xref: /netbsd-src/share/man/man9/mutex.9 (revision b1c86f5f087524e68db12794ee9c3e3da1ab17a0)
1.\"	$NetBSD: mutex.9,v 1.22 2010/02/16 19:21:30 rmind Exp $
2.\"
3.\" Copyright (c) 2007, 2009 The NetBSD Foundation, Inc.
4.\" All rights reserved.
5.\"
6.\" This code is derived from software contributed to The NetBSD Foundation
7.\" by Andrew Doran.
8.\"
9.\" Redistribution and use in source and binary forms, with or without
10.\" modification, are permitted provided that the following conditions
11.\" are met:
12.\" 1. Redistributions of source code must retain the above copyright
13.\"    notice, this list of conditions and the following disclaimer.
14.\" 2. Redistributions in binary form must reproduce the above copyright
15.\"    notice, this list of conditions and the following disclaimer in the
16.\"    documentation and/or other materials provided with the distribution.
17.\"
18.\" THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
19.\" ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
20.\" TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
21.\" PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
22.\" BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
23.\" CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
24.\" SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
25.\" INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
26.\" CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
27.\" ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
28.\" POSSIBILITY OF SUCH DAMAGE.
29.\"
30.Dd February 16, 2010
31.Dt MUTEX 9
32.Os
33.Sh NAME
34.Nm mutex ,
35.Nm mutex_init ,
36.Nm mutex_destroy ,
37.Nm mutex_enter ,
38.Nm mutex_exit ,
39.Nm mutex_owned ,
40.Nm mutex_spin_enter ,
41.Nm mutex_spin_exit ,
42.Nm mutex_tryenter
43.Nd mutual exclusion primitives
44.Sh SYNOPSIS
45.In sys/mutex.h
46.Ft void
47.Fn mutex_init "kmutex_t *mtx" "kmutex_type_t type" "int ipl"
48.Ft void
49.Fn mutex_destroy "kmutex_t *mtx"
50.Ft void
51.Fn mutex_enter "kmutex_t *mtx"
52.Ft void
53.Fn mutex_exit "kmutex_t *mtx"
54.Ft int
55.Fn mutex_owned "kmutex_t *mtx"
56.Ft void
57.Fn mutex_spin_enter "kmutex_t *mtx"
58.Ft void
59.Fn mutex_spin_exit "kmutex_t *mtx"
60.Ft int
61.Fn mutex_tryenter "kmutex_t *mtx"
62.Pp
63.Cd "options DIAGNOSTIC"
64.Cd "options LOCKDEBUG"
65.Sh DESCRIPTION
66Mutexes are used in the kernel to implement mutual exclusion among LWPs
67(lightweight processes) and interrupt handlers.
68.Pp
69The
70.Vt kmutex_t
71type provides storage for the mutex object.
72This should be treated as an opaque object and not examined directly by
73consumers.
74.Pp
75Mutexes replace the
76.Xr spl 9
77system traditionally used to provide synchronization between interrupt
78handlers and LWPs.
79.Sh OPTIONS
80.Bl -tag -width abcd
81.It Cd "options DIAGNOSTIC"
82.Pp
83Kernels compiled with the
84.Dv DIAGNOSTIC
85option perform basic sanity checks on mutex operations.
86.It Cd "options LOCKDEBUG"
87.Pp
88Kernels compiled with the
89.Dv LOCKDEBUG
90option perform potentially CPU intensive sanity checks
91on mutex operations.
92.El
93.Sh FUNCTIONS
94.Bl -tag -width abcd
95.It Fn mutex_init "mtx" "type" "ipl"
96.Pp
97Dynamically initialize a mutex for use.
98.Pp
99No other operations can be performed on a mutex until it has been initialized.
100Once initialized, all types of mutex are manipulated using the same interface.
101Note that
102.Fn mutex_init
103may block in order to allocate memory.
104.Pp
105The
106.Fa type
107argument must be given as
108.Dv MUTEX_DEFAULT .
109Other constants are defined but are for low-level system use and are not
110an endorsed, stable part of the interface.
111.Pp
112The type of mutex returned depends on the
113.Fa ipl
114argument:
115.Bl -tag -width abcd
116.It IPL_NONE, or one of the IPL_SOFT* constants
117.Pp
118An adaptive mutex will be returned.
119Adaptive mutexes provide mutual exclusion between LWPs,
120and between LWPs and soft interrupt handlers.
121.Pp
122Adaptive mutexes cannot be acquired from a hardware interrupt handler.
123An LWP may either sleep or busy-wait when attempting to acquire
124an adaptive mutex that is already held.
125.It IPL_VM, IPL_SCHED, IPL_HIGH
126.Pp
127A spin mutex will be returned.
128Spin mutexes provide mutual exclusion between LWPs, and between LWPs
129and interrupt handlers.
130.Pp
131The
132.Fa ipl
133argument is used to pass a system interrupt priority level (IPL)
134that will block all interrupt handlers that may try to acquire the mutex.
135.Pp
136LWPs that own spin mutexes may not sleep, and therefore must not
137try to acquire adaptive mutexes or other sleep locks.
138.Pp
139A processor will always busy-wait when attempting to acquire
140a spin mutex that is already held.
141.El
142.Pp
143See
144.Xr spl 9
145for further information on interrupt priority levels (IPLs).
146.Pp
147.It Fn mutex_destroy "mtx"
148.Pp
149Release resources used by a mutex.
150The mutex may not be used after it has been destroyed.
151.Fn mutex_destroy
152may block in order to free memory.
153.It Fn mutex_enter "mtx"
154.Pp
155Acquire a mutex.
156If the mutex is already held, the caller will block and not return until the
157mutex is acquired.
158.Pp
159Mutexes and other types of locks must always be acquired in a
160consistent order with respect to each other.
161Otherwise, the potential for system deadlock exists.
162.Pp
163Adaptive mutexes and other types of lock that can sleep may
164not be acquired while a spin mutex is held by the caller.
165.It Fn mutex_exit "mtx"
166.Pp
167Release a mutex.
168The mutex must have been previously acquired by the caller.
169Mutexes may be released out of order as needed.
170.It Fn mutex_owned "mtx"
171.Pp
172For adaptive mutexes, return non-zero if the current LWP holds the mutex.
173For spin mutexes, return non-zero if the mutex is held, potentially by the
174current processor.
175Otherwise, return zero.
176.Pp
177.Fn mutex_owned
178is provided for making diagnostic checks to verify that a lock is held.
179For example:
180.Bd -literal
181	KASSERT(mutex_owned(\*[Am]driver_lock));
182.Ed
183.Pp
184It should not be used to make locking decisions at run time, or to
185verify that a lock is not held.
186.It Fn mutex_spin_enter "mtx"
187.Pp
188Equivalent to
189.Fn mutex_enter ,
190but may only be used when it is known that
191.Ar mtx
192is a spin mutex.
193On some architectures, this can substantially reduce the cost of acquring
194a spin mutex.
195.It Fn mutex_spin_exit "mtx"
196.Pp
197Equivalent to
198.Fn mutex_exit ,
199but may only be used when it is known that
200.Ar mtx
201is a spin mutex.
202On some architectures, this can substantially reduce the cost of releasing
203a spin mutex.
204.It Fn mutex_tryenter "mtx"
205.Pp
206Try to acquire a mutex, but do not block if the mutex is already held.
207Returns non-zero if the mutex was acquired, or zero if the mutex was
208already held.
209.Pp
210.Fn mutex_tryenter
211can be used as an optimization when acquiring locks in the wrong order.
212For example, in a setting where the convention is that
213.Dv first_lock
214must be acquired before
215.Dv second_lock ,
216the following can be used to optimistically lock in reverse order:
217.Bd -literal
218	/* We hold second_lock, but not first_lock. */
219	KASSERT(mutex_owned(\*[Am]second_lock));
220
221	if (!mutex_tryenter(\*[Am]first_lock)) {
222		/* Failed to get it - lock in the correct order. */
223		mutex_exit(\*[Am]second_lock);
224		mutex_enter(\*[Am]first_lock);
225		mutex_enter(\*[Am]second_lock);
226
227		/*
228		 * We may need to recheck any conditions the code
229		 * path depends on, as we released second_lock
230		 * briefly.
231		 */
232	}
233.Ed
234.El
235.Sh CODE REFERENCES
236This section describes places within the
237.Nx
238source tree where code implementing mutexes can be found.
239All pathnames are relative to
240.Pa /usr/src .
241.Pp
242The core of the mutex implementation is in
243.Pa sys/kern/kern_mutex.c .
244.Pp
245The header file
246.Pa sys/sys/mutex.h
247describes the public interface, and interfaces that machine-dependent
248code must provide to support mutexes.
249.Sh SEE ALSO
250.Xr atomic_ops 3 ,
251.Xr membar_ops 3 ,
252.Xr lockstat 8 ,
253.Xr condvar 9 ,
254.Xr kpreempt 9 ,
255.Xr rwlock 9 ,
256.Xr spl 9
257.Pp
258.Rs
259.%A Jim Mauro
260.%A Richard McDougall
261.%T Solaris Internals: Core Kernel Architecture
262.%I Prentice Hall
263.%D 2001
264.%O ISBN 0-13-022496-0
265.Re
266.Sh HISTORY
267The mutex primitives first appeared in
268.Nx 5.0 .
269