1 .. SPDX-License-Identifier: GPL-2.0 1 .. SPDX-License-Identifier: GPL-2.0
2 .. include:: <isonum.txt> 2 .. include:: <isonum.txt>
3 3
4 .. _driverapi_pm_devices: 4 .. _driverapi_pm_devices:
5 5
6 ============================== 6 ==============================
7 Device Power Management Basics 7 Device Power Management Basics
8 ============================== 8 ==============================
9 9
10 :Copyright: |copy| 2010-2011 Rafael J. Wysocki< 10 :Copyright: |copy| 2010-2011 Rafael J. Wysocki <rjw@sisk.pl>, Novell Inc.
11 :Copyright: |copy| 2010 Alan Stern <stern@rowla 11 :Copyright: |copy| 2010 Alan Stern <stern@rowland.harvard.edu>
12 :Copyright: |copy| 2016 Intel Corporation 12 :Copyright: |copy| 2016 Intel Corporation
13 13
14 :Author: Rafael J. Wysocki <rafael.j.wysocki@in 14 :Author: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
15 15
16 16
17 Most of the code in Linux is device drivers, s 17 Most of the code in Linux is device drivers, so most of the Linux power
18 management (PM) code is also driver-specific. 18 management (PM) code is also driver-specific. Most drivers will do very
19 little; others, especially for platforms with 19 little; others, especially for platforms with small batteries (like cell
20 phones), will do a lot. 20 phones), will do a lot.
21 21
22 This writeup gives an overview of how drivers 22 This writeup gives an overview of how drivers interact with system-wide
23 power management goals, emphasizing the models 23 power management goals, emphasizing the models and interfaces that are
24 shared by everything that hooks up to the driv 24 shared by everything that hooks up to the driver model core. Read it as
25 background for the domain-specific work you'd 25 background for the domain-specific work you'd do with any specific driver.
26 26
27 27
28 Two Models for Device Power Management 28 Two Models for Device Power Management
29 ====================================== 29 ======================================
30 30
31 Drivers will use one or both of these models t 31 Drivers will use one or both of these models to put devices into low-power
32 states: 32 states:
33 33
34 System Sleep model: 34 System Sleep model:
35 35
36 Drivers can enter low-power states as 36 Drivers can enter low-power states as part of entering system-wide
37 low-power states like "suspend" (also 37 low-power states like "suspend" (also known as "suspend-to-RAM"), or
38 (mostly for systems with disks) "hiber 38 (mostly for systems with disks) "hibernation" (also known as
39 "suspend-to-disk"). 39 "suspend-to-disk").
40 40
41 This is something that device, bus, an 41 This is something that device, bus, and class drivers collaborate on
42 by implementing various role-specific 42 by implementing various role-specific suspend and resume methods to
43 cleanly power down hardware and softwa 43 cleanly power down hardware and software subsystems, then reactivate
44 them without loss of data. 44 them without loss of data.
45 45
46 Some drivers can manage hardware wakeu 46 Some drivers can manage hardware wakeup events, which make the system
47 leave the low-power state. This featu 47 leave the low-power state. This feature may be enabled or disabled
48 using the relevant :file:`/sys/devices 48 using the relevant :file:`/sys/devices/.../power/wakeup` file (for
49 Ethernet drivers the ioctl interface u 49 Ethernet drivers the ioctl interface used by ethtool may also be used
50 for this purpose); enabling it may cos 50 for this purpose); enabling it may cost some power usage, but let the
51 whole system enter low-power states mo 51 whole system enter low-power states more often.
52 52
53 Runtime Power Management model: 53 Runtime Power Management model:
54 54
55 Devices may also be put into low-power 55 Devices may also be put into low-power states while the system is
56 running, independently of other power 56 running, independently of other power management activity in principle.
57 However, devices are not generally ind 57 However, devices are not generally independent of each other (for
58 example, a parent device cannot be sus 58 example, a parent device cannot be suspended unless all of its child
59 devices have been suspended). Moreove 59 devices have been suspended). Moreover, depending on the bus type the
60 device is on, it may be necessary to c 60 device is on, it may be necessary to carry out some bus-specific
61 operations on the device for this purp 61 operations on the device for this purpose. Devices put into low power
62 states at run time may require special 62 states at run time may require special handling during system-wide power
63 transitions (suspend or hibernation). 63 transitions (suspend or hibernation).
64 64
65 For these reasons not only the device 65 For these reasons not only the device driver itself, but also the
66 appropriate subsystem (bus type, devic 66 appropriate subsystem (bus type, device type or device class) driver and
67 the PM core are involved in runtime po 67 the PM core are involved in runtime power management. As in the system
68 sleep power management case, they need 68 sleep power management case, they need to collaborate by implementing
69 various role-specific suspend and resu 69 various role-specific suspend and resume methods, so that the hardware
70 is cleanly powered down and reactivate 70 is cleanly powered down and reactivated without data or service loss.
71 71
72 There's not a lot to be said about those low-p 72 There's not a lot to be said about those low-power states except that they are
73 very system-specific, and often device-specifi 73 very system-specific, and often device-specific. Also, that if enough devices
74 have been put into low-power states (at runtim 74 have been put into low-power states (at runtime), the effect may be very similar
75 to entering some system-wide low-power state ( 75 to entering some system-wide low-power state (system sleep) ... and that
76 synergies exist, so that several drivers using 76 synergies exist, so that several drivers using runtime PM might put the system
77 into a state where even deeper power saving op 77 into a state where even deeper power saving options are available.
78 78
79 Most suspended devices will have quiesced all 79 Most suspended devices will have quiesced all I/O: no more DMA or IRQs (except
80 for wakeup events), no more data read or writt 80 for wakeup events), no more data read or written, and requests from upstream
81 drivers are no longer accepted. A given bus o 81 drivers are no longer accepted. A given bus or platform may have different
82 requirements though. 82 requirements though.
83 83
84 Examples of hardware wakeup events include an 84 Examples of hardware wakeup events include an alarm from a real time clock,
85 network wake-on-LAN packets, keyboard or mouse 85 network wake-on-LAN packets, keyboard or mouse activity, and media insertion
86 or removal (for PCMCIA, MMC/SD, USB, and so on 86 or removal (for PCMCIA, MMC/SD, USB, and so on).
87 87
88 Interfaces for Entering System Sleep States 88 Interfaces for Entering System Sleep States
89 =========================================== 89 ===========================================
90 90
91 There are programming interfaces provided for 91 There are programming interfaces provided for subsystems (bus type, device type,
92 device class) and device drivers to allow them 92 device class) and device drivers to allow them to participate in the power
93 management of devices they are concerned with. 93 management of devices they are concerned with. These interfaces cover both
94 system sleep and runtime power management. 94 system sleep and runtime power management.
95 95
96 96
97 Device Power Management Operations 97 Device Power Management Operations
98 ---------------------------------- 98 ----------------------------------
99 99
100 Device power management operations, at the sub 100 Device power management operations, at the subsystem level as well as at the
101 device driver level, are implemented by defini 101 device driver level, are implemented by defining and populating objects of type
102 struct dev_pm_ops defined in :file:`include/li 102 struct dev_pm_ops defined in :file:`include/linux/pm.h`. The roles of the
103 methods included in it will be explained in wh 103 methods included in it will be explained in what follows. For now, it should be
104 sufficient to remember that the last three met 104 sufficient to remember that the last three methods are specific to runtime power
105 management while the remaining ones are used d 105 management while the remaining ones are used during system-wide power
106 transitions. 106 transitions.
107 107
108 There also is a deprecated "old" or "legacy" i 108 There also is a deprecated "old" or "legacy" interface for power management
109 operations available at least for some subsyst 109 operations available at least for some subsystems. This approach does not use
110 struct dev_pm_ops objects and it is suitable o 110 struct dev_pm_ops objects and it is suitable only for implementing system
111 sleep power management methods in a limited wa 111 sleep power management methods in a limited way. Therefore it is not described
112 in this document, so please refer directly to 112 in this document, so please refer directly to the source code for more
113 information about it. 113 information about it.
114 114
115 115
116 Subsystem-Level Methods 116 Subsystem-Level Methods
117 ----------------------- 117 -----------------------
118 118
119 The core methods to suspend and resume devices 119 The core methods to suspend and resume devices reside in
120 struct dev_pm_ops pointed to by the :c:member: 120 struct dev_pm_ops pointed to by the :c:member:`ops` member of
121 struct dev_pm_domain, or by the :c:member:`pm` 121 struct dev_pm_domain, or by the :c:member:`pm` member of struct bus_type,
122 struct device_type and struct class. They are 122 struct device_type and struct class. They are mostly of interest to the
123 people writing infrastructure for platforms an 123 people writing infrastructure for platforms and buses, like PCI or USB, or
124 device type and device class drivers. They al 124 device type and device class drivers. They also are relevant to the writers of
125 device drivers whose subsystems (PM domains, d 125 device drivers whose subsystems (PM domains, device types, device classes and
126 bus types) don't provide all power management 126 bus types) don't provide all power management methods.
127 127
128 Bus drivers implement these methods as appropr 128 Bus drivers implement these methods as appropriate for the hardware and the
129 drivers using it; PCI works differently from U 129 drivers using it; PCI works differently from USB, and so on. Not many people
130 write subsystem-level drivers; most driver cod 130 write subsystem-level drivers; most driver code is a "device driver" that builds
131 on top of bus-specific framework code. 131 on top of bus-specific framework code.
132 132
133 For more information on these driver calls, se 133 For more information on these driver calls, see the description later;
134 they are called in phases for every device, re 134 they are called in phases for every device, respecting the parent-child
135 sequencing in the driver model tree. 135 sequencing in the driver model tree.
136 136
137 137
138 :file:`/sys/devices/.../power/wakeup` files 138 :file:`/sys/devices/.../power/wakeup` files
139 ------------------------------------------- 139 -------------------------------------------
140 140
141 All device objects in the driver model contain 141 All device objects in the driver model contain fields that control the handling
142 of system wakeup events (hardware signals that 142 of system wakeup events (hardware signals that can force the system out of a
143 sleep state). These fields are initialized by 143 sleep state). These fields are initialized by bus or device driver code using
144 :c:func:`device_set_wakeup_capable()` and :c:f 144 :c:func:`device_set_wakeup_capable()` and :c:func:`device_set_wakeup_enable()`,
145 defined in :file:`include/linux/pm_wakeup.h`. 145 defined in :file:`include/linux/pm_wakeup.h`.
146 146
147 The :c:member:`power.can_wakeup` flag just rec 147 The :c:member:`power.can_wakeup` flag just records whether the device (and its
148 driver) can physically support wakeup events. 148 driver) can physically support wakeup events. The
149 :c:func:`device_set_wakeup_capable()` routine 149 :c:func:`device_set_wakeup_capable()` routine affects this flag. The
150 :c:member:`power.wakeup` field is a pointer to 150 :c:member:`power.wakeup` field is a pointer to an object of type
151 struct wakeup_source used for controlling whet 151 struct wakeup_source used for controlling whether or not the device should use
152 its system wakeup mechanism and for notifying 152 its system wakeup mechanism and for notifying the PM core of system wakeup
153 events signaled by the device. This object is 153 events signaled by the device. This object is only present for wakeup-capable
154 devices (i.e. devices whose :c:member:`can_wak 154 devices (i.e. devices whose :c:member:`can_wakeup` flags are set) and is created
155 (or removed) by :c:func:`device_set_wakeup_cap 155 (or removed) by :c:func:`device_set_wakeup_capable()`.
156 156
157 Whether or not a device is capable of issuing 157 Whether or not a device is capable of issuing wakeup events is a hardware
158 matter, and the kernel is responsible for keep 158 matter, and the kernel is responsible for keeping track of it. By contrast,
159 whether or not a wakeup-capable device should 159 whether or not a wakeup-capable device should issue wakeup events is a policy
160 decision, and it is managed by user space thro 160 decision, and it is managed by user space through a sysfs attribute: the
161 :file:`power/wakeup` file. User space can wri 161 :file:`power/wakeup` file. User space can write the "enabled" or "disabled"
162 strings to it to indicate whether or not, resp 162 strings to it to indicate whether or not, respectively, the device is supposed
163 to signal system wakeup. This file is only pr 163 to signal system wakeup. This file is only present if the
164 :c:member:`power.wakeup` object exists for the 164 :c:member:`power.wakeup` object exists for the given device and is created (or
165 removed) along with that object, by :c:func:`d 165 removed) along with that object, by :c:func:`device_set_wakeup_capable()`.
166 Reads from the file will return the correspond 166 Reads from the file will return the corresponding string.
167 167
168 The initial value in the :file:`power/wakeup` 168 The initial value in the :file:`power/wakeup` file is "disabled" for the
169 majority of devices; the major exceptions are 169 majority of devices; the major exceptions are power buttons, keyboards, and
170 Ethernet adapters whose WoL (wake-on-LAN) feat 170 Ethernet adapters whose WoL (wake-on-LAN) feature has been set up with ethtool.
171 It should also default to "enabled" for device 171 It should also default to "enabled" for devices that don't generate wakeup
172 requests on their own but merely forward wakeu 172 requests on their own but merely forward wakeup requests from one bus to another
173 (like PCI Express ports). 173 (like PCI Express ports).
174 174
175 The :c:func:`device_may_wakeup()` routine retu 175 The :c:func:`device_may_wakeup()` routine returns true only if the
176 :c:member:`power.wakeup` object exists and the 176 :c:member:`power.wakeup` object exists and the corresponding :file:`power/wakeup`
177 file contains the "enabled" string. This info 177 file contains the "enabled" string. This information is used by subsystems,
178 like the PCI bus type code, to see whether or 178 like the PCI bus type code, to see whether or not to enable the devices' wakeup
179 mechanisms. If device wakeup mechanisms are e 179 mechanisms. If device wakeup mechanisms are enabled or disabled directly by
180 drivers, they also should use :c:func:`device_ 180 drivers, they also should use :c:func:`device_may_wakeup()` to decide what to do
181 during a system sleep transition. Device driv 181 during a system sleep transition. Device drivers, however, are not expected to
182 call :c:func:`device_set_wakeup_enable()` dire 182 call :c:func:`device_set_wakeup_enable()` directly in any case.
183 183
184 It ought to be noted that system wakeup is con 184 It ought to be noted that system wakeup is conceptually different from "remote
185 wakeup" used by runtime power management, alth 185 wakeup" used by runtime power management, although it may be supported by the
186 same physical mechanism. Remote wakeup is a f 186 same physical mechanism. Remote wakeup is a feature allowing devices in
187 low-power states to trigger specific interrupt 187 low-power states to trigger specific interrupts to signal conditions in which
188 they should be put into the full-power state. 188 they should be put into the full-power state. Those interrupts may or may not
189 be used to signal system wakeup events, depend 189 be used to signal system wakeup events, depending on the hardware design. On
190 some systems it is impossible to trigger them 190 some systems it is impossible to trigger them from system sleep states. In any
191 case, remote wakeup should always be enabled f 191 case, remote wakeup should always be enabled for runtime power management for
192 all devices and drivers that support it. 192 all devices and drivers that support it.
193 193
194 194
195 :file:`/sys/devices/.../power/control` files 195 :file:`/sys/devices/.../power/control` files
196 -------------------------------------------- 196 --------------------------------------------
197 197
198 Each device in the driver model has a flag to 198 Each device in the driver model has a flag to control whether it is subject to
199 runtime power management. This flag, :c:membe 199 runtime power management. This flag, :c:member:`runtime_auto`, is initialized
200 by the bus type (or generally subsystem) code 200 by the bus type (or generally subsystem) code using :c:func:`pm_runtime_allow()`
201 or :c:func:`pm_runtime_forbid()`; the default 201 or :c:func:`pm_runtime_forbid()`; the default is to allow runtime power
202 management. 202 management.
203 203
204 The setting can be adjusted by user space by w 204 The setting can be adjusted by user space by writing either "on" or "auto" to
205 the device's :file:`power/control` sysfs file. 205 the device's :file:`power/control` sysfs file. Writing "auto" calls
206 :c:func:`pm_runtime_allow()`, setting the flag 206 :c:func:`pm_runtime_allow()`, setting the flag and allowing the device to be
207 runtime power-managed by its driver. Writing 207 runtime power-managed by its driver. Writing "on" calls
208 :c:func:`pm_runtime_forbid()`, clearing the fl 208 :c:func:`pm_runtime_forbid()`, clearing the flag, returning the device to full
209 power if it was in a low-power state, and prev 209 power if it was in a low-power state, and preventing the
210 device from being runtime power-managed. User 210 device from being runtime power-managed. User space can check the current value
211 of the :c:member:`runtime_auto` flag by readin 211 of the :c:member:`runtime_auto` flag by reading that file.
212 212
213 The device's :c:member:`runtime_auto` flag has 213 The device's :c:member:`runtime_auto` flag has no effect on the handling of
214 system-wide power transitions. In particular, 214 system-wide power transitions. In particular, the device can (and in the
215 majority of cases should and will) be put into 215 majority of cases should and will) be put into a low-power state during a
216 system-wide transition to a sleep state even t 216 system-wide transition to a sleep state even though its :c:member:`runtime_auto`
217 flag is clear. 217 flag is clear.
218 218
219 For more information about the runtime power m 219 For more information about the runtime power management framework, refer to
220 Documentation/power/runtime_pm.rst. 220 Documentation/power/runtime_pm.rst.
221 221
222 222
223 Calling Drivers to Enter and Leave System Slee 223 Calling Drivers to Enter and Leave System Sleep States
224 ============================================== 224 ======================================================
225 225
226 When the system goes into a sleep state, each 226 When the system goes into a sleep state, each device's driver is asked to
227 suspend the device by putting it into a state 227 suspend the device by putting it into a state compatible with the target
228 system state. That's usually some version of 228 system state. That's usually some version of "off", but the details are
229 system-specific. Also, wakeup-enabled devices 229 system-specific. Also, wakeup-enabled devices will usually stay partly
230 functional in order to wake the system. 230 functional in order to wake the system.
231 231
232 When the system leaves that low-power state, t 232 When the system leaves that low-power state, the device's driver is asked to
233 resume it by returning it to full power. The 233 resume it by returning it to full power. The suspend and resume operations
234 always go together, and both are multi-phase o 234 always go together, and both are multi-phase operations.
235 235
236 For simple drivers, suspend might quiesce the 236 For simple drivers, suspend might quiesce the device using class code
237 and then turn its hardware as "off" as possibl 237 and then turn its hardware as "off" as possible during suspend_noirq. The
238 matching resume calls would then completely re 238 matching resume calls would then completely reinitialize the hardware
239 before reactivating its class I/O queues. 239 before reactivating its class I/O queues.
240 240
241 More power-aware drivers might prepare the dev 241 More power-aware drivers might prepare the devices for triggering system wakeup
242 events. 242 events.
243 243
244 244
245 Call Sequence Guarantees 245 Call Sequence Guarantees
246 ------------------------ 246 ------------------------
247 247
248 To ensure that bridges and similar links needi 248 To ensure that bridges and similar links needing to talk to a device are
249 available when the device is suspended or resu 249 available when the device is suspended or resumed, the device hierarchy is
250 walked in a bottom-up order to suspend devices 250 walked in a bottom-up order to suspend devices. A top-down order is
251 used to resume those devices. 251 used to resume those devices.
252 252
253 The ordering of the device hierarchy is define 253 The ordering of the device hierarchy is defined by the order in which devices
254 get registered: a child can never be register 254 get registered: a child can never be registered, probed or resumed before
255 its parent; and can't be removed or suspended 255 its parent; and can't be removed or suspended after that parent.
256 256
257 The policy is that the device hierarchy should 257 The policy is that the device hierarchy should match hardware bus topology.
258 [Or at least the control bus, for devices whic 258 [Or at least the control bus, for devices which use multiple busses.]
259 In particular, this means that a device regist 259 In particular, this means that a device registration may fail if the parent of
260 the device is suspending (i.e. has been chosen 260 the device is suspending (i.e. has been chosen by the PM core as the next
261 device to suspend) or has already suspended, a 261 device to suspend) or has already suspended, as well as after all of the other
262 devices have been suspended. Device drivers m 262 devices have been suspended. Device drivers must be prepared to cope with such
263 situations. 263 situations.
264 264
265 265
266 System Power Management Phases 266 System Power Management Phases
267 ------------------------------ 267 ------------------------------
268 268
269 Suspending or resuming the system is done in s 269 Suspending or resuming the system is done in several phases. Different phases
270 are used for suspend-to-idle, shallow (standby 270 are used for suspend-to-idle, shallow (standby), and deep ("suspend-to-RAM")
271 sleep states and the hibernation state ("suspe 271 sleep states and the hibernation state ("suspend-to-disk"). Each phase involves
272 executing callbacks for every device before th 272 executing callbacks for every device before the next phase begins. Not all
273 buses or classes support all these callbacks a 273 buses or classes support all these callbacks and not all drivers use all the
274 callbacks. The various phases always run afte 274 callbacks. The various phases always run after tasks have been frozen and
275 before they are unfrozen. Furthermore, the `` 275 before they are unfrozen. Furthermore, the ``*_noirq`` phases run at a time
276 when IRQ handlers have been disabled (except f 276 when IRQ handlers have been disabled (except for those marked with the
277 IRQF_NO_SUSPEND flag). 277 IRQF_NO_SUSPEND flag).
278 278
279 All phases use PM domain, bus, type, class or 279 All phases use PM domain, bus, type, class or driver callbacks (that is, methods
280 defined in ``dev->pm_domain->ops``, ``dev->bus 280 defined in ``dev->pm_domain->ops``, ``dev->bus->pm``, ``dev->type->pm``,
281 ``dev->class->pm`` or ``dev->driver->pm``). T 281 ``dev->class->pm`` or ``dev->driver->pm``). These callbacks are regarded by the
282 PM core as mutually exclusive. Moreover, PM d 282 PM core as mutually exclusive. Moreover, PM domain callbacks always take
283 precedence over all of the other callbacks and 283 precedence over all of the other callbacks and, for example, type callbacks take
284 precedence over bus, class and driver callback 284 precedence over bus, class and driver callbacks. To be precise, the following
285 rules are used to determine which callback to 285 rules are used to determine which callback to execute in the given phase:
286 286
287 1. If ``dev->pm_domain`` is present, the 287 1. If ``dev->pm_domain`` is present, the PM core will choose the callback
288 provided by ``dev->pm_domain->ops`` fo 288 provided by ``dev->pm_domain->ops`` for execution.
289 289
290 2. Otherwise, if both ``dev->type`` and ` 290 2. Otherwise, if both ``dev->type`` and ``dev->type->pm`` are present, the
291 callback provided by ``dev->type->pm`` 291 callback provided by ``dev->type->pm`` will be chosen for execution.
292 292
293 3. Otherwise, if both ``dev->class`` and 293 3. Otherwise, if both ``dev->class`` and ``dev->class->pm`` are present,
294 the callback provided by ``dev->class- 294 the callback provided by ``dev->class->pm`` will be chosen for
295 execution. 295 execution.
296 296
297 4. Otherwise, if both ``dev->bus`` and `` 297 4. Otherwise, if both ``dev->bus`` and ``dev->bus->pm`` are present, the
298 callback provided by ``dev->bus->pm`` 298 callback provided by ``dev->bus->pm`` will be chosen for execution.
299 299
300 This allows PM domains and device types to ove 300 This allows PM domains and device types to override callbacks provided by bus
301 types or device classes if necessary. 301 types or device classes if necessary.
302 302
303 The PM domain, type, class and bus callbacks m 303 The PM domain, type, class and bus callbacks may in turn invoke device- or
304 driver-specific methods stored in ``dev->drive 304 driver-specific methods stored in ``dev->driver->pm``, but they don't have to do
305 that. 305 that.
306 306
307 If the subsystem callback chosen for execution 307 If the subsystem callback chosen for execution is not present, the PM core will
308 execute the corresponding method from the ``de 308 execute the corresponding method from the ``dev->driver->pm`` set instead if
309 there is one. 309 there is one.
310 310
311 311
312 Entering System Suspend 312 Entering System Suspend
313 ----------------------- 313 -----------------------
314 314
315 When the system goes into the freeze, standby 315 When the system goes into the freeze, standby or memory sleep state,
316 the phases are: ``prepare``, ``suspend``, ``su 316 the phases are: ``prepare``, ``suspend``, ``suspend_late``, ``suspend_noirq``.
317 317
318 1. The ``prepare`` phase is meant to prev 318 1. The ``prepare`` phase is meant to prevent races by preventing new
319 devices from being registered; the PM 319 devices from being registered; the PM core would never know that all the
320 children of a device had been suspende 320 children of a device had been suspended if new children could be
321 registered at will. [By contrast, fro 321 registered at will. [By contrast, from the PM core's perspective,
322 devices may be unregistered at any tim 322 devices may be unregistered at any time.] Unlike the other
323 suspend-related phases, during the ``p 323 suspend-related phases, during the ``prepare`` phase the device
324 hierarchy is traversed top-down. 324 hierarchy is traversed top-down.
325 325
326 After the ``->prepare`` callback metho 326 After the ``->prepare`` callback method returns, no new children may be
327 registered below the device. The meth 327 registered below the device. The method may also prepare the device or
328 driver in some way for the upcoming sy 328 driver in some way for the upcoming system power transition, but it
329 should not put the device into a low-p 329 should not put the device into a low-power state. Moreover, if the
330 device supports runtime power manageme 330 device supports runtime power management, the ``->prepare`` callback
331 method must not update its state in ca 331 method must not update its state in case it is necessary to resume it
332 from runtime suspend later on. 332 from runtime suspend later on.
333 333
334 For devices supporting runtime power m 334 For devices supporting runtime power management, the return value of the
335 prepare callback can be used to indica 335 prepare callback can be used to indicate to the PM core that it may
336 safely leave the device in runtime sus 336 safely leave the device in runtime suspend (if runtime-suspended
337 already), provided that all of the dev 337 already), provided that all of the device's descendants are also left in
338 runtime suspend. Namely, if the prepa 338 runtime suspend. Namely, if the prepare callback returns a positive
339 number and that happens for all of the 339 number and that happens for all of the descendants of the device too,
340 and all of them (including the device 340 and all of them (including the device itself) are runtime-suspended, the
341 PM core will skip the ``suspend``, ``s 341 PM core will skip the ``suspend``, ``suspend_late`` and
342 ``suspend_noirq`` phases as well as al 342 ``suspend_noirq`` phases as well as all of the corresponding phases of
343 the subsequent device resume for all o 343 the subsequent device resume for all of these devices. In that case,
344 the ``->complete`` callback will be th 344 the ``->complete`` callback will be the next one invoked after the
345 ``->prepare`` callback and is entirely 345 ``->prepare`` callback and is entirely responsible for putting the
346 device into a consistent state as appr 346 device into a consistent state as appropriate.
347 347
348 Note that this direct-complete procedu 348 Note that this direct-complete procedure applies even if the device is
349 disabled for runtime PM; only the runt 349 disabled for runtime PM; only the runtime-PM status matters. It follows
350 that if a device has system-sleep call 350 that if a device has system-sleep callbacks but does not support runtime
351 PM, then its prepare callback must nev 351 PM, then its prepare callback must never return a positive value. This
352 is because all such devices are initia 352 is because all such devices are initially set to runtime-suspended with
353 runtime PM disabled. 353 runtime PM disabled.
354 354
355 This feature also can be controlled by 355 This feature also can be controlled by device drivers by using the
356 ``DPM_FLAG_NO_DIRECT_COMPLETE`` and `` 356 ``DPM_FLAG_NO_DIRECT_COMPLETE`` and ``DPM_FLAG_SMART_PREPARE`` driver
357 power management flags. [Typically, t 357 power management flags. [Typically, they are set at the time the driver
358 is probed against the device in questi 358 is probed against the device in question by passing them to the
359 :c:func:`dev_pm_set_driver_flags` help 359 :c:func:`dev_pm_set_driver_flags` helper function.] If the first of
360 these flags is set, the PM core will n 360 these flags is set, the PM core will not apply the direct-complete
361 procedure described above to the given 361 procedure described above to the given device and, consequenty, to any
362 of its ancestors. The second flag, wh 362 of its ancestors. The second flag, when set, informs the middle layer
363 code (bus types, device types, PM doma 363 code (bus types, device types, PM domains, classes) that it should take
364 the return value of the ``->prepare`` 364 the return value of the ``->prepare`` callback provided by the driver
365 into account and it may only return a 365 into account and it may only return a positive value from its own
366 ``->prepare`` callback if the driver's 366 ``->prepare`` callback if the driver's one also has returned a positive
367 value. 367 value.
368 368
369 2. The ``->suspend`` methods should quies 369 2. The ``->suspend`` methods should quiesce the device to stop it from
370 performing I/O. They also may save th 370 performing I/O. They also may save the device registers and put it into
371 the appropriate low-power state, depen 371 the appropriate low-power state, depending on the bus type the device is
372 on, and they may enable wakeup events. 372 on, and they may enable wakeup events.
373 373
374 However, for devices supporting runtim 374 However, for devices supporting runtime power management, the
375 ``->suspend`` methods provided by subs 375 ``->suspend`` methods provided by subsystems (bus types and PM domains
376 in particular) must follow an addition 376 in particular) must follow an additional rule regarding what can be done
377 to the devices before their drivers' ` 377 to the devices before their drivers' ``->suspend`` methods are called.
378 Namely, they may resume the devices fr 378 Namely, they may resume the devices from runtime suspend by
379 calling :c:func:`pm_runtime_resume` fo 379 calling :c:func:`pm_runtime_resume` for them, if that is necessary, but
380 they must not update the state of the 380 they must not update the state of the devices in any other way at that
381 time (in case the drivers need to resu 381 time (in case the drivers need to resume the devices from runtime
382 suspend in their ``->suspend`` methods 382 suspend in their ``->suspend`` methods). In fact, the PM core prevents
383 subsystems or drivers from putting dev 383 subsystems or drivers from putting devices into runtime suspend at
384 these times by calling :c:func:`pm_run 384 these times by calling :c:func:`pm_runtime_get_noresume` before issuing
385 the ``->prepare`` callback (and callin 385 the ``->prepare`` callback (and calling :c:func:`pm_runtime_put` after
386 issuing the ``->complete`` callback). 386 issuing the ``->complete`` callback).
387 387
388 3. For a number of devices it is convenie 388 3. For a number of devices it is convenient to split suspend into the
389 "quiesce device" and "save device stat 389 "quiesce device" and "save device state" phases, in which cases
390 ``suspend_late`` is meant to do the la 390 ``suspend_late`` is meant to do the latter. It is always executed after
391 runtime power management has been disa 391 runtime power management has been disabled for the device in question.
392 392
393 4. The ``suspend_noirq`` phase occurs aft 393 4. The ``suspend_noirq`` phase occurs after IRQ handlers have been disabled,
394 which means that the driver's interrup 394 which means that the driver's interrupt handler will not be called while
395 the callback method is running. The ` 395 the callback method is running. The ``->suspend_noirq`` methods should
396 save the values of the device's regist 396 save the values of the device's registers that weren't saved previously
397 and finally put the device into the ap 397 and finally put the device into the appropriate low-power state.
398 398
399 The majority of subsystems and device 399 The majority of subsystems and device drivers need not implement this
400 callback. However, bus types allowing 400 callback. However, bus types allowing devices to share interrupt
401 vectors, like PCI, generally need it; 401 vectors, like PCI, generally need it; otherwise a driver might encounter
402 an error during the suspend phase by f 402 an error during the suspend phase by fielding a shared interrupt
403 generated by some other device after i 403 generated by some other device after its own device had been set to low
404 power. 404 power.
405 405
406 At the end of these phases, drivers should hav 406 At the end of these phases, drivers should have stopped all I/O transactions
407 (DMA, IRQs), saved enough state that they can 407 (DMA, IRQs), saved enough state that they can re-initialize or restore previous
408 state (as needed by the hardware), and placed 408 state (as needed by the hardware), and placed the device into a low-power state.
409 On many platforms they will gate off one or mo 409 On many platforms they will gate off one or more clock sources; sometimes they
410 will also switch off power supplies or reduce 410 will also switch off power supplies or reduce voltages. [Drivers supporting
411 runtime PM may already have performed some or 411 runtime PM may already have performed some or all of these steps.]
412 412
413 If :c:func:`device_may_wakeup()` returns ``tru 413 If :c:func:`device_may_wakeup()` returns ``true``, the device should be
414 prepared for generating hardware wakeup signal 414 prepared for generating hardware wakeup signals to trigger a system wakeup event
415 when the system is in the sleep state. For ex 415 when the system is in the sleep state. For example, :c:func:`enable_irq_wake()`
416 might identify GPIO signals hooked up to a swi 416 might identify GPIO signals hooked up to a switch or other external hardware,
417 and :c:func:`pci_enable_wake()` does something 417 and :c:func:`pci_enable_wake()` does something similar for the PCI PME signal.
418 418
419 If any of these callbacks returns an error, th 419 If any of these callbacks returns an error, the system won't enter the desired
420 low-power state. Instead, the PM core will un 420 low-power state. Instead, the PM core will unwind its actions by resuming all
421 the devices that were suspended. 421 the devices that were suspended.
422 422
423 423
424 Leaving System Suspend 424 Leaving System Suspend
425 ---------------------- 425 ----------------------
426 426
427 When resuming from freeze, standby or memory s 427 When resuming from freeze, standby or memory sleep, the phases are:
428 ``resume_noirq``, ``resume_early``, ``resume`` 428 ``resume_noirq``, ``resume_early``, ``resume``, ``complete``.
429 429
430 1. The ``->resume_noirq`` callback method 430 1. The ``->resume_noirq`` callback methods should perform any actions
431 needed before the driver's interrupt h 431 needed before the driver's interrupt handlers are invoked. This
432 generally means undoing the actions of 432 generally means undoing the actions of the ``suspend_noirq`` phase. If
433 the bus type permits devices to share 433 the bus type permits devices to share interrupt vectors, like PCI, the
434 method should bring the device and its 434 method should bring the device and its driver into a state in which the
435 driver can recognize if the device is 435 driver can recognize if the device is the source of incoming interrupts,
436 if any, and handle them correctly. 436 if any, and handle them correctly.
437 437
438 For example, the PCI bus type's ``->pm 438 For example, the PCI bus type's ``->pm.resume_noirq()`` puts the device
439 into the full-power state (D0 in the P 439 into the full-power state (D0 in the PCI terminology) and restores the
440 standard configuration registers of th 440 standard configuration registers of the device. Then it calls the
441 device driver's ``->pm.resume_noirq()` 441 device driver's ``->pm.resume_noirq()`` method to perform device-specific
442 actions. 442 actions.
443 443
444 2. The ``->resume_early`` methods should 444 2. The ``->resume_early`` methods should prepare devices for the execution
445 of the resume methods. This generally 445 of the resume methods. This generally involves undoing the actions of
446 the preceding ``suspend_late`` phase. 446 the preceding ``suspend_late`` phase.
447 447
448 3. The ``->resume`` methods should bring 448 3. The ``->resume`` methods should bring the device back to its operating
449 state, so that it can perform normal I 449 state, so that it can perform normal I/O. This generally involves
450 undoing the actions of the ``suspend`` 450 undoing the actions of the ``suspend`` phase.
451 451
452 4. The ``complete`` phase should undo the 452 4. The ``complete`` phase should undo the actions of the ``prepare`` phase.
453 For this reason, unlike the other resu 453 For this reason, unlike the other resume-related phases, during the
454 ``complete`` phase the device hierarch 454 ``complete`` phase the device hierarchy is traversed bottom-up.
455 455
456 Note, however, that new children may b 456 Note, however, that new children may be registered below the device as
457 soon as the ``->resume`` callbacks occ 457 soon as the ``->resume`` callbacks occur; it's not necessary to wait
458 until the ``complete`` phase runs. 458 until the ``complete`` phase runs.
459 459
460 Moreover, if the preceding ``->prepare 460 Moreover, if the preceding ``->prepare`` callback returned a positive
461 number, the device may have been left 461 number, the device may have been left in runtime suspend throughout the
462 whole system suspend and resume (its ` 462 whole system suspend and resume (its ``->suspend``, ``->suspend_late``,
463 ``->suspend_noirq``, ``->resume_noirq` 463 ``->suspend_noirq``, ``->resume_noirq``,
464 ``->resume_early``, and ``->resume`` c 464 ``->resume_early``, and ``->resume`` callbacks may have been
465 skipped). In that case, the ``->compl 465 skipped). In that case, the ``->complete`` callback is entirely
466 responsible for putting the device int 466 responsible for putting the device into a consistent state after system
467 suspend if necessary. [For example, i 467 suspend if necessary. [For example, it may need to queue up a runtime
468 resume request for the device for this 468 resume request for the device for this purpose.] To check if that is
469 the case, the ``->complete`` callback 469 the case, the ``->complete`` callback can consult the device's
470 ``power.direct_complete`` flag. If th 470 ``power.direct_complete`` flag. If that flag is set when the
471 ``->complete`` callback is being run t 471 ``->complete`` callback is being run then the direct-complete mechanism
472 was used, and special actions may be r 472 was used, and special actions may be required to make the device work
473 correctly afterward. 473 correctly afterward.
474 474
475 At the end of these phases, drivers should be 475 At the end of these phases, drivers should be as functional as they were before
476 suspending: I/O can be performed using DMA and 476 suspending: I/O can be performed using DMA and IRQs, and the relevant clocks are
477 gated on. 477 gated on.
478 478
479 However, the details here may again be platfor 479 However, the details here may again be platform-specific. For example,
480 some systems support multiple "run" states, an 480 some systems support multiple "run" states, and the mode in effect at
481 the end of resume might not be the one which p 481 the end of resume might not be the one which preceded suspension.
482 That means availability of certain clocks or p 482 That means availability of certain clocks or power supplies changed,
483 which could easily affect how a driver works. 483 which could easily affect how a driver works.
484 484
485 Drivers need to be able to handle hardware whi 485 Drivers need to be able to handle hardware which has been reset since all of the
486 suspend methods were called, for example by co 486 suspend methods were called, for example by complete reinitialization.
487 This may be the hardest part, and the one most 487 This may be the hardest part, and the one most protected by NDA'd documents
488 and chip errata. It's simplest if the hardwar 488 and chip errata. It's simplest if the hardware state hasn't changed since
489 the suspend was carried out, but that can only 489 the suspend was carried out, but that can only be guaranteed if the target
490 system sleep entered was suspend-to-idle. For 490 system sleep entered was suspend-to-idle. For the other system sleep states
491 that may not be the case (and usually isn't fo 491 that may not be the case (and usually isn't for ACPI-defined system sleep
492 states, like S3). 492 states, like S3).
493 493
494 Drivers must also be prepared to notice that t 494 Drivers must also be prepared to notice that the device has been removed
495 while the system was powered down, whenever th 495 while the system was powered down, whenever that's physically possible.
496 PCMCIA, MMC, USB, Firewire, SCSI, and even IDE 496 PCMCIA, MMC, USB, Firewire, SCSI, and even IDE are common examples of busses
497 where common Linux platforms will see such rem 497 where common Linux platforms will see such removal. Details of how drivers
498 will notice and handle such removals are curre 498 will notice and handle such removals are currently bus-specific, and often
499 involve a separate thread. 499 involve a separate thread.
500 500
501 These callbacks may return an error value, but 501 These callbacks may return an error value, but the PM core will ignore such
502 errors since there's nothing it can do about t 502 errors since there's nothing it can do about them other than printing them in
503 the system log. 503 the system log.
504 504
505 505
506 Entering Hibernation 506 Entering Hibernation
507 -------------------- 507 --------------------
508 508
509 Hibernating the system is more complicated tha 509 Hibernating the system is more complicated than putting it into sleep states,
510 because it involves creating and saving a syst 510 because it involves creating and saving a system image. Therefore there are
511 more phases for hibernation, with a different 511 more phases for hibernation, with a different set of callbacks. These phases
512 always run after tasks have been frozen and en 512 always run after tasks have been frozen and enough memory has been freed.
513 513
514 The general procedure for hibernation is to qu 514 The general procedure for hibernation is to quiesce all devices ("freeze"),
515 create an image of the system memory while eve 515 create an image of the system memory while everything is stable, reactivate all
516 devices ("thaw"), write the image to permanent 516 devices ("thaw"), write the image to permanent storage, and finally shut down
517 the system ("power off"). The phases used to 517 the system ("power off"). The phases used to accomplish this are: ``prepare``,
518 ``freeze``, ``freeze_late``, ``freeze_noirq``, 518 ``freeze``, ``freeze_late``, ``freeze_noirq``, ``thaw_noirq``, ``thaw_early``,
519 ``thaw``, ``complete``, ``prepare``, ``powerof 519 ``thaw``, ``complete``, ``prepare``, ``poweroff``, ``poweroff_late``,
520 ``poweroff_noirq``. 520 ``poweroff_noirq``.
521 521
522 1. The ``prepare`` phase is discussed in 522 1. The ``prepare`` phase is discussed in the "Entering System Suspend"
523 section above. 523 section above.
524 524
525 2. The ``->freeze`` methods should quiesc 525 2. The ``->freeze`` methods should quiesce the device so that it doesn't
526 generate IRQs or DMA, and they may nee 526 generate IRQs or DMA, and they may need to save the values of device
527 registers. However the device does no 527 registers. However the device does not have to be put in a low-power
528 state, and to save time it's best not 528 state, and to save time it's best not to do so. Also, the device should
529 not be prepared to generate wakeup eve 529 not be prepared to generate wakeup events.
530 530
531 3. The ``freeze_late`` phase is analogous 531 3. The ``freeze_late`` phase is analogous to the ``suspend_late`` phase
532 described earlier, except that the dev 532 described earlier, except that the device should not be put into a
533 low-power state and should not be allo 533 low-power state and should not be allowed to generate wakeup events.
534 534
535 4. The ``freeze_noirq`` phase is analogou 535 4. The ``freeze_noirq`` phase is analogous to the ``suspend_noirq`` phase
536 discussed earlier, except again that t 536 discussed earlier, except again that the device should not be put into
537 a low-power state and should not be al 537 a low-power state and should not be allowed to generate wakeup events.
538 538
539 At this point the system image is created. Al 539 At this point the system image is created. All devices should be inactive and
540 the contents of memory should remain undisturb 540 the contents of memory should remain undisturbed while this happens, so that the
541 image forms an atomic snapshot of the system s 541 image forms an atomic snapshot of the system state.
542 542
543 5. The ``thaw_noirq`` phase is analogous 543 5. The ``thaw_noirq`` phase is analogous to the ``resume_noirq`` phase
544 discussed earlier. The main differenc 544 discussed earlier. The main difference is that its methods can assume
545 the device is in the same state as at 545 the device is in the same state as at the end of the ``freeze_noirq``
546 phase. 546 phase.
547 547
548 6. The ``thaw_early`` phase is analogous 548 6. The ``thaw_early`` phase is analogous to the ``resume_early`` phase
549 described above. Its methods should u 549 described above. Its methods should undo the actions of the preceding
550 ``freeze_late``, if necessary. 550 ``freeze_late``, if necessary.
551 551
552 7. The ``thaw`` phase is analogous to the 552 7. The ``thaw`` phase is analogous to the ``resume`` phase discussed
553 earlier. Its methods should bring the 553 earlier. Its methods should bring the device back to an operating
554 state, so that it can be used for savi 554 state, so that it can be used for saving the image if necessary.
555 555
556 8. The ``complete`` phase is discussed in 556 8. The ``complete`` phase is discussed in the "Leaving System Suspend"
557 section above. 557 section above.
558 558
559 At this point the system image is saved, and t 559 At this point the system image is saved, and the devices then need to be
560 prepared for the upcoming system shutdown. Th 560 prepared for the upcoming system shutdown. This is much like suspending them
561 before putting the system into the suspend-to- 561 before putting the system into the suspend-to-idle, shallow or deep sleep state,
562 and the phases are similar. 562 and the phases are similar.
563 563
564 9. The ``prepare`` phase is discussed abo 564 9. The ``prepare`` phase is discussed above.
565 565
566 10. The ``poweroff`` phase is analogous to 566 10. The ``poweroff`` phase is analogous to the ``suspend`` phase.
567 567
568 11. The ``poweroff_late`` phase is analogo 568 11. The ``poweroff_late`` phase is analogous to the ``suspend_late`` phase.
569 569
570 12. The ``poweroff_noirq`` phase is analog 570 12. The ``poweroff_noirq`` phase is analogous to the ``suspend_noirq`` phase.
571 571
572 The ``->poweroff``, ``->poweroff_late`` and `` 572 The ``->poweroff``, ``->poweroff_late`` and ``->poweroff_noirq`` callbacks
573 should do essentially the same things as the ` 573 should do essentially the same things as the ``->suspend``, ``->suspend_late``
574 and ``->suspend_noirq`` callbacks, respectivel 574 and ``->suspend_noirq`` callbacks, respectively. A notable difference is
575 that they need not store the device register v 575 that they need not store the device register values, because the registers
576 should already have been stored during the ``f 576 should already have been stored during the ``freeze``, ``freeze_late`` or
577 ``freeze_noirq`` phases. Also, on many machin 577 ``freeze_noirq`` phases. Also, on many machines the firmware will power-down
578 the entire system, so it is not necessary for 578 the entire system, so it is not necessary for the callback to put the device in
579 a low-power state. 579 a low-power state.
580 580
581 581
582 Leaving Hibernation 582 Leaving Hibernation
583 ------------------- 583 -------------------
584 584
585 Resuming from hibernation is, again, more comp 585 Resuming from hibernation is, again, more complicated than resuming from a sleep
586 state in which the contents of main memory are 586 state in which the contents of main memory are preserved, because it requires
587 a system image to be loaded into memory and th 587 a system image to be loaded into memory and the pre-hibernation memory contents
588 to be restored before control can be passed ba 588 to be restored before control can be passed back to the image kernel.
589 589
590 Although in principle the image might be loade 590 Although in principle the image might be loaded into memory and the
591 pre-hibernation memory contents restored by th 591 pre-hibernation memory contents restored by the boot loader, in practice this
592 can't be done because boot loaders aren't smar 592 can't be done because boot loaders aren't smart enough and there is no
593 established protocol for passing the necessary 593 established protocol for passing the necessary information. So instead, the
594 boot loader loads a fresh instance of the kern 594 boot loader loads a fresh instance of the kernel, called "the restore kernel",
595 into memory and passes control to it in the us 595 into memory and passes control to it in the usual way. Then the restore kernel
596 reads the system image, restores the pre-hiber 596 reads the system image, restores the pre-hibernation memory contents, and passes
597 control to the image kernel. Thus two differe 597 control to the image kernel. Thus two different kernel instances are involved
598 in resuming from hibernation. In fact, the re 598 in resuming from hibernation. In fact, the restore kernel may be completely
599 different from the image kernel: a different c 599 different from the image kernel: a different configuration and even a different
600 version. This has important consequences for 600 version. This has important consequences for device drivers and their
601 subsystems. 601 subsystems.
602 602
603 To be able to load the system image into memor 603 To be able to load the system image into memory, the restore kernel needs to
604 include at least a subset of device drivers al 604 include at least a subset of device drivers allowing it to access the storage
605 medium containing the image, although it doesn 605 medium containing the image, although it doesn't need to include all of the
606 drivers present in the image kernel. After th 606 drivers present in the image kernel. After the image has been loaded, the
607 devices managed by the boot kernel need to be 607 devices managed by the boot kernel need to be prepared for passing control back
608 to the image kernel. This is very similar to 608 to the image kernel. This is very similar to the initial steps involved in
609 creating a system image, and it is accomplishe 609 creating a system image, and it is accomplished in the same way, using
610 ``prepare``, ``freeze``, and ``freeze_noirq`` 610 ``prepare``, ``freeze``, and ``freeze_noirq`` phases. However, the devices
611 affected by these phases are only those having 611 affected by these phases are only those having drivers in the restore kernel;
612 other devices will still be in whatever state 612 other devices will still be in whatever state the boot loader left them.
613 613
614 Should the restoration of the pre-hibernation 614 Should the restoration of the pre-hibernation memory contents fail, the restore
615 kernel would go through the "thawing" procedur 615 kernel would go through the "thawing" procedure described above, using the
616 ``thaw_noirq``, ``thaw_early``, ``thaw``, and 616 ``thaw_noirq``, ``thaw_early``, ``thaw``, and ``complete`` phases, and then
617 continue running normally. This happens only 617 continue running normally. This happens only rarely. Most often the
618 pre-hibernation memory contents are restored s 618 pre-hibernation memory contents are restored successfully and control is passed
619 to the image kernel, which then becomes respon 619 to the image kernel, which then becomes responsible for bringing the system back
620 to the working state. 620 to the working state.
621 621
622 To achieve this, the image kernel must restore 622 To achieve this, the image kernel must restore the devices' pre-hibernation
623 functionality. The operation is much like wak 623 functionality. The operation is much like waking up from a sleep state (with
624 the memory contents preserved), although it in 624 the memory contents preserved), although it involves different phases:
625 ``restore_noirq``, ``restore_early``, ``restor 625 ``restore_noirq``, ``restore_early``, ``restore``, ``complete``.
626 626
627 1. The ``restore_noirq`` phase is analogo 627 1. The ``restore_noirq`` phase is analogous to the ``resume_noirq`` phase.
628 628
629 2. The ``restore_early`` phase is analogo 629 2. The ``restore_early`` phase is analogous to the ``resume_early`` phase.
630 630
631 3. The ``restore`` phase is analogous to 631 3. The ``restore`` phase is analogous to the ``resume`` phase.
632 632
633 4. The ``complete`` phase is discussed ab 633 4. The ``complete`` phase is discussed above.
634 634
635 The main difference from ``resume[_early|_noir 635 The main difference from ``resume[_early|_noirq]`` is that
636 ``restore[_early|_noirq]`` must assume the dev 636 ``restore[_early|_noirq]`` must assume the device has been accessed and
637 reconfigured by the boot loader or the restore 637 reconfigured by the boot loader or the restore kernel. Consequently, the state
638 of the device may be different from the state 638 of the device may be different from the state remembered from the ``freeze``,
639 ``freeze_late`` and ``freeze_noirq`` phases. 639 ``freeze_late`` and ``freeze_noirq`` phases. The device may even need to be
640 reset and completely re-initialized. In many 640 reset and completely re-initialized. In many cases this difference doesn't
641 matter, so the ``->resume[_early|_noirq]`` and 641 matter, so the ``->resume[_early|_noirq]`` and ``->restore[_early|_norq]``
642 method pointers can be set to the same routine 642 method pointers can be set to the same routines. Nevertheless, different
643 callback pointers are used in case there is a 643 callback pointers are used in case there is a situation where it actually does
644 matter. 644 matter.
645 645
646 646
647 Power Management Notifiers 647 Power Management Notifiers
648 ========================== 648 ==========================
649 649
650 There are some operations that cannot be carri 650 There are some operations that cannot be carried out by the power management
651 callbacks discussed above, because the callbac 651 callbacks discussed above, because the callbacks occur too late or too early.
652 To handle these cases, subsystems and device d 652 To handle these cases, subsystems and device drivers may register power
653 management notifiers that are called before ta 653 management notifiers that are called before tasks are frozen and after they have
654 been thawed. Generally speaking, the PM notif 654 been thawed. Generally speaking, the PM notifiers are suitable for performing
655 actions that either require user space to be a 655 actions that either require user space to be available, or at least won't
656 interfere with user space. 656 interfere with user space.
657 657
658 For details refer to Documentation/driver-api/ 658 For details refer to Documentation/driver-api/pm/notifiers.rst.
659 659
660 660
661 Device Low-Power (suspend) States 661 Device Low-Power (suspend) States
662 ================================= 662 =================================
663 663
664 Device low-power states aren't standard. One 664 Device low-power states aren't standard. One device might only handle
665 "on" and "off", while another might support a 665 "on" and "off", while another might support a dozen different versions of
666 "on" (how many engines are active?), plus a st 666 "on" (how many engines are active?), plus a state that gets back to "on"
667 faster than from a full "off". 667 faster than from a full "off".
668 668
669 Some buses define rules about what different s 669 Some buses define rules about what different suspend states mean. PCI
670 gives one example: after the suspend sequence 670 gives one example: after the suspend sequence completes, a non-legacy
671 PCI device may not perform DMA or issue IRQs, 671 PCI device may not perform DMA or issue IRQs, and any wakeup events it
672 issues would be issued through the PME# bus si 672 issues would be issued through the PME# bus signal. Plus, there are
673 several PCI-standard device states, some of wh 673 several PCI-standard device states, some of which are optional.
674 674
675 In contrast, integrated system-on-chip process 675 In contrast, integrated system-on-chip processors often use IRQs as the
676 wakeup event sources (so drivers would call :c 676 wakeup event sources (so drivers would call :c:func:`enable_irq_wake`) and
677 might be able to treat DMA completion as a wak 677 might be able to treat DMA completion as a wakeup event (sometimes DMA can stay
678 active too, it'd only be the CPU and some peri 678 active too, it'd only be the CPU and some peripherals that sleep).
679 679
680 Some details here may be platform-specific. S 680 Some details here may be platform-specific. Systems may have devices that
681 can be fully active in certain sleep states, s 681 can be fully active in certain sleep states, such as an LCD display that's
682 refreshed using DMA while most of the system i 682 refreshed using DMA while most of the system is sleeping lightly ... and
683 its frame buffer might even be updated by a DS 683 its frame buffer might even be updated by a DSP or other non-Linux CPU while
684 the Linux control processor stays idle. 684 the Linux control processor stays idle.
685 685
686 Moreover, the specific actions taken may depen 686 Moreover, the specific actions taken may depend on the target system state.
687 One target system state might allow a given de 687 One target system state might allow a given device to be very operational;
688 another might require a hard shut down with re 688 another might require a hard shut down with re-initialization on resume.
689 And two different target systems might use the 689 And two different target systems might use the same device in different
690 ways; the aforementioned LCD might be active i 690 ways; the aforementioned LCD might be active in one product's "standby",
691 but a different product using the same SOC mig 691 but a different product using the same SOC might work differently.
692 692
693 693
694 Device Power Management Domains 694 Device Power Management Domains
695 =============================== 695 ===============================
696 696
697 Sometimes devices share reference clocks or ot 697 Sometimes devices share reference clocks or other power resources. In those
698 cases it generally is not possible to put devi 698 cases it generally is not possible to put devices into low-power states
699 individually. Instead, a set of devices shari 699 individually. Instead, a set of devices sharing a power resource can be put
700 into a low-power state together at the same ti 700 into a low-power state together at the same time by turning off the shared
701 power resource. Of course, they also need to 701 power resource. Of course, they also need to be put into the full-power state
702 together, by turning the shared power resource 702 together, by turning the shared power resource on. A set of devices with this
703 property is often referred to as a power domai 703 property is often referred to as a power domain. A power domain may also be
704 nested inside another power domain. The nested 704 nested inside another power domain. The nested domain is referred to as the
705 sub-domain of the parent domain. 705 sub-domain of the parent domain.
706 706
707 Support for power domains is provided through 707 Support for power domains is provided through the :c:member:`pm_domain` field of
708 struct device. This field is a pointer to an 708 struct device. This field is a pointer to an object of type
709 struct dev_pm_domain, defined in :file:`includ 709 struct dev_pm_domain, defined in :file:`include/linux/pm.h`, providing a set
710 of power management callbacks analogous to the 710 of power management callbacks analogous to the subsystem-level and device driver
711 callbacks that are executed for the given devi 711 callbacks that are executed for the given device during all power transitions,
712 instead of the respective subsystem-level call 712 instead of the respective subsystem-level callbacks. Specifically, if a
713 device's :c:member:`pm_domain` pointer is not 713 device's :c:member:`pm_domain` pointer is not NULL, the ``->suspend()`` callback
714 from the object pointed to by it will be execu 714 from the object pointed to by it will be executed instead of its subsystem's
715 (e.g. bus type's) ``->suspend()`` callback and 715 (e.g. bus type's) ``->suspend()`` callback and analogously for all of the
716 remaining callbacks. In other words, power ma 716 remaining callbacks. In other words, power management domain callbacks, if
717 defined for the given device, always take prec 717 defined for the given device, always take precedence over the callbacks provided
718 by the device's subsystem (e.g. bus type). 718 by the device's subsystem (e.g. bus type).
719 719
720 The support for device power management domain 720 The support for device power management domains is only relevant to platforms
721 needing to use the same device driver power ma 721 needing to use the same device driver power management callbacks in many
722 different power domain configurations and want 722 different power domain configurations and wanting to avoid incorporating the
723 support for power domains into subsystem-level 723 support for power domains into subsystem-level callbacks, for example by
724 modifying the platform bus type. Other platfo 724 modifying the platform bus type. Other platforms need not implement it or take
725 it into account in any way. 725 it into account in any way.
726 726
727 Devices may be defined as IRQ-safe which indic 727 Devices may be defined as IRQ-safe which indicates to the PM core that their
728 runtime PM callbacks may be invoked with disab 728 runtime PM callbacks may be invoked with disabled interrupts (see
729 Documentation/power/runtime_pm.rst for more in 729 Documentation/power/runtime_pm.rst for more information). If an
730 IRQ-safe device belongs to a PM domain, the ru 730 IRQ-safe device belongs to a PM domain, the runtime PM of the domain will be
731 disallowed, unless the domain itself is define 731 disallowed, unless the domain itself is defined as IRQ-safe. However, it
732 makes sense to define a PM domain as IRQ-safe 732 makes sense to define a PM domain as IRQ-safe only if all the devices in it
733 are IRQ-safe. Moreover, if an IRQ-safe domain 733 are IRQ-safe. Moreover, if an IRQ-safe domain has a parent domain, the runtime
734 PM of the parent is only allowed if the parent 734 PM of the parent is only allowed if the parent itself is IRQ-safe too with the
735 additional restriction that all child domains 735 additional restriction that all child domains of an IRQ-safe parent must also
736 be IRQ-safe. 736 be IRQ-safe.
737 737
738 738
739 Runtime Power Management 739 Runtime Power Management
740 ======================== 740 ========================
741 741
742 Many devices are able to dynamically power dow 742 Many devices are able to dynamically power down while the system is still
743 running. This feature is useful for devices th 743 running. This feature is useful for devices that are not being used, and
744 can offer significant power savings on a runni 744 can offer significant power savings on a running system. These devices
745 often support a range of runtime power states, 745 often support a range of runtime power states, which might use names such
746 as "off", "sleep", "idle", "active", and so on 746 as "off", "sleep", "idle", "active", and so on. Those states will in some
747 cases (like PCI) be partially constrained by t 747 cases (like PCI) be partially constrained by the bus the device uses, and will
748 usually include hardware states that are also 748 usually include hardware states that are also used in system sleep states.
749 749
750 A system-wide power transition can be started 750 A system-wide power transition can be started while some devices are in low
751 power states due to runtime power management. 751 power states due to runtime power management. The system sleep PM callbacks
752 should recognize such situations and react to 752 should recognize such situations and react to them appropriately, but the
753 necessary actions are subsystem-specific. 753 necessary actions are subsystem-specific.
754 754
755 In some cases the decision may be made at the 755 In some cases the decision may be made at the subsystem level while in other
756 cases the device driver may be left to decide. 756 cases the device driver may be left to decide. In some cases it may be
757 desirable to leave a suspended device in that 757 desirable to leave a suspended device in that state during a system-wide power
758 transition, but in other cases the device must 758 transition, but in other cases the device must be put back into the full-power
759 state temporarily, for example so that its sys 759 state temporarily, for example so that its system wakeup capability can be
760 disabled. This all depends on the hardware an 760 disabled. This all depends on the hardware and the design of the subsystem and
761 device driver in question. 761 device driver in question.
762 762
763 If it is necessary to resume a device from run 763 If it is necessary to resume a device from runtime suspend during a system-wide
764 transition into a sleep state, that can be don 764 transition into a sleep state, that can be done by calling
765 :c:func:`pm_runtime_resume` from the ``->suspe 765 :c:func:`pm_runtime_resume` from the ``->suspend`` callback (or the ``->freeze``
766 or ``->poweroff`` callback for transitions rel 766 or ``->poweroff`` callback for transitions related to hibernation) of either the
767 device's driver or its subsystem (for example, 767 device's driver or its subsystem (for example, a bus type or a PM domain).
768 However, subsystems must not otherwise change 768 However, subsystems must not otherwise change the runtime status of devices
769 from their ``->prepare`` and ``->suspend`` cal 769 from their ``->prepare`` and ``->suspend`` callbacks (or equivalent) *before*
770 invoking device drivers' ``->suspend`` callbac 770 invoking device drivers' ``->suspend`` callbacks (or equivalent).
771 771
772 .. _smart_suspend_flag: 772 .. _smart_suspend_flag:
773 773
774 The ``DPM_FLAG_SMART_SUSPEND`` Driver Flag 774 The ``DPM_FLAG_SMART_SUSPEND`` Driver Flag
775 ------------------------------------------ 775 ------------------------------------------
776 776
777 Some bus types and PM domains have a policy to 777 Some bus types and PM domains have a policy to resume all devices from runtime
778 suspend upfront in their ``->suspend`` callbac 778 suspend upfront in their ``->suspend`` callbacks, but that may not be really
779 necessary if the device's driver can cope with 779 necessary if the device's driver can cope with runtime-suspended devices.
780 The driver can indicate this by setting ``DPM_ 780 The driver can indicate this by setting ``DPM_FLAG_SMART_SUSPEND`` in
781 :c:member:`power.driver_flags` at probe time, 781 :c:member:`power.driver_flags` at probe time, with the assistance of the
782 :c:func:`dev_pm_set_driver_flags` helper routi 782 :c:func:`dev_pm_set_driver_flags` helper routine.
783 783
784 Setting that flag causes the PM core and middl 784 Setting that flag causes the PM core and middle-layer code
785 (bus types, PM domains etc.) to skip the ``->s 785 (bus types, PM domains etc.) to skip the ``->suspend_late`` and
786 ``->suspend_noirq`` callbacks provided by the 786 ``->suspend_noirq`` callbacks provided by the driver if the device remains in
787 runtime suspend throughout those phases of the 787 runtime suspend throughout those phases of the system-wide suspend (and
788 similarly for the "freeze" and "poweroff" part 788 similarly for the "freeze" and "poweroff" parts of system hibernation).
789 [Otherwise the same driver 789 [Otherwise the same driver
790 callback might be executed twice in a row for 790 callback might be executed twice in a row for the same device, which would not
791 be valid in general.] If the middle-layer sys 791 be valid in general.] If the middle-layer system-wide PM callbacks are present
792 for the device then they are responsible for s 792 for the device then they are responsible for skipping these driver callbacks;
793 if not then the PM core skips them. The subsy 793 if not then the PM core skips them. The subsystem callback routines can
794 determine whether they need to skip the driver 794 determine whether they need to skip the driver callbacks by testing the return
795 value from the :c:func:`dev_pm_skip_suspend` h 795 value from the :c:func:`dev_pm_skip_suspend` helper function.
796 796
797 In addition, with ``DPM_FLAG_SMART_SUSPEND`` s 797 In addition, with ``DPM_FLAG_SMART_SUSPEND`` set, the driver's ``->thaw_noirq``
798 and ``->thaw_early`` callbacks are skipped in 798 and ``->thaw_early`` callbacks are skipped in hibernation if the device remained
799 in runtime suspend throughout the preceding "f 799 in runtime suspend throughout the preceding "freeze" transition. Again, if the
800 middle-layer callbacks are present for the dev 800 middle-layer callbacks are present for the device, they are responsible for
801 doing this, otherwise the PM core takes care o 801 doing this, otherwise the PM core takes care of it.
802 802
803 803
804 The ``DPM_FLAG_MAY_SKIP_RESUME`` Driver Flag 804 The ``DPM_FLAG_MAY_SKIP_RESUME`` Driver Flag
805 -------------------------------------------- 805 --------------------------------------------
806 806
807 During system-wide resume from a sleep state i 807 During system-wide resume from a sleep state it's easiest to put devices into
808 the full-power state, as explained in Document 808 the full-power state, as explained in Documentation/power/runtime_pm.rst.
809 [Refer to that document for more information r 809 [Refer to that document for more information regarding this particular issue as
810 well as for information on the device runtime 810 well as for information on the device runtime power management framework in
811 general.] However, it often is desirable to l 811 general.] However, it often is desirable to leave devices in suspend after
812 system transitions to the working state, espec 812 system transitions to the working state, especially if those devices had been in
813 runtime suspend before the preceding system-wi 813 runtime suspend before the preceding system-wide suspend (or analogous)
814 transition. 814 transition.
815 815
816 To that end, device drivers can use the ``DPM_ 816 To that end, device drivers can use the ``DPM_FLAG_MAY_SKIP_RESUME`` flag to
817 indicate to the PM core and middle-layer code 817 indicate to the PM core and middle-layer code that they allow their "noirq" and
818 "early" resume callbacks to be skipped if the 818 "early" resume callbacks to be skipped if the device can be left in suspend
819 after system-wide PM transitions to the workin 819 after system-wide PM transitions to the working state. Whether or not that is
820 the case generally depends on the state of the 820 the case generally depends on the state of the device before the given system
821 suspend-resume cycle and on the type of the sy 821 suspend-resume cycle and on the type of the system transition under way.
822 In particular, the "thaw" and "restore" transi 822 In particular, the "thaw" and "restore" transitions related to hibernation are
823 not affected by ``DPM_FLAG_MAY_SKIP_RESUME`` a 823 not affected by ``DPM_FLAG_MAY_SKIP_RESUME`` at all. [All callbacks are
824 issued during the "restore" transition regardl 824 issued during the "restore" transition regardless of the flag settings,
825 and whether or not any driver callbacks 825 and whether or not any driver callbacks
826 are skipped during the "thaw" transition depen 826 are skipped during the "thaw" transition depends whether or not the
827 ``DPM_FLAG_SMART_SUSPEND`` flag is set (see `a 827 ``DPM_FLAG_SMART_SUSPEND`` flag is set (see `above <smart_suspend_flag_>`_).
828 In addition, a device is not allowed to remain 828 In addition, a device is not allowed to remain in runtime suspend if any of its
829 children will be returned to full power.] 829 children will be returned to full power.]
830 830
831 The ``DPM_FLAG_MAY_SKIP_RESUME`` flag is taken 831 The ``DPM_FLAG_MAY_SKIP_RESUME`` flag is taken into account in combination with
832 the :c:member:`power.may_skip_resume` status b 832 the :c:member:`power.may_skip_resume` status bit set by the PM core during the
833 "suspend" phase of suspend-type transitions. 833 "suspend" phase of suspend-type transitions. If the driver or the middle layer
834 has a reason to prevent the driver's "noirq" a 834 has a reason to prevent the driver's "noirq" and "early" resume callbacks from
835 being skipped during the subsequent system res 835 being skipped during the subsequent system resume transition, it should
836 clear :c:member:`power.may_skip_resume` in its 836 clear :c:member:`power.may_skip_resume` in its ``->suspend``, ``->suspend_late``
837 or ``->suspend_noirq`` callback. [Note that t 837 or ``->suspend_noirq`` callback. [Note that the drivers setting
838 ``DPM_FLAG_SMART_SUSPEND`` need to clear :c:me 838 ``DPM_FLAG_SMART_SUSPEND`` need to clear :c:member:`power.may_skip_resume` in
839 their ``->suspend`` callback in case the other 839 their ``->suspend`` callback in case the other two are skipped.]
840 840
841 Setting the :c:member:`power.may_skip_resume` 841 Setting the :c:member:`power.may_skip_resume` status bit along with the
842 ``DPM_FLAG_MAY_SKIP_RESUME`` flag is necessary 842 ``DPM_FLAG_MAY_SKIP_RESUME`` flag is necessary, but generally not sufficient,
843 for the driver's "noirq" and "early" resume ca 843 for the driver's "noirq" and "early" resume callbacks to be skipped. Whether or
844 not they should be skipped can be determined b 844 not they should be skipped can be determined by evaluating the
845 :c:func:`dev_pm_skip_resume` helper function. 845 :c:func:`dev_pm_skip_resume` helper function.
846 846
847 If that function returns ``true``, the driver' 847 If that function returns ``true``, the driver's "noirq" and "early" resume
848 callbacks should be skipped and the device's r 848 callbacks should be skipped and the device's runtime PM status will be set to
849 "suspended" by the PM core. Otherwise, if the 849 "suspended" by the PM core. Otherwise, if the device was runtime-suspended
850 during the preceding system-wide suspend trans 850 during the preceding system-wide suspend transition and its
851 ``DPM_FLAG_SMART_SUSPEND`` is set, its runtime 851 ``DPM_FLAG_SMART_SUSPEND`` is set, its runtime PM status will be set to
852 "active" by the PM core. [Hence, the drivers 852 "active" by the PM core. [Hence, the drivers that do not set
853 ``DPM_FLAG_SMART_SUSPEND`` should not expect t 853 ``DPM_FLAG_SMART_SUSPEND`` should not expect the runtime PM status of their
854 devices to be changed from "suspended" to "act 854 devices to be changed from "suspended" to "active" by the PM core during
855 system-wide resume-type transitions.] 855 system-wide resume-type transitions.]
856 856
857 If the ``DPM_FLAG_MAY_SKIP_RESUME`` flag is no 857 If the ``DPM_FLAG_MAY_SKIP_RESUME`` flag is not set for a device, but
858 ``DPM_FLAG_SMART_SUSPEND`` is set and the driv 858 ``DPM_FLAG_SMART_SUSPEND`` is set and the driver's "late" and "noirq" suspend
859 callbacks are skipped, its system-wide "noirq" 859 callbacks are skipped, its system-wide "noirq" and "early" resume callbacks, if
860 present, are invoked as usual and the device's 860 present, are invoked as usual and the device's runtime PM status is set to
861 "active" by the PM core before enabling runtim 861 "active" by the PM core before enabling runtime PM for it. In that case, the
862 driver must be prepared to cope with the invoc 862 driver must be prepared to cope with the invocation of its system-wide resume
863 callbacks back-to-back with its ``->runtime_su 863 callbacks back-to-back with its ``->runtime_suspend`` one (without the
864 intervening ``->runtime_resume`` and system-wi 864 intervening ``->runtime_resume`` and system-wide suspend callbacks) and the
865 final state of the device must reflect the "ac 865 final state of the device must reflect the "active" runtime PM status in that
866 case. [Note that this is not a problem at all 866 case. [Note that this is not a problem at all if the driver's
867 ``->suspend_late`` callback pointer points to 867 ``->suspend_late`` callback pointer points to the same function as its
868 ``->runtime_suspend`` one and its ``->resume_e 868 ``->runtime_suspend`` one and its ``->resume_early`` callback pointer points to
869 the same function as the ``->runtime_resume`` 869 the same function as the ``->runtime_resume`` one, while none of the other
870 system-wide suspend-resume callbacks of the dr 870 system-wide suspend-resume callbacks of the driver are present, for example.]
871 871
872 Likewise, if ``DPM_FLAG_MAY_SKIP_RESUME`` is s 872 Likewise, if ``DPM_FLAG_MAY_SKIP_RESUME`` is set for a device, its driver's
873 system-wide "noirq" and "early" resume callbac 873 system-wide "noirq" and "early" resume callbacks may be skipped while its "late"
874 and "noirq" suspend callbacks may have been ex 874 and "noirq" suspend callbacks may have been executed (in principle, regardless
875 of whether or not ``DPM_FLAG_SMART_SUSPEND`` i 875 of whether or not ``DPM_FLAG_SMART_SUSPEND`` is set). In that case, the driver
876 needs to be able to cope with the invocation o 876 needs to be able to cope with the invocation of its ``->runtime_resume``
877 callback back-to-back with its "late" and "noi 877 callback back-to-back with its "late" and "noirq" suspend ones. [For instance,
878 that is not a concern if the driver sets both 878 that is not a concern if the driver sets both ``DPM_FLAG_SMART_SUSPEND`` and
879 ``DPM_FLAG_MAY_SKIP_RESUME`` and uses the same 879 ``DPM_FLAG_MAY_SKIP_RESUME`` and uses the same pair of suspend/resume callback
880 functions for runtime PM and system-wide suspe 880 functions for runtime PM and system-wide suspend/resume.]
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.