xref: /freebsd/share/man/man9/sx.9 (revision f6c0136c7fb87ab8277221a306291e386fe944fb)
1.\"
2.\" Copyright (C) 2001 Jason Evans <jasone@FreeBSD.org>.  All rights reserved.
3.\"
4.\" Redistribution and use in source and binary forms, with or without
5.\" modification, are permitted provided that the following conditions
6.\" are met:
7.\" 1. Redistributions of source code must retain the above copyright
8.\"    notice(s), this list of conditions and the following disclaimer as
9.\"    the first lines of this file unmodified other than the possible
10.\"    addition of one or more copyright notices.
11.\" 2. Redistributions in binary form must reproduce the above copyright
12.\"    notice(s), this list of conditions and the following disclaimer in the
13.\"    documentation and/or other materials provided with the distribution.
14.\"
15.\" THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDER(S) ``AS IS'' AND ANY
16.\" EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
17.\" WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
18.\" DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY
19.\" DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
20.\" (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
21.\" SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
22.\" CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH
25.\" DAMAGE.
26.\"
27.\" $FreeBSD$
28.\"
29.Dd February 1, 2006
30.Dt SX 9
31.Os
32.Sh NAME
33.Nm sx ,
34.Nm sx_init ,
35.Nm sx_destroy ,
36.Nm sx_slock ,
37.Nm sx_xlock ,
38.Nm sx_try_slock ,
39.Nm sx_try_xlock ,
40.Nm sx_sunlock ,
41.Nm sx_xunlock ,
42.Nm sx_unlock ,
43.Nm sx_try_upgrade ,
44.Nm sx_downgrade ,
45.Nm sx_sleep ,
46.Nm sx_xlocked ,
47.Nm sx_assert ,
48.Nm SX_SYSINIT
49.Nd kernel shared/exclusive lock
50.Sh SYNOPSIS
51.In sys/param.h
52.In sys/lock.h
53.In sys/sx.h
54.Ft void
55.Fn sx_init "struct sx *sx" "const char *description"
56.Ft void
57.Fn sx_destroy "struct sx *sx"
58.Ft void
59.Fn sx_slock "struct sx *sx"
60.Ft void
61.Fn sx_xlock "struct sx *sx"
62.Ft int
63.Fn sx_try_slock "struct sx *sx"
64.Ft int
65.Fn sx_try_xlock "struct sx *sx"
66.Ft void
67.Fn sx_sunlock "struct sx *sx"
68.Ft void
69.Fn sx_xunlock "struct sx *sx"
70.Ft void
71.Fn sx_unlock "struct sx *sx"
72.Ft int
73.Fn sx_try_upgrade "struct sx *sx"
74.Ft void
75.Fn sx_downgrade "struct sx *sx"
76.Ft int
77.Fn sx_sleep "void *chan" "struct sx *sx" "int priority" "const char *wmesg" "int timo"
78.Ft int
79.Fn sx_xlocked "struct sx *sx"
80.Pp
81.Cd "options INVARIANTS"
82.Cd "options INVARIANT_SUPPORT"
83.Ft void
84.Fn sx_assert "struct sx *sx" "int what"
85.In sys/kernel.h
86.Fn SX_SYSINIT "name" "struct sx *sx" "const char *description"
87.Sh DESCRIPTION
88Shared/exclusive locks are used to protect data that are read far more often
89than they are written.
90Mutexes are inherently more efficient than shared/exclusive locks, so
91shared/exclusive locks should be used prudently.
92.Pp
93Shared/exclusive locks are created with
94.Fn sx_init ,
95where
96.Fa sx
97is a pointer to space for a
98.Vt struct sx ,
99and
100.Fa description
101is a pointer to a null-terminated character string that describes the
102shared/exclusive lock.
103Shared/exclusive locks are destroyed with
104.Fn sx_destroy .
105Threads acquire and release a shared lock by calling
106.Fn sx_slock
107or
108.Fn sx_try_slock
109and
110.Fn sx_sunlock
111or
112.Fn sx_unlock .
113Threads acquire and release an exclusive lock by calling
114.Fn sx_xlock
115or
116.Fn sx_try_xlock
117and
118.Fn sx_xunlock
119or
120.Fn sx_unlock .
121A thread can attempt to upgrade a currently held shared lock to an exclusive
122lock by calling
123.Fn sx_try_upgrade .
124A thread that has an exclusive lock can downgrade it to a shared lock by
125calling
126.Fn sx_downgrade .
127.Pp
128.Fn sx_try_slock
129and
130.Fn sx_try_xlock
131will return 0 if the shared/exclusive lock cannot be acquired immediately;
132otherwise the shared/exclusive lock will be acquired and a non-zero value will
133be returned.
134.Pp
135.Fn sx_try_upgrade
136will return 0 if the shared lock cannot be upgraded to an exclusive lock
137immediately; otherwise the exclusive lock will be acquired and a non-zero value
138will be returned.
139.Pp
140A thread can atomically release a shared/exclusive lock while waiting for an
141event by calling
142.Fn sx_sleep .
143For more details on the parameters to this function,
144see
145.Xr sleep 9 .
146.Pp
147When compiled with
148.Cd "options INVARIANTS"
149and
150.Cd "options INVARIANT_SUPPORT" ,
151the
152.Fn sx_assert
153function tests
154.Fa sx
155for the assertions specified in
156.Fa what ,
157and panics if they are not met.
158The following assertions are supported:
159.Bl -tag -width ".Dv SX_UNLOCKED"
160.It Dv SX_LOCKED
161Assert that the current thread has either a shared or an exclusive lock on the
162.Vt sx
163lock pointed to by the first argument.
164.It Dv SX_SLOCKED
165Assert that the current thread has a shared lock on the
166.Vt sx
167lock pointed to by
168the first argument.
169.It Dv SX_XLOCKED
170Assert that the current thread has an exclusive lock on the
171.Vt sx
172lock pointed to
173by the first argument.
174.It Dv SX_UNLOCKED
175Assert that the current thread has no lock on the
176.Vt sx
177lock pointed to
178by the first argument.
179.El
180.Pp
181.Fn sx_xlocked
182will return non-zero if the current thread holds the exclusive lock;
183otherwise, it will return zero.
184.Pp
185For ease of programming,
186.Fn sx_unlock
187is provided as a macro frontend to the respective functions,
188.Fn sx_sunlock
189and
190.Fn sx_xunlock .
191Algorithms that are aware of what state the lock is in should use either
192of the two specific functions for a minor performance benefit.
193.Pp
194The
195.Fn SX_SYSINIT
196macro is used to generate a call to the
197.Fn sx_sysinit
198routine at system startup in order to initialize a given
199.Fa sx
200lock.
201The parameters are the same as
202.Fn sx_init
203but with an additional argument,
204.Fa name ,
205that is used in generating unique variable names for the related
206structures associated with the lock and the sysinit routine.
207.Pp
208A thread may not hold both a shared lock and an exclusive lock on the same
209lock simultaneously;
210attempting to do so will result in deadlock.
211.Sh CONTEXT
212A thread may hold a shared or exclusive lock on an
213.Nm
214lock while sleeping.
215As a result, an
216.Nm
217lock may not be acquired while holding a mutex.
218Otherwise, if one thread slept while holding an
219.Nm
220lock while another thread blocked on the same
221.Nm
222lock after acquiring a mutex, then the second thread would effectively
223end up sleeping while holding a mutex, which is not allowed.
224.Sh SEE ALSO
225.Xr mutex 9 ,
226.Xr panic 9 ,
227.Xr rwlock 9 ,
228.Xr sema 9
229.Sh BUGS
230Currently there is no way to assert that a lock is not held.
231This is not possible in the
232.No non- Ns Dv WITNESS
233case for asserting that this thread
234does not hold a shared lock.
235In the
236.No non- Ns Dv WITNESS
237case, the
238.Dv SX_LOCKED
239and
240.Dv SX_SLOCKED
241assertions merely check that some thread holds a shared lock.
242They do not ensure that the current thread holds a shared lock.
243