blob: 3bf6818c8cf58dd386e03a2c6e085718f58d0004 [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 Stern8e4ceb32009-12-07 13:01:37 -05005 November 10, 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
32built with CONFIG_USB_SUSPEND enabled. System PM support is present
33only if the kernel was built with CONFIG_SUSPEND or CONFIG_HIBERNATION
34enabled.
35
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
110relevant attribute files are: wakeup, level, and autosuspend.
111
112 power/wakeup
113
114 This file is empty if the device does not support
115 remote wakeup. Otherwise the file contains either the
116 word "enabled" or the word "disabled", and you can
117 write those words to the file. The setting determines
118 whether or not remote wakeup will be enabled when the
119 device is next suspended. (If the setting is changed
120 while the device is suspended, the change won't take
121 effect until the following suspend.)
122
123 power/level
124
Alan Stern8e4ceb32009-12-07 13:01:37 -0500125 This file contains one of two words: "on" or "auto".
126 You can write those words to the file to change the
127 device's setting.
Alan Sterncd38c1e2007-10-10 16:24:56 -0400128
129 "on" means that the device should be resumed and
130 autosuspend is not allowed. (Of course, system
131 suspends are still allowed.)
132
133 "auto" is the normal state in which the kernel is
134 allowed to autosuspend and autoresume the device.
135
Alan Stern8e4ceb32009-12-07 13:01:37 -0500136 (In kernels up to 2.6.32, you could also specify
137 "suspend", meaning that the device should remain
138 suspended and autoresume was not allowed. This
139 setting is no longer supported.)
Alan Sterncd38c1e2007-10-10 16:24:56 -0400140
141 power/autosuspend
142
143 This file contains an integer value, which is the
144 number of seconds the device should remain idle before
145 the kernel will autosuspend it (the idle-delay time).
146 The default is 2. 0 means to autosuspend as soon as
Alan Sternbaf67742009-12-08 15:49:48 -0500147 the device becomes idle, and negative values mean
148 never to autosuspend. You can write a number to the
149 file to change the autosuspend idle-delay time.
Alan Sterncd38c1e2007-10-10 16:24:56 -0400150
151Writing "-1" to power/autosuspend and writing "on" to power/level do
152essentially the same thing -- they both prevent the device from being
153autosuspended. Yes, this is a redundancy in the API.
154
155(In 2.6.21 writing "0" to power/autosuspend would prevent the device
156from being autosuspended; the behavior was changed in 2.6.22. The
157power/autosuspend attribute did not exist prior to 2.6.21, and the
158power/level attribute did not exist prior to 2.6.22.)
159
160
161 Changing the default idle-delay time
162 ------------------------------------
163
164The default autosuspend idle-delay time is controlled by a module
165parameter in usbcore. You can specify the value when usbcore is
166loaded. For example, to set it to 5 seconds instead of 2 you would
167do:
168
169 modprobe usbcore autosuspend=5
170
171Equivalently, you could add to /etc/modprobe.conf a line saying:
172
173 options usbcore autosuspend=5
174
175Some distributions load the usbcore module very early during the boot
176process, by means of a program or script running from an initramfs
177image. To alter the parameter value you would have to rebuild that
178image.
179
180If usbcore is compiled into the kernel rather than built as a loadable
181module, you can add
182
183 usbcore.autosuspend=5
184
185to the kernel's boot command line.
186
187Finally, the parameter value can be changed while the system is
188running. If you do:
189
190 echo 5 >/sys/module/usbcore/parameters/autosuspend
191
192then each new USB device will have its autosuspend idle-delay
193initialized to 5. (The idle-delay values for already existing devices
194will not be affected.)
195
196Setting the initial default idle-delay to -1 will prevent any
197autosuspend of any USB device. This is a simple alternative to
198disabling CONFIG_USB_SUSPEND and rebuilding the kernel, and it has the
199added benefit of allowing you to enable autosuspend for selected
200devices.
201
202
203 Warnings
204 --------
205
206The USB specification states that all USB devices must support power
207management. Nevertheless, the sad fact is that many devices do not
208support it very well. You can suspend them all right, but when you
209try to resume them they disconnect themselves from the USB bus or
210they stop working entirely. This seems to be especially prevalent
211among printers and scanners, but plenty of other types of device have
212the same deficiency.
213
214For this reason, by default the kernel disables autosuspend (the
215power/level attribute is initialized to "on") for all devices other
216than hubs. Hubs, at least, appear to be reasonably well-behaved in
217this regard.
218
219(In 2.6.21 and 2.6.22 this wasn't the case. Autosuspend was enabled
220by default for almost all USB devices. A number of people experienced
221problems as a result.)
222
223This means that non-hub devices won't be autosuspended unless the user
224or a program explicitly enables it. As of this writing there aren't
225any widespread programs which will do this; we hope that in the near
226future device managers such as HAL will take on this added
227responsibility. In the meantime you can always carry out the
228necessary operations by hand or add them to a udev script. You can
229also change the idle-delay time; 2 seconds is not the best choice for
230every device.
231
232Sometimes it turns out that even when a device does work okay with
233autosuspend there are still problems. For example, there are
234experimental patches adding autosuspend support to the usbhid driver,
235which manages keyboards and mice, among other things. Tests with a
236number of keyboards showed that typing on a suspended keyboard, while
237causing the keyboard to do a remote wakeup all right, would
238nonetheless frequently result in lost keystrokes. Tests with mice
239showed that some of them would issue a remote-wakeup request in
240response to button presses but not to motion, and some in response to
241neither.
242
243The kernel will not prevent you from enabling autosuspend on devices
244that can't handle it. It is even possible in theory to damage a
245device by suspending it at the wrong time -- for example, suspending a
246USB hard disk might cause it to spin down without parking the heads.
247(Highly unlikely, but possible.) Take care.
248
249
250 The driver interface for Power Management
251 -----------------------------------------
252
253The requirements for a USB driver to support external power management
254are pretty modest; the driver need only define
255
256 .suspend
257 .resume
258 .reset_resume
259
260methods in its usb_driver structure, and the reset_resume method is
261optional. The methods' jobs are quite simple:
262
263 The suspend method is called to warn the driver that the
264 device is going to be suspended. If the driver returns a
265 negative error code, the suspend will be aborted. Normally
266 the driver will return 0, in which case it must cancel all
267 outstanding URBs (usb_kill_urb()) and not submit any more.
268
269 The resume method is called to tell the driver that the
270 device has been resumed and the driver can return to normal
271 operation. URBs may once more be submitted.
272
273 The reset_resume method is called to tell the driver that
274 the device has been resumed and it also has been reset.
275 The driver should redo any necessary device initialization,
276 since the device has probably lost most or all of its state
277 (although the interfaces will be in the same altsettings as
278 before the suspend).
279
Alan Stern3c886c52007-11-16 11:58:15 -0500280If the device is disconnected or powered down while it is suspended,
281the disconnect method will be called instead of the resume or
282reset_resume method. This is also quite likely to happen when
283waking up from hibernation, as many systems do not maintain suspend
284current to the USB host controllers during hibernation. (It's
285possible to work around the hibernation-forces-disconnect problem by
286using the USB Persist facility.)
287
Alan Sterncd38c1e2007-10-10 16:24:56 -0400288The reset_resume method is used by the USB Persist facility (see
289Documentation/usb/persist.txt) and it can also be used under certain
290circumstances when CONFIG_USB_PERSIST is not enabled. Currently, if a
291device is reset during a resume and the driver does not have a
292reset_resume method, the driver won't receive any notification about
293the resume. Later kernels will call the driver's disconnect method;
2942.6.23 doesn't do this.
295
296USB drivers are bound to interfaces, so their suspend and resume
297methods get called when the interfaces are suspended or resumed. In
298principle one might want to suspend some interfaces on a device (i.e.,
299force the drivers for those interface to stop all activity) without
300suspending the other interfaces. The USB core doesn't allow this; all
301interfaces are suspended when the device itself is suspended and all
302interfaces are resumed when the device is resumed. It isn't possible
303to suspend or resume some but not all of a device's interfaces. The
304closest you can come is to unbind the interfaces' drivers.
305
306
307 The driver interface for autosuspend and autoresume
308 ---------------------------------------------------
309
310To support autosuspend and autoresume, a driver should implement all
311three of the methods listed above. In addition, a driver indicates
312that it supports autosuspend by setting the .supports_autosuspend flag
313in its usb_driver structure. It is then responsible for informing the
314USB core whenever one of its interfaces becomes busy or idle. The
Alan Stern8e4ceb32009-12-07 13:01:37 -0500315driver does so by calling these six functions:
Alan Sterncd38c1e2007-10-10 16:24:56 -0400316
317 int usb_autopm_get_interface(struct usb_interface *intf);
318 void usb_autopm_put_interface(struct usb_interface *intf);
Alan Stern9ac39f22008-11-12 16:19:49 -0500319 int usb_autopm_get_interface_async(struct usb_interface *intf);
320 void usb_autopm_put_interface_async(struct usb_interface *intf);
Alan Stern8e4ceb32009-12-07 13:01:37 -0500321 void usb_autopm_get_interface_no_resume(struct usb_interface *intf);
322 void usb_autopm_put_interface_no_suspend(struct usb_interface *intf);
Alan Sterncd38c1e2007-10-10 16:24:56 -0400323
324The functions work by maintaining a counter in the usb_interface
325structure. When intf->pm_usage_count is > 0 then the interface is
326deemed to be busy, and the kernel will not autosuspend the interface's
327device. When intf->pm_usage_count is <= 0 then the interface is
328considered to be idle, and the kernel may autosuspend the device.
329
330(There is a similar pm_usage_count field in struct usb_device,
331associated with the device itself rather than any of its interfaces.
332This field is used only by the USB core.)
333
Alan Stern8e4ceb32009-12-07 13:01:37 -0500334Drivers must not modify intf->pm_usage_count directly; its value
335should be changed only be using the functions listed above. Drivers
336are responsible for insuring that the overall change to pm_usage_count
337during their lifetime balances out to 0 (it may be necessary for the
338disconnect method to call usb_autopm_put_interface() one or more times
339to fulfill this requirement). The first two routines use the PM mutex
340in struct usb_device for mutual exclusion; drivers using the async
Alan Stern9ac39f22008-11-12 16:19:49 -0500341routines are responsible for their own synchronization and mutual
342exclusion.
Alan Sterncd38c1e2007-10-10 16:24:56 -0400343
344 usb_autopm_get_interface() increments pm_usage_count and
345 attempts an autoresume if the new value is > 0 and the
346 device is suspended.
347
348 usb_autopm_put_interface() decrements pm_usage_count and
349 attempts an autosuspend if the new value is <= 0 and the
350 device isn't suspended.
351
Alan Stern9ac39f22008-11-12 16:19:49 -0500352 usb_autopm_get_interface_async() and
353 usb_autopm_put_interface_async() do almost the same things as
354 their non-async counterparts. The differences are: they do
355 not acquire the PM mutex, and they use a workqueue to do their
356 jobs. As a result they can be called in an atomic context,
357 such as an URB's completion handler, but when they return the
358 device will not generally not yet be in the desired state.
359
Alan Stern8e4ceb32009-12-07 13:01:37 -0500360 usb_autopm_get_interface_no_resume() and
361 usb_autopm_put_interface_no_suspend() merely increment or
362 decrement the pm_usage_count value; they do not attempt to
363 carry out an autoresume or an autosuspend. Hence they can be
364 called in an atomic context.
Geoff Levand81ab5b82008-09-20 14:41:47 -0700365
Alan Sterncd38c1e2007-10-10 16:24:56 -0400366The conventional usage pattern is that a driver calls
367usb_autopm_get_interface() in its open routine and
368usb_autopm_put_interface() in its close or release routine. But
369other patterns are possible.
370
371The autosuspend attempts mentioned above will often fail for one
372reason or another. For example, the power/level attribute might be
373set to "on", or another interface in the same device might not be
374idle. This is perfectly normal. If the reason for failure was that
375the device hasn't been idle for long enough, a delayed workqueue
376routine is automatically set up to carry out the operation when the
377autosuspend idle-delay has expired.
378
Alan Sternbaf67742009-12-08 15:49:48 -0500379Autoresume attempts also can fail, although failure would mean that
380the device is no longer present or operating properly. Unlike
381autosuspend, there's no delay for an autoresume.
Alan Sterncd38c1e2007-10-10 16:24:56 -0400382
383
384 Other parts of the driver interface
385 -----------------------------------
386
387Sometimes a driver needs to make sure that remote wakeup is enabled
388during autosuspend. For example, there's not much point
389autosuspending a keyboard if the user can't cause the keyboard to do a
390remote wakeup by typing on it. If the driver sets
391intf->needs_remote_wakeup to 1, the kernel won't autosuspend the
392device if remote wakeup isn't available or has been disabled through
393the power/wakeup attribute. (If the device is already autosuspended,
394though, setting this flag won't cause the kernel to autoresume it.
395Normally a driver would set this flag in its probe method, at which
396time the device is guaranteed not to be autosuspended.)
397
Alan Stern8e4ceb32009-12-07 13:01:37 -0500398The synchronous usb_autopm_* routines have to run in a sleepable
399process context; they must not be called from an interrupt handler or
400while holding a spinlock. In fact, the entire autosuspend mechanism
401is not well geared toward interrupt-driven operation. However there
402is one thing a driver can do in an interrupt handler:
Alan Sterncd38c1e2007-10-10 16:24:56 -0400403
404 usb_mark_last_busy(struct usb_device *udev);
405
406This sets udev->last_busy to the current time. udev->last_busy is the
407field used for idle-delay calculations; updating it will cause any
408pending autosuspend to be moved back. The usb_autopm_* routines will
409also set the last_busy field to the current time.
410
411Calling urb_mark_last_busy() from within an URB completion handler is
412subject to races: The kernel may have just finished deciding the
413device has been idle for long enough but not yet gotten around to
414calling the driver's suspend method. The driver would have to be
415responsible for synchronizing its suspend method with its URB
416completion handler and causing the autosuspend to fail with -EBUSY if
417an URB had completed too recently.
418
419External suspend calls should never be allowed to fail in this way,
420only autosuspend calls. The driver can tell them apart by checking
Alan Sternfb34d532009-11-13 11:53:59 -0500421the PM_EVENT_AUTO bit in the message.event argument to the suspend
422method; this bit will be set for internal PM events (autosuspend) and
423clear for external PM events.
Alan Sterncd38c1e2007-10-10 16:24:56 -0400424
425Many of the ingredients in the autosuspend framework are oriented
426towards interfaces: The usb_interface structure contains the
427pm_usage_cnt field, and the usb_autopm_* routines take an interface
428pointer as their argument. But somewhat confusingly, a few of the
Alan Sternfb34d532009-11-13 11:53:59 -0500429pieces (i.e., usb_mark_last_busy()) use the usb_device structure
430instead. Drivers need to keep this straight; they can call
Alan Sterncd38c1e2007-10-10 16:24:56 -0400431interface_to_usbdev() to find the device structure for a given
432interface.
433
434
435 Locking requirements
436 --------------------
437
438All three suspend/resume methods are always called while holding the
439usb_device's PM mutex. For external events -- but not necessarily for
440autosuspend or autoresume -- the device semaphore (udev->dev.sem) will
441also be held. This implies that external suspend/resume events are
442mutually exclusive with calls to probe, disconnect, pre_reset, and
443post_reset; the USB core guarantees that this is true of internal
444suspend/resume events as well.
445
446If a driver wants to block all suspend/resume calls during some
Oliver Neukumf0fa7462008-07-04 10:10:53 +0200447critical section, it can simply acquire udev->pm_mutex. Note that
448calls to resume may be triggered indirectly. Block IO due to memory
449allocations can make the vm subsystem resume a device. Thus while
450holding this lock you must not allocate memory with GFP_KERNEL or
451GFP_NOFS.
452
Alan Sterncd38c1e2007-10-10 16:24:56 -0400453Alternatively, if the critical section might call some of the
454usb_autopm_* routines, the driver can avoid deadlock by doing:
455
456 down(&udev->dev.sem);
457 rc = usb_autopm_get_interface(intf);
458
459and at the end of the critical section:
460
461 if (!rc)
462 usb_autopm_put_interface(intf);
463 up(&udev->dev.sem);
464
465Holding the device semaphore will block all external PM calls, and the
466usb_autopm_get_interface() will prevent any internal PM calls, even if
467it fails. (Exercise: Why?)
468
469The rules for locking order are:
470
471 Never acquire any device semaphore while holding any PM mutex.
472
473 Never acquire udev->pm_mutex while holding the PM mutex for
474 a device that isn't a descendant of udev.
475
476In other words, PM mutexes should only be acquired going up the device
477tree, and they should be acquired only after locking all the device
478semaphores you need to hold. These rules don't matter to drivers very
479much; they usually affect just the USB core.
480
481Still, drivers do need to be careful. For example, many drivers use a
482private mutex to synchronize their normal I/O activities with their
483disconnect method. Now if the driver supports autosuspend then it
484must call usb_autopm_put_interface() from somewhere -- maybe from its
485close method. It should make the call while holding the private mutex,
486since a driver shouldn't call any of the usb_autopm_* functions for an
487interface from which it has been unbound.
488
489But the usb_autpm_* routines always acquire the device's PM mutex, and
490consequently the locking order has to be: private mutex first, PM
491mutex second. Since the suspend method is always called with the PM
492mutex held, it mustn't try to acquire the private mutex. It has to
493synchronize with the driver's I/O activities in some other way.
494
495
496 Interaction between dynamic PM and system PM
497 --------------------------------------------
498
499Dynamic power management and system power management can interact in
500a couple of ways.
501
502Firstly, a device may already be manually suspended or autosuspended
503when a system suspend occurs. Since system suspends are supposed to
504be as transparent as possible, the device should remain suspended
505following the system resume. The 2.6.23 kernel obeys this principle
506for manually suspended devices but not for autosuspended devices; they
507do get resumed when the system wakes up. (Presumably they will be
508autosuspended again after their idle-delay time expires.) In later
509kernels this behavior will be fixed.
510
511(There is an exception. If a device would undergo a reset-resume
512instead of a normal resume, and the device is enabled for remote
513wakeup, then the reset-resume takes place even if the device was
514already suspended when the system suspend began. The justification is
515that a reset-resume is a kind of remote-wakeup event. Or to put it
516another way, a device which needs a reset won't be able to generate
517normal remote-wakeup signals, so it ought to be resumed immediately.)
518
519Secondly, a dynamic power-management event may occur as a system
520suspend is underway. The window for this is short, since system
521suspends don't take long (a few seconds usually), but it can happen.
522For example, a suspended device may send a remote-wakeup signal while
523the system is suspending. The remote wakeup may succeed, which would
524cause the system suspend to abort. If the remote wakeup doesn't
525succeed, it may still remain active and thus cause the system to
526resume as soon as the system suspend is complete. Or the remote
527wakeup may fail and get lost. Which outcome occurs depends on timing
528and on the hardware and firmware design.