xref: /freebsd/share/man/man9/sx.9 (revision 84ee9401a3fc8d3c22424266f421a928989cd692)
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_try_upgrade ,
43.Nm sx_downgrade ,
44.Nm sx_assert ,
45.Nm sx_unlock ,
46.Nm sx_xlocked ,
47.Nm SX_SYSINIT
48.Nd kernel shared/exclusive lock
49.Sh SYNOPSIS
50.In sys/param.h
51.In sys/lock.h
52.In sys/sx.h
53.Ft void
54.Fn sx_init "struct sx *sx" "const char *description"
55.Ft void
56.Fn sx_destroy "struct sx *sx"
57.Ft void
58.Fn sx_slock "struct sx *sx"
59.Ft void
60.Fn sx_xlock "struct sx *sx"
61.Ft int
62.Fn sx_try_slock "struct sx *sx"
63.Ft int
64.Fn sx_try_xlock "struct sx *sx"
65.Ft void
66.Fn sx_sunlock "struct sx *sx"
67.Ft void
68.Fn sx_xunlock "struct sx *sx"
69.Ft int
70.Fn sx_try_upgrade "struct sx *sx"
71.Ft void
72.Fn sx_downgrade "struct sx *sx"
73.Ft void
74.Fn sx_assert "struct sx *sx" "int what"
75.Ft int
76.Fn sx_xlocked "struct sx *sx"
77.\"
78.Ss Nm Ss utility macros
79.Fn sx_unlock "struct sx *sx"
80.Fn SX_SYSINIT "name" "struct sx *sx" "const char *description"
81.\"
82.Ss Kernel options
83.Cd "options INVARIANTS"
84.Cd "options INVARIANT_SUPPORT"
85.Sh DESCRIPTION
86Shared/exclusive locks are used to protect data that are read far more often
87than they are written.
88Mutexes are inherently more efficient than shared/exclusive locks, so
89shared/exclusive locks should be used prudently.
90.Pp
91Shared/exclusive locks are created with
92.Fn sx_init ,
93where
94.Fa sx
95is a pointer to space for a
96.Vt struct sx ,
97and
98.Fa description
99is a pointer to a null-terminated character string that describes the
100shared/exclusive lock.
101Shared/exclusive locks are destroyed with
102.Fn sx_destroy .
103Threads acquire and release a shared lock by calling
104.Fn sx_slock
105or
106.Fn sx_try_slock
107and
108.Fn sx_sunlock
109or
110.Fn sx_unlock .
111Threads acquire and release an exclusive lock by calling
112.Fn sx_xlock
113or
114.Fn sx_try_xlock
115and
116.Fn sx_xunlock
117or
118.Fn sx_unlock .
119A thread can attempt to upgrade a currently held shared lock to an exclusive
120lock by calling
121.Fn sx_try_upgrade .
122A thread that has an exclusive lock can downgrade it to a shared lock by
123calling
124.Fn sx_downgrade .
125.Pp
126.Fn sx_try_slock
127and
128.Fn sx_try_xlock
129will return 0 if the shared/exclusive lock cannot be acquired immediately;
130otherwise the shared/exclusive lock will be acquired and a non-zero value will
131be returned.
132.Pp
133.Fn sx_try_upgrade
134will return 0 if the shared lock cannot be upgraded to an exclusive lock
135immediately; otherwise the exclusive lock will be acquired and a non-zero value
136will be returned.
137.Pp
138When compiled with
139.Cd "options INVARIANTS"
140and
141.Cd "options INVARIANT_SUPPORT" ,
142the
143.Fn sx_assert
144function tests
145.Fa sx
146for the assertions specified in
147.Fa what ,
148and panics if they are not met.
149The following assertions are supported:
150.Bl -tag -width ".Dv SX_UNLOCKED"
151.It Dv SX_LOCKED
152Assert that the current thread has either a shared or an exclusive lock on the
153.Vt sx
154lock pointed to by the first argument.
155.It Dv SX_SLOCKED
156Assert that the current thread has a shared lock on the
157.Vt sx
158lock pointed to by
159the first argument.
160.It Dv SX_XLOCKED
161Assert that the current thread has an exclusive lock on the
162.Vt sx
163lock pointed to
164by the first argument.
165.It Dv SX_UNLOCKED
166Assert that the current thread has no lock on the
167.Vt sx
168lock pointed to
169by the first argument.
170.El
171.Pp
172.Fn sx_xlocked
173will return non-zero if the current process holds the exclusive lock;
174otherwise, it will return zero.
175.Pp
176For ease of programming,
177.Fn sx_unlock
178is provided as a macro frontend to the respective functions,
179.Fn sx_sunlock
180and
181.Fn sx_xunlock .
182Algorithms that are aware of what state the lock is in should use either
183of the two specific functions for a minor performance benefit.
184.Pp
185The
186.Fn SX_SYSINIT
187macro is used to generate a call to the
188.Fn sx_sysinit
189routine at system startup in order to initialize a given
190.Fa sx
191lock.
192The parameters are the same as
193.Fn sx_init
194but with an additional argument,
195.Fa name ,
196that is used in generating unique variable names for the related
197structures associated with the lock and the sysinit routine.
198.Pp
199A thread may not hold both a shared lock and an exclusive lock on the same
200lock simultaneously;
201attempting to do so will result in deadlock.
202.Sh CONTEXT
203A thread may hold a shared or exclusive lock on an
204.Nm
205lock while sleeping.
206As a result, an
207.Nm
208lock may not be acquired while holding a mutex.
209Otherwise, if one thread slept while holding an
210.Nm
211lock while another thread blocked on the same
212.Nm
213lock after acquiring a mutex, then the second thread would effectively
214end up sleeping while holding a mutex, which is not allowed.
215.Sh SEE ALSO
216.Xr mutex 9 ,
217.Xr panic 9 ,
218.Xr rwlock 9 ,
219.Xr sema 9
220.Sh BUGS
221Currently there is no way to assert that a lock is not held.
222This is not possible in the
223.No non- Ns Dv WITNESS
224case for asserting that this thread
225does not hold a shared lock.
226In the
227.No non- Ns Dv WITNESS
228case, the
229.Dv SX_LOCKED
230and
231.Dv SX_SLOCKED
232assertions merely check that some thread holds a shared lock.
233They do not ensure that the current thread holds a shared lock.
234