (and in the
majority of cases should and will) be put into a low-power state during a
system-wide transition to a sleep state even though its
flag is clear.
For more information about the runtime power management framework, refer to
Calling Drivers to Enter and Leave System Sleep States¶
When the system goes into a sleep state, each device’s driver is asked to suspend the device by putting it into a state compatible with the target system state. That’s usually some version of “off”, but the details are system-specific. Also, wakeup-enabled devices will usually stay partly functional in order to wake the system.
When the system leaves that low-power state, the device’s driver is asked to resume it by returning it to full power. The suspend and resume operations always go together, and both are multi-phase operations.
For simple drivers, suspend might quiesce the device using class code and then turn its hardware as “off” as possible during suspend_noirq. The matching resume calls would then completely reinitialize the hardware before reactivating its class I/O queues.
More power-aware drivers might prepare the devices for triggering system wakeup events.
Call Sequence Guarantees¶
To ensure that bridges and similar links needing to talk to a device are available when the device is suspended or resumed, the device hierarchy is walked in a bottom-up order to suspend devices. A top-down order is used to resume those devices.
The ordering of the device hierarchy is defined by the order in which devices get registered: a child can never be registered, probed or resumed before its parent; and can’t be removed or suspended after that parent.
The policy is that the device hierarchy should match hardware bus topology. [Or at least the control bus, for devices which use multiple busses.] In particular, this means that a device registration may fail if the parent of the device is suspending (i.e. has been chosen by the PM core as the next device to suspend) or has already suspended, as well as after all of the other devices have been suspended. Device drivers must be prepared to cope with such situations.
System Power Management Phases¶
Suspending or resuming the system is done in several phases. Different phases
are used for suspend-to-idle, shallow (standby), and deep (“suspend-to-RAM”)
sleep states and the hibernation state (“suspend-to-disk”). Each phase involves
executing callbacks for every device before the next phase begins. Not all
buses or classes support all these callbacks and not all drivers use all the
callbacks. The various phases always run after tasks have been frozen and
before they are unfrozen. Furthermore, the
*_noirq phases run at a time
when IRQ handlers have been disabled (except for those marked with the
All phases use PM domain, bus, type, class or driver callbacks (that is, methods
dev->driver->pm). These callbacks are regarded by the
PM core as mutually exclusive. Moreover, PM domain callbacks always take
precedence over all of the other callbacks and, for example, type callbacks take
precedence over bus, class and driver callbacks. To be precise, the following
rules are used to determine which callback to execute in the given phase:
dev->pm_domainis present, the PM core will choose the callback provided by
- Otherwise, if both
dev->type->pmare present, the callback provided by
dev->type->pmwill be chosen for execution.
- Otherwise, if both
dev->class->pmare present, the callback provided by
dev->class->pmwill be chosen for execution.
- Otherwise, if both
dev->bus->pmare present, the callback provided by
dev->bus->pmwill be chosen for execution.
This allows PM domains and device types to override callbacks provided by bus types or device classes if necessary.
The PM domain, type, class and bus callbacks may in turn invoke device- or
driver-specific methods stored in
dev->driver->pm, but they don’t have to do
If the subsystem callback chosen for execution is not present, the PM core will
execute the corresponding method from the
dev->driver->pm set instead if
there is one.
Entering System Suspend¶
When the system goes into the freeze, standby or memory sleep state,
the phases are:
preparephase is meant to prevent races by preventing new devices from being registered; the PM core would never know that all the children of a device had been suspended if new children could be registered at will. [By contrast, from the PM core’s perspective, devices may be unregistered at any time.] Unlike the other suspend-related phases, during the
preparephase the device hierarchy is traversed top-down.
->preparecallback method returns, no new children may be registered below the device. The method may also prepare the device or driver in some way for the upcoming system power transition, but it should not put the device into a low-power state.
For devices supporting runtime power management, the return value of the prepare callback can be used to indicate to the PM core that it may safely leave the device in runtime suspend (if runtime-suspended already), provided that all of the device’s descendants are also left in runtime suspend. Namely, if the prepare callback returns a positive number and that happens for all of the descendants of the device too, and all of them (including the device itself) are runtime-suspended, the PM core will skip the
suspend_noirqphases as well as all of the corresponding phases of the subsequent device resume for all of these devices. In that case, the
->completecallback will be invoked directly after the
->preparecallback and is entirely responsible for putting the device into a consistent state as appropriate.
Note that this direct-complete procedure applies even if the device is disabled for runtime PM; only the runtime-PM status matters. It follows that if a device has system-sleep callbacks but does not support runtime PM, then its prepare callback must never return a positive value. This is because all such devices are initially set to runtime-suspended with runtime PM disabled.
->suspendmethods should quiesce the device to stop it from performing I/O. They also may save the device registers and put it into the appropriate low-power state, depending on the bus type the device is on, and they may enable wakeup events.
For a number of devices it is convenient to split suspend into the “quiesce device” and “save device state” phases, in which cases
suspend_lateis meant to do the latter. It is always executed after runtime power management has been disabled for the device in question.
suspend_noirqphase occurs after IRQ handlers have been disabled, which means that the driver’s interrupt handler will not be called while the callback method is running. The
->suspend_noirqmethods should save the values of the device’s registers that weren’t saved previously and finally put the device into the appropriate low-power state.
The majority of subsystems and device drivers need not implement this callback. However, bus types allowing devices to share interrupt vectors, like PCI, generally need it; otherwise a driver might encounter an error during the suspend phase by fielding a shared interrupt generated by some other device after its own device had been set to low power.
At the end of these phases, drivers should have stopped all I/O transactions (DMA, IRQs), saved enough state that they can re-initialize or restore previous state (as needed by the hardware), and placed the device into a low-power state. On many platforms they will gate off one or more clock sources; sometimes they will also switch off power supplies or reduce voltages. [Drivers supporting runtime PM may already have performed some or all of these steps.]
true, the device should be
prepared for generating hardware wakeup signals to trigger a system wakeup event
when the system is in the sleep state. For example,
might identify GPIO signals hooked up to a switch or other external hardware,
pci_enable_wake() does something similar for the PCI PME signal.
If any of these callbacks returns an error, the system won’t enter the desired low-power state. Instead, the PM core will unwind its actions by resuming all the devices that were suspended.
Leaving System Suspend¶
When resuming from freeze, standby or memory sleep, the phases are:
->resume_noirqcallback methods should perform any actions needed before the driver’s interrupt handlers are invoked. This generally means undoing the actions of the
suspend_noirqphase. If the bus type permits devices to share interrupt vectors, like PCI, the method should bring the device and its driver into a state in which the driver can recognize if the device is the source of incoming interrupts, if any, and handle them correctly.
For example, the PCI bus type’s
->pm.resume_noirq()puts the device into the full-power state (D0 in the PCI terminology) and restores the standard configuration registers of the device. Then it calls the device driver’s
->pm.resume_noirq()method to perform device-specific actions.
->resume_earlymethods should prepare devices for the execution of the resume methods. This generally involves undoing the actions of the preceding
->resumemethods should bring the device back to its operating state, so that it can perform normal I/O. This generally involves undoing the actions of the
completephase should undo the actions of the
preparephase. For this reason, unlike the other resume-related phases, during the
completephase the device hierarchy is traversed bottom-up.
Note, however, that new children may be registered below the device as soon as the
->resumecallbacks occur; it’s not necessary to wait until the
completephase with that.
Moreover, if the preceding
->preparecallback returned a positive number, the device may have been left in runtime suspend throughout the whole system suspend and resume (the
suspend_noirqphases of system suspend and the
resumephases of system resume may have been skipped for it). In that case, the
->completecallback is entirely responsible for putting the device into a consistent state after system suspend if necessary. [For example, it may need to queue up a runtime resume request for the device for this purpose.] To check if that is the case, the
->completecallback can consult the device’s
power.direct_completeflag. Namely, if that flag is set when the
->completecallback is being run, it has been called directly after the preceding
->prepareand special actions may be required to make the device work correctly afterward.
At the end of these phases, drivers should be as functional as they were before suspending: I/O can be performed using DMA and IRQs, and the relevant clocks are gated on.
However, the details here may again be platform-specific. For example, some systems support multiple “run” states, and the mode in effect at the end of resume might not be the one which preceded suspension. That means availability of certain clocks or power supplies changed, which could easily affect how a driver works.
Drivers need to be able to handle hardware which has been reset since all of the suspend methods were called, for example by complete reinitialization. This may be the hardest part, and the one most protected by NDA’d documents and chip errata. It’s simplest if the hardware state hasn’t changed since the suspend was carried out, but that can only be guaranteed if the target system sleep entered was suspend-to-idle. For the other system sleep states that may not be the case (and usually isn’t for ACPI-defined system sleep states, like S3).
Drivers must also be prepared to notice that the device has been removed while the system was powered down, whenever that’s physically possible. PCMCIA, MMC, USB, Firewire, SCSI, and even IDE are common examples of busses where common Linux platforms will see such removal. Details of how drivers will notice and handle such removals are currently bus-specific, and often involve a separate thread.
These callbacks may return an error value, but the PM core will ignore such errors since there’s nothing it can do about them other than printing them in the system log.
Hibernating the system is more complicated than putting it into sleep states, because it involves creating and saving a system image. Therefore there are more phases for hibernation, with a different set of callbacks. These phases always run after tasks have been frozen and enough memory has been freed.
The general procedure for hibernation is to quiesce all devices (“freeze”),
create an image of the system memory while everything is stable, reactivate all
devices (“thaw”), write the image to permanent storage, and finally shut down
the system (“power off”). The phases used to accomplish this are:
preparephase is discussed in the “Entering System Suspend” section above.
->freezemethods should quiesce the device so that it doesn’t generate IRQs or DMA, and they may need to save the values of device registers. However the device does not have to be put in a low-power state, and to save time it’s best not to do so. Also, the device should not be prepared to generate wakeup events.
freeze_latephase is analogous to the
suspend_latephase described earlier, except that the device should not be put into a low-power state and should not be allowed to generate wakeup events.
freeze_noirqphase is analogous to the
suspend_noirqphase discussed earlier, except again that the device should not be put into a low-power state and should not be allowed to generate wakeup events.
At this point the system image is created. All devices should be inactive and the contents of memory should remain undisturbed while this happens, so that the image forms an atomic snapshot of the system state.
thaw_noirqphase is analogous to the
resume_noirqphase discussed earlier. The main difference is that its methods can assume the device is in the same state as at the end of the
thaw_earlyphase is analogous to the
resume_earlyphase described above. Its methods should undo the actions of the preceding
freeze_late, if necessary.
thawphase is analogous to the
resumephase discussed earlier. Its methods should bring the device back to an operating state, so that it can be used for saving the image if necessary.
completephase is discussed in the “Leaving System Suspend” section above.
At this point the system image is saved, and the devices then need to be prepared for the upcoming system shutdown. This is much like suspending them before putting the system into the suspend-to-idle, shallow or deep sleep state, and the phases are similar.
preparephase is discussed above.
poweroffphase is analogous to the
poweroff_latephase is analogous to the
poweroff_noirqphase is analogous to the
should do essentially the same things as the
->suspend_noirq callbacks, respectively. The only notable difference is
that they need not store the device register values, because the registers
should already have been stored during the
Resuming from hibernation is, again, more complicated than resuming from a sleep state in which the contents of main memory are preserved, because it requires a system image to be loaded into memory and the pre-hibernation memory contents to be restored before control can be passed back to the image kernel.
Although in principle the image might be loaded into memory and the pre-hibernation memory contents restored by the boot loader, in practice this can’t be done because boot loaders aren’t smart enough and there is no established protocol for passing the necessary information. So instead, the boot loader loads a fresh instance of the kernel, called “the restore kernel”, into memory and passes control to it in the usual way. Then the restore kernel reads the system image, restores the pre-hibernation memory contents, and passes control to the image kernel. Thus two different kernel instances are involved in resuming from hibernation. In fact, the restore kernel may be completely different from the image kernel: a different configuration and even a different version. This has important consequences for device drivers and their subsystems.
To be able to load the system image into memory, the restore kernel needs to
include at least a subset of device drivers allowing it to access the storage
medium containing the image, although it doesn’t need to include all of the
drivers present in the image kernel. After the image has been loaded, the
devices managed by the boot kernel need to be prepared for passing control back
to the image kernel. This is very similar to the initial steps involved in
creating a system image, and it is accomplished in the same way, using
freeze_noirq phases. However, the devices
affected by these phases are only those having drivers in the restore kernel;
other devices will still be in whatever state the boot loader left them.
Should the restoration of the pre-hibernation memory contents fail, the restore
kernel would go through the “thawing” procedure described above, using the
complete phases, and then
continue running normally. This happens only rarely. Most often the
pre-hibernation memory contents are restored successfully and control is passed
to the image kernel, which then becomes responsible for bringing the system back
to the working state.
To achieve this, the image kernel must restore the devices’ pre-hibernation
functionality. The operation is much like waking up from a sleep state (with
the memory contents preserved), although it involves different phases:
restore_noirqphase is analogous to the
restore_earlyphase is analogous to the
restorephase is analogous to the
completephase is discussed above.
The main difference from
resume[_early|_noirq] is that
restore[_early|_noirq] must assume the device has been accessed and
reconfigured by the boot loader or the restore kernel. Consequently, the state
of the device may be different from the state remembered from the
freeze_noirq phases. The device may even need to be
reset and completely re-initialized. In many cases this difference doesn’t
matter, so the
method pointers can be set to the same routines. Nevertheless, different
callback pointers are used in case there is a situation where it actually does
Power Management Notifiers¶
There are some operations that cannot be carried out by the power management callbacks discussed above, because the callbacks occur too late or too early. To handle these cases, subsystems and device drivers may register power management notifiers that are called before tasks are frozen and after they have been thawed. Generally speaking, the PM notifiers are suitable for performing actions that either require user space to be available, or at least won’t interfere with user space.
For details refer to Suspend/Hibernation Notifiers.
Device Low-Power (suspend) States¶
Device low-power states aren’t standard. One device might only handle “on” and “off”, while another might support a dozen different versions of “on” (how many engines are active?), plus a state that gets back to “on” faster than from a full “off”.
Some buses define rules about what different suspend states mean. PCI gives one example: after the suspend sequence completes, a non-legacy PCI device may not perform DMA or issue IRQs, and any wakeup events it issues would be issued through the PME# bus signal. Plus, there are several PCI-standard device states, some of which are optional.
In contrast, integrated system-on-chip processors often use IRQs as the
wakeup event sources (so drivers would call
might be able to treat DMA completion as a wakeup event (sometimes DMA can stay
active too, it’d only be the CPU and some peripherals that sleep).
Some details here may be platform-specific. Systems may have devices that can be fully active in certain sleep states, such as an LCD display that’s refreshed using DMA while most of the system is sleeping lightly ... and its frame buffer might even be updated by a DSP or other non-Linux CPU while the Linux control processor stays idle.
Moreover, the specific actions taken may depend on the target system state. One target system state might allow a given device to be very operational; another might require a hard shut down with re-initialization on resume. And two different target systems might use the same device in different ways; the aforementioned LCD might be active in one product’s “standby”, but a different product using the same SOC might work differently.
Device Power Management Domains¶
Sometimes devices share reference clocks or other power resources. In those cases it generally is not possible to put devices into low-power states individually. Instead, a set of devices sharing a power resource can be put into a low-power state together at the same time by turning off the shared power resource. Of course, they also need to be put into the full-power state together, by turning the shared power resource on. A set of devices with this property is often referred to as a power domain. A power domain may also be nested inside another power domain. The nested domain is referred to as the sub-domain of the parent domain.
Support for power domains is provided through the
pm_domain field of
struct device. This field is a pointer to an object of type
struct dev_pm_domain, defined in
include/linux/pm.h`, providing a set
of power management callbacks analogous to the subsystem-level and device driver
callbacks that are executed for the given device during all power transitions,
instead of the respective subsystem-level callbacks. Specifically, if a
pm_domain pointer is not NULL, the
from the object pointed to by it will be executed instead of its subsystem’s
(e.g. bus type’s)
->suspend() callback and analogously for all of the
remaining callbacks. In other words, power management domain callbacks, if
defined for the given device, always take precedence over the callbacks provided
by the device’s subsystem (e.g. bus type).
The support for device power management domains is only relevant to platforms needing to use the same device driver power management callbacks in many different power domain configurations and wanting to avoid incorporating the support for power domains into subsystem-level callbacks, for example by modifying the platform bus type. Other platforms need not implement it or take it into account in any way.
Devices may be defined as IRQ-safe which indicates to the PM core that their
runtime PM callbacks may be invoked with disabled interrupts (see
Documentation/power/runtime_pm.txt for more information). If an
IRQ-safe device belongs to a PM domain, the runtime PM of the domain will be
disallowed, unless the domain itself is defined as IRQ-safe. However, it
makes sense to define a PM domain as IRQ-safe only if all the devices in it
are IRQ-safe. Moreover, if an IRQ-safe domain has a parent domain, the runtime
PM of the parent is only allowed if the parent itself is IRQ-safe too with the
additional restriction that all child domains of an IRQ-safe parent must also
Runtime Power Management¶
Many devices are able to dynamically power down while the system is still running. This feature is useful for devices that are not being used, and can offer significant power savings on a running system. These devices often support a range of runtime power states, which might use names such as “off”, “sleep”, “idle”, “active”, and so on. Those states will in some cases (like PCI) be partially constrained by the bus the device uses, and will usually include hardware states that are also used in system sleep states.
A system-wide power transition can be started while some devices are in low power states due to runtime power management. The system sleep PM callbacks should recognize such situations and react to them appropriately, but the necessary actions are subsystem-specific.
In some cases the decision may be made at the subsystem level while in other cases the device driver may be left to decide. In some cases it may be desirable to leave a suspended device in that state during a system-wide power transition, but in other cases the device must be put back into the full-power state temporarily, for example so that its system wakeup capability can be disabled. This all depends on the hardware and the design of the subsystem and device driver in question.
During system-wide resume from a sleep state it’s easiest to put devices into
the full-power state, as explained in
Refer to that document for more information regarding this particular issue as
well as for information on the device runtime power management framework in