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 ,
44 .Nm mtx_trylock_spin ,
45 .Nm mtx_trylock_spin_flags ,
48 .Nm mtx_unlock_flags ,
49 .Nm mtx_unlock_spin_flags ,
56 .Nd kernel synchronization primitives
62 .Fn mtx_init "struct mtx *mutex" "const char *name" "const char *type" "int opts"
64 .Fn mtx_destroy "struct mtx *mutex"
66 .Fn mtx_lock "struct mtx *mutex"
68 .Fn mtx_lock_spin "struct mtx *mutex"
70 .Fn mtx_lock_flags "struct mtx *mutex" "int flags"
72 .Fn mtx_lock_spin_flags "struct mtx *mutex" "int flags"
74 .Fn mtx_trylock "struct mtx *mutex"
76 .Fn mtx_trylock_flags "struct mtx *mutex" "int flags"
78 .Fn mtx_trylock_spin "struct mtx *mutex"
80 .Fn mtx_trylock_spin_flags "struct mtx *mutex" "int flags"
82 .Fn mtx_unlock "struct mtx *mutex"
84 .Fn mtx_unlock_spin "struct mtx *mutex"
86 .Fn mtx_unlock_flags "struct mtx *mutex" "int flags"
88 .Fn mtx_unlock_spin_flags "struct mtx *mutex" "int flags"
90 .Fn mtx_sleep "void *chan" "struct mtx *mtx" "int priority" "const char *wmesg" "int timo"
92 .Fn mtx_initialized "const struct mtx *mutex"
94 .Fn mtx_owned "const struct mtx *mutex"
96 .Fn mtx_recursed "const struct mtx *mutex"
98 .Cd "options INVARIANTS"
99 .Cd "options INVARIANT_SUPPORT"
101 .Fn mtx_assert "const struct mtx *mutex" "int what"
103 .Fn MTX_SYSINIT "name" "struct mtx *mtx" "const char *description" "int opts"
105 Mutexes are the most basic and primary method of thread synchronization.
106 The major design considerations for mutexes are:
109 Acquiring and releasing uncontested mutexes should be as cheap
112 They must have the information and storage space to support
113 priority propagation.
115 A thread must be able to recursively acquire a mutex,
116 provided that the mutex is initialized to support recursion.
119 There are currently two flavors of mutexes, those that context switch
120 when they block and those that do not.
124 mutexes will context switch when they are already held.
126 they may spin for some amount
127 of time before context switching.
128 It is important to remember that since a thread may be preempted at any time,
129 the possible context switch introduced by acquiring a mutex is guaranteed
130 to not break anything that is not already broken.
132 Mutexes which do not context switch are
135 These should only be used to protect data shared with primary interrupt
137 This includes interrupt filters and low level scheduling code.
138 In all architectures both acquiring and releasing of a
139 uncontested spin mutex is more expensive than the same operation
141 In order to protect an interrupt service routine from blocking
142 against itself all interrupts are either blocked or deferred on a processor
143 while holding a spin lock.
144 It is permissible to hold multiple spin mutexes.
146 Once a spin mutex has been acquired it is not permissible to acquire a
149 The storage needed to implement a mutex is provided by a
151 In general this should be treated as an opaque object and
152 referenced only with the mutex primitives.
156 function must be used to initialize a mutex
157 before it can be passed to any of the other mutex functions.
160 option is used to identify the lock in debugging output etc.
163 option is used by the witness code to classify a mutex when doing checks
170 is used in its place.
171 The pointer passed in as
175 is saved rather than the data it points to.
176 The data pointed to must remain stable
177 until the mutex is destroyed.
180 argument is used to set the type of mutex.
181 It may contain either
186 See below for additional initialization options.
187 It is not permissible to pass the same
191 multiple times without intervening calls to
198 mutual exclusion lock
199 on behalf of the currently running kernel thread.
200 If another kernel thread is holding the mutex,
201 the caller will be disconnected from the CPU
202 until the mutex is available
203 (i.e., it will block).
209 mutual exclusion lock
210 on behalf of the currently running kernel thread.
211 If another kernel thread is holding the mutex,
212 the caller will spin until the mutex becomes available.
213 Interrupts are disabled during the spin and remain disabled
214 following the acquiring of the lock.
216 It is possible for the same thread to recursively acquire a mutex
217 with no ill effects, provided that the
221 during the initialization of the mutex.
226 .Fn mtx_lock_spin_flags
231 lock, respectively, and also accept a
234 In both cases, the only flags presently available for lock acquires are
240 bit is turned on in the
244 tracing is being done,
245 it will be silenced during the lock acquire.
248 bit is turned on in the
250 argument, then the mutex can be acquired recursively.
256 functions attempt to acquire a
260 mutex, respectively, pointed to by
262 If the mutex cannot be immediately acquired, the functions will return 0,
263 otherwise the mutex will be acquired and a non-zero value will be returned.
266 .Fn mtx_trylock_flags
268 .Fn mtx_trylock_spin_flags
269 functions have the same behavior as
273 respectively, but should be used when the caller desires to pass in a
276 Presently, the only valid value in the
282 and its effects are identical to those described for
290 mutual exclusion lock.
291 The current thread may be preempted if a higher priority thread is waiting
298 mutual exclusion lock.
303 .Fn mtx_unlock_spin_flags
304 functions behave in exactly the same way as do the standard mutex
305 unlock routines above, while also allowing a
307 argument which may specify
311 is identical to its behavior in the mutex lock routines.
315 function is used to destroy
317 so the data associated with it may be freed
318 or otherwise overwritten.
319 Any mutex which is destroyed
320 must previously have been initialized with
322 It is permissible to have a single hold count
323 on a mutex when it is destroyed.
324 It is not permissible to hold the mutex recursively,
325 or have another thread blocked on the mutex
326 when it is destroyed.
330 function is used to atomically release
332 while waiting for an event.
333 For more details on the parameters to this function,
339 function returns non-zero if
341 has been initialized and zero otherwise.
345 function returns non-zero
346 if the current thread holds
348 If the current thread does not hold
354 function returns non-zero if the
357 This check should only be made if the running thread already owns
362 function allows assertions specified in
366 If the assertions are not true and the kernel is compiled with
367 .Cd "options INVARIANTS"
369 .Cd "options INVARIANT_SUPPORT" ,
370 the kernel will panic.
371 Currently the following assertions are supported:
372 .Bl -tag -width MA_NOTRECURSED
374 Assert that the current thread
376 pointed to by the first argument.
378 Assert that the current thread
379 does not hold the mutex
380 pointed to by the first argument.
382 Assert that the current thread has recursed on the mutex
383 pointed to by the first argument.
384 This assertion is only valid in conjunction with
386 .It Dv MA_NOTRECURSED
387 Assert that the current thread has not recursed on the mutex
388 pointed to by the first argument.
389 This assertion is only valid in conjunction with
395 macro is used to generate a call to the
397 routine at system startup in order to initialize a given mutex lock.
398 The parameters are the same as
400 but with an additional argument,
402 that is used in generating unique variable names for the related structures associated with the lock and the sysinit routine.
403 .Ss The Default Mutex Type
404 Most kernel code should use the default lock type,
406 The default lock type will allow the thread
407 to be disconnected from the CPU
408 if the lock is already held by another thread.
410 may treat the lock as a short term spin lock
411 under some circumstances.
412 However, it is always safe to use these forms of locks
413 in an interrupt thread
414 without fear of deadlock
415 against an interrupted thread on the same CPU.
416 .Ss The Spin Mutex Type
419 mutex will not relinquish the CPU
420 when it cannot immediately get the requested lock,
421 but will loop, waiting for the mutex to be released by another CPU.
422 This could result in deadlock
423 if another thread interrupted the thread which held a mutex
424 and then tried to acquire the mutex.
425 For this reason spin locks disable all interrupts on the local CPU.
427 Spin locks are fairly specialized locks
428 that are intended to be held for very short periods of time.
429 Their primary purpose is to protect portions of the code
430 that implement other synchronization primitives such as default mutexes,
431 thread scheduling, and interrupt threads.
432 .Ss Initialization Options
433 The options passed in the
437 specify the mutex type.
442 options is required and only one of those two options may be specified.
443 The possibilities are:
444 .Bl -tag -width MTX_NOWITNESS
447 will always allow the current thread to be suspended
448 to avoid deadlock conditions against interrupt threads.
449 The implementation of this lock type
450 may spin for a while before suspending the current thread.
453 will never relinquish the CPU.
454 All interrupts are disabled on the local CPU
455 while any spin lock is held.
457 Specifies that the initialized mutex is allowed to recurse.
458 This bit must be present if the mutex is permitted to recurse.
465 that is, attempting to acquire an already-owned mutex fails.
467 Do not log any mutex operations for this lock.
473 Witness should not log messages about duplicate locks being acquired.
475 Do not profile this lock.
477 .Ss Lock and Unlock Flags
478 The flags passed to the
480 .Fn mtx_lock_spin_flags ,
481 .Fn mtx_unlock_flags ,
483 .Fn mtx_unlock_spin_flags
484 functions provide some basic options to the caller,
485 and are often used only under special circumstances to modify lock or
487 Standard locking and unlocking should be performed with the
494 Only if a flag is required should the corresponding
495 flags-accepting routines be used.
497 Options that modify mutex behavior:
498 .Bl -tag -width MTX_QUIET
500 This option is used to quiet logging messages during individual mutex
502 This can be used to trim superfluous logging messages for debugging purposes.
507 must be acquired, it must be acquired prior to acquiring
509 Put another way: it is impossible to acquire
511 non-recursively while
512 holding another mutex.
513 It is possible to acquire other mutexes while holding
515 and it is possible to acquire
517 recursively while holding other mutexes.
519 Sleeping while holding a mutex (except for
522 and should be avoided.
523 There are numerous assertions which will fail if this is attempted.
524 .Ss Functions Which Access Memory in Userspace
525 No mutexes should be held (except for
527 across functions which
528 access memory in userspace, such as
534 No locks are needed when calling these functions.
537 .Xr LOCK_PROFILING 9 ,
547 functions appeared in
553 function was added in