| blob | 41a7faad75c32317f53028c39382966db0fe8e18 |
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 November 25, 2007
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 "struct sx *sx"
92 .Pp
93 .Cd "options INVARIANTS"
94 .Cd "options INVARIANT_SUPPORT"
95 .Ft void
96 .Fn sx_assert "struct sx *sx" "int what"
97 .In sys/kernel.h
98 .Fn SX_SYSINIT "name" "struct sx *sx" "const char *description"
99 .Sh DESCRIPTION
100 Shared/exclusive locks are used to protect data that are read far more often
101 than they are written.
102 Shared/exclusive locks do not implement priority propagation like mutexes and
103 reader/writer locks to prevent priority inversions, so
104 shared/exclusive locks should be used prudently.
105 .Pp
106 Shared/exclusive locks are created with either
107 .Fn sx_init
108 or
109 .Fn sx_init_flags
110 where
111 .Fa sx
112 is a pointer to space for a
113 .Vt struct sx ,
114 and
115 .Fa description
116 is a pointer to a null-terminated character string that describes the
117 shared/exclusive lock.
118 The
119 .Fa opts
120 argument to
121 .Fn sx_init_flags
122 specifies a set of optional flags to alter the behavior of
123 .Fa sx .
124 It contains one or more of the following flags:
125 .Bl -tag -width SX_ADAPTIVESPIN
126 .It Dv SX_ADAPTIVESPIN
127 If the kernel is compiled with
128 .Cd "options ADAPTIVE_SX" ,
129 then lock operations for
130 .Fa sx
131 will spin instead of sleeping while an exclusive lock holder is executing on
132 another CPU.
133 .It Dv SX_DUPOK
134 Witness should not log messages about duplicate locks being acquired.
135 .It Dv SX_NOWITNESS
136 Instruct
137 .Xr witness 4
138 to ignore this lock.
139 .It Dv SX_NOPROFILE
140 Do not profile this lock.
141 .It Dv SX_RECURSE
142 Allow threads to recursively acquire exclusive locks for
143 .Fa sx .
144 .It Dv SX_QUIET
145 Do not log any operations for this lock via
146 .Xr ktr 4 .
147 .El
148 .Pp
149 Shared/exclusive locks are destroyed with
150 .Fn sx_destroy .
151 The lock
152 .Fa sx
153 must not be locked by any thread when it is destroyed.
154 .Pp
155 Threads acquire and release a shared lock by calling
156 .Fn sx_slock ,
157 .Fn sx_slock_sig
158 or
159 .Fn sx_try_slock
160 and
161 .Fn sx_sunlock
162 or
163 .Fn sx_unlock .
164 Threads acquire and release an exclusive lock by calling
165 .Fn sx_xlock ,
166 .Fn sx_xlock_sig
167 or
168 .Fn sx_try_xlock
169 and
170 .Fn sx_xunlock
171 or
172 .Fn sx_unlock .
173 A thread can attempt to upgrade a currently held shared lock to an exclusive
174 lock by calling
175 .Fn sx_try_upgrade .
176 A thread that has an exclusive lock can downgrade it to a shared lock by
177 calling
178 .Fn sx_downgrade .
179 .Pp
180 .Fn sx_try_slock
181 and
182 .Fn sx_try_xlock
183 will return 0 if the shared/exclusive lock cannot be acquired immediately;
184 otherwise the shared/exclusive lock will be acquired and a non-zero value will
185 be returned.
186 .Pp
187 .Fn sx_try_upgrade
188 will return 0 if the shared lock cannot be upgraded to an exclusive lock
189 immediately; otherwise the exclusive lock will be acquired and a non-zero value
190 will be returned.
191 .Pp
192 .Fn sx_slock_sig
193 and
194 .Fn sx_xlock_sig
195 do the same as their normal versions but performing an interruptible sleep.
196 They return a non-zero value if the sleep has been interrupted by a signal
197 or an interrupt, otherwise 0.
198 .Pp
199 A thread can atomically release a shared/exclusive lock while waiting for an
200 event by calling
201 .Fn sx_sleep .
202 For more details on the parameters to this function,
203 see
204 .Xr sleep 9 .
205 .Pp
206 When compiled with
207 .Cd "options INVARIANTS"
208 and
209 .Cd "options INVARIANT_SUPPORT" ,
210 the
211 .Fn sx_assert
212 function tests
213 .Fa sx
214 for the assertions specified in
215 .Fa what ,
216 and panics if they are not met.
217 One of the following assertions must be specified:
218 .Bl -tag -width ".Dv SA_UNLOCKED"
219 .It Dv SA_LOCKED
220 Assert that the current thread has either a shared or an exclusive lock on the
221 .Vt sx
222 lock pointed to by the first argument.
223 .It Dv SA_SLOCKED
224 Assert that the current thread has a shared lock on the
225 .Vt sx
226 lock pointed to by
227 the first argument.
228 .It Dv SA_XLOCKED
229 Assert that the current thread has an exclusive lock on the
230 .Vt sx
231 lock pointed to
232 by the first argument.
233 .It Dv SA_UNLOCKED
234 Assert that the current thread has no lock on the
235 .Vt sx
236 lock pointed to
237 by the first argument.
238 .El
239 .Pp
240 In addition, one of the following optional assertions may be included with
241 either an
242 .Dv SA_LOCKED ,
243 .Dv SA_SLOCKED ,
244 or
245 .Dv SA_XLOCKED
246 assertion:
247 .Bl -tag -width ".Dv SA_NOTRECURSED"
248 .It Dv SA_RECURSED
249 Assert that the current thread has a recursed lock on
250 .Fa sx .
251 .It Dv SA_NOTRECURSED
252 Assert that the current thread does not have a recursed lock on
253 .Fa sx .
254 .El
255 .Pp
256 .Fn sx_xholder
257 will return a pointer to the thread which currently holds an exclusive lock on
258 .Fa sx .
259 If no thread holds an exclusive lock on
260 .Fa sx ,
261 then
262 .Dv NULL
263 is returned instead.
264 .Pp
265 .Fn sx_xlocked
266 will return non-zero if the current thread holds the exclusive lock;
267 otherwise, it will return zero.
268 .Pp
269 For ease of programming,
270 .Fn sx_unlock
271 is provided as a macro frontend to the respective functions,
272 .Fn sx_sunlock
273 and
274 .Fn sx_xunlock .
275 Algorithms that are aware of what state the lock is in should use either
276 of the two specific functions for a minor performance benefit.
277 .Pp
278 The
279 .Fn SX_SYSINIT
280 macro is used to generate a call to the
281 .Fn sx_sysinit
282 routine at system startup in order to initialize a given
283 .Fa sx
284 lock.
285 The parameters are the same as
286 .Fn sx_init
287 but with an additional argument,
288 .Fa name ,
289 that is used in generating unique variable names for the related
290 structures associated with the lock and the sysinit routine.
291 .Pp
292 A thread may not hold both a shared lock and an exclusive lock on the same
293 lock simultaneously;
294 attempting to do so will result in deadlock.
295 .Sh CONTEXT
296 A thread may hold a shared or exclusive lock on an
297 .Nm
298 lock while sleeping.
299 As a result, an
300 .Nm
301 lock may not be acquired while holding a mutex.
302 Otherwise, if one thread slept while holding an
303 .Nm
304 lock while another thread blocked on the same
305 .Nm
306 lock after acquiring a mutex, then the second thread would effectively
307 end up sleeping while holding a mutex, which is not allowed.
308 .Sh SEE ALSO
309 .Xr locking 9 ,
310 .Xr lock 9 ,
311 .Xr mutex 9 ,
312 .Xr panic 9 ,
313 .Xr rwlock 9 ,
314 .Xr sema 9
315 .Sh BUGS
316 Currently there is no way to assert that a lock is not held.
317 This is not possible in the
318 .No non- Ns Dv WITNESS
319 case for asserting that this thread
320 does not hold a shared lock.
321 In the
322 .No non- Ns Dv WITNESS
323 case, the
324 .Dv SA_LOCKED
325 and
326 .Dv SA_SLOCKED
327 assertions merely check that some thread holds a shared lock.
328 They do not ensure that the current thread holds a shared lock.