1SHMCTL(2) Linux Programmer's Manual SHMCTL(2)
2
3
4
6 shmctl - System V shared memory control
7
9 #include <sys/shm.h>
10
11 int shmctl(int shmid, int cmd, struct shmid_ds *buf);
12
14 shmctl() performs the control operation specified by cmd on the Sys‐
15 tem V shared memory segment whose identifier is given in shmid.
16
17 The buf argument is a pointer to a shmid_ds structure, defined in
18 <sys/shm.h> as follows:
19
20 struct shmid_ds {
21 struct ipc_perm shm_perm; /* Ownership and permissions */
22 size_t shm_segsz; /* Size of segment (bytes) */
23 time_t shm_atime; /* Last attach time */
24 time_t shm_dtime; /* Last detach time */
25 time_t shm_ctime; /* Creation time/time of last
26 modification via shmctl() */
27 pid_t shm_cpid; /* PID of creator */
28 pid_t shm_lpid; /* PID of last shmat(2)/shmdt(2) */
29 shmatt_t shm_nattch; /* No. of current attaches */
30 ...
31 };
32
33 The fields of the shmid_ds structure are as follows:
34
35 shm_perm This is an ipc_perm structure (see below) that specifies
36 the access permissions on the shared memory segment.
37
38 shm_segsz Size in bytes of the shared memory segment.
39
40 shm_atime Time of the last shmat(2) system call that attached this
41 segment.
42
43 shm_dtime Time of the last shmdt(2) system call that detached tgis
44 segment.
45
46 shm_ctime Time of creation of segment or time of the last shmctl()
47 IPC_SET operation.
48
49 shm_cpid ID of the process that created the shared memory segment.
50
51 shm_lpid ID of the last process that executed a shmat(2) or shmdt(2)
52 system call on this segment.
53
54 shm_nattch Number of processes that have this segment attached.
55
56 The ipc_perm structure is defined as follows (the highlighted fields
57 are settable using IPC_SET):
58
59 struct ipc_perm {
60 key_t __key; /* Key supplied to shmget(2) */
61 uid_t uid; /* Effective UID of owner */
62 gid_t gid; /* Effective GID of owner */
63 uid_t cuid; /* Effective UID of creator */
64 gid_t cgid; /* Effective GID of creator */
65 unsigned short mode; /* Permissions + SHM_DEST and
66 SHM_LOCKED flags */
67 unsigned short __seq; /* Sequence number */
68 };
69
70 The least significant 9 bits of the mode field of the ipc_perm struc‐
71 ture define the access permissions for the shared memory segment. The
72 permission bits are as follows:
73
74 0400 Read by user
75 0200 Write by user
76 0040 Read by group
77 0020 Write by group
78 0004 Read by others
79 0002 Write by others
80
81 Bits 0100, 0010, and 0001 (the execute bits) are unused by the system.
82 (It is not necessary to have execute permission on a segment in order
83 to perform a shmat(2) call with the SHM_EXEC flag.)
84
85 Valid values for cmd are:
86
87 IPC_STAT
88 Copy information from the kernel data structure associated with
89 shmid into the shmid_ds structure pointed to by buf. The caller
90 must have read permission on the shared memory segment.
91
92 IPC_SET
93 Write the values of some members of the shmid_ds structure
94 pointed to by buf to the kernel data structure associated with
95 this shared memory segment, updating also its shm_ctime member.
96
97 The following fields are updated: shm_perm.uid, shm_perm.gid,
98 and (the least significant 9 bits of) shm_perm.mode.
99
100 The effective UID of the calling process must match the owner
101 (shm_perm.uid) or creator (shm_perm.cuid) of the shared memory
102 segment, or the caller must be privileged.
103
104 IPC_RMID
105 Mark the segment to be destroyed. The segment will actually be
106 destroyed only after the last process detaches it (i.e., when
107 the shm_nattch member of the associated structure shmid_ds is
108 zero). The caller must be the owner or creator of the segment,
109 or be privileged. The buf argument is ignored.
110
111 If a segment has been marked for destruction, then the (nonstan‐
112 dard) SHM_DEST flag of the shm_perm.mode field in the associated
113 data structure retrieved by IPC_STAT will be set.
114
115 The caller must ensure that a segment is eventually destroyed;
116 otherwise its pages that were faulted in will remain in memory
117 or swap.
118
119 See also the description of /proc/sys/kernel/shm_rmid_forced in
120 proc(5).
121
122 IPC_INFO (Linux-specific)
123 Return information about system-wide shared memory limits and
124 parameters in the structure pointed to by buf. This structure
125 is of type shminfo (thus, a cast is required), defined in
126 <sys/shm.h> if the _GNU_SOURCE feature test macro is defined:
127
128 struct shminfo {
129 unsigned long shmmax; /* Maximum segment size */
130 unsigned long shmmin; /* Minimum segment size;
131 always 1 */
132 unsigned long shmmni; /* Maximum number of segments */
133 unsigned long shmseg; /* Maximum number of segments
134 that a process can attach;
135 unused within kernel */
136 unsigned long shmall; /* Maximum number of pages of
137 shared memory, system-wide */
138 };
139
140 The shmmni, shmmax, and shmall settings can be changed via /proc
141 files of the same name; see proc(5) for details.
142
143 SHM_INFO (Linux-specific)
144 Return a shm_info structure whose fields contain information
145 about system resources consumed by shared memory. This struc‐
146 ture is defined in <sys/shm.h> if the _GNU_SOURCE feature test
147 macro is defined:
148
149 struct shm_info {
150 int used_ids; /* # of currently existing
151 segments */
152 unsigned long shm_tot; /* Total number of shared
153 memory pages */
154 unsigned long shm_rss; /* # of resident shared
155 memory pages */
156 unsigned long shm_swp; /* # of swapped shared
157 memory pages */
158 unsigned long swap_attempts;
159 /* Unused since Linux 2.4 */
160 unsigned long swap_successes;
161 /* Unused since Linux 2.4 */
162 };
163
164 SHM_STAT (Linux-specific)
165 Return a shmid_ds structure as for IPC_STAT. However, the shmid
166 argument is not a segment identifier, but instead an index into
167 the kernel's internal array that maintains information about all
168 shared memory segments on the system.
169
170 SHM_STAT_ANY (Linux-specific, since Linux 4.17)
171 Return a shmid_ds structure as for SHM_STAT. However,
172 shm_perm.mode is not checked for read access for shmid, meaning
173 that any user can employ this operation (just as any user may
174 read /proc/sysvipc/shm to obtain the same information).
175
176 The caller can prevent or allow swapping of a shared memory segment
177 with the following cmd values:
178
179 SHM_LOCK (Linux-specific)
180 Prevent swapping of the shared memory segment. The caller must
181 fault in any pages that are required to be present after locking
182 is enabled. If a segment has been locked, then the (nonstan‐
183 dard) SHM_LOCKED flag of the shm_perm.mode field in the associ‐
184 ated data structure retrieved by IPC_STAT will be set.
185
186 SHM_UNLOCK (Linux-specific)
187 Unlock the segment, allowing it to be swapped out.
188
189 In kernels before 2.6.10, only a privileged process could employ
190 SHM_LOCK and SHM_UNLOCK. Since kernel 2.6.10, an unprivileged process
191 can employ these operations if its effective UID matches the owner or
192 creator UID of the segment, and (for SHM_LOCK) the amount of memory to
193 be locked falls within the RLIMIT_MEMLOCK resource limit (see setr‐
194 limit(2)).
195
197 A successful IPC_INFO or SHM_INFO operation returns the index of the
198 highest used entry in the kernel's internal array recording information
199 about all shared memory segments. (This information can be used with
200 repeated SHM_STAT or SHM_STAT_ANY operations to obtain information
201 about all shared memory segments on the system.) A successful SHM_STAT
202 operation returns the identifier of the shared memory segment whose in‐
203 dex was given in shmid. Other operations return 0 on success.
204
205 On error, -1 is returned, and errno is set to indicate the error.
206
208 EACCES IPC_STAT or SHM_STAT is requested and shm_perm.mode does not al‐
209 low read access for shmid, and the calling process does not have
210 the CAP_IPC_OWNER capability in the user namespace that governs
211 its IPC namespace.
212
213 EFAULT The argument cmd has value IPC_SET or IPC_STAT but the address
214 pointed to by buf isn't accessible.
215
216 EIDRM shmid points to a removed identifier.
217
218 EINVAL shmid is not a valid identifier, or cmd is not a valid command.
219 Or: for a SHM_STAT or SHM_STAT_ANY operation, the index value
220 specified in shmid referred to an array slot that is currently
221 unused.
222
223 ENOMEM (In kernels since 2.6.9), SHM_LOCK was specified and the size of
224 the to-be-locked segment would mean that the total bytes in
225 locked shared memory segments would exceed the limit for the
226 real user ID of the calling process. This limit is defined by
227 the RLIMIT_MEMLOCK soft resource limit (see setrlimit(2)).
228
229 EOVERFLOW
230 IPC_STAT is attempted, and the GID or UID value is too large to
231 be stored in the structure pointed to by buf.
232
233 EPERM IPC_SET or IPC_RMID is attempted, and the effective user ID of
234 the calling process is not that of the creator (found in
235 shm_perm.cuid), or the owner (found in shm_perm.uid), and the
236 process was not privileged (Linux: did not have the CAP_SYS_AD‐
237 MIN capability).
238
239 Or (in kernels before 2.6.9), SHM_LOCK or SHM_UNLOCK was speci‐
240 fied, but the process was not privileged (Linux: did not have
241 the CAP_IPC_LOCK capability). (Since Linux 2.6.9, this error
242 can also occur if the RLIMIT_MEMLOCK is 0 and the caller is not
243 privileged.)
244
246 POSIX.1-2001, POSIX.1-2008, SVr4.
247
249 The IPC_INFO, SHM_STAT, and SHM_INFO operations are used by the ipcs(1)
250 program to provide information on allocated resources. In the future,
251 these may modified or moved to a /proc filesystem interface.
252
253 Linux permits a process to attach (shmat(2)) a shared memory segment
254 that has already been marked for deletion using shmctl(IPC_RMID). This
255 feature is not available on other UNIX implementations; portable appli‐
256 cations should avoid relying on it.
257
258 Various fields in a struct shmid_ds were typed as short under Linux 2.2
259 and have become long under Linux 2.4. To take advantage of this, a re‐
260 compilation under glibc-2.1.91 or later should suffice. (The kernel
261 distinguishes old and new calls by an IPC_64 flag in cmd.)
262
264 mlock(2), setrlimit(2), shmget(2), shmop(2), capabilities(7),
265 sysvipc(7)
266
268 This page is part of release 5.13 of the Linux man-pages project. A
269 description of the project, information about reporting bugs, and the
270 latest version of this page, can be found at
271 https://www.kernel.org/doc/man-pages/.
272
273
274
275Linux 2021-03-22 SHMCTL(2)