blob: b29d8e56cf287b7ab561328c414e6df4c86237df [file] [log] [blame]
Alan Sterncd38c1e2007-10-10 16:24:56 -04001 Power Management for USB
2
3 Alan Stern <stern@rowland.harvard.edu>
4
Alan Stern9bbdf1e2010-01-08 12:57:28 -05005 December 11, 2009
Alan Sterncd38c1e2007-10-10 16:24:56 -04006
7
8
9 What is Power Management?
10 -------------------------
11
12Power Management (PM) is the practice of saving energy by suspending
13parts of a computer system when they aren't being used. While a
14component is "suspended" it is in a nonfunctional low-power state; it
15might even be turned off completely. A suspended component can be
16"resumed" (returned to a functional full-power state) when the kernel
17needs to use it. (There also are forms of PM in which components are
18placed in a less functional but still usable state instead of being
19suspended; an example would be reducing the CPU's clock rate. This
20document will not discuss those other forms.)
21
22When the parts being suspended include the CPU and most of the rest of
23the system, we speak of it as a "system suspend". When a particular
24device is turned off while the system as a whole remains running, we
25call it a "dynamic suspend" (also known as a "runtime suspend" or
26"selective suspend"). This document concentrates mostly on how
27dynamic PM is implemented in the USB subsystem, although system PM is
28covered to some extent (see Documentation/power/*.txt for more
29information about system PM).
30
31Note: Dynamic PM support for USB is present only if the kernel was
Alan Stern9bbdf1e2010-01-08 12:57:28 -050032built with CONFIG_USB_SUSPEND enabled (which depends on
33CONFIG_PM_RUNTIME). System PM support is present only if the kernel
34was built with CONFIG_SUSPEND or CONFIG_HIBERNATION enabled.
Alan Sterncd38c1e2007-10-10 16:24:56 -040035
36
37 What is Remote Wakeup?
38 ----------------------
39
40When a device has been suspended, it generally doesn't resume until
41the computer tells it to. Likewise, if the entire computer has been
42suspended, it generally doesn't resume until the user tells it to, say
43by pressing a power button or opening the cover.
44
45However some devices have the capability of resuming by themselves, or
46asking the kernel to resume them, or even telling the entire computer
47to resume. This capability goes by several names such as "Wake On
48LAN"; we will refer to it generically as "remote wakeup". When a
49device is enabled for remote wakeup and it is suspended, it may resume
50itself (or send a request to be resumed) in response to some external
51event. Examples include a suspended keyboard resuming when a key is
52pressed, or a suspended USB hub resuming when a device is plugged in.
53
54
55 When is a USB device idle?
56 --------------------------
57
58A device is idle whenever the kernel thinks it's not busy doing
59anything important and thus is a candidate for being suspended. The
60exact definition depends on the device's driver; drivers are allowed
61to declare that a device isn't idle even when there's no actual
62communication taking place. (For example, a hub isn't considered idle
63unless all the devices plugged into that hub are already suspended.)
64In addition, a device isn't considered idle so long as a program keeps
65its usbfs file open, whether or not any I/O is going on.
66
67If a USB device has no driver, its usbfs file isn't open, and it isn't
68being accessed through sysfs, then it definitely is idle.
69
70
71 Forms of dynamic PM
72 -------------------
73
Alan Sternbaf67742009-12-08 15:49:48 -050074Dynamic suspends occur when the kernel decides to suspend an idle
75device. This is called "autosuspend" for short. In general, a device
76won't be autosuspended unless it has been idle for some minimum period
77of time, the so-called idle-delay time.
Alan Sterncd38c1e2007-10-10 16:24:56 -040078
79Of course, nothing the kernel does on its own initiative should
80prevent the computer or its devices from working properly. If a
81device has been autosuspended and a program tries to use it, the
82kernel will automatically resume the device (autoresume). For the
83same reason, an autosuspended device will usually have remote wakeup
84enabled, if the device supports remote wakeup.
85
86It is worth mentioning that many USB drivers don't support
87autosuspend. In fact, at the time of this writing (Linux 2.6.23) the
88only drivers which do support it are the hub driver, kaweth, asix,
89usblp, usblcd, and usb-skeleton (which doesn't count). If a
90non-supporting driver is bound to a device, the device won't be
91autosuspended. In effect, the kernel pretends the device is never
92idle.
93
94We can categorize power management events in two broad classes:
95external and internal. External events are those triggered by some
96agent outside the USB stack: system suspend/resume (triggered by
Alan Sternbaf67742009-12-08 15:49:48 -050097userspace), manual dynamic resume (also triggered by userspace), and
98remote wakeup (triggered by the device). Internal events are those
99triggered within the USB stack: autosuspend and autoresume. Note that
100all dynamic suspend events are internal; external agents are not
101allowed to issue dynamic suspends.
Alan Sterncd38c1e2007-10-10 16:24:56 -0400102
103
104 The user interface for dynamic PM
105 ---------------------------------
106
107The user interface for controlling dynamic PM is located in the power/
108subdirectory of each USB device's sysfs directory, that is, in
109/sys/bus/usb/devices/.../power/ where "..." is the device's ID. The
Alan Sterna9030982010-04-02 13:22:16 -0400110relevant attribute files are: wakeup, control, and autosuspend.
111(There may also be a file named "level"; this file was deprecated
112as of the 2.6.35 kernel and replaced by the "control" file.)
Alan Sterncd38c1e2007-10-10 16:24:56 -0400113
114 power/wakeup
115
116 This file is empty if the device does not support
117 remote wakeup. Otherwise the file contains either the
118 word "enabled" or the word "disabled", and you can
119 write those words to the file. The setting determines
120 whether or not remote wakeup will be enabled when the
121 device is next suspended. (If the setting is changed
122 while the device is suspended, the change won't take
123 effect until the following suspend.)
124
Alan Sterna9030982010-04-02 13:22:16 -0400125 power/control
Alan Sterncd38c1e2007-10-10 16:24:56 -0400126
Alan Stern8e4ceb32009-12-07 13:01:37 -0500127 This file contains one of two words: "on" or "auto".
128 You can write those words to the file to change the
129 device's setting.
Alan Sterncd38c1e2007-10-10 16:24:56 -0400130
131 "on" means that the device should be resumed and
132 autosuspend is not allowed. (Of course, system
133 suspends are still allowed.)
134
135 "auto" is the normal state in which the kernel is
136 allowed to autosuspend and autoresume the device.
137
Alan Stern8e4ceb32009-12-07 13:01:37 -0500138 (In kernels up to 2.6.32, you could also specify
139 "suspend", meaning that the device should remain
140 suspended and autoresume was not allowed. This
141 setting is no longer supported.)
Alan Sterncd38c1e2007-10-10 16:24:56 -0400142
143 power/autosuspend
144
145 This file contains an integer value, which is the
146 number of seconds the device should remain idle before
147 the kernel will autosuspend it (the idle-delay time).
148 The default is 2. 0 means to autosuspend as soon as
Alan Sternbaf67742009-12-08 15:49:48 -0500149 the device becomes idle, and negative values mean
150 never to autosuspend. You can write a number to the
151 file to change the autosuspend idle-delay time.
Alan Sterncd38c1e2007-10-10 16:24:56 -0400152
Alan Sterna9030982010-04-02 13:22:16 -0400153Writing "-1" to power/autosuspend and writing "on" to power/control do
Alan Sterncd38c1e2007-10-10 16:24:56 -0400154essentially the same thing -- they both prevent the device from being
155autosuspended. Yes, this is a redundancy in the API.
156
157(In 2.6.21 writing "0" to power/autosuspend would prevent the device
158from being autosuspended; the behavior was changed in 2.6.22. The
159power/autosuspend attribute did not exist prior to 2.6.21, and the
Alan Sterna9030982010-04-02 13:22:16 -0400160power/level attribute did not exist prior to 2.6.22. power/control
161was added in 2.6.34.)
Alan Sterncd38c1e2007-10-10 16:24:56 -0400162
163
164 Changing the default idle-delay time
165 ------------------------------------
166
167The default autosuspend idle-delay time is controlled by a module
168parameter in usbcore. You can specify the value when usbcore is
169loaded. For example, to set it to 5 seconds instead of 2 you would
170do:
171
172 modprobe usbcore autosuspend=5
173
174Equivalently, you could add to /etc/modprobe.conf a line saying:
175
176 options usbcore autosuspend=5
177
178Some distributions load the usbcore module very early during the boot
179process, by means of a program or script running from an initramfs
180image. To alter the parameter value you would have to rebuild that
181image.
182
183If usbcore is compiled into the kernel rather than built as a loadable
184module, you can add
185
186 usbcore.autosuspend=5
187
188to the kernel's boot command line.
189
190Finally, the parameter value can be changed while the system is
191running. If you do:
192
193 echo 5 >/sys/module/usbcore/parameters/autosuspend
194
195then each new USB device will have its autosuspend idle-delay
196initialized to 5. (The idle-delay values for already existing devices
197will not be affected.)
198
199Setting the initial default idle-delay to -1 will prevent any
200autosuspend of any USB device. This is a simple alternative to
201disabling CONFIG_USB_SUSPEND and rebuilding the kernel, and it has the
202added benefit of allowing you to enable autosuspend for selected
203devices.
204
205
206 Warnings
207 --------
208
209The USB specification states that all USB devices must support power
210management. Nevertheless, the sad fact is that many devices do not
211support it very well. You can suspend them all right, but when you
212try to resume them they disconnect themselves from the USB bus or
213they stop working entirely. This seems to be especially prevalent
214among printers and scanners, but plenty of other types of device have
215the same deficiency.
216
217For this reason, by default the kernel disables autosuspend (the
Alan Sterna9030982010-04-02 13:22:16 -0400218power/control attribute is initialized to "on") for all devices other
Alan Sterncd38c1e2007-10-10 16:24:56 -0400219than hubs. Hubs, at least, appear to be reasonably well-behaved in
220this regard.
221
222(In 2.6.21 and 2.6.22 this wasn't the case. Autosuspend was enabled
223by default for almost all USB devices. A number of people experienced
224problems as a result.)
225
226This means that non-hub devices won't be autosuspended unless the user
227or a program explicitly enables it. As of this writing there aren't
228any widespread programs which will do this; we hope that in the near
229future device managers such as HAL will take on this added
230responsibility. In the meantime you can always carry out the
231necessary operations by hand or add them to a udev script. You can
232also change the idle-delay time; 2 seconds is not the best choice for
233every device.
234
Alan Stern088f7fe2010-01-08 12:56:54 -0500235If a driver knows that its device has proper suspend/resume support,
236it can enable autosuspend all by itself. For example, the video
237driver for a laptop's webcam might do this, since these devices are
238rarely used and so should normally be autosuspended.
239
Alan Sterncd38c1e2007-10-10 16:24:56 -0400240Sometimes it turns out that even when a device does work okay with
241autosuspend there are still problems. For example, there are
242experimental patches adding autosuspend support to the usbhid driver,
243which manages keyboards and mice, among other things. Tests with a
244number of keyboards showed that typing on a suspended keyboard, while
245causing the keyboard to do a remote wakeup all right, would
246nonetheless frequently result in lost keystrokes. Tests with mice
247showed that some of them would issue a remote-wakeup request in
248response to button presses but not to motion, and some in response to
249neither.
250
251The kernel will not prevent you from enabling autosuspend on devices
252that can't handle it. It is even possible in theory to damage a
253device by suspending it at the wrong time -- for example, suspending a
254USB hard disk might cause it to spin down without parking the heads.
255(Highly unlikely, but possible.) Take care.
256
257
258 The driver interface for Power Management
259 -----------------------------------------
260
261The requirements for a USB driver to support external power management
262are pretty modest; the driver need only define
263
264 .suspend
265 .resume
266 .reset_resume
267
268methods in its usb_driver structure, and the reset_resume method is
269optional. The methods' jobs are quite simple:
270
271 The suspend method is called to warn the driver that the
272 device is going to be suspended. If the driver returns a
273 negative error code, the suspend will be aborted. Normally
274 the driver will return 0, in which case it must cancel all
275 outstanding URBs (usb_kill_urb()) and not submit any more.
276
277 The resume method is called to tell the driver that the
278 device has been resumed and the driver can return to normal
279 operation. URBs may once more be submitted.
280
281 The reset_resume method is called to tell the driver that
282 the device has been resumed and it also has been reset.
283 The driver should redo any necessary device initialization,
284 since the device has probably lost most or all of its state
285 (although the interfaces will be in the same altsettings as
286 before the suspend).
287
Alan Stern3c886c52007-11-16 11:58:15 -0500288If the device is disconnected or powered down while it is suspended,
289the disconnect method will be called instead of the resume or
290reset_resume method. This is also quite likely to happen when
291waking up from hibernation, as many systems do not maintain suspend
292current to the USB host controllers during hibernation. (It's
293possible to work around the hibernation-forces-disconnect problem by
294using the USB Persist facility.)
295
Alan Sterncd38c1e2007-10-10 16:24:56 -0400296The reset_resume method is used by the USB Persist facility (see
297Documentation/usb/persist.txt) and it can also be used under certain
298circumstances when CONFIG_USB_PERSIST is not enabled. Currently, if a
299device is reset during a resume and the driver does not have a
300reset_resume method, the driver won't receive any notification about
301the resume. Later kernels will call the driver's disconnect method;
3022.6.23 doesn't do this.
303
304USB drivers are bound to interfaces, so their suspend and resume
305methods get called when the interfaces are suspended or resumed. In
306principle one might want to suspend some interfaces on a device (i.e.,
307force the drivers for those interface to stop all activity) without
308suspending the other interfaces. The USB core doesn't allow this; all
309interfaces are suspended when the device itself is suspended and all
310interfaces are resumed when the device is resumed. It isn't possible
311to suspend or resume some but not all of a device's interfaces. The
312closest you can come is to unbind the interfaces' drivers.
313
314
315 The driver interface for autosuspend and autoresume
316 ---------------------------------------------------
317
318To support autosuspend and autoresume, a driver should implement all
319three of the methods listed above. In addition, a driver indicates
320that it supports autosuspend by setting the .supports_autosuspend flag
321in its usb_driver structure. It is then responsible for informing the
322USB core whenever one of its interfaces becomes busy or idle. The
Alan Stern8e4ceb32009-12-07 13:01:37 -0500323driver does so by calling these six functions:
Alan Sterncd38c1e2007-10-10 16:24:56 -0400324
325 int usb_autopm_get_interface(struct usb_interface *intf);
326 void usb_autopm_put_interface(struct usb_interface *intf);
Alan Stern9ac39f22008-11-12 16:19:49 -0500327 int usb_autopm_get_interface_async(struct usb_interface *intf);
328 void usb_autopm_put_interface_async(struct usb_interface *intf);
Alan Stern8e4ceb32009-12-07 13:01:37 -0500329 void usb_autopm_get_interface_no_resume(struct usb_interface *intf);
330 void usb_autopm_put_interface_no_suspend(struct usb_interface *intf);
Alan Sterncd38c1e2007-10-10 16:24:56 -0400331
Alan Stern9bbdf1e2010-01-08 12:57:28 -0500332The functions work by maintaining a usage counter in the
333usb_interface's embedded device structure. When the counter is > 0
334then the interface is deemed to be busy, and the kernel will not
335autosuspend the interface's device. When the usage counter is = 0
336then the interface is considered to be idle, and the kernel may
337autosuspend the device.
Alan Sterncd38c1e2007-10-10 16:24:56 -0400338
Alan Stern9bbdf1e2010-01-08 12:57:28 -0500339(There is a similar usage counter field in struct usb_device,
Alan Sterncd38c1e2007-10-10 16:24:56 -0400340associated with the device itself rather than any of its interfaces.
Alan Stern9bbdf1e2010-01-08 12:57:28 -0500341This counter is used only by the USB core.)
Alan Sterncd38c1e2007-10-10 16:24:56 -0400342
Alan Stern9bbdf1e2010-01-08 12:57:28 -0500343Drivers need not be concerned about balancing changes to the usage
344counter; the USB core will undo any remaining "get"s when a driver
345is unbound from its interface. As a corollary, drivers must not call
346any of the usb_autopm_* functions after their diconnect() routine has
347returned.
Alan Sterncd38c1e2007-10-10 16:24:56 -0400348
Alan Stern9bbdf1e2010-01-08 12:57:28 -0500349Drivers using the async routines are responsible for their own
350synchronization and mutual exclusion.
Alan Sterncd38c1e2007-10-10 16:24:56 -0400351
Alan Stern9bbdf1e2010-01-08 12:57:28 -0500352 usb_autopm_get_interface() increments the usage counter and
353 does an autoresume if the device is suspended. If the
354 autoresume fails, the counter is decremented back.
355
356 usb_autopm_put_interface() decrements the usage counter and
357 attempts an autosuspend if the new value is = 0.
Alan Sterncd38c1e2007-10-10 16:24:56 -0400358
Alan Stern9ac39f22008-11-12 16:19:49 -0500359 usb_autopm_get_interface_async() and
360 usb_autopm_put_interface_async() do almost the same things as
Alan Stern9bbdf1e2010-01-08 12:57:28 -0500361 their non-async counterparts. The big difference is that they
362 use a workqueue to do the resume or suspend part of their
Alan Stern9ac39f22008-11-12 16:19:49 -0500363 jobs. As a result they can be called in an atomic context,
364 such as an URB's completion handler, but when they return the
Alan Stern9bbdf1e2010-01-08 12:57:28 -0500365 device will generally not yet be in the desired state.
Alan Stern9ac39f22008-11-12 16:19:49 -0500366
Alan Stern8e4ceb32009-12-07 13:01:37 -0500367 usb_autopm_get_interface_no_resume() and
368 usb_autopm_put_interface_no_suspend() merely increment or
Alan Stern9bbdf1e2010-01-08 12:57:28 -0500369 decrement the usage counter; they do not attempt to carry out
370 an autoresume or an autosuspend. Hence they can be called in
371 an atomic context.
Geoff Levand81ab5b82008-09-20 14:41:47 -0700372
Alan Stern9bbdf1e2010-01-08 12:57:28 -0500373The simplest usage pattern is that a driver calls
Alan Sterncd38c1e2007-10-10 16:24:56 -0400374usb_autopm_get_interface() in its open routine and
Alan Stern9bbdf1e2010-01-08 12:57:28 -0500375usb_autopm_put_interface() in its close or release routine. But other
376patterns are possible.
Alan Sterncd38c1e2007-10-10 16:24:56 -0400377
378The autosuspend attempts mentioned above will often fail for one
Alan Sterna9030982010-04-02 13:22:16 -0400379reason or another. For example, the power/control attribute might be
Alan Sterncd38c1e2007-10-10 16:24:56 -0400380set to "on", or another interface in the same device might not be
381idle. This is perfectly normal. If the reason for failure was that
Alan Stern9bbdf1e2010-01-08 12:57:28 -0500382the device hasn't been idle for long enough, a timer is scheduled to
383carry out the operation automatically when the autosuspend idle-delay
384has expired.
Alan Sterncd38c1e2007-10-10 16:24:56 -0400385
Alan Sternbaf67742009-12-08 15:49:48 -0500386Autoresume attempts also can fail, although failure would mean that
387the device is no longer present or operating properly. Unlike
Alan Stern9bbdf1e2010-01-08 12:57:28 -0500388autosuspend, there's no idle-delay for an autoresume.
Alan Sterncd38c1e2007-10-10 16:24:56 -0400389
390
391 Other parts of the driver interface
392 -----------------------------------
393
Alan Stern088f7fe2010-01-08 12:56:54 -0500394Drivers can enable autosuspend for their devices by calling
395
396 usb_enable_autosuspend(struct usb_device *udev);
397
398in their probe() routine, if they know that the device is capable of
399suspending and resuming correctly. This is exactly equivalent to
Alan Sterna9030982010-04-02 13:22:16 -0400400writing "auto" to the device's power/control attribute. Likewise,
Alan Stern088f7fe2010-01-08 12:56:54 -0500401drivers can disable autosuspend by calling
402
403 usb_disable_autosuspend(struct usb_device *udev);
404
Alan Sterna9030982010-04-02 13:22:16 -0400405This is exactly the same as writing "on" to the power/control attribute.
Alan Stern088f7fe2010-01-08 12:56:54 -0500406
Alan Sterncd38c1e2007-10-10 16:24:56 -0400407Sometimes a driver needs to make sure that remote wakeup is enabled
408during autosuspend. For example, there's not much point
409autosuspending a keyboard if the user can't cause the keyboard to do a
410remote wakeup by typing on it. If the driver sets
411intf->needs_remote_wakeup to 1, the kernel won't autosuspend the
412device if remote wakeup isn't available or has been disabled through
413the power/wakeup attribute. (If the device is already autosuspended,
414though, setting this flag won't cause the kernel to autoresume it.
415Normally a driver would set this flag in its probe method, at which
416time the device is guaranteed not to be autosuspended.)
417
Alan Stern9bbdf1e2010-01-08 12:57:28 -0500418If a driver does its I/O asynchronously in interrupt context, it
419should call usb_autopm_get_interface_async() before starting output and
420usb_autopm_put_interface_async() when the output queue drains. When
421it receives an input event, it should call
Alan Sterncd38c1e2007-10-10 16:24:56 -0400422
423 usb_mark_last_busy(struct usb_device *udev);
424
Alan Stern9bbdf1e2010-01-08 12:57:28 -0500425in the event handler. This sets udev->last_busy to the current time.
426udev->last_busy is the field used for idle-delay calculations;
427updating it will cause any pending autosuspend to be moved back. Most
428of the usb_autopm_* routines will also set the last_busy field to the
429current time.
Alan Sterncd38c1e2007-10-10 16:24:56 -0400430
Alan Stern9bbdf1e2010-01-08 12:57:28 -0500431Asynchronous operation is always subject to races. For example, a
432driver may call one of the usb_autopm_*_interface_async() routines at
433a time when the core has just finished deciding the device has been
434idle for long enough but not yet gotten around to calling the driver's
435suspend method. The suspend method must be responsible for
436synchronizing with the output request routine and the URB completion
437handler; it should cause autosuspends to fail with -EBUSY if the
438driver needs to use the device.
Alan Sterncd38c1e2007-10-10 16:24:56 -0400439
440External suspend calls should never be allowed to fail in this way,
441only autosuspend calls. The driver can tell them apart by checking
Alan Sternfb34d532009-11-13 11:53:59 -0500442the PM_EVENT_AUTO bit in the message.event argument to the suspend
443method; this bit will be set for internal PM events (autosuspend) and
444clear for external PM events.
Alan Sterncd38c1e2007-10-10 16:24:56 -0400445
Alan Sterncd38c1e2007-10-10 16:24:56 -0400446
Alan Stern9bbdf1e2010-01-08 12:57:28 -0500447 Mutual exclusion
448 ----------------
Alan Sterncd38c1e2007-10-10 16:24:56 -0400449
Alan Stern9bbdf1e2010-01-08 12:57:28 -0500450For external events -- but not necessarily for autosuspend or
451autoresume -- the device semaphore (udev->dev.sem) will be held when a
452suspend or resume method is called. This implies that external
453suspend/resume events are mutually exclusive with calls to probe,
454disconnect, pre_reset, and post_reset; the USB core guarantees that
455this is true of autosuspend/autoresume events as well.
Alan Sterncd38c1e2007-10-10 16:24:56 -0400456
457If a driver wants to block all suspend/resume calls during some
Alan Stern9bbdf1e2010-01-08 12:57:28 -0500458critical section, the best way is to lock the device and call
459usb_autopm_get_interface() (and do the reverse at the end of the
460critical section). Holding the device semaphore will block all
461external PM calls, and the usb_autopm_get_interface() will prevent any
462internal PM calls, even if it fails. (Exercise: Why?)
Alan Sterncd38c1e2007-10-10 16:24:56 -0400463
464
465 Interaction between dynamic PM and system PM
466 --------------------------------------------
467
468Dynamic power management and system power management can interact in
469a couple of ways.
470
Alan Stern9bbdf1e2010-01-08 12:57:28 -0500471Firstly, a device may already be autosuspended when a system suspend
472occurs. Since system suspends are supposed to be as transparent as
473possible, the device should remain suspended following the system
474resume. But this theory may not work out well in practice; over time
475the kernel's behavior in this regard has changed.
Alan Sterncd38c1e2007-10-10 16:24:56 -0400476
477Secondly, a dynamic power-management event may occur as a system
478suspend is underway. The window for this is short, since system
479suspends don't take long (a few seconds usually), but it can happen.
480For example, a suspended device may send a remote-wakeup signal while
481the system is suspending. The remote wakeup may succeed, which would
482cause the system suspend to abort. If the remote wakeup doesn't
483succeed, it may still remain active and thus cause the system to
484resume as soon as the system suspend is complete. Or the remote
485wakeup may fail and get lost. Which outcome occurs depends on timing
486and on the hardware and firmware design.