Debian Codegraph @

upstream/debian-unstable/man2/shmop.2

raw
1 .\" Copyright 1993, Giorgio Ciucci <[email protected]>
2 .\" Copyright 2020, Michael Kerrisk <[email protected]>
3 .\" Copyright, the authors of the Linux man-pages project
4 .\"
5 .\" SPDX-License-Identifier: Linux-man-pages-copyleft
6 .\"
7 .TH SHMOP 2 2026-02-08 "Linux man-pages 6.18"
8 .SH NAME
9 shmat, shmdt \- System V shared memory operations
10 .SH LIBRARY
11 Standard C library
12 .RI ( libc ,\~ \-lc )
13 .SH SYNOPSIS
14 .nf
15 .B #include <sys/shm.h>
16 .P
17 .BI "void *shmat(int " shmid ", const void *_Nullable " shmaddr ", int " shmflg );
18 .BI "int shmdt(const void *" shmaddr );
19 .fi
20 .SH DESCRIPTION
21 .SS shmat()
22 .BR shmat ()
23 attaches the System\ V shared memory segment identified by
24 .I shmid
25 to the address space of the calling process.
26 The attaching address is specified by
27 .I shmaddr
28 with one of the following criteria:
29 .IP \[bu] 3
30 If
31 .I shmaddr
32 is NULL,
33 the system chooses a suitable (unused) page-aligned address to attach
34 the segment.
35 .IP \[bu]
36 If
37 .I shmaddr
38 isn't NULL
39 and
40 .B SHM_RND
41 is specified in
42 .IR shmflg ,
43 the attach occurs at the address equal to
44 .I shmaddr
45 rounded down to the nearest multiple of
46 .BR SHMLBA .
47 .IP \[bu]
48 Otherwise,
49 .I shmaddr
50 must be a page-aligned address at which the attach occurs.
51 .P
52 In addition to
53 .BR SHM_RND ,
54 the following flags may be specified in the
55 .I shmflg
56 bit-mask argument:
57 .TP
58 .BR SHM_EXEC " (Linux-specific; since Linux 2.6.9)"
59 Allow the contents of the segment to be executed.
60 The caller must have execute permission on the segment.
61 .TP
62 .B SHM_RDONLY
63 Attach the segment for read-only access.
64 The process must have read permission for the segment.
65 If this flag is not specified,
66 the segment is attached for read and write access,
67 and the process must have read and write permission for the segment.
68 There is no notion of a write-only shared memory segment.
69 .TP
70 .BR SHM_REMAP " (Linux-specific)"
71 This flag specifies
72 that the mapping of the segment should replace
73 any existing mapping in the range starting at
74 .I shmaddr
75 and continuing for the size of the segment.
76 (Normally, an
77 .B EINVAL
78 error would result if a mapping already exists in this address range.)
79 In this case,
80 .I shmaddr
81 must not be NULL.
82 .P
83 The
84 .BR brk (2)
85 value of the calling process is not altered by the attach.
86 The segment will automatically be detached at process exit.
87 The same segment may be attached as a read and as a read-write
88 one, and more than once, in the process's address space.
89 .P
90 A successful
91 .BR shmat ()
92 call updates the members of the
93 .I shmid_ds
94 structure (see
95 .BR shmctl (2))
96 associated with the shared memory segment as follows:
97 .IP \[bu] 3
98 .I shm_atime
99 is set to the current time.
100 .IP \[bu]
101 .I shm_lpid
102 is set to the process-ID of the calling process.
103 .IP \[bu]
104 .I shm_nattch
105 is incremented by one.
106 .\"
107 .SS shmdt()
108 .BR shmdt ()
109 detaches the shared memory segment located at the address specified by
110 .I shmaddr
111 from the address space of the calling process.
112 The to-be-detached segment must be currently
113 attached with
114 .I shmaddr
115 equal to the value returned by the attaching
116 .BR shmat ()
117 call.
118 .P
119 On a successful
120 .BR shmdt ()
121 call, the system updates the members of the
122 .I shmid_ds
123 structure associated with the shared memory segment as follows:
124 .IP \[bu] 3
125 .I shm_dtime
126 is set to the current time.
127 .IP \[bu]
128 .I shm_lpid
129 is set to the process-ID of the calling process.
130 .IP \[bu]
131 .I shm_nattch
132 is decremented by one.
133 If it becomes 0 and the segment is marked for deletion,
134 the segment is deleted.
135 .SH RETURN VALUE
136 On success,
137 .BR shmat ()
138 returns the address of the attached shared memory segment;
139 on error,
140 .I (void\~*)\~\-1
141 is returned, and
142 .I errno
143 is set to indicate the error.
144 .P
145 On success,
146 .BR shmdt ()
147 returns 0;
148 on error \-1 is returned, and
149 .I errno
150 is set to indicate the error.
151 .SH ERRORS
152 .BR shmat ()
153 can fail with one of the following errors:
154 .TP
155 .B EACCES
156 The calling process does not have the required permissions for
157 the requested attach type, and does not have the
158 .B CAP_IPC_OWNER
159 capability in the user namespace that governs its IPC namespace.
160 .TP
161 .B EIDRM
162 .I shmid
163 points to a removed identifier.
164 .TP
165 .B EINVAL
166 Invalid
167 .I shmid
168 value, unaligned (i.e., not page-aligned and
169 .B SHM_RND
170 was not specified)
171 or invalid
172 .I shmaddr
173 value, or can't attach segment at
174 .IR shmaddr ,
175 or
176 .B SHM_REMAP
177 was specified and
178 .I shmaddr
179 was NULL.
180 .TP
181 .B ENOMEM
182 Could not allocate memory for the descriptor or for the page tables.
183 .P
184 .BR shmdt ()
185 can fail with one of the following errors:
186 .TP
187 .B EINVAL
188 There is no shared memory segment attached at
189 .IR shmaddr ;
190 or,
191 .\" The following since Linux 2.6.17-rc1:
192 .I shmaddr
193 is not aligned on a page boundary.
194 .SH STANDARDS
195 POSIX.1-2008.
196 .SH HISTORY
197 POSIX.1-2001, SVr4.
198 .\" SVr4 documents an additional error condition EMFILE.
199 .P
200 In SVID 3 (or perhaps earlier),
201 the type of the
202 .I shmaddr
203 argument was changed from
204 .I char\ *
205 into
206 .IR "const void\ *" ,
207 and the returned type of
208 .BR shmat ()
209 from
210 .I char\ *
211 into
212 .IR "void\ *" .
213 .P
214 POSIX.1-2024
215 changed the error code from
216 .I (void\ *)\ \-1
217 to
218 .BR SHM_FAILED .
219 .SH NOTES
220 After a
221 .BR fork (2),
222 the child inherits the attached shared memory segments.
223 .P
224 After an
225 .BR execve (2),
226 all attached shared memory segments are detached from the process.
227 .P
228 Upon
229 .BR _exit (2),
230 all attached shared memory segments are detached from the process.
231 .P
232 Using
233 .BR shmat ()
234 with
235 .I shmaddr
236 equal to NULL
237 is the preferred, portable way of attaching a shared memory segment.
238 Be aware that the shared memory segment attached in this way
239 may be attached at different addresses in different processes.
240 Therefore, any pointers maintained within the shared memory must be
241 made relative (typically to the starting address of the segment),
242 rather than absolute.
243 .P
244 On Linux, it is possible to attach a shared memory segment even if it
245 is already marked to be deleted.
246 However, POSIX.1 does not specify this behavior and
247 many other implementations do not support it.
248 .P
249 The following system parameter affects
250 .BR shmat ():
251 .TP
252 .B SHMLBA
253 Segment low boundary address multiple.
254 When explicitly specifying an attach address in a call to
255 .BR shmat (),
256 the caller should ensure that the address is a multiple of this value.
257 This is necessary on some architectures,
258 in order either to ensure good CPU cache performance or to ensure that
259 different attaches of the same segment have consistent views
260 within the CPU cache.
261 .B SHMLBA
262 is normally some multiple of the system page size.
263 (On many Linux architectures,
264 .B SHMLBA
265 is the same as the system page size.)
266 .P
267 The implementation places no intrinsic per-process limit on the
268 number of shared memory segments
269 .RB ( SHMSEG ).
270 .SH EXAMPLES
271 The two programs shown below exchange a string using a shared memory segment.
272 Further details about the programs are given below.
273 First, we show a shell session demonstrating their use.
274 .P
275 In one terminal window, we run the "reader" program,
276 which creates a System V shared memory segment and a System V semaphore set.
277 The program prints out the IDs of the created objects,
278 and then waits for the semaphore to change value.
279 .P
280 .in +4n
281 .EX
282 .RB $ " ./svshm_string_read" ;
283 shmid = 1114194
284 semid = 15
285 .EE
286 .in
287 .P
288 In another terminal window, we run the "writer" program.
289 The "writer" program takes three command-line arguments:
290 the IDs of the shared memory segment and semaphore set created
291 by the "reader", and a string.
292 It attaches the existing shared memory segment,
293 copies the string to the shared memory, and modifies the semaphore value.
294 .P
295 .in +4n
296 .EX
297 .RB $ " ./svshm_string_write 1114194 15 \[aq]Hello, world\[aq]" ;
298 .EE
299 .in
300 .P
301 Returning to the terminal where the "reader" is running,
302 we see that the program has ceased waiting on the semaphore
303 and has printed the string that was copied into the
304 shared memory segment by the writer:
305 .P
306 .in +4n
307 .EX
308 Hello, world
309 .EE
310 .in
311 .\"
312 .SS Program source: svshm_string.h
313 The following header file is included by the "reader" and "writer" programs:
314 .P
315 .in +4n
316 .\" SRC BEGIN (svshm_string.h)
317 .EX
318 /* svshm_string.h
319 \&
320 Licensed under GNU General Public License v2 or later.
321 */
322 #ifndef SVSHM_STRING_H
323 #define SVSHM_STRING_H
324 \&
325 #include <err.h>
326 #include <stdio.h>
327 #include <stdlib.h>
328 #include <sys/sem.h>
329 \&
330 union semun { /* Used in calls to semctl() */
331 int val;
332 struct semid_ds *buf;
333 unsigned short *array;
334 #if defined(__linux__)
335 struct seminfo *__buf;
336 #endif
337 };
338 \&
339 #define MEM_SIZE 4096
340 \&
341 #endif // include guard
342 .EE
343 .\" SRC END
344 .in
345 .\"
346 .SS Program source: svshm_string_read.c
347 The "reader" program creates a shared memory segment and a semaphore set
348 containing one semaphore.
349 It then attaches the shared memory object into its address space
350 and initializes the semaphore value to 1.
351 Finally, the program waits for the semaphore value to become 0,
352 and afterwards prints the string that has been copied into the
353 shared memory segment by the "writer".
354 .P
355 .in +4n
356 .\" SRC BEGIN (svshm_string_read.c)
357 .EX
358 /* svshm_string_read.c
359 \&
360 Licensed under GNU General Public License v2 or later.
361 */
362 #include <err.h>
363 #include <stdio.h>
364 #include <stdlib.h>
365 #include <sys/ipc.h>
366 #include <sys/sem.h>
367 #include <sys/shm.h>
368 \&
369 #include "svshm_string.h"
370 \&
371 int
372 main(void)
373 {
374 int semid, shmid;
375 char *addr;
376 union semun arg, dummy;
377 struct sembuf sop;
378 \&
379 /* Create shared memory and semaphore set containing one
380 semaphore. */
381 \&
382 shmid = shmget(IPC_PRIVATE, MEM_SIZE, IPC_CREAT | 0600);
383 if (shmid == \-1)
384 err(EXIT_FAILURE, "shmget");
385 \&
386 semid = semget(IPC_PRIVATE, 1, IPC_CREAT | 0600);
387 if (semid == \-1)
388 err(EXIT_FAILURE, "semget");
389 \&
390 /* Attach shared memory into our address space. */
391 \&
392 addr = shmat(shmid, NULL, SHM_RDONLY);
393 if (addr == (void *) \-1)
394 err(EXIT_FAILURE, "shmat");
395 \&
396 /* Initialize semaphore 0 in set with value 1. */
397 \&
398 arg.val = 1;
399 if (semctl(semid, 0, SETVAL, arg) == \-1)
400 err(EXIT_FAILURE, "semctl");
401 \&
402 printf("shmid = %d\[rs]n", shmid);
403 printf("semid = %d\[rs]n", semid);
404 \&
405 /* Wait for semaphore value to become 0. */
406 \&
407 sop.sem_num = 0;
408 sop.sem_op = 0;
409 sop.sem_flg = 0;
410 \&
411 if (semop(semid, &sop, 1) == \-1)
412 err(EXIT_FAILURE, "semop");
413 \&
414 /* Print the string from shared memory. */
415 \&
416 printf("%s\[rs]n", addr);
417 \&
418 /* Remove shared memory and semaphore set. */
419 \&
420 if (shmctl(shmid, IPC_RMID, NULL) == \-1)
421 err(EXIT_FAILURE, "shmctl");
422 if (semctl(semid, 0, IPC_RMID, dummy) == \-1)
423 err(EXIT_FAILURE, "semctl");
424 \&
425 exit(EXIT_SUCCESS);
426 }
427 .EE
428 .\" SRC END
429 .in
430 .\"
431 .SS Program source: svshm_string_write.c
432 The writer program takes three command-line arguments:
433 the IDs of the shared memory segment and semaphore set
434 that have already been created by the "reader", and a string.
435 It attaches the shared memory segment into its address space,
436 and then decrements the semaphore value to 0 in order to inform the
437 "reader" that it can now examine the contents of the shared memory.
438 .P
439 .in +4n
440 .\" SRC BEGIN (svshm_string_write.c)
441 .EX
442 /* svshm_string_write.c
443 \&
444 Licensed under GNU General Public License v2 or later.
445 */
446 #include <err.h>
447 #include <stdio.h>
448 #include <stdlib.h>
449 #include <string.h>
450 #include <sys/sem.h>
451 #include <sys/shm.h>
452 \&
453 #include "svshm_string.h"
454 \&
455 int
456 main(int argc, char *argv[])
457 {
458 int semid, shmid;
459 char *addr;
460 size_t size;
461 struct sembuf sop;
462 \&
463 if (argc != 4) {
464 fprintf(stderr, "Usage: %s shmid semid string\[rs]n", argv[0]);
465 exit(EXIT_FAILURE);
466 }
467 \&
468 size = strlen(argv[3]) + 1; /* +1 to include trailing \[aq]\[rs]0\[aq] */
469 if (size > MEM_SIZE) {
470 fprintf(stderr, "String is too big!\[rs]n");
471 exit(EXIT_FAILURE);
472 }
473 \&
474 /* Get object IDs from command\-line. */
475 \&
476 shmid = atoi(argv[1]);
477 semid = atoi(argv[2]);
478 \&
479 /* Attach shared memory into our address space and copy string
480 (including trailing null byte) into memory. */
481 \&
482 addr = shmat(shmid, NULL, 0);
483 if (addr == (void *) \-1)
484 err(EXIT_FAILURE, "shmat");
485 \&
486 memcpy(addr, argv[3], size);
487 \&
488 /* Decrement semaphore to 0. */
489 \&
490 sop.sem_num = 0;
491 sop.sem_op = \-1;
492 sop.sem_flg = 0;
493 \&
494 if (semop(semid, &sop, 1) == \-1)
495 err(EXIT_FAILURE, "semop");
496 \&
497 exit(EXIT_SUCCESS);
498 }
499 .EE
500 .\" SRC END
501 .in
502 .SH SEE ALSO
503 .BR brk (2),
504 .BR mmap (2),
505 .BR shmctl (2),
506 .BR shmget (2),
507 .BR capabilities (7),
508 .BR shm_overview (7),
509 .BR sysvipc (7)