Based on kernel version 2.6.26. Page generated on 2008-07-16 21:13 EST.
1 Power Management for USB 2 3 Alan Stern <stern[AT]rowland.harvard[DOT]edu> 4 5 October 5, 2007 6 7 8 9 What is Power Management? 10 ------------------------- 11 12 Power Management (PM) is the practice of saving energy by suspending 13 parts of a computer system when they aren't being used. While a 14 component is "suspended" it is in a nonfunctional low-power state; it 15 might even be turned off completely. A suspended component can be 16 "resumed" (returned to a functional full-power state) when the kernel 17 needs to use it. (There also are forms of PM in which components are 18 placed in a less functional but still usable state instead of being 19 suspended; an example would be reducing the CPU's clock rate. This 20 document will not discuss those other forms.) 21 22 When the parts being suspended include the CPU and most of the rest of 23 the system, we speak of it as a "system suspend". When a particular 24 device is turned off while the system as a whole remains running, we 25 call it a "dynamic suspend" (also known as a "runtime suspend" or 26 "selective suspend"). This document concentrates mostly on how 27 dynamic PM is implemented in the USB subsystem, although system PM is 28 covered to some extent (see Documentation/power/*.txt for more 29 information about system PM). 30 31 Note: Dynamic PM support for USB is present only if the kernel was 32 built with CONFIG_USB_SUSPEND enabled. System PM support is present 33 only if the kernel was built with CONFIG_SUSPEND or CONFIG_HIBERNATION 34 enabled. 35 36 37 What is Remote Wakeup? 38 ---------------------- 39 40 When a device has been suspended, it generally doesn't resume until 41 the computer tells it to. Likewise, if the entire computer has been 42 suspended, it generally doesn't resume until the user tells it to, say 43 by pressing a power button or opening the cover. 44 45 However some devices have the capability of resuming by themselves, or 46 asking the kernel to resume them, or even telling the entire computer 47 to resume. This capability goes by several names such as "Wake On 48 LAN"; we will refer to it generically as "remote wakeup". When a 49 device is enabled for remote wakeup and it is suspended, it may resume 50 itself (or send a request to be resumed) in response to some external 51 event. Examples include a suspended keyboard resuming when a key is 52 pressed, or a suspended USB hub resuming when a device is plugged in. 53 54 55 When is a USB device idle? 56 -------------------------- 57 58 A device is idle whenever the kernel thinks it's not busy doing 59 anything important and thus is a candidate for being suspended. The 60 exact definition depends on the device's driver; drivers are allowed 61 to declare that a device isn't idle even when there's no actual 62 communication taking place. (For example, a hub isn't considered idle 63 unless all the devices plugged into that hub are already suspended.) 64 In addition, a device isn't considered idle so long as a program keeps 65 its usbfs file open, whether or not any I/O is going on. 66 67 If a USB device has no driver, its usbfs file isn't open, and it isn't 68 being accessed through sysfs, then it definitely is idle. 69 70 71 Forms of dynamic PM 72 ------------------- 73 74 Dynamic suspends can occur in two ways: manual and automatic. 75 "Manual" means that the user has told the kernel to suspend a device, 76 whereas "automatic" means that the kernel has decided all by itself to 77 suspend a device. Automatic suspend is called "autosuspend" for 78 short. In general, a device won't be autosuspended unless it has been 79 idle for some minimum period of time, the so-called idle-delay time. 80 81 Of course, nothing the kernel does on its own initiative should 82 prevent the computer or its devices from working properly. If a 83 device has been autosuspended and a program tries to use it, the 84 kernel will automatically resume the device (autoresume). For the 85 same reason, an autosuspended device will usually have remote wakeup 86 enabled, if the device supports remote wakeup. 87 88 It is worth mentioning that many USB drivers don't support 89 autosuspend. In fact, at the time of this writing (Linux 2.6.23) the 90 only drivers which do support it are the hub driver, kaweth, asix, 91 usblp, usblcd, and usb-skeleton (which doesn't count). If a 92 non-supporting driver is bound to a device, the device won't be 93 autosuspended. In effect, the kernel pretends the device is never 94 idle. 95 96 We can categorize power management events in two broad classes: 97 external and internal. External events are those triggered by some 98 agent outside the USB stack: system suspend/resume (triggered by 99 userspace), manual dynamic suspend/resume (also triggered by 100 userspace), and remote wakeup (triggered by the device). Internal 101 events are those triggered within the USB stack: autosuspend and 102 autoresume. 103 104 105 The user interface for dynamic PM 106 --------------------------------- 107 108 The user interface for controlling dynamic PM is located in the power/ 109 subdirectory of each USB device's sysfs directory, that is, in 110 /sys/bus/usb/devices/.../power/ where "..." is the device's ID. The 111 relevant attribute files are: wakeup, level, and autosuspend. 112 113 power/wakeup 114 115 This file is empty if the device does not support 116 remote wakeup. Otherwise the file contains either the 117 word "enabled" or the word "disabled", and you can 118 write those words to the file. The setting determines 119 whether or not remote wakeup will be enabled when the 120 device is next suspended. (If the setting is changed 121 while the device is suspended, the change won't take 122 effect until the following suspend.) 123 124 power/level 125 126 This file contains one of three words: "on", "auto", 127 or "suspend". You can write those words to the file 128 to change the device's setting. 129 130 "on" means that the device should be resumed and 131 autosuspend is not allowed. (Of course, system 132 suspends are still allowed.) 133 134 "auto" is the normal state in which the kernel is 135 allowed to autosuspend and autoresume the device. 136 137 "suspend" means that the device should remain 138 suspended, and autoresume is not allowed. (But remote 139 wakeup may still be allowed, since it is controlled 140 separately by the power/wakeup attribute.) 141 142 power/autosuspend 143 144 This file contains an integer value, which is the 145 number of seconds the device should remain idle before 146 the kernel will autosuspend it (the idle-delay time). 147 The default is 2. 0 means to autosuspend as soon as 148 the device becomes idle, and -1 means never to 149 autosuspend. You can write a number to the file to 150 change the autosuspend idle-delay time. 151 152 Writing "-1" to power/autosuspend and writing "on" to power/level do 153 essentially the same thing -- they both prevent the device from being 154 autosuspended. Yes, this is a redundancy in the API. 155 156 (In 2.6.21 writing "0" to power/autosuspend would prevent the device 157 from being autosuspended; the behavior was changed in 2.6.22. The 158 power/autosuspend attribute did not exist prior to 2.6.21, and the 159 power/level attribute did not exist prior to 2.6.22.) 160 161 162 Changing the default idle-delay time 163 ------------------------------------ 164 165 The default autosuspend idle-delay time is controlled by a module 166 parameter in usbcore. You can specify the value when usbcore is 167 loaded. For example, to set it to 5 seconds instead of 2 you would 168 do: 169 170 modprobe usbcore autosuspend=5 171 172 Equivalently, you could add to /etc/modprobe.conf a line saying: 173 174 options usbcore autosuspend=5 175 176 Some distributions load the usbcore module very early during the boot 177 process, by means of a program or script running from an initramfs 178 image. To alter the parameter value you would have to rebuild that 179 image. 180 181 If usbcore is compiled into the kernel rather than built as a loadable 182 module, you can add 183 184 usbcore.autosuspend=5 185 186 to the kernel's boot command line. 187 188 Finally, the parameter value can be changed while the system is 189 running. If you do: 190 191 echo 5 >/sys/module/usbcore/parameters/autosuspend 192 193 then each new USB device will have its autosuspend idle-delay 194 initialized to 5. (The idle-delay values for already existing devices 195 will not be affected.) 196 197 Setting the initial default idle-delay to -1 will prevent any 198 autosuspend of any USB device. This is a simple alternative to 199 disabling CONFIG_USB_SUSPEND and rebuilding the kernel, and it has the 200 added benefit of allowing you to enable autosuspend for selected 201 devices. 202 203 204 Warnings 205 -------- 206 207 The USB specification states that all USB devices must support power 208 management. Nevertheless, the sad fact is that many devices do not 209 support it very well. You can suspend them all right, but when you 210 try to resume them they disconnect themselves from the USB bus or 211 they stop working entirely. This seems to be especially prevalent 212 among printers and scanners, but plenty of other types of device have 213 the same deficiency. 214 215 For this reason, by default the kernel disables autosuspend (the 216 power/level attribute is initialized to "on") for all devices other 217 than hubs. Hubs, at least, appear to be reasonably well-behaved in 218 this regard. 219 220 (In 2.6.21 and 2.6.22 this wasn't the case. Autosuspend was enabled 221 by default for almost all USB devices. A number of people experienced 222 problems as a result.) 223 224 This means that non-hub devices won't be autosuspended unless the user 225 or a program explicitly enables it. As of this writing there aren't 226 any widespread programs which will do this; we hope that in the near 227 future device managers such as HAL will take on this added 228 responsibility. In the meantime you can always carry out the 229 necessary operations by hand or add them to a udev script. You can 230 also change the idle-delay time; 2 seconds is not the best choice for 231 every device. 232 233 Sometimes it turns out that even when a device does work okay with 234 autosuspend there are still problems. For example, there are 235 experimental patches adding autosuspend support to the usbhid driver, 236 which manages keyboards and mice, among other things. Tests with a 237 number of keyboards showed that typing on a suspended keyboard, while 238 causing the keyboard to do a remote wakeup all right, would 239 nonetheless frequently result in lost keystrokes. Tests with mice 240 showed that some of them would issue a remote-wakeup request in 241 response to button presses but not to motion, and some in response to 242 neither. 243 244 The kernel will not prevent you from enabling autosuspend on devices 245 that can't handle it. It is even possible in theory to damage a 246 device by suspending it at the wrong time -- for example, suspending a 247 USB hard disk might cause it to spin down without parking the heads. 248 (Highly unlikely, but possible.) Take care. 249 250 251 The driver interface for Power Management 252 ----------------------------------------- 253 254 The requirements for a USB driver to support external power management 255 are pretty modest; the driver need only define 256 257 .suspend 258 .resume 259 .reset_resume 260 261 methods in its usb_driver structure, and the reset_resume method is 262 optional. The methods' jobs are quite simple: 263 264 The suspend method is called to warn the driver that the 265 device is going to be suspended. If the driver returns a 266 negative error code, the suspend will be aborted. Normally 267 the driver will return 0, in which case it must cancel all 268 outstanding URBs (usb_kill_urb()) and not submit any more. 269 270 The resume method is called to tell the driver that the 271 device has been resumed and the driver can return to normal 272 operation. URBs may once more be submitted. 273 274 The reset_resume method is called to tell the driver that 275 the device has been resumed and it also has been reset. 276 The driver should redo any necessary device initialization, 277 since the device has probably lost most or all of its state 278 (although the interfaces will be in the same altsettings as 279 before the suspend). 280 281 If the device is disconnected or powered down while it is suspended, 282 the disconnect method will be called instead of the resume or 283 reset_resume method. This is also quite likely to happen when 284 waking up from hibernation, as many systems do not maintain suspend 285 current to the USB host controllers during hibernation. (It's 286 possible to work around the hibernation-forces-disconnect problem by 287 using the USB Persist facility.) 288 289 The reset_resume method is used by the USB Persist facility (see 290 Documentation/usb/persist.txt) and it can also be used under certain 291 circumstances when CONFIG_USB_PERSIST is not enabled. Currently, if a 292 device is reset during a resume and the driver does not have a 293 reset_resume method, the driver won't receive any notification about 294 the resume. Later kernels will call the driver's disconnect method; 295 2.6.23 doesn't do this. 296 297 USB drivers are bound to interfaces, so their suspend and resume 298 methods get called when the interfaces are suspended or resumed. In 299 principle one might want to suspend some interfaces on a device (i.e., 300 force the drivers for those interface to stop all activity) without 301 suspending the other interfaces. The USB core doesn't allow this; all 302 interfaces are suspended when the device itself is suspended and all 303 interfaces are resumed when the device is resumed. It isn't possible 304 to suspend or resume some but not all of a device's interfaces. The 305 closest you can come is to unbind the interfaces' drivers. 306 307 308 The driver interface for autosuspend and autoresume 309 --------------------------------------------------- 310 311 To support autosuspend and autoresume, a driver should implement all 312 three of the methods listed above. In addition, a driver indicates 313 that it supports autosuspend by setting the .supports_autosuspend flag 314 in its usb_driver structure. It is then responsible for informing the 315 USB core whenever one of its interfaces becomes busy or idle. The 316 driver does so by calling these three functions: 317 318 int usb_autopm_get_interface(struct usb_interface *intf); 319 void usb_autopm_put_interface(struct usb_interface *intf); 320 int usb_autopm_set_interface(struct usb_interface *intf); 321 322 The functions work by maintaining a counter in the usb_interface 323 structure. When intf->pm_usage_count is > 0 then the interface is 324 deemed to be busy, and the kernel will not autosuspend the interface's 325 device. When intf->pm_usage_count is <= 0 then the interface is 326 considered to be idle, and the kernel may autosuspend the device. 327 328 (There is a similar pm_usage_count field in struct usb_device, 329 associated with the device itself rather than any of its interfaces. 330 This field is used only by the USB core.) 331 332 The driver owns intf->pm_usage_count; it can modify the value however 333 and whenever it likes. A nice aspect of the usb_autopm_* routines is 334 that the changes they make are protected by the usb_device structure's 335 PM mutex (udev->pm_mutex); however drivers may change pm_usage_count 336 without holding the mutex. 337 338 usb_autopm_get_interface() increments pm_usage_count and 339 attempts an autoresume if the new value is > 0 and the 340 device is suspended. 341 342 usb_autopm_put_interface() decrements pm_usage_count and 343 attempts an autosuspend if the new value is <= 0 and the 344 device isn't suspended. 345 346 usb_autopm_set_interface() leaves pm_usage_count alone. 347 It attempts an autoresume if the value is > 0 and the device 348 is suspended, and it attempts an autosuspend if the value is 349 <= 0 and the device isn't suspended. 350 351 There also are a couple of utility routines drivers can use: 352 353 usb_autopm_enable() sets pm_usage_cnt to 1 and then calls 354 usb_autopm_set_interface(), which will attempt an autoresume. 355 356 usb_autopm_disable() sets pm_usage_cnt to 0 and then calls 357 usb_autopm_set_interface(), which will attempt an autosuspend. 358 359 The conventional usage pattern is that a driver calls 360 usb_autopm_get_interface() in its open routine and 361 usb_autopm_put_interface() in its close or release routine. But 362 other patterns are possible. 363 364 The autosuspend attempts mentioned above will often fail for one 365 reason or another. For example, the power/level attribute might be 366 set to "on", or another interface in the same device might not be 367 idle. This is perfectly normal. If the reason for failure was that 368 the device hasn't been idle for long enough, a delayed workqueue 369 routine is automatically set up to carry out the operation when the 370 autosuspend idle-delay has expired. 371 372 Autoresume attempts also can fail. This will happen if power/level is 373 set to "suspend" or if the device doesn't manage to resume properly. 374 Unlike autosuspend, there's no delay for an autoresume. 375 376 377 Other parts of the driver interface 378 ----------------------------------- 379 380 Sometimes a driver needs to make sure that remote wakeup is enabled 381 during autosuspend. For example, there's not much point 382 autosuspending a keyboard if the user can't cause the keyboard to do a 383 remote wakeup by typing on it. If the driver sets 384 intf->needs_remote_wakeup to 1, the kernel won't autosuspend the 385 device if remote wakeup isn't available or has been disabled through 386 the power/wakeup attribute. (If the device is already autosuspended, 387 though, setting this flag won't cause the kernel to autoresume it. 388 Normally a driver would set this flag in its probe method, at which 389 time the device is guaranteed not to be autosuspended.) 390 391 The usb_autopm_* routines have to run in a sleepable process context; 392 they must not be called from an interrupt handler or while holding a 393 spinlock. In fact, the entire autosuspend mechanism is not well geared 394 toward interrupt-driven operation. However there is one thing a 395 driver can do in an interrupt handler: 396 397 usb_mark_last_busy(struct usb_device *udev); 398 399 This sets udev->last_busy to the current time. udev->last_busy is the 400 field used for idle-delay calculations; updating it will cause any 401 pending autosuspend to be moved back. The usb_autopm_* routines will 402 also set the last_busy field to the current time. 403 404 Calling urb_mark_last_busy() from within an URB completion handler is 405 subject to races: The kernel may have just finished deciding the 406 device has been idle for long enough but not yet gotten around to 407 calling the driver's suspend method. The driver would have to be 408 responsible for synchronizing its suspend method with its URB 409 completion handler and causing the autosuspend to fail with -EBUSY if 410 an URB had completed too recently. 411 412 External suspend calls should never be allowed to fail in this way, 413 only autosuspend calls. The driver can tell them apart by checking 414 udev->auto_pm; this flag will be set to 1 for internal PM events 415 (autosuspend or autoresume) and 0 for external PM events. 416 417 Many of the ingredients in the autosuspend framework are oriented 418 towards interfaces: The usb_interface structure contains the 419 pm_usage_cnt field, and the usb_autopm_* routines take an interface 420 pointer as their argument. But somewhat confusingly, a few of the 421 pieces (usb_mark_last_busy() and udev->auto_pm) use the usb_device 422 structure instead. Drivers need to keep this straight; they can call 423 interface_to_usbdev() to find the device structure for a given 424 interface. 425 426 427 Locking requirements 428 -------------------- 429 430 All three suspend/resume methods are always called while holding the 431 usb_device's PM mutex. For external events -- but not necessarily for 432 autosuspend or autoresume -- the device semaphore (udev->dev.sem) will 433 also be held. This implies that external suspend/resume events are 434 mutually exclusive with calls to probe, disconnect, pre_reset, and 435 post_reset; the USB core guarantees that this is true of internal 436 suspend/resume events as well. 437 438 If a driver wants to block all suspend/resume calls during some 439 critical section, it can simply acquire udev->pm_mutex. 440 Alternatively, if the critical section might call some of the 441 usb_autopm_* routines, the driver can avoid deadlock by doing: 442 443 down(&udev->dev.sem); 444 rc = usb_autopm_get_interface(intf); 445 446 and at the end of the critical section: 447 448 if (!rc) 449 usb_autopm_put_interface(intf); 450 up(&udev->dev.sem); 451 452 Holding the device semaphore will block all external PM calls, and the 453 usb_autopm_get_interface() will prevent any internal PM calls, even if 454 it fails. (Exercise: Why?) 455 456 The rules for locking order are: 457 458 Never acquire any device semaphore while holding any PM mutex. 459 460 Never acquire udev->pm_mutex while holding the PM mutex for 461 a device that isn't a descendant of udev. 462 463 In other words, PM mutexes should only be acquired going up the device 464 tree, and they should be acquired only after locking all the device 465 semaphores you need to hold. These rules don't matter to drivers very 466 much; they usually affect just the USB core. 467 468 Still, drivers do need to be careful. For example, many drivers use a 469 private mutex to synchronize their normal I/O activities with their 470 disconnect method. Now if the driver supports autosuspend then it 471 must call usb_autopm_put_interface() from somewhere -- maybe from its 472 close method. It should make the call while holding the private mutex, 473 since a driver shouldn't call any of the usb_autopm_* functions for an 474 interface from which it has been unbound. 475 476 But the usb_autpm_* routines always acquire the device's PM mutex, and 477 consequently the locking order has to be: private mutex first, PM 478 mutex second. Since the suspend method is always called with the PM 479 mutex held, it mustn't try to acquire the private mutex. It has to 480 synchronize with the driver's I/O activities in some other way. 481 482 483 Interaction between dynamic PM and system PM 484 -------------------------------------------- 485 486 Dynamic power management and system power management can interact in 487 a couple of ways. 488 489 Firstly, a device may already be manually suspended or autosuspended 490 when a system suspend occurs. Since system suspends are supposed to 491 be as transparent as possible, the device should remain suspended 492 following the system resume. The 2.6.23 kernel obeys this principle 493 for manually suspended devices but not for autosuspended devices; they 494 do get resumed when the system wakes up. (Presumably they will be 495 autosuspended again after their idle-delay time expires.) In later 496 kernels this behavior will be fixed. 497 498 (There is an exception. If a device would undergo a reset-resume 499 instead of a normal resume, and the device is enabled for remote 500 wakeup, then the reset-resume takes place even if the device was 501 already suspended when the system suspend began. The justification is 502 that a reset-resume is a kind of remote-wakeup event. Or to put it 503 another way, a device which needs a reset won't be able to generate 504 normal remote-wakeup signals, so it ought to be resumed immediately.) 505 506 Secondly, a dynamic power-management event may occur as a system 507 suspend is underway. The window for this is short, since system 508 suspends don't take long (a few seconds usually), but it can happen. 509 For example, a suspended device may send a remote-wakeup signal while 510 the system is suspending. The remote wakeup may succeed, which would 511 cause the system suspend to abort. If the remote wakeup doesn't 512 succeed, it may still remain active and thus cause the system to 513 resume as soon as the system suspend is complete. Or the remote 514 wakeup may fail and get lost. Which outcome occurs depends on timing 515 and on the hardware and firmware design. 516 517 More interestingly, a device might undergo a manual resume or 518 autoresume during system suspend. With current kernels this shouldn't 519 happen, because manual resumes must be initiated by userspace and 520 autoresumes happen in response to I/O requests, but all user processes 521 and I/O should be quiescent during a system suspend -- thanks to the 522 freezer. However there are plans to do away with the freezer, which 523 would mean these things would become possible. If and when this comes 524 about, the USB core will carefully arrange matters so that either type 525 of resume will block until the entire system has resumed.