• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1rfkill - RF switch subsystem support
2====================================
3
41 Introduction
52 Implementation details
63 Kernel driver guidelines
73.1 wireless device drivers
83.2 platform/switch drivers
93.3 input device drivers
104 Kernel API
115 Userspace support
12
13
141. Introduction:
15
16The rfkill switch subsystem exists to add a generic interface to circuitry that
17can enable or disable the signal output of a wireless *transmitter* of any
18type.  By far, the most common use is to disable radio-frequency transmitters.
19
20Note that disabling the signal output means that the the transmitter is to be
21made to not emit any energy when "blocked".  rfkill is not about blocking data
22transmissions, it is about blocking energy emission.
23
24The rfkill subsystem offers support for keys and switches often found on
25laptops to enable wireless devices like WiFi and Bluetooth, so that these keys
26and switches actually perform an action in all wireless devices of a given type
27attached to the system.
28
29The buttons to enable and disable the wireless transmitters are important in
30situations where the user is for example using his laptop on a location where
31radio-frequency transmitters _must_ be disabled (e.g. airplanes).
32
33Because of this requirement, userspace support for the keys should not be made
34mandatory.  Because userspace might want to perform some additional smarter
35tasks when the key is pressed, rfkill provides userspace the possibility to
36take over the task to handle the key events.
37
38===============================================================================
392: Implementation details
40
41The rfkill subsystem is composed of various components: the rfkill class, the
42rfkill-input module (an input layer handler), and some specific input layer
43events.
44
45The rfkill class provides kernel drivers with an interface that allows them to
46know when they should enable or disable a wireless network device transmitter.
47This is enabled by the CONFIG_RFKILL Kconfig option.
48
49The rfkill class support makes sure userspace will be notified of all state
50changes on rfkill devices through uevents.  It provides a notification chain
51for interested parties in the kernel to also get notified of rfkill state
52changes in other drivers.  It creates several sysfs entries which can be used
53by userspace.  See section "Userspace support".
54
55The rfkill-input module provides the kernel with the ability to implement a
56basic response when the user presses a key or button (or toggles a switch)
57related to rfkill functionality.  It is an in-kernel implementation of default
58policy of reacting to rfkill-related input events and neither mandatory nor
59required for wireless drivers to operate.  It is enabled by the
60CONFIG_RFKILL_INPUT Kconfig option.
61
62rfkill-input is a rfkill-related events input layer handler.  This handler will
63listen to all rfkill key events and will change the rfkill state of the
64wireless devices accordingly.  With this option enabled userspace could either
65do nothing or simply perform monitoring tasks.
66
67The rfkill-input module also provides EPO (emergency power-off) functionality
68for all wireless transmitters.  This function cannot be overridden, and it is
69always active.  rfkill EPO is related to *_RFKILL_ALL input layer events.
70
71
72Important terms for the rfkill subsystem:
73
74In order to avoid confusion, we avoid the term "switch" in rfkill when it is
75referring to an electronic control circuit that enables or disables a
76transmitter.  We reserve it for the physical device a human manipulates
77(which is an input device, by the way):
78
79rfkill switch:
80
81	A physical device a human manipulates.  Its state can be perceived by
82	the kernel either directly (through a GPIO pin, ACPI GPE) or by its
83	effect on a rfkill line of a wireless device.
84
85rfkill controller:
86
87	A hardware circuit that controls the state of a rfkill line, which a
88	kernel driver can interact with *to modify* that state (i.e. it has
89	either write-only or read/write access).
90
91rfkill line:
92
93	An input channel (hardware or software) of a wireless device, which
94	causes a wireless transmitter to stop emitting energy (BLOCK) when it
95	is active.  Point of view is extremely important here: rfkill lines are
96	always seen from the PoV of a wireless device (and its driver).
97
98soft rfkill line/software rfkill line:
99
100	A rfkill line the wireless device driver can directly change the state
101	of.  Related to rfkill_state RFKILL_STATE_SOFT_BLOCKED.
102
103hard rfkill line/hardware rfkill line:
104
105	A rfkill line that works fully in hardware or firmware, and that cannot
106	be overridden by the kernel driver.  The hardware device or the
107	firmware just exports its status to the driver, but it is read-only.
108	Related to rfkill_state RFKILL_STATE_HARD_BLOCKED.
109
110The enum rfkill_state describes the rfkill state of a transmitter:
111
112When a rfkill line or rfkill controller is in the RFKILL_STATE_UNBLOCKED state,
113the wireless transmitter (radio TX circuit for example) is *enabled*.  When the
114it is in the RFKILL_STATE_SOFT_BLOCKED or RFKILL_STATE_HARD_BLOCKED, the
115wireless transmitter is to be *blocked* from operating.
116
117RFKILL_STATE_SOFT_BLOCKED indicates that a call to toggle_radio() can change
118that state.  RFKILL_STATE_HARD_BLOCKED indicates that a call to toggle_radio()
119will not be able to change the state and will return with a suitable error if
120attempts are made to set the state to RFKILL_STATE_UNBLOCKED.
121
122RFKILL_STATE_HARD_BLOCKED is used by drivers to signal that the device is
123locked in the BLOCKED state by a hardwire rfkill line (typically an input pin
124that, when active, forces the transmitter to be disabled) which the driver
125CANNOT override.
126
127Full rfkill functionality requires two different subsystems to cooperate: the
128input layer and the rfkill class.  The input layer issues *commands* to the
129entire system requesting that devices registered to the rfkill class change
130state.  The way this interaction happens is not complex, but it is not obvious
131either:
132
133Kernel Input layer:
134
135	* Generates KEY_WWAN, KEY_WLAN, KEY_BLUETOOTH, SW_RFKILL_ALL, and
136	  other such events when the user presses certain keys, buttons, or
137	  toggles certain physical switches.
138
139	THE INPUT LAYER IS NEVER USED TO PROPAGATE STATUS, NOTIFICATIONS OR THE
140	KIND OF STUFF AN ON-SCREEN-DISPLAY APPLICATION WOULD REPORT.  It is
141	used to issue *commands* for the system to change behaviour, and these
142	commands may or may not be carried out by some kernel driver or
143	userspace application.  It follows that doing user feedback based only
144	on input events is broken, as there is no guarantee that an input event
145	will be acted upon.
146
147	Most wireless communication device drivers implementing rfkill
148	functionality MUST NOT generate these events, and have no reason to
149	register themselves with the input layer.  Doing otherwise is a common
150	misconception.  There is an API to propagate rfkill status change
151	information, and it is NOT the input layer.
152
153rfkill class:
154
155	* Calls a hook in a driver to effectively change the wireless
156	  transmitter state;
157	* Keeps track of the wireless transmitter state (with help from
158	  the driver);
159	* Generates userspace notifications (uevents) and a call to a
160	  notification chain (kernel) when there is a wireless transmitter
161	  state change;
162	* Connects a wireless communications driver with the common rfkill
163	  control system, which, for example, allows actions such as
164	  "switch all bluetooth devices offline" to be carried out by
165	  userspace or by rfkill-input.
166
167	THE RFKILL CLASS NEVER ISSUES INPUT EVENTS.  THE RFKILL CLASS DOES
168	NOT LISTEN TO INPUT EVENTS.  NO DRIVER USING THE RFKILL CLASS SHALL
169	EVER LISTEN TO, OR ACT ON RFKILL INPUT EVENTS.  Doing otherwise is
170	a layering violation.
171
172	Most wireless data communication drivers in the kernel have just to
173	implement the rfkill class API to work properly.  Interfacing to the
174	input layer is not often required (and is very often a *bug*) on
175	wireless drivers.
176
177	Platform drivers often have to attach to the input layer to *issue*
178	(but never to listen to) rfkill events for rfkill switches, and also to
179	the rfkill class to export a control interface for the platform rfkill
180	controllers to the rfkill subsystem.  This does NOT mean the rfkill
181	switch is attached to a rfkill class (doing so is almost always wrong).
182	It just means the same kernel module is the driver for different
183	devices (rfkill switches and rfkill controllers).
184
185
186Userspace input handlers (uevents) or kernel input handlers (rfkill-input):
187
188	* Implements the policy of what should happen when one of the input
189	  layer events related to rfkill operation is received.
190	* Uses the sysfs interface (userspace) or private rfkill API calls
191	  to tell the devices registered with the rfkill class to change
192	  their state (i.e. translates the input layer event into real
193	  action).
194
195	* rfkill-input implements EPO by handling EV_SW SW_RFKILL_ALL 0
196	  (power off all transmitters) in a special way: it ignores any
197	  overrides and local state cache and forces all transmitters to the
198	  RFKILL_STATE_SOFT_BLOCKED state (including those which are already
199	  supposed to be BLOCKED).
200	* rfkill EPO will remain active until rfkill-input receives an
201	  EV_SW SW_RFKILL_ALL 1 event.  While the EPO is active, transmitters
202	  are locked in the blocked state (rfkill will refuse to unblock them).
203	* rfkill-input implements different policies that the user can
204	  select for handling EV_SW SW_RFKILL_ALL 1.  It will unlock rfkill,
205	  and either do nothing (leave transmitters blocked, but now unlocked),
206	  restore the transmitters to their state before the EPO, or unblock
207	  them all.
208
209Userspace uevent handler or kernel platform-specific drivers hooked to the
210rfkill notifier chain:
211
212	* Taps into the rfkill notifier chain or to KOBJ_CHANGE uevents,
213	  in order to know when a device that is registered with the rfkill
214	  class changes state;
215	* Issues feedback notifications to the user;
216	* In the rare platforms where this is required, synthesizes an input
217	  event to command all *OTHER* rfkill devices to also change their
218	  statues when a specific rfkill device changes state.
219
220
221===============================================================================
2223: Kernel driver guidelines
223
224Remember: point-of-view is everything for a driver that connects to the rfkill
225subsystem.  All the details below must be measured/perceived from the point of
226view of the specific driver being modified.
227
228The first thing one needs to know is whether his driver should be talking to
229the rfkill class or to the input layer.  In rare cases (platform drivers), it
230could happen that you need to do both, as platform drivers often handle a
231variety of devices in the same driver.
232
233Do not mistake input devices for rfkill controllers.  The only type of "rfkill
234switch" device that is to be registered with the rfkill class are those
235directly controlling the circuits that cause a wireless transmitter to stop
236working (or the software equivalent of them), i.e. what we call a rfkill
237controller.  Every other kind of "rfkill switch" is just an input device and
238MUST NOT be registered with the rfkill class.
239
240A driver should register a device with the rfkill class when ALL of the
241following conditions are met (they define a rfkill controller):
242
2431. The device is/controls a data communications wireless transmitter;
244
2452. The kernel can interact with the hardware/firmware to CHANGE the wireless
246   transmitter state (block/unblock TX operation);
247
2483. The transmitter can be made to not emit any energy when "blocked":
249   rfkill is not about blocking data transmissions, it is about blocking
250   energy emission;
251
252A driver should register a device with the input subsystem to issue
253rfkill-related events (KEY_WLAN, KEY_BLUETOOTH, KEY_WWAN, KEY_WIMAX,
254SW_RFKILL_ALL, etc) when ALL of the folowing conditions are met:
255
2561. It is directly related to some physical device the user interacts with, to
257   command the O.S./firmware/hardware to enable/disable a data communications
258   wireless transmitter.
259
260   Examples of the physical device are: buttons, keys and switches the user
261   will press/touch/slide/switch to enable or disable the wireless
262   communication device.
263
2642. It is NOT slaved to another device, i.e. there is no other device that
265   issues rfkill-related input events in preference to this one.
266
267   Please refer to the corner cases and examples section for more details.
268
269When in doubt, do not issue input events.  For drivers that should generate
270input events in some platforms, but not in others (e.g. b43), the best solution
271is to NEVER generate input events in the first place.  That work should be
272deferred to a platform-specific kernel module (which will know when to generate
273events through the rfkill notifier chain) or to userspace.  This avoids the
274usual maintenance problems with DMI whitelisting.
275
276
277Corner cases and examples:
278====================================
279
2801. If the device is an input device that, because of hardware or firmware,
281causes wireless transmitters to be blocked regardless of the kernel's will, it
282is still just an input device, and NOT to be registered with the rfkill class.
283
2842. If the wireless transmitter switch control is read-only, it is an input
285device and not to be registered with the rfkill class (and maybe not to be made
286an input layer event source either, see below).
287
2883. If there is some other device driver *closer* to the actual hardware the
289user interacted with (the button/switch/key) to issue an input event, THAT is
290the device driver that should be issuing input events.
291
292E.g:
293  [RFKILL slider switch] -- [GPIO hardware] -- [WLAN card rf-kill input]
294                           (platform driver)    (wireless card driver)
295
296The user is closer to the RFKILL slide switch plaform driver, so the driver
297which must issue input events is the platform driver looking at the GPIO
298hardware, and NEVER the wireless card driver (which is just a slave).  It is
299very likely that there are other leaves than just the WLAN card rf-kill input
300(e.g. a bluetooth card, etc)...
301
302On the other hand, some embedded devices do this:
303
304  [RFKILL slider switch] -- [WLAN card rf-kill input]
305                             (wireless card driver)
306
307In this situation, the wireless card driver *could* register itself as an input
308device and issue rf-kill related input events... but in order to AVOID the need
309for DMI whitelisting, the wireless card driver does NOT do it.  Userspace (HAL)
310or a platform driver (that exists only on these embedded devices) will do the
311dirty job of issuing the input events.
312
313
314COMMON MISTAKES in kernel drivers, related to rfkill:
315====================================
316
3171. NEVER confuse input device keys and buttons with input device switches.
318
319  1a. Switches are always set or reset.  They report the current state
320      (on position or off position).
321
322  1b. Keys and buttons are either in the pressed or not-pressed state, and
323      that's it.  A "button" that latches down when you press it, and
324      unlatches when you press it again is in fact a switch as far as input
325      devices go.
326
327Add the SW_* events you need for switches, do NOT try to emulate a button using
328KEY_* events just because there is no such SW_* event yet.  Do NOT try to use,
329for example, KEY_BLUETOOTH when you should be using SW_BLUETOOTH instead.
330
3312. Input device switches (sources of EV_SW events) DO store their current state
332(so you *must* initialize it by issuing a gratuitous input layer event on
333driver start-up and also when resuming from sleep), and that state CAN be
334queried from userspace through IOCTLs.  There is no sysfs interface for this,
335but that doesn't mean you should break things trying to hook it to the rfkill
336class to get a sysfs interface :-)
337
3383. Do not issue *_RFKILL_ALL events by default, unless you are sure it is the
339correct event for your switch/button.  These events are emergency power-off
340events when they are trying to turn the transmitters off.  An example of an
341input device which SHOULD generate *_RFKILL_ALL events is the wireless-kill
342switch in a laptop which is NOT a hotkey, but a real sliding/rocker switch.
343An example of an input device which SHOULD NOT generate *_RFKILL_ALL events by
344default, is any sort of hot key that is type-specific (e.g. the one for WLAN).
345
346
3473.1 Guidelines for wireless device drivers
348------------------------------------------
349
350(in this text, rfkill->foo means the foo field of struct rfkill).
351
3521. Each independent transmitter in a wireless device (usually there is only one
353transmitter per device) should have a SINGLE rfkill class attached to it.
354
3552. If the device does not have any sort of hardware assistance to allow the
356driver to rfkill the device, the driver should emulate it by taking all actions
357required to silence the transmitter.
358
3593. If it is impossible to silence the transmitter (i.e. it still emits energy,
360even if it is just in brief pulses, when there is no data to transmit and there
361is no hardware support to turn it off) do NOT lie to the users.  Do not attach
362it to a rfkill class.  The rfkill subsystem does not deal with data
363transmission, it deals with energy emission.  If the transmitter is emitting
364energy, it is not blocked in rfkill terms.
365
3664. It doesn't matter if the device has multiple rfkill input lines affecting
367the same transmitter, their combined state is to be exported as a single state
368per transmitter (see rule 1).
369
370This rule exists because users of the rfkill subsystem expect to get (and set,
371when possible) the overall transmitter rfkill state, not of a particular rfkill
372line.
373
3745. The wireless device driver MUST NOT leave the transmitter enabled during
375suspend and hibernation unless:
376
377	5.1. The transmitter has to be enabled for some sort of functionality
378	like wake-on-wireless-packet or autonomous packed forwarding in a mesh
379	network, and that functionality is enabled for this suspend/hibernation
380	cycle.
381
382AND
383
384	5.2. The device was not on a user-requested BLOCKED state before
385	the suspend (i.e. the driver must NOT unblock a device, not even
386	to support wake-on-wireless-packet or remain in the mesh).
387
388In other words, there is absolutely no allowed scenario where a driver can
389automatically take action to unblock a rfkill controller (obviously, this deals
390with scenarios where soft-blocking or both soft and hard blocking is happening.
391Scenarios where hardware rfkill lines are the only ones blocking the
392transmitter are outside of this rule, since the wireless device driver does not
393control its input hardware rfkill lines in the first place).
394
3956. During resume, rfkill will try to restore its previous state.
396
3977. After a rfkill class is suspended, it will *not* call rfkill->toggle_radio
398until it is resumed.
399
400
401Example of a WLAN wireless driver connected to the rfkill subsystem:
402--------------------------------------------------------------------
403
404A certain WLAN card has one input pin that causes it to block the transmitter
405and makes the status of that input pin available (only for reading!) to the
406kernel driver.  This is a hard rfkill input line (it cannot be overridden by
407the kernel driver).
408
409The card also has one PCI register that, if manipulated by the driver, causes
410it to block the transmitter.  This is a soft rfkill input line.
411
412It has also a thermal protection circuitry that shuts down its transmitter if
413the card overheats, and makes the status of that protection available (only for
414reading!) to the kernel driver.  This is also a hard rfkill input line.
415
416If either one of these rfkill lines are active, the transmitter is blocked by
417the hardware and forced offline.
418
419The driver should allocate and attach to its struct device *ONE* instance of
420the rfkill class (there is only one transmitter).
421
422It can implement the get_state() hook, and return RFKILL_STATE_HARD_BLOCKED if
423either one of its two hard rfkill input lines are active.  If the two hard
424rfkill lines are inactive, it must return RFKILL_STATE_SOFT_BLOCKED if its soft
425rfkill input line is active.  Only if none of the rfkill input lines are
426active, will it return RFKILL_STATE_UNBLOCKED.
427
428Since the device has a hardware rfkill line, it IS subject to state changes
429external to rfkill.  Therefore, the driver must make sure that it calls
430rfkill_force_state() to keep the status always up-to-date, and it must do a
431rfkill_force_state() on resume from sleep.
432
433Every time the driver gets a notification from the card that one of its rfkill
434lines changed state (polling might be needed on badly designed cards that don't
435generate interrupts for such events), it recomputes the rfkill state as per
436above, and calls rfkill_force_state() to update it.
437
438The driver should implement the toggle_radio() hook, that:
439
4401. Returns an error if one of the hardware rfkill lines are active, and the
441caller asked for RFKILL_STATE_UNBLOCKED.
442
4432. Activates the soft rfkill line if the caller asked for state
444RFKILL_STATE_SOFT_BLOCKED.  It should do this even if one of the hard rfkill
445lines are active, effectively double-blocking the transmitter.
446
4473. Deactivates the soft rfkill line if none of the hardware rfkill lines are
448active and the caller asked for RFKILL_STATE_UNBLOCKED.
449
450===============================================================================
4514: Kernel API
452
453To build a driver with rfkill subsystem support, the driver should depend on
454(or select) the Kconfig symbol RFKILL; it should _not_ depend on RKFILL_INPUT.
455
456The hardware the driver talks to may be write-only (where the current state
457of the hardware is unknown), or read-write (where the hardware can be queried
458about its current state).
459
460The rfkill class will call the get_state hook of a device every time it needs
461to know the *real* current state of the hardware.  This can happen often, but
462it does not do any polling, so it is not enough on hardware that is subject
463to state changes outside of the rfkill subsystem.
464
465Therefore, calling rfkill_force_state() when a state change happens is
466mandatory when the device has a hardware rfkill line, or when something else
467like the firmware could cause its state to be changed without going through the
468rfkill class.
469
470Some hardware provides events when its status changes.  In these cases, it is
471best for the driver to not provide a get_state hook, and instead register the
472rfkill class *already* with the correct status, and keep it updated using
473rfkill_force_state() when it gets an event from the hardware.
474
475rfkill_force_state() must be used on the device resume handlers to update the
476rfkill status, should there be any chance of the device status changing during
477the sleep.
478
479There is no provision for a statically-allocated rfkill struct.  You must
480use rfkill_allocate() to allocate one.
481
482You should:
483	- rfkill_allocate()
484	- modify rfkill fields (flags, name)
485	- modify state to the current hardware state (THIS IS THE ONLY TIME
486	  YOU CAN ACCESS state DIRECTLY)
487	- rfkill_register()
488
489The only way to set a device to the RFKILL_STATE_HARD_BLOCKED state is through
490a suitable return of get_state() or through rfkill_force_state().
491
492When a device is in the RFKILL_STATE_HARD_BLOCKED state, the only way to switch
493it to a different state is through a suitable return of get_state() or through
494rfkill_force_state().
495
496If toggle_radio() is called to set a device to state RFKILL_STATE_SOFT_BLOCKED
497when that device is already at the RFKILL_STATE_HARD_BLOCKED state, it should
498not return an error.  Instead, it should try to double-block the transmitter,
499so that its state will change from RFKILL_STATE_HARD_BLOCKED to
500RFKILL_STATE_SOFT_BLOCKED should the hardware blocking cease.
501
502Please refer to the source for more documentation.
503
504===============================================================================
5055: Userspace support
506
507rfkill devices issue uevents (with an action of "change"), with the following
508environment variables set:
509
510RFKILL_NAME
511RFKILL_STATE
512RFKILL_TYPE
513
514The ABI for these variables is defined by the sysfs attributes.  It is best
515to take a quick look at the source to make sure of the possible values.
516
517It is expected that HAL will trap those, and bridge them to DBUS, etc.  These
518events CAN and SHOULD be used to give feedback to the user about the rfkill
519status of the system.
520
521Input devices may issue events that are related to rfkill.  These are the
522various KEY_* events and SW_* events supported by rfkill-input.c.
523
524******IMPORTANT******
525When rfkill-input is ACTIVE, userspace is NOT TO CHANGE THE STATE OF AN RFKILL
526SWITCH IN RESPONSE TO AN INPUT EVENT also handled by rfkill-input, unless it
527has set to true the user_claim attribute for that particular switch.  This rule
528is *absolute*; do NOT violate it.
529******IMPORTANT******
530
531Userspace must not assume it is the only source of control for rfkill switches.
532Their state CAN and WILL change due to firmware actions, direct user actions,
533and the rfkill-input EPO override for *_RFKILL_ALL.
534
535When rfkill-input is not active, userspace must initiate a rfkill status
536change by writing to the "state" attribute in order for anything to happen.
537
538Take particular care to implement EV_SW SW_RFKILL_ALL properly.  When that
539switch is set to OFF, *every* rfkill device *MUST* be immediately put into the
540RFKILL_STATE_SOFT_BLOCKED state, no questions asked.
541
542The following sysfs entries will be created:
543
544	name: Name assigned by driver to this key (interface or driver name).
545	type: Name of the key type ("wlan", "bluetooth", etc).
546	state: Current state of the transmitter
547		0: RFKILL_STATE_SOFT_BLOCKED
548			transmitter is forced off, but one can override it
549			by a write to the state attribute;
550		1: RFKILL_STATE_UNBLOCKED
551			transmiter is NOT forced off, and may operate if
552			all other conditions for such operation are met
553			(such as interface is up and configured, etc);
554		2: RFKILL_STATE_HARD_BLOCKED
555			transmitter is forced off by something outside of
556			the driver's control.  One cannot set a device to
557			this state through writes to the state attribute;
558	claim: 1: Userspace handles events, 0: Kernel handles events
559
560Both the "state" and "claim" entries are also writable. For the "state" entry
561this means that when 1 or 0 is written, the device rfkill state (if not yet in
562the requested state), will be will be toggled accordingly.
563
564For the "claim" entry writing 1 to it means that the kernel no longer handles
565key events even though RFKILL_INPUT input was enabled. When "claim" has been
566set to 0, userspace should make sure that it listens for the input events or
567check the sysfs "state" entry regularly to correctly perform the required tasks
568when the rkfill key is pressed.
569
570A note about input devices and EV_SW events:
571
572In order to know the current state of an input device switch (like
573SW_RFKILL_ALL), you will need to use an IOCTL.  That information is not
574available through sysfs in a generic way at this time, and it is not available
575through the rfkill class AT ALL.
576