1SEMOP(2) Linux Programmer's Manual SEMOP(2)
2
6 semop, semtimedop - System V semaphore operations
7
9 #include <sys/types.h>
10 #include <sys/ipc.h>
11 #include <sys/sem.h>
12 int semop(int semid, struct sembuf *sops, size_t nsops);
14 int semtimedop(int semid, struct sembuf *sops, size_t nsops,
16 const struct timespec *timeout);
17 Feature Test Macro Requirements for glibc (see feature_test_macros(7)):
19 semtimedop(): _GNU_SOURCE
21
23 Each semaphore in a System V semaphore set has the following associated
24 values:
25 unsigned short semval; /* semaphore value */
27 unsigned short semzcnt; /* # waiting for zero */
28 unsigned short semncnt; /* # waiting for increase */
29 pid_t sempid; /* PID of process that last
30 semop() performs operations on selected semaphores in the set indicated
32 by semid. Each of the nsops elements in the array pointed to by sops
33 is a structure that specifies an operation to be performed on a single
34 semaphore. The elements of this structure are of type struct sembuf,
35 containing the following members:
36 unsigned short sem_num; /* semaphore number */
38 short sem_op; /* semaphore operation */
39 short sem_flg; /* operation flags */
40 Flags recognized in sem_flg are IPC_NOWAIT and SEM_UNDO. If an opera‐
42 tion specifies SEM_UNDO, it will be automatically undone when the
43 process terminates.
44 The set of operations contained in sops is performed in array order,
46 and atomically, that is, the operations are performed either as a com‐
47 plete unit, or not at all. The behavior of the system call if not all
48 operations can be performed immediately depends on the presence of the
49 IPC_NOWAIT flag in the individual sem_flg fields, as noted below.
50 Each operation is performed on the sem_num-th semaphore of the sema‐
52 phore set, where the first semaphore of the set is numbered 0. There
53 are three types of operation, distinguished by the value of sem_op.
54 If sem_op is a positive integer, the operation adds this value to the
56 semaphore value (semval). Furthermore, if SEM_UNDO is specified for
57 this operation, the system subtracts the value sem_op from the sema‐
58 phore adjustment (semadj) value for this semaphore. This operation can
59 always proceed—it never forces a thread to wait. The calling process
60 must have alter permission on the semaphore set.
61 If sem_op is zero, the process must have read permission on the sema‐
63 phore set. This is a "wait-for-zero" operation: if semval is zero, the
64 operation can immediately proceed. Otherwise, if IPC_NOWAIT is speci‐
65 fied in sem_flg, semop() fails with errno set to EAGAIN (and none of
66 the operations in sops is performed). Otherwise, semzcnt (the count of
67 threads waiting until this semaphore's value becomes zero) is incre‐
68 mented by one and the thread sleeps until one of the following occurs:
69 • semval becomes 0, at which time the value of semzcnt is decremented.
71 • The semaphore set is removed: semop() fails, with errno set to EIDRM.
73 • The calling thread catches a signal: the value of semzcnt is decre‐
75 mented and semop() fails, with errno set to EINTR.
76 If sem_op is less than zero, the process must have alter permission on
78 the semaphore set. If semval is greater than or equal to the absolute
79 value of sem_op, the operation can proceed immediately: the absolute
80 value of sem_op is subtracted from semval, and, if SEM_UNDO is speci‐
81 fied for this operation, the system adds the absolute value of sem_op
82 to the semaphore adjustment (semadj) value for this semaphore. If the
83 absolute value of sem_op is greater than semval, and IPC_NOWAIT is
84 specified in sem_flg, semop() fails, with errno set to EAGAIN (and none
85 of the operations in sops is performed). Otherwise, semncnt (the
86 counter of threads waiting for this semaphore's value to increase) is
87 incremented by one and the thread sleeps until one of the following oc‐
88 curs:
89 • semval becomes greater than or equal to the absolute value of sem_op:
91 the operation now proceeds, as described above.
92 • The semaphore set is removed from the system: semop() fails, with er‐
94 rno set to EIDRM.
95 • The calling thread catches a signal: the value of semncnt is decre‐
97 mented and semop() fails, with errno set to EINTR.
98 On successful completion, the sempid value for each semaphore specified
100 in the array pointed to by sops is set to the caller's process ID. In
101 addition, the sem_otime is set to the current time.
102 semtimedop()
104 semtimedop() behaves identically to semop() except that in those cases
105 where the calling thread would sleep, the duration of that sleep is
106 limited by the amount of elapsed time specified by the timespec struc‐
107 ture whose address is passed in the timeout argument. (This sleep in‐
108 terval will be rounded up to the system clock granularity, and kernel
109 scheduling delays mean that the interval may overrun by a small
110 amount.) If the specified time limit has been reached, semtimedop()
111 fails with errno set to EAGAIN (and none of the operations in sops is
112 performed). If the timeout argument is NULL, then semtimedop() behaves
113 exactly like semop().
114 Note that if semtimedop() is interrupted by a signal, causing the call
116 to fail with the error EINTR, the contents of timeout are left un‐
117 changed.
118
120 If successful, semop() and semtimedop() return 0; otherwise they return
121 -1 with errno indicating the error.
122
124 On failure, errno is set to one of the following:
125 E2BIG The argument nsops is greater than SEMOPM, the maximum number of
127 operations allowed per system call.
128 EACCES The calling process does not have the permissions required to
130 perform the specified semaphore operations, and does not have
131 the CAP_IPC_OWNER capability in the user namespace that governs
132 its IPC namespace.
133 EAGAIN An operation could not proceed immediately and either IPC_NOWAIT
135 was specified in sem_flg or the time limit specified in timeout
136 expired.
137 EFAULT An address specified in either the sops or the timeout argument
139 isn't accessible.
140 EFBIG For some operation the value of sem_num is less than 0 or
142 greater than or equal to the number of semaphores in the set.
143 EIDRM The semaphore set was removed.
145 EINTR While blocked in this system call, the thread caught a signal;
147 see signal(7).
148 EINVAL The semaphore set doesn't exist, or semid is less than zero, or
150 nsops has a nonpositive value.
151 ENOMEM The sem_flg of some operation specified SEM_UNDO and the system
153 does not have enough memory to allocate the undo structure.
154 ERANGE For some operation sem_op+semval is greater than SEMVMX, the im‐
156 plementation dependent maximum value for semval.
157
159 semtimedop() first appeared in Linux 2.5.52, and was subsequently back‐
160 ported into kernel 2.4.22. Glibc support for semtimedop() first ap‐
161 peared in version 2.3.3.
162
164 POSIX.1-2001, POSIX.1-2008, SVr4.
165
167 The inclusion of <sys/types.h> and <sys/ipc.h> isn't required on Linux
168 or by any version of POSIX. However, some old implementations required
169 the inclusion of these header files, and the SVID also documented their
170 inclusion. Applications intended to be portable to such old systems
171 may need to include these header files.
172 The sem_undo structures of a process aren't inherited by the child pro‐
174 duced by fork(2), but they are inherited across an execve(2) system
175 call.
176 semop() is never automatically restarted after being interrupted by a
178 signal handler, regardless of the setting of the SA_RESTART flag when
179 establishing a signal handler.
180 A semaphore adjustment (semadj) value is a per-process, per-semaphore
182 integer that is the negated sum of all operations performed on a sema‐
183 phore specifying the SEM_UNDO flag. Each process has a list of semadj
184 values—one value for each semaphore on which it has operated using
185 SEM_UNDO. When a process terminates, each of its per-semaphore semadj
186 values is added to the corresponding semaphore, thus undoing the effect
187 of that process's operations on the semaphore (but see BUGS below).
188 When a semaphore's value is directly set using the SETVAL or SETALL re‐
189 quest to semctl(2), the corresponding semadj values in all processes
190 are cleared. The clone(2) CLONE_SYSVSEM flag allows more than one
191 process to share a semadj list; see clone(2) for details.
192 The semval, sempid, semzcnt, and semnct values for a semaphore can all
194 be retrieved using appropriate semctl(2) calls.
195 Semaphore limits
197 The following limits on semaphore set resources affect the semop()
198 call:
199 SEMOPM Maximum number of operations allowed for one semop() call. Be‐
201 fore Linux 3.19, the default value for this limit was 32. Since
202 Linux 3.19, the default value is 500. On Linux, this limit can
203 be read and modified via the third field of /proc/sys/ker‐
204 nel/sem. Note: this limit should not be raised above 1000, be‐
205 cause of the risk of that semop() fails due to kernel memory
206 fragmentation when allocating memory to copy the sops array.
207 SEMVMX Maximum allowable value for semval: implementation dependent
209 (32767).
210 The implementation has no intrinsic limits for the adjust on exit maxi‐
212 mum value (SEMAEM), the system wide maximum number of undo structures
213 (SEMMNU) and the per-process maximum number of undo entries system pa‐
214 rameters.
215
217 When a process terminates, its set of associated semadj structures is
218 used to undo the effect of all of the semaphore operations it performed
219 with the SEM_UNDO flag. This raises a difficulty: if one (or more) of
220 these semaphore adjustments would result in an attempt to decrease a
221 semaphore's value below zero, what should an implementation do? One
222 possible approach would be to block until all the semaphore adjustments
223 could be performed. This is however undesirable since it could force
224 process termination to block for arbitrarily long periods. Another
225 possibility is that such semaphore adjustments could be ignored alto‐
226 gether (somewhat analogously to failing when IPC_NOWAIT is specified
227 for a semaphore operation). Linux adopts a third approach: decreasing
228 the semaphore value as far as possible (i.e., to zero) and allowing
229 process termination to proceed immediately.
230 In kernels 2.6.x, x <= 10, there is a bug that in some circumstances
232 prevents a thread that is waiting for a semaphore value to become zero
233 from being woken up when the value does actually become zero. This bug
234 is fixed in kernel 2.6.11.
235
237 The following code segment uses semop() to atomically wait for the
238 value of semaphore 0 to become zero, and then increment the semaphore
239 value by one.
240 struct sembuf sops[2];
242 int semid;
243 /* Code to set semid omitted */
245 sops[0].sem_num = 0; /* Operate on semaphore 0 */
247 sops[0].sem_op = 0; /* Wait for value to equal 0 */
248 sops[0].sem_flg = 0;
249 sops[1].sem_num = 0; /* Operate on semaphore 0 */
251 sops[1].sem_op = 1; /* Increment value by one */
252 sops[1].sem_flg = 0;
253 if (semop(semid, sops, 2) == -1) {
255 perror("semop");
256 exit(EXIT_FAILURE);
257 }
258 A further example of the use of semop() can be found in shmop(2).
260
262 clone(2), semctl(2), semget(2), sigaction(2), capabilities(7),
263 sem_overview(7), sysvipc(7), time(7)
264
266 This page is part of release 5.10 of the Linux man-pages project. A
267 description of the project, information about reporting bugs, and the
268 latest version of this page, can be found at
269 https://www.kernel.org/doc/man-pages/.
270Linux 2020-04-11 SEMOP(2)