2 .\" Copyright (c) 1998 Berkeley Software Design, Inc. All rights reserved.
4 .\" Redistribution and use in source and binary forms, with or without
5 .\" modification, are permitted provided that the following conditions
7 .\" 1. Redistributions of source code must retain the above copyright
8 .\" notice, this list of conditions and the following disclaimer.
9 .\" 2. Redistributions in binary form must reproduce the above copyright
10 .\" notice, this list of conditions and the following disclaimer in the
11 .\" documentation and/or other materials provided with the distribution.
12 .\" 3. Berkeley Software Design Inc's name may not be used to endorse or
13 .\" promote products derived from this software without specific prior
14 .\" written permission.
16 .\" THIS SOFTWARE IS PROVIDED BY BERKELEY SOFTWARE DESIGN INC ``AS IS'' AND
17 .\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 .\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 .\" ARE DISCLAIMED. IN NO EVENT SHALL BERKELEY SOFTWARE DESIGN INC BE LIABLE
20 .\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 .\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 .\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 .\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 .\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28 .\" from BSDI $Id: mutex.4,v 1.1.2.3 1998/04/27 22:53:13 ewv Exp $
41 .Nm mtx_lock_spin_flags ,
43 .Nm mtx_trylock_flags ,
46 .Nm mtx_unlock_flags ,
47 .Nm mtx_unlock_spin_flags ,
54 .Nd kernel synchronization primitives
60 .Fn mtx_init "struct mtx *mutex" "const char *name" "const char *type" "int opts"
62 .Fn mtx_destroy "struct mtx *mutex"
64 .Fn mtx_lock "struct mtx *mutex"
66 .Fn mtx_lock_spin "struct mtx *mutex"
68 .Fn mtx_lock_flags "struct mtx *mutex" "int flags"
70 .Fn mtx_lock_spin_flags "struct mtx *mutex" "int flags"
72 .Fn mtx_trylock "struct mtx *mutex"
74 .Fn mtx_trylock_flags "struct mtx *mutex" "int flags"
76 .Fn mtx_unlock "struct mtx *mutex"
78 .Fn mtx_unlock_spin "struct mtx *mutex"
80 .Fn mtx_unlock_flags "struct mtx *mutex" "int flags"
82 .Fn mtx_unlock_spin_flags "struct mtx *mutex" "int flags"
84 .Fn mtx_sleep "void *chan" "struct mtx *mtx" "int priority" "const char *wmesg" "int timo"
86 .Fn mtx_initialized "const struct mtx *mutex"
88 .Fn mtx_owned "const struct mtx *mutex"
90 .Fn mtx_recursed "const struct mtx *mutex"
92 .Cd "options INVARIANTS"
93 .Cd "options INVARIANT_SUPPORT"
95 .Fn mtx_assert "const struct mtx *mutex" "int what"
97 .Fn MTX_SYSINIT "name" "struct mtx *mtx" "const char *description" "int opts"
99 Mutexes are the most basic and primary method of thread synchronization.
100 The major design considerations for mutexes are:
103 Acquiring and releasing uncontested mutexes should be as cheap
106 They must have the information and storage space to support
107 priority propagation.
109 A thread must be able to recursively acquire a mutex,
110 provided that the mutex is initialized to support recursion.
113 There are currently two flavors of mutexes, those that context switch
114 when they block and those that do not.
118 mutexes will context switch when they are already held.
120 they may spin for some amount
121 of time before context switching.
122 It is important to remember that since a thread may be preempted at any time,
123 the possible context switch introduced by acquiring a mutex is guaranteed
124 to not break anything that is not already broken.
126 Mutexes which do not context switch are
129 These should only be used to protect data shared with primary interrupt
131 This includes interrupt filters and low level scheduling code.
132 In all architectures both acquiring and releasing of a
133 uncontested spin mutex is more expensive than the same operation
135 In order to protect an interrupt service routine from blocking
136 against itself all interrupts are either blocked or deferred on a processor
137 while holding a spin lock.
138 It is permissible to hold multiple spin mutexes.
140 Once a spin mutex has been acquired it is not permissible to acquire a
143 The storage needed to implement a mutex is provided by a
145 In general this should be treated as an opaque object and
146 referenced only with the mutex primitives.
150 function must be used to initialize a mutex
151 before it can be passed to any of the other mutex functions.
154 option is used to identify the lock in debugging output etc.
157 option is used by the witness code to classify a mutex when doing checks
164 is used in its place.
165 The pointer passed in as
169 is saved rather than the data it points to.
170 The data pointed to must remain stable
171 until the mutex is destroyed.
174 argument is used to set the type of mutex.
175 It may contain either
180 If the kernel has been compiled with
181 .Cd "option INVARIANTS" ,
185 has not been initialized multiple times without intervening calls to
190 See below for additional initialization options.
196 mutual exclusion lock
197 on behalf of the currently running kernel thread.
198 If another kernel thread is holding the mutex,
199 the caller will be disconnected from the CPU
200 until the mutex is available
201 (i.e., it will block).
207 mutual exclusion lock
208 on behalf of the currently running kernel thread.
209 If another kernel thread is holding the mutex,
210 the caller will spin until the mutex becomes available.
211 Interrupts are disabled during the spin and remain disabled
212 following the acquiring of the lock.
214 It is possible for the same thread to recursively acquire a mutex
215 with no ill effects, provided that the
219 during the initialization of the mutex.
224 .Fn mtx_lock_spin_flags
229 lock, respectively, and also accept a
232 In both cases, the only flags presently available for lock acquires are
238 bit is turned on in the
242 tracing is being done,
243 it will be silenced during the lock acquire.
246 bit is turned on in the
248 argument, then the mutex can be acquired recursively.
252 attempts to acquire the
256 If the mutex cannot be immediately acquired
259 otherwise the mutex will be acquired
260 and a non-zero value will be returned.
263 .Fn mtx_trylock_flags
264 function has the same behavior as
266 but should be used when the caller desires to pass in a
269 Presently, the only valid value in the
273 and its effects are identical to those described for
281 mutual exclusion lock.
282 The current thread may be preempted if a higher priority thread is waiting
289 mutual exclusion lock.
294 .Fn mtx_unlock_spin_flags
295 functions behave in exactly the same way as do the standard mutex
296 unlock routines above, while also allowing a
298 argument which may specify
302 is identical to its behavior in the mutex lock routines.
306 function is used to destroy
308 so the data associated with it may be freed
309 or otherwise overwritten.
310 Any mutex which is destroyed
311 must previously have been initialized with
313 It is permissible to have a single hold count
314 on a mutex when it is destroyed.
315 It is not permissible to hold the mutex recursively,
316 or have another thread blocked on the mutex
317 when it is destroyed.
321 function is used to atomically release
323 while waiting for an event.
324 For more details on the parameters to this function,
330 function returns non-zero if
332 has been initialized and zero otherwise.
336 function returns non-zero
337 if the current thread holds
339 If the current thread does not hold
345 function returns non-zero if the
348 This check should only be made if the running thread already owns
353 function allows assertions specified in
357 If the assertions are not true and the kernel is compiled with
358 .Cd "options INVARIANTS"
360 .Cd "options INVARIANT_SUPPORT" ,
361 the kernel will panic.
362 Currently the following assertions are supported:
363 .Bl -tag -width MA_NOTRECURSED
365 Assert that the current thread
367 pointed to by the first argument.
369 Assert that the current thread
370 does not hold the mutex
371 pointed to by the first argument.
373 Assert that the current thread has recursed on the mutex
374 pointed to by the first argument.
375 This assertion is only valid in conjunction with
377 .It Dv MA_NOTRECURSED
378 Assert that the current thread has not recursed on the mutex
379 pointed to by the first argument.
380 This assertion is only valid in conjunction with
386 macro is used to generate a call to the
388 routine at system startup in order to initialize a given mutex lock.
389 The parameters are the same as
391 but with an additional argument,
393 that is used in generating unique variable names for the related structures associated with the lock and the sysinit routine.
394 .Ss The Default Mutex Type
395 Most kernel code should use the default lock type,
397 The default lock type will allow the thread
398 to be disconnected from the CPU
399 if the lock is already held by another thread.
401 may treat the lock as a short term spin lock
402 under some circumstances.
403 However, it is always safe to use these forms of locks
404 in an interrupt thread
405 without fear of deadlock
406 against an interrupted thread on the same CPU.
407 .Ss The Spin Mutex Type
410 mutex will not relinquish the CPU
411 when it cannot immediately get the requested lock,
412 but will loop, waiting for the mutex to be released by another CPU.
413 This could result in deadlock
414 if another thread interrupted the thread which held a mutex
415 and then tried to acquire the mutex.
416 For this reason spin locks disable all interrupts on the local CPU.
418 Spin locks are fairly specialized locks
419 that are intended to be held for very short periods of time.
420 Their primary purpose is to protect portions of the code
421 that implement other synchronization primitives such as default mutexes,
422 thread scheduling, and interrupt threads.
423 .Ss Initialization Options
424 The options passed in the
428 specify the mutex type.
433 options is required and only one of those two options may be specified.
434 The possibilities are:
435 .Bl -tag -width MTX_NOWITNESS
438 will always allow the current thread to be suspended
439 to avoid deadlock conditions against interrupt threads.
440 The implementation of this lock type
441 may spin for a while before suspending the current thread.
444 will never relinquish the CPU.
445 All interrupts are disabled on the local CPU
446 while any spin lock is held.
448 Specifies that the initialized mutex is allowed to recurse.
449 This bit must be present if the mutex is permitted to recurse.
451 Do not log any mutex operations for this lock.
457 Witness should not log messages about duplicate locks being acquired.
459 Do not profile this lock.
461 Do not check for double-init.
463 .Ss Lock and Unlock Flags
464 The flags passed to the
466 .Fn mtx_lock_spin_flags ,
467 .Fn mtx_unlock_flags ,
469 .Fn mtx_unlock_spin_flags
470 functions provide some basic options to the caller,
471 and are often used only under special circumstances to modify lock or
473 Standard locking and unlocking should be performed with the
480 Only if a flag is required should the corresponding
481 flags-accepting routines be used.
483 Options that modify mutex behavior:
484 .Bl -tag -width MTX_QUIET
486 This option is used to quiet logging messages during individual mutex
488 This can be used to trim superfluous logging messages for debugging purposes.
493 must be acquired, it must be acquired prior to acquiring
495 Put another way: it is impossible to acquire
497 non-recursively while
498 holding another mutex.
499 It is possible to acquire other mutexes while holding
501 and it is possible to acquire
503 recursively while holding other mutexes.
505 Sleeping while holding a mutex (except for
508 and should be avoided.
509 There are numerous assertions which will fail if this is attempted.
510 .Ss Functions Which Access Memory in Userspace
511 No mutexes should be held (except for
513 across functions which
514 access memory in userspace, such as
520 No locks are needed when calling these functions.
523 .Xr LOCK_PROFILING 9 ,
533 functions appeared in