xref: /freebsd/share/man/man9/sx.9 (revision ec65e4f8d0654361df5e97d4de3518edebf76b46)
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 March 28, 2016
30.Dt SX 9
31.Os
32.Sh NAME
33.Nm sx ,
34.Nm sx_init ,
35.Nm sx_init_flags ,
36.Nm sx_destroy ,
37.Nm sx_slock ,
38.Nm sx_xlock ,
39.Nm sx_slock_sig ,
40.Nm sx_xlock_sig ,
41.Nm sx_try_slock ,
42.Nm sx_try_xlock ,
43.Nm sx_sunlock ,
44.Nm sx_xunlock ,
45.Nm sx_unlock ,
46.Nm sx_try_upgrade ,
47.Nm sx_downgrade ,
48.Nm sx_sleep ,
49.Nm sx_xholder ,
50.Nm sx_xlocked ,
51.Nm sx_assert ,
52.Nm SX_SYSINIT
53.Nd kernel shared/exclusive lock
54.Sh SYNOPSIS
55.In sys/param.h
56.In sys/lock.h
57.In sys/sx.h
58.Ft void
59.Fn sx_init "struct sx *sx" "const char *description"
60.Ft void
61.Fn sx_init_flags "struct sx *sx" "const char *description" "int opts"
62.Ft void
63.Fn sx_destroy "struct sx *sx"
64.Ft void
65.Fn sx_slock "struct sx *sx"
66.Ft void
67.Fn sx_xlock "struct sx *sx"
68.Ft int
69.Fn sx_slock_sig "struct sx *sx"
70.Ft int
71.Fn sx_xlock_sig "struct sx *sx"
72.Ft int
73.Fn sx_try_slock "struct sx *sx"
74.Ft int
75.Fn sx_try_xlock "struct sx *sx"
76.Ft void
77.Fn sx_sunlock "struct sx *sx"
78.Ft void
79.Fn sx_xunlock "struct sx *sx"
80.Ft void
81.Fn sx_unlock "struct sx *sx"
82.Ft int
83.Fn sx_try_upgrade "struct sx *sx"
84.Ft void
85.Fn sx_downgrade "struct sx *sx"
86.Ft int
87.Fn sx_sleep "void *chan" "struct sx *sx" "int priority" "const char *wmesg" "int timo"
88.Ft "struct thread *"
89.Fn sx_xholder "struct sx *sx"
90.Ft int
91.Fn sx_xlocked "const struct sx *sx"
92.Pp
93.Cd "options INVARIANTS"
94.Cd "options INVARIANT_SUPPORT"
95.Ft void
96.Fn sx_assert "const struct sx *sx" "int what"
97.In sys/kernel.h
98.Fn SX_SYSINIT "name" "struct sx *sx" "const char *description"
99.Sh DESCRIPTION
100Shared/exclusive locks are used to protect data that are read far more often
101than they are written.
102Shared/exclusive locks do not implement priority propagation like mutexes and
103reader/writer locks to prevent priority inversions, so
104shared/exclusive locks should be used prudently.
105.Pp
106Shared/exclusive locks are created with either
107.Fn sx_init
108or
109.Fn sx_init_flags
110where
111.Fa sx
112is a pointer to space for a
113.Vt struct sx ,
114and
115.Fa description
116is a pointer to a null-terminated character string that describes the
117shared/exclusive lock.
118The
119.Fa opts
120argument to
121.Fn sx_init_flags
122specifies a set of optional flags to alter the behavior of
123.Fa sx .
124It contains one or more of the following flags:
125.Bl -tag -width SX_NOADAPTIVE
126.It Dv SX_NOADAPTIVE
127Disable adaptive spinning, rather than sleeping, for lock operations
128while an exclusive lock holder is executing on another CPU.
129Adaptive spinning is the default unless the kernel is compiled with
130.Cd "options NO_ADAPTIVE_SX" .
131.It Dv SX_DUPOK
132Witness should not log messages about duplicate locks being acquired.
133.It Dv SX_NOWITNESS
134Instruct
135.Xr witness 4
136to ignore this lock.
137.It Dv SX_NOPROFILE
138Do not profile this lock.
139.It Dv SX_RECURSE
140Allow threads to recursively acquire exclusive locks for
141.Fa sx .
142.It Dv SX_QUIET
143Do not log any operations for this lock via
144.Xr ktr 4 .
145.It Dv SX_NEW
146If the kernel has been compiled with
147.Cd "options INVARIANTS" ,
148.Fn sx_init
149will assert that the
150.Fa sx
151has not been initialized multiple times without intervening calls to
152.Fn sx_destroy
153unless this option is specified.
154.El
155.Pp
156Shared/exclusive locks are destroyed with
157.Fn sx_destroy .
158The lock
159.Fa sx
160must not be locked by any thread when it is destroyed.
161.Pp
162Threads acquire and release a shared lock by calling
163.Fn sx_slock ,
164.Fn sx_slock_sig
165or
166.Fn sx_try_slock
167and
168.Fn sx_sunlock
169or
170.Fn sx_unlock .
171Threads acquire and release an exclusive lock by calling
172.Fn sx_xlock ,
173.Fn sx_xlock_sig
174or
175.Fn sx_try_xlock
176and
177.Fn sx_xunlock
178or
179.Fn sx_unlock .
180A thread can attempt to upgrade a currently held shared lock to an exclusive
181lock by calling
182.Fn sx_try_upgrade .
183A thread that has an exclusive lock can downgrade it to a shared lock by
184calling
185.Fn sx_downgrade .
186.Pp
187.Fn sx_try_slock
188and
189.Fn sx_try_xlock
190will return 0 if the shared/exclusive lock cannot be acquired immediately;
191otherwise the shared/exclusive lock will be acquired and a non-zero value will
192be returned.
193.Pp
194.Fn sx_try_upgrade
195will return 0 if the shared lock cannot be upgraded to an exclusive lock
196immediately; otherwise the exclusive lock will be acquired and a non-zero value
197will be returned.
198.Pp
199.Fn sx_slock_sig
200and
201.Fn sx_xlock_sig
202do the same as their normal versions but performing an interruptible sleep.
203They return a non-zero value if the sleep has been interrupted by a signal
204or an interrupt, otherwise 0.
205.Pp
206A thread can atomically release a shared/exclusive lock while waiting for an
207event by calling
208.Fn sx_sleep .
209For more details on the parameters to this function,
210see
211.Xr sleep 9 .
212.Pp
213When compiled with
214.Cd "options INVARIANTS"
215and
216.Cd "options INVARIANT_SUPPORT" ,
217the
218.Fn sx_assert
219function tests
220.Fa sx
221for the assertions specified in
222.Fa what ,
223and panics if they are not met.
224One of the following assertions must be specified:
225.Bl -tag -width ".Dv SA_UNLOCKED"
226.It Dv SA_LOCKED
227Assert that the current thread has either a shared or an exclusive lock on the
228.Vt sx
229lock pointed to by the first argument.
230.It Dv SA_SLOCKED
231Assert that the current thread has a shared lock on the
232.Vt sx
233lock pointed to by
234the first argument.
235.It Dv SA_XLOCKED
236Assert that the current thread has an exclusive lock on the
237.Vt sx
238lock pointed to
239by the first argument.
240.It Dv SA_UNLOCKED
241Assert that the current thread has no lock on the
242.Vt sx
243lock pointed to
244by the first argument.
245.El
246.Pp
247In addition, one of the following optional assertions may be included with
248either an
249.Dv SA_LOCKED ,
250.Dv SA_SLOCKED ,
251or
252.Dv SA_XLOCKED
253assertion:
254.Bl -tag -width ".Dv SA_NOTRECURSED"
255.It Dv SA_RECURSED
256Assert that the current thread has a recursed lock on
257.Fa sx .
258.It Dv SA_NOTRECURSED
259Assert that the current thread does not have a recursed lock on
260.Fa sx .
261.El
262.Pp
263.Fn sx_xholder
264will return a pointer to the thread which currently holds an exclusive lock on
265.Fa sx .
266If no thread holds an exclusive lock on
267.Fa sx ,
268then
269.Dv NULL
270is returned instead.
271.Pp
272.Fn sx_xlocked
273will return non-zero if the current thread holds the exclusive lock;
274otherwise, it will return zero.
275.Pp
276For ease of programming,
277.Fn sx_unlock
278is provided as a macro frontend to the respective functions,
279.Fn sx_sunlock
280and
281.Fn sx_xunlock .
282Algorithms that are aware of what state the lock is in should use either
283of the two specific functions for a minor performance benefit.
284.Pp
285The
286.Fn SX_SYSINIT
287macro is used to generate a call to the
288.Fn sx_sysinit
289routine at system startup in order to initialize a given
290.Fa sx
291lock.
292The parameters are the same as
293.Fn sx_init
294but with an additional argument,
295.Fa name ,
296that is used in generating unique variable names for the related
297structures associated with the lock and the sysinit routine.
298.Pp
299A thread may not hold both a shared lock and an exclusive lock on the same
300lock simultaneously;
301attempting to do so will result in deadlock.
302.Sh CONTEXT
303A thread may hold a shared or exclusive lock on an
304.Nm
305lock while sleeping.
306As a result, an
307.Nm
308lock may not be acquired while holding a mutex.
309Otherwise, if one thread slept while holding an
310.Nm
311lock while another thread blocked on the same
312.Nm
313lock after acquiring a mutex, then the second thread would effectively
314end up sleeping while holding a mutex, which is not allowed.
315.Sh SEE ALSO
316.Xr lock 9 ,
317.Xr locking 9 ,
318.Xr mutex 9 ,
319.Xr panic 9 ,
320.Xr rwlock 9 ,
321.Xr sema 9
322.Sh BUGS
323A kernel without
324.Dv WITNESS
325cannot assert whether the current thread does or does not hold a shared lock.
326.Dv SA_LOCKED
327and
328.Dv SA_SLOCKED
329can only assert that
330.Em any
331thread holds a shared lock.
332They cannot ensure that the current thread holds a shared lock.
333Further,
334.Dv SA_UNLOCKED
335can only assert that the current thread does not hold an exclusive lock.
336