power-management.txt 21 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483
  1. Power Management for USB
  2. Alan Stern <stern@rowland.harvard.edu>
  3. December 11, 2009
  4. What is Power Management?
  5. -------------------------
  6. Power Management (PM) is the practice of saving energy by suspending
  7. parts of a computer system when they aren't being used. While a
  8. component is "suspended" it is in a nonfunctional low-power state; it
  9. might even be turned off completely. A suspended component can be
  10. "resumed" (returned to a functional full-power state) when the kernel
  11. needs to use it. (There also are forms of PM in which components are
  12. placed in a less functional but still usable state instead of being
  13. suspended; an example would be reducing the CPU's clock rate. This
  14. document will not discuss those other forms.)
  15. When the parts being suspended include the CPU and most of the rest of
  16. the system, we speak of it as a "system suspend". When a particular
  17. device is turned off while the system as a whole remains running, we
  18. call it a "dynamic suspend" (also known as a "runtime suspend" or
  19. "selective suspend"). This document concentrates mostly on how
  20. dynamic PM is implemented in the USB subsystem, although system PM is
  21. covered to some extent (see Documentation/power/*.txt for more
  22. information about system PM).
  23. Note: Dynamic PM support for USB is present only if the kernel was
  24. built with CONFIG_USB_SUSPEND enabled (which depends on
  25. CONFIG_PM_RUNTIME). System PM support is present only if the kernel
  26. was built with CONFIG_SUSPEND or CONFIG_HIBERNATION enabled.
  27. What is Remote Wakeup?
  28. ----------------------
  29. When a device has been suspended, it generally doesn't resume until
  30. the computer tells it to. Likewise, if the entire computer has been
  31. suspended, it generally doesn't resume until the user tells it to, say
  32. by pressing a power button or opening the cover.
  33. However some devices have the capability of resuming by themselves, or
  34. asking the kernel to resume them, or even telling the entire computer
  35. to resume. This capability goes by several names such as "Wake On
  36. LAN"; we will refer to it generically as "remote wakeup". When a
  37. device is enabled for remote wakeup and it is suspended, it may resume
  38. itself (or send a request to be resumed) in response to some external
  39. event. Examples include a suspended keyboard resuming when a key is
  40. pressed, or a suspended USB hub resuming when a device is plugged in.
  41. When is a USB device idle?
  42. --------------------------
  43. A device is idle whenever the kernel thinks it's not busy doing
  44. anything important and thus is a candidate for being suspended. The
  45. exact definition depends on the device's driver; drivers are allowed
  46. to declare that a device isn't idle even when there's no actual
  47. communication taking place. (For example, a hub isn't considered idle
  48. unless all the devices plugged into that hub are already suspended.)
  49. In addition, a device isn't considered idle so long as a program keeps
  50. its usbfs file open, whether or not any I/O is going on.
  51. If a USB device has no driver, its usbfs file isn't open, and it isn't
  52. being accessed through sysfs, then it definitely is idle.
  53. Forms of dynamic PM
  54. -------------------
  55. Dynamic suspends occur when the kernel decides to suspend an idle
  56. device. This is called "autosuspend" for short. In general, a device
  57. won't be autosuspended unless it has been idle for some minimum period
  58. of time, the so-called idle-delay time.
  59. Of course, nothing the kernel does on its own initiative should
  60. prevent the computer or its devices from working properly. If a
  61. device has been autosuspended and a program tries to use it, the
  62. kernel will automatically resume the device (autoresume). For the
  63. same reason, an autosuspended device will usually have remote wakeup
  64. enabled, if the device supports remote wakeup.
  65. It is worth mentioning that many USB drivers don't support
  66. autosuspend. In fact, at the time of this writing (Linux 2.6.23) the
  67. only drivers which do support it are the hub driver, kaweth, asix,
  68. usblp, usblcd, and usb-skeleton (which doesn't count). If a
  69. non-supporting driver is bound to a device, the device won't be
  70. autosuspended. In effect, the kernel pretends the device is never
  71. idle.
  72. We can categorize power management events in two broad classes:
  73. external and internal. External events are those triggered by some
  74. agent outside the USB stack: system suspend/resume (triggered by
  75. userspace), manual dynamic resume (also triggered by userspace), and
  76. remote wakeup (triggered by the device). Internal events are those
  77. triggered within the USB stack: autosuspend and autoresume. Note that
  78. all dynamic suspend events are internal; external agents are not
  79. allowed to issue dynamic suspends.
  80. The user interface for dynamic PM
  81. ---------------------------------
  82. The user interface for controlling dynamic PM is located in the power/
  83. subdirectory of each USB device's sysfs directory, that is, in
  84. /sys/bus/usb/devices/.../power/ where "..." is the device's ID. The
  85. relevant attribute files are: wakeup, level, and autosuspend.
  86. power/wakeup
  87. This file is empty if the device does not support
  88. remote wakeup. Otherwise the file contains either the
  89. word "enabled" or the word "disabled", and you can
  90. write those words to the file. The setting determines
  91. whether or not remote wakeup will be enabled when the
  92. device is next suspended. (If the setting is changed
  93. while the device is suspended, the change won't take
  94. effect until the following suspend.)
  95. power/level
  96. This file contains one of two words: "on" or "auto".
  97. You can write those words to the file to change the
  98. device's setting.
  99. "on" means that the device should be resumed and
  100. autosuspend is not allowed. (Of course, system
  101. suspends are still allowed.)
  102. "auto" is the normal state in which the kernel is
  103. allowed to autosuspend and autoresume the device.
  104. (In kernels up to 2.6.32, you could also specify
  105. "suspend", meaning that the device should remain
  106. suspended and autoresume was not allowed. This
  107. setting is no longer supported.)
  108. power/autosuspend
  109. This file contains an integer value, which is the
  110. number of seconds the device should remain idle before
  111. the kernel will autosuspend it (the idle-delay time).
  112. The default is 2. 0 means to autosuspend as soon as
  113. the device becomes idle, and negative values mean
  114. never to autosuspend. You can write a number to the
  115. file to change the autosuspend idle-delay time.
  116. Writing "-1" to power/autosuspend and writing "on" to power/level do
  117. essentially the same thing -- they both prevent the device from being
  118. autosuspended. Yes, this is a redundancy in the API.
  119. (In 2.6.21 writing "0" to power/autosuspend would prevent the device
  120. from being autosuspended; the behavior was changed in 2.6.22. The
  121. power/autosuspend attribute did not exist prior to 2.6.21, and the
  122. power/level attribute did not exist prior to 2.6.22.)
  123. Changing the default idle-delay time
  124. ------------------------------------
  125. The default autosuspend idle-delay time is controlled by a module
  126. parameter in usbcore. You can specify the value when usbcore is
  127. loaded. For example, to set it to 5 seconds instead of 2 you would
  128. do:
  129. modprobe usbcore autosuspend=5
  130. Equivalently, you could add to /etc/modprobe.conf a line saying:
  131. options usbcore autosuspend=5
  132. Some distributions load the usbcore module very early during the boot
  133. process, by means of a program or script running from an initramfs
  134. image. To alter the parameter value you would have to rebuild that
  135. image.
  136. If usbcore is compiled into the kernel rather than built as a loadable
  137. module, you can add
  138. usbcore.autosuspend=5
  139. to the kernel's boot command line.
  140. Finally, the parameter value can be changed while the system is
  141. running. If you do:
  142. echo 5 >/sys/module/usbcore/parameters/autosuspend
  143. then each new USB device will have its autosuspend idle-delay
  144. initialized to 5. (The idle-delay values for already existing devices
  145. will not be affected.)
  146. Setting the initial default idle-delay to -1 will prevent any
  147. autosuspend of any USB device. This is a simple alternative to
  148. disabling CONFIG_USB_SUSPEND and rebuilding the kernel, and it has the
  149. added benefit of allowing you to enable autosuspend for selected
  150. devices.
  151. Warnings
  152. --------
  153. The USB specification states that all USB devices must support power
  154. management. Nevertheless, the sad fact is that many devices do not
  155. support it very well. You can suspend them all right, but when you
  156. try to resume them they disconnect themselves from the USB bus or
  157. they stop working entirely. This seems to be especially prevalent
  158. among printers and scanners, but plenty of other types of device have
  159. the same deficiency.
  160. For this reason, by default the kernel disables autosuspend (the
  161. power/level attribute is initialized to "on") for all devices other
  162. than hubs. Hubs, at least, appear to be reasonably well-behaved in
  163. this regard.
  164. (In 2.6.21 and 2.6.22 this wasn't the case. Autosuspend was enabled
  165. by default for almost all USB devices. A number of people experienced
  166. problems as a result.)
  167. This means that non-hub devices won't be autosuspended unless the user
  168. or a program explicitly enables it. As of this writing there aren't
  169. any widespread programs which will do this; we hope that in the near
  170. future device managers such as HAL will take on this added
  171. responsibility. In the meantime you can always carry out the
  172. necessary operations by hand or add them to a udev script. You can
  173. also change the idle-delay time; 2 seconds is not the best choice for
  174. every device.
  175. If a driver knows that its device has proper suspend/resume support,
  176. it can enable autosuspend all by itself. For example, the video
  177. driver for a laptop's webcam might do this, since these devices are
  178. rarely used and so should normally be autosuspended.
  179. Sometimes it turns out that even when a device does work okay with
  180. autosuspend there are still problems. For example, there are
  181. experimental patches adding autosuspend support to the usbhid driver,
  182. which manages keyboards and mice, among other things. Tests with a
  183. number of keyboards showed that typing on a suspended keyboard, while
  184. causing the keyboard to do a remote wakeup all right, would
  185. nonetheless frequently result in lost keystrokes. Tests with mice
  186. showed that some of them would issue a remote-wakeup request in
  187. response to button presses but not to motion, and some in response to
  188. neither.
  189. The kernel will not prevent you from enabling autosuspend on devices
  190. that can't handle it. It is even possible in theory to damage a
  191. device by suspending it at the wrong time -- for example, suspending a
  192. USB hard disk might cause it to spin down without parking the heads.
  193. (Highly unlikely, but possible.) Take care.
  194. The driver interface for Power Management
  195. -----------------------------------------
  196. The requirements for a USB driver to support external power management
  197. are pretty modest; the driver need only define
  198. .suspend
  199. .resume
  200. .reset_resume
  201. methods in its usb_driver structure, and the reset_resume method is
  202. optional. The methods' jobs are quite simple:
  203. The suspend method is called to warn the driver that the
  204. device is going to be suspended. If the driver returns a
  205. negative error code, the suspend will be aborted. Normally
  206. the driver will return 0, in which case it must cancel all
  207. outstanding URBs (usb_kill_urb()) and not submit any more.
  208. The resume method is called to tell the driver that the
  209. device has been resumed and the driver can return to normal
  210. operation. URBs may once more be submitted.
  211. The reset_resume method is called to tell the driver that
  212. the device has been resumed and it also has been reset.
  213. The driver should redo any necessary device initialization,
  214. since the device has probably lost most or all of its state
  215. (although the interfaces will be in the same altsettings as
  216. before the suspend).
  217. If the device is disconnected or powered down while it is suspended,
  218. the disconnect method will be called instead of the resume or
  219. reset_resume method. This is also quite likely to happen when
  220. waking up from hibernation, as many systems do not maintain suspend
  221. current to the USB host controllers during hibernation. (It's
  222. possible to work around the hibernation-forces-disconnect problem by
  223. using the USB Persist facility.)
  224. The reset_resume method is used by the USB Persist facility (see
  225. Documentation/usb/persist.txt) and it can also be used under certain
  226. circumstances when CONFIG_USB_PERSIST is not enabled. Currently, if a
  227. device is reset during a resume and the driver does not have a
  228. reset_resume method, the driver won't receive any notification about
  229. the resume. Later kernels will call the driver's disconnect method;
  230. 2.6.23 doesn't do this.
  231. USB drivers are bound to interfaces, so their suspend and resume
  232. methods get called when the interfaces are suspended or resumed. In
  233. principle one might want to suspend some interfaces on a device (i.e.,
  234. force the drivers for those interface to stop all activity) without
  235. suspending the other interfaces. The USB core doesn't allow this; all
  236. interfaces are suspended when the device itself is suspended and all
  237. interfaces are resumed when the device is resumed. It isn't possible
  238. to suspend or resume some but not all of a device's interfaces. The
  239. closest you can come is to unbind the interfaces' drivers.
  240. The driver interface for autosuspend and autoresume
  241. ---------------------------------------------------
  242. To support autosuspend and autoresume, a driver should implement all
  243. three of the methods listed above. In addition, a driver indicates
  244. that it supports autosuspend by setting the .supports_autosuspend flag
  245. in its usb_driver structure. It is then responsible for informing the
  246. USB core whenever one of its interfaces becomes busy or idle. The
  247. driver does so by calling these six functions:
  248. int usb_autopm_get_interface(struct usb_interface *intf);
  249. void usb_autopm_put_interface(struct usb_interface *intf);
  250. int usb_autopm_get_interface_async(struct usb_interface *intf);
  251. void usb_autopm_put_interface_async(struct usb_interface *intf);
  252. void usb_autopm_get_interface_no_resume(struct usb_interface *intf);
  253. void usb_autopm_put_interface_no_suspend(struct usb_interface *intf);
  254. The functions work by maintaining a usage counter in the
  255. usb_interface's embedded device structure. When the counter is > 0
  256. then the interface is deemed to be busy, and the kernel will not
  257. autosuspend the interface's device. When the usage counter is = 0
  258. then the interface is considered to be idle, and the kernel may
  259. autosuspend the device.
  260. (There is a similar usage counter field in struct usb_device,
  261. associated with the device itself rather than any of its interfaces.
  262. This counter is used only by the USB core.)
  263. Drivers need not be concerned about balancing changes to the usage
  264. counter; the USB core will undo any remaining "get"s when a driver
  265. is unbound from its interface. As a corollary, drivers must not call
  266. any of the usb_autopm_* functions after their diconnect() routine has
  267. returned.
  268. Drivers using the async routines are responsible for their own
  269. synchronization and mutual exclusion.
  270. usb_autopm_get_interface() increments the usage counter and
  271. does an autoresume if the device is suspended. If the
  272. autoresume fails, the counter is decremented back.
  273. usb_autopm_put_interface() decrements the usage counter and
  274. attempts an autosuspend if the new value is = 0.
  275. usb_autopm_get_interface_async() and
  276. usb_autopm_put_interface_async() do almost the same things as
  277. their non-async counterparts. The big difference is that they
  278. use a workqueue to do the resume or suspend part of their
  279. jobs. As a result they can be called in an atomic context,
  280. such as an URB's completion handler, but when they return the
  281. device will generally not yet be in the desired state.
  282. usb_autopm_get_interface_no_resume() and
  283. usb_autopm_put_interface_no_suspend() merely increment or
  284. decrement the usage counter; they do not attempt to carry out
  285. an autoresume or an autosuspend. Hence they can be called in
  286. an atomic context.
  287. The simplest usage pattern is that a driver calls
  288. usb_autopm_get_interface() in its open routine and
  289. usb_autopm_put_interface() in its close or release routine. But other
  290. patterns are possible.
  291. The autosuspend attempts mentioned above will often fail for one
  292. reason or another. For example, the power/level attribute might be
  293. set to "on", or another interface in the same device might not be
  294. idle. This is perfectly normal. If the reason for failure was that
  295. the device hasn't been idle for long enough, a timer is scheduled to
  296. carry out the operation automatically when the autosuspend idle-delay
  297. has expired.
  298. Autoresume attempts also can fail, although failure would mean that
  299. the device is no longer present or operating properly. Unlike
  300. autosuspend, there's no idle-delay for an autoresume.
  301. Other parts of the driver interface
  302. -----------------------------------
  303. Drivers can enable autosuspend for their devices by calling
  304. usb_enable_autosuspend(struct usb_device *udev);
  305. in their probe() routine, if they know that the device is capable of
  306. suspending and resuming correctly. This is exactly equivalent to
  307. writing "auto" to the device's power/level attribute. Likewise,
  308. drivers can disable autosuspend by calling
  309. usb_disable_autosuspend(struct usb_device *udev);
  310. This is exactly the same as writing "on" to the power/level attribute.
  311. Sometimes a driver needs to make sure that remote wakeup is enabled
  312. during autosuspend. For example, there's not much point
  313. autosuspending a keyboard if the user can't cause the keyboard to do a
  314. remote wakeup by typing on it. If the driver sets
  315. intf->needs_remote_wakeup to 1, the kernel won't autosuspend the
  316. device if remote wakeup isn't available or has been disabled through
  317. the power/wakeup attribute. (If the device is already autosuspended,
  318. though, setting this flag won't cause the kernel to autoresume it.
  319. Normally a driver would set this flag in its probe method, at which
  320. time the device is guaranteed not to be autosuspended.)
  321. If a driver does its I/O asynchronously in interrupt context, it
  322. should call usb_autopm_get_interface_async() before starting output and
  323. usb_autopm_put_interface_async() when the output queue drains. When
  324. it receives an input event, it should call
  325. usb_mark_last_busy(struct usb_device *udev);
  326. in the event handler. This sets udev->last_busy to the current time.
  327. udev->last_busy is the field used for idle-delay calculations;
  328. updating it will cause any pending autosuspend to be moved back. Most
  329. of the usb_autopm_* routines will also set the last_busy field to the
  330. current time.
  331. Asynchronous operation is always subject to races. For example, a
  332. driver may call one of the usb_autopm_*_interface_async() routines at
  333. a time when the core has just finished deciding the device has been
  334. idle for long enough but not yet gotten around to calling the driver's
  335. suspend method. The suspend method must be responsible for
  336. synchronizing with the output request routine and the URB completion
  337. handler; it should cause autosuspends to fail with -EBUSY if the
  338. driver needs to use the device.
  339. External suspend calls should never be allowed to fail in this way,
  340. only autosuspend calls. The driver can tell them apart by checking
  341. the PM_EVENT_AUTO bit in the message.event argument to the suspend
  342. method; this bit will be set for internal PM events (autosuspend) and
  343. clear for external PM events.
  344. Mutual exclusion
  345. ----------------
  346. For external events -- but not necessarily for autosuspend or
  347. autoresume -- the device semaphore (udev->dev.sem) will be held when a
  348. suspend or resume method is called. This implies that external
  349. suspend/resume events are mutually exclusive with calls to probe,
  350. disconnect, pre_reset, and post_reset; the USB core guarantees that
  351. this is true of autosuspend/autoresume events as well.
  352. If a driver wants to block all suspend/resume calls during some
  353. critical section, the best way is to lock the device and call
  354. usb_autopm_get_interface() (and do the reverse at the end of the
  355. critical section). Holding the device semaphore will block all
  356. external PM calls, and the usb_autopm_get_interface() will prevent any
  357. internal PM calls, even if it fails. (Exercise: Why?)
  358. Interaction between dynamic PM and system PM
  359. --------------------------------------------
  360. Dynamic power management and system power management can interact in
  361. a couple of ways.
  362. Firstly, a device may already be autosuspended when a system suspend
  363. occurs. Since system suspends are supposed to be as transparent as
  364. possible, the device should remain suspended following the system
  365. resume. But this theory may not work out well in practice; over time
  366. the kernel's behavior in this regard has changed.
  367. Secondly, a dynamic power-management event may occur as a system
  368. suspend is underway. The window for this is short, since system
  369. suspends don't take long (a few seconds usually), but it can happen.
  370. For example, a suspended device may send a remote-wakeup signal while
  371. the system is suspending. The remote wakeup may succeed, which would
  372. cause the system suspend to abort. If the remote wakeup doesn't
  373. succeed, it may still remain active and thus cause the system to
  374. resume as soon as the system suspend is complete. Or the remote
  375. wakeup may fail and get lost. Which outcome occurs depends on timing
  376. and on the hardware and firmware design.