1FORK(3P) POSIX Programmer's Manual FORK(3P)
2
6 This manual page is part of the POSIX Programmer's Manual. The Linux
7 implementation of this interface may differ (consult the corresponding
8 Linux manual page for details of Linux behavior), or the interface may
9 not be implemented on Linux.
10
12 fork — create a new process
13
15 #include <unistd.h>
16 pid_t fork(void);
18
20 The fork() function shall create a new process. The new process (child
21 process) shall be an exact copy of the calling process (parent process)
22 except as detailed below:
23 * The child process shall have a unique process ID.
25 * The child process ID also shall not match any active process group
27 ID.
28 * The child process shall have a different parent process ID, which
30 shall be the process ID of the calling process.
31 * The child process shall have its own copy of the parent's file
33 descriptors. Each of the child's file descriptors shall refer to
34 the same open file description with the corresponding file descrip‐
35 tor of the parent.
36 * The child process shall have its own copy of the parent's open
38 directory streams. Each open directory stream in the child process
39 may share directory stream positioning with the corresponding
40 directory stream of the parent.
41 * The child process shall have its own copy of the parent's message
43 catalog descriptors.
44 * The child process values of tms_utime, tms_stime, tms_cutime, and
46 tms_cstime shall be set to 0.
47 * The time left until an alarm clock signal shall be reset to zero,
49 and the alarm, if any, shall be canceled; see alarm().
50 * All semadj values shall be cleared.
52 * File locks set by the parent process shall not be inherited by the
54 child process.
55 * The set of signals pending for the child process shall be initial‐
57 ized to the empty set.
58 * Interval timers shall be reset in the child process.
60 * Any semaphores that are open in the parent process shall also be
62 open in the child process.
63 * The child process shall not inherit any address space memory locks
65 established by the parent process via calls to mlockall() or
66 mlock().
67 * Memory mappings created in the parent shall be retained in the
69 child process. MAP_PRIVATE mappings inherited from the parent shall
70 also be MAP_PRIVATE mappings in the child, and any modifications to
71 the data in these mappings made by the parent prior to calling
72 fork() shall be visible to the child. Any modifications to the data
73 in MAP_PRIVATE mappings made by the parent after fork() returns
74 shall be visible only to the parent. Modifications to the data in
75 MAP_PRIVATE mappings made by the child shall be visible only to the
76 child.
77 * For the SCHED_FIFO and SCHED_RR scheduling policies, the child
79 process shall inherit the policy and priority settings of the par‐
80 ent process during a fork() function. For other scheduling poli‐
81 cies, the policy and priority settings on fork() are implementa‐
82 tion-defined.
83 * Per-process timers created by the parent shall not be inherited by
85 the child process.
86 * The child process shall have its own copy of the message queue
88 descriptors of the parent. Each of the message descriptors of the
89 child shall refer to the same open message queue description as the
90 corresponding message descriptor of the parent.
91 * No asynchronous input or asynchronous output operations shall be
93 inherited by the child process. Any use of asynchronous control
94 blocks created by the parent produces undefined behavior.
95 * A process shall be created with a single thread. If a multi-
97 threaded process calls fork(), the new process shall contain a
98 replica of the calling thread and its entire address space, possi‐
99 bly including the states of mutexes and other resources. Conse‐
100 quently, to avoid errors, the child process may only execute async-
101 signal-safe operations until such time as one of the exec functions
102 is called.
103 When the application calls fork() from a signal handler and any of
105 the fork handlers registered by pthread_atfork() calls a function
106 that is not async-signal-safe, the behavior is undefined.
107 * If the Trace option and the Trace Inherit option are both sup‐
109 ported:
110 If the calling process was being traced in a trace stream that had
112 its inheritance policy set to POSIX_TRACE_INHERITED, the child
113 process shall be traced into that trace stream, and the child
114 process shall inherit the parent's mapping of trace event names to
115 trace event type identifiers. If the trace stream in which the
116 calling process was being traced had its inheritance policy set to
117 POSIX_TRACE_CLOSE_FOR_CHILD, the child process shall not be traced
118 into that trace stream. The inheritance policy is set by a call to
119 the posix_trace_attr_setinherited() function.
120 * If the Trace option is supported, but the Trace Inherit option is
122 not supported:
123 The child process shall not be traced into any of the trace streams
125 of its parent process.
126 * If the Trace option is supported, the child process of a trace con‐
128 troller process shall not control the trace streams controlled by
129 its parent process.
130 * The initial value of the CPU-time clock of the child process shall
132 be set to zero.
133 * The initial value of the CPU-time clock of the single thread of the
135 child process shall be set to zero.
136 All other process characteristics defined by POSIX.1‐2008 shall be the
138 same in the parent and child processes. The inheritance of process
139 characteristics not defined by POSIX.1‐2008 is unspecified by
140 POSIX.1‐2008.
141 After fork(), both the parent and the child processes shall be capable
143 of executing independently before either one terminates.
144
146 Upon successful completion, fork() shall return 0 to the child process
147 and shall return the process ID of the child process to the parent
148 process. Both processes shall continue to execute from the fork() func‐
149 tion. Otherwise, -1 shall be returned to the parent process, no child
150 process shall be created, and errno shall be set to indicate the error.
151
153 The fork() function shall fail if:
154 EAGAIN The system lacked the necessary resources to create another
156 process, or the system-imposed limit on the total number of pro‐
157 cesses under execution system-wide or by a single user
158 {CHILD_MAX} would be exceeded.
159 The fork() function may fail if:
161 ENOMEM Insufficient storage space is available.
163 The following sections are informative.
165
167 None.
168
170 None.
171
173 Many historical implementations have timing windows where a signal sent
174 to a process group (for example, an interactive SIGINT) just prior to
175 or during execution of fork() is delivered to the parent following the
176 fork() but not to the child because the fork() code clears the child's
177 set of pending signals. This volume of POSIX.1‐2017 does not require,
178 or even permit, this behavior. However, it is pragmatic to expect that
179 problems of this nature may continue to exist in implementations that
180 appear to conform to this volume of POSIX.1‐2017 and pass available
181 verification suites. This behavior is only a consequence of the imple‐
182 mentation failing to make the interval between signal generation and
183 delivery totally invisible. From the application's perspective, a
184 fork() call should appear atomic. A signal that is generated prior to
185 the fork() should be delivered prior to the fork(). A signal sent to
186 the process group after the fork() should be delivered to both parent
187 and child. The implementation may actually initialize internal data
188 structures corresponding to the child's set of pending signals to
189 include signals sent to the process group during the fork(). Since the
190 fork() call can be considered as atomic from the application's perspec‐
191 tive, the set would be initialized as empty and such signals would have
192 arrived after the fork(); see also <signal.h>.
193 One approach that has been suggested to address the problem of signal
195 inheritance across fork() is to add an [EINTR] error, which would be
196 returned when a signal is detected during the call. While this is
197 preferable to losing signals, it was not considered an optimal solu‐
198 tion. Although it is not recommended for this purpose, such an error
199 would be an allowable extension for an implementation.
200 The [ENOMEM] error value is reserved for those implementations that
202 detect and distinguish such a condition. This condition occurs when an
203 implementation detects that there is not enough memory to create the
204 process. This is intended to be returned when [EAGAIN] is inappropriate
205 because there can never be enough memory (either primary or secondary
206 storage) to perform the operation. Since fork() duplicates an existing
207 process, this must be a condition where there is sufficient memory for
208 one such process, but not for two. Many historical implementations
209 actually return [ENOMEM] due to temporary lack of memory, a case that
210 is not generally distinct from [EAGAIN] from the perspective of a con‐
211 forming application.
212 Part of the reason for including the optional error [ENOMEM] is because
214 the SVID specifies it and it should be reserved for the error condition
215 specified there. The condition is not applicable on many implementa‐
216 tions.
217 IEEE Std 1003.1‐1988 neglected to require concurrent execution of the
219 parent and child of fork(). A system that single-threads processes was
220 clearly not intended and is considered an unacceptable ``toy implemen‐
221 tation'' of this volume of POSIX.1‐2017. The only objection antici‐
222 pated to the phrase ``executing independently'' is testability, but
223 this assertion should be testable. Such tests require that both the
224 parent and child can block on a detectable action of the other, such as
225 a write to a pipe or a signal. An interactive exchange of such actions
226 should be possible for the system to conform to the intent of this vol‐
227 ume of POSIX.1‐2017.
228 The [EAGAIN] error exists to warn applications that such a condition
230 might occur. Whether it occurs or not is not in any practical sense
231 under the control of the application because the condition is usually a
232 consequence of the user's use of the system, not of the application's
233 code. Thus, no application can or should rely upon its occurrence under
234 any circumstances, nor should the exact semantics of what concept of
235 ``user'' is used be of concern to the application developer. Valida‐
236 tion writers should be cognizant of this limitation.
237 There are two reasons why POSIX programmers call fork(). One reason is
239 to create a new thread of control within the same program (which was
240 originally only possible in POSIX by creating a new process); the other
241 is to create a new process running a different program. In the latter
242 case, the call to fork() is soon followed by a call to one of the exec
243 functions.
244 The general problem with making fork() work in a multi-threaded world
246 is what to do with all of the threads. There are two alternatives. One
247 is to copy all of the threads into the new process. This causes the
248 programmer or implementation to deal with threads that are suspended on
249 system calls or that might be about to execute system calls that should
250 not be executed in the new process. The other alternative is to copy
251 only the thread that calls fork(). This creates the difficulty that
252 the state of process-local resources is usually held in process memory.
253 If a thread that is not calling fork() holds a resource, that resource
254 is never released in the child process because the thread whose job it
255 is to release the resource does not exist in the child process.
256 When a programmer is writing a multi-threaded program, the first
258 described use of fork(), creating new threads in the same program, is
259 provided by the pthread_create() function. The fork() function is thus
260 used only to run new programs, and the effects of calling functions
261 that require certain resources between the call to fork() and the call
262 to an exec function are undefined.
263 The addition of the forkall() function to the standard was considered
265 and rejected. The forkall() function lets all the threads in the parent
266 be duplicated in the child. This essentially duplicates the state of
267 the parent in the child. This allows threads in the child to continue
268 processing and allows locks and the state to be preserved without
269 explicit pthread_atfork() code. The calling process has to ensure that
270 the threads processing state that is shared between the parent and
271 child (that is, file descriptors or MAP_SHARED memory) behaves properly
272 after forkall(). For example, if a thread is reading a file descriptor
273 in the parent when forkall() is called, then two threads (one in the
274 parent and one in the child) are reading the file descriptor after the
275 forkall(). If this is not desired behavior, the parent process has to
276 synchronize with such threads before calling forkall().
277 While the fork() function is async-signal-safe, there is no way for an
279 implementation to determine whether the fork handlers established by
280 pthread_atfork() are async-signal-safe. The fork handlers may attempt
281 to execute portions of the implementation that are not async-signal-
282 safe, such as those that are protected by mutexes, leading to a dead‐
283 lock condition. It is therefore undefined for the fork handlers to
284 execute functions that are not async-signal-safe when fork() is called
285 from a signal handler.
286 When forkall() is called, threads, other than the calling thread, that
288 are in functions that can return with an [EINTR] error may have those
289 functions return [EINTR] if the implementation cannot ensure that the
290 function behaves correctly in the parent and child. In particular,
291 pthread_cond_wait() and pthread_cond_timedwait() need to return in
292 order to ensure that the condition has not changed. These functions
293 can be awakened by a spurious condition wakeup rather than returning
294 [EINTR].
295
297 None.
298
300 alarm(), exec, fcntl(), posix_trace_attr_getinherited(),
301 posix_trace_eventid_equal(), pthread_atfork(), semop(), signal(),
302 times()
303 The Base Definitions volume of POSIX.1‐2017, Section 4.12, Memory Syn‐
305 chronization, <sys_types.h>, <unistd.h>
306
308 Portions of this text are reprinted and reproduced in electronic form
309 from IEEE Std 1003.1-2017, Standard for Information Technology -- Por‐
310 table Operating System Interface (POSIX), The Open Group Base Specifi‐
311 cations Issue 7, 2018 Edition, Copyright (C) 2018 by the Institute of
312 Electrical and Electronics Engineers, Inc and The Open Group. In the
313 event of any discrepancy between this version and the original IEEE and
314 The Open Group Standard, the original IEEE and The Open Group Standard
315 is the referee document. The original Standard can be obtained online
316 at http://www.opengroup.org/unix/online.html .
317 Any typographical or formatting errors that appear in this page are
319 most likely to have been introduced during the conversion of the source
320 files to man page format. To report such errors, see https://www.ker‐
321 nel.org/doc/man-pages/reporting_bugs.html .
322IEEE/The Open Group 2017 FORK(3P)