1FORK(P) POSIX Programmer's Manual FORK(P)
2
6 fork - create a new process
7
9 #include <unistd.h>
10 pid_t fork(void);
12
15 The fork() function shall create a new process. The new process (child
16 process) shall be an exact copy of the calling process (parent process)
17 except as detailed below:
18 * The child process shall have a unique process ID.
20 * The child process ID also shall not match any active process group
22 ID.
23 * The child process shall have a different parent process ID, which
25 shall be the process ID of the calling process.
26 * The child process shall have its own copy of the parent's file
28 descriptors. Each of the child's file descriptors shall refer to
29 the same open file description with the corresponding file descrip‐
30 tor of the parent.
31 * The child process shall have its own copy of the parent's open
33 directory streams. Each open directory stream in the child process
34 may share directory stream positioning with the corresponding direc‐
35 tory stream of the parent.
36 * The child process shall have its own copy of the parent's message
38 catalog descriptors.
39 * The child process' values of tms_utime, tms_stime, tms_cutime, and
41 tms_cstime shall be set to 0.
42 * The time left until an alarm clock signal shall be reset to zero,
44 and the alarm, if any, shall be canceled; see alarm() .
45 * All semadj values shall be cleared.
47 * File locks set by the parent process shall not be inherited by the
49 child process.
50 * The set of signals pending for the child process shall be initial‐
52 ized to the empty set.
53 * Interval timers shall be reset in the child process.
55 * Any semaphores that are open in the parent process shall also be
57 open in the child process.
58 * The child process shall not inherit any address space memory locks
60 established by the parent process via calls to mlockall() or
61 mlock().
62 * Memory mappings created in the parent shall be retained in the child
64 process. MAP_PRIVATE mappings inherited from the parent shall also
65 be MAP_PRIVATE mappings in the child, and any modifications to the
66 data in these mappings made by the parent prior to calling fork()
67 shall be visible to the child. Any modifications to the data in
68 MAP_PRIVATE mappings made by the parent after fork() returns shall
69 be visible only to the parent. Modifications to the data in MAP_PRI‐
70 VATE mappings made by the child shall be visible only to the child.
71 * For the SCHED_FIFO and SCHED_RR scheduling policies, the child
73 process shall inherit the policy and priority settings of the parent
74 process during a fork() function. For other scheduling policies, the
75 policy and priority settings on fork() are implementation-defined.
76 * Per-process timers created by the parent shall not be inherited by
78 the child process.
79 * The child process shall have its own copy of the message queue
81 descriptors of the parent. Each of the message descriptors of the
82 child shall refer to the same open message queue description as the
83 corresponding message descriptor of the parent.
84 * No asynchronous input or asynchronous output operations shall be
86 inherited by the child process.
87 * A process shall be created with a single thread. If a multi-threaded
89 process calls fork(), the new process shall contain a replica of the
90 calling thread and its entire address space, possibly including the
91 states of mutexes and other resources. Consequently, to avoid
92 errors, the child process may only execute async-signal-safe opera‐
93 tions until such time as one of the exec functions is called.
94 Fork handlers may be established by means of the pthread_atfork()
95 function in order to maintain application invariants across fork()
96 calls.
97 When the application calls fork() from a signal handler and any of the
99 fork handlers registered by pthread_atfork() calls a function that is
100 not asynch-signal-safe, the behavior is undefined.
101 * If the Trace option and the Trace Inherit option are both supported:
103 If the calling process was being traced in a trace stream that had its
105 inheritance policy set to POSIX_TRACE_INHERITED, the child process
106 shall be traced into that trace stream, and the child process shall
107 inherit the parent's mapping of trace event names to trace event type
108 identifiers. If the trace stream in which the calling process was being
109 traced had its inheritance policy set to POSIX_TRACE_CLOSE_FOR_CHILD,
110 the child process shall not be traced into that trace stream. The
111 inheritance policy is set by a call to the posix_trace_attr_setinher‐
112 ited() function.
113 * If the Trace option is supported, but the Trace Inherit option is
115 not supported:
116 The child process shall not be traced into any of the trace streams of
118 its parent process.
119 * If the Trace option is supported, the child process of a trace con‐
121 troller process shall not control the trace streams controlled by
122 its parent process.
123 * The initial value of the CPU-time clock of the child process shall
125 be set to zero.
126 * The initial value of the CPU-time clock of the single thread of the
128 child process shall be set to zero.
129 All other process characteristics defined by IEEE Std 1003.1-2001 shall
131 be the same in the parent and child processes. The inheritance of
132 process characteristics not defined by IEEE Std 1003.1-2001 is unspeci‐
133 fied by IEEE Std 1003.1-2001.
134 After fork(), both the parent and the child processes shall be capable
136 of executing independently before either one terminates.
137
139 Upon successful completion, fork() shall return 0 to the child process
140 and shall return the process ID of the child process to the parent
141 process. Both processes shall continue to execute from the fork() func‐
142 tion. Otherwise, -1 shall be returned to the parent process, no child
143 process shall be created, and errno shall be set to indicate the error.
144
146 The fork() function shall fail if:
147 EAGAIN The system lacked the necessary resources to create another
149 process, or the system-imposed limit on the total number of pro‐
150 cesses under execution system-wide or by a single user
151 {CHILD_MAX} would be exceeded.
152 The fork() function may fail if:
155 ENOMEM Insufficient storage space is available.
157 The following sections are informative.
160
162 None.
163
165 None.
166
168 Many historical implementations have timing windows where a signal sent
169 to a process group (for example, an interactive SIGINT) just prior to
170 or during execution of fork() is delivered to the parent following the
171 fork() but not to the child because the fork() code clears the child's
172 set of pending signals. This volume of IEEE Std 1003.1-2001 does not
173 require, or even permit, this behavior. However, it is pragmatic to
174 expect that problems of this nature may continue to exist in implemen‐
175 tations that appear to conform to this volume of IEEE Std 1003.1-2001
176 and pass available verification suites. This behavior is only a conse‐
177 quence of the implementation failing to make the interval between sig‐
178 nal generation and delivery totally invisible. From the application's
179 perspective, a fork() call should appear atomic. A signal that is gen‐
180 erated prior to the fork() should be delivered prior to the fork(). A
181 signal sent to the process group after the fork() should be delivered
182 to both parent and child. The implementation may actually initialize
183 internal data structures corresponding to the child's set of pending
184 signals to include signals sent to the process group during the fork().
185 Since the fork() call can be considered as atomic from the applica‐
186 tion's perspective, the set would be initialized as empty and such sig‐
187 nals would have arrived after the fork(); see also <signal.h>.
188 One approach that has been suggested to address the problem of signal
190 inheritance across fork() is to add an [EINTR] error, which would be
191 returned when a signal is detected during the call. While this is
192 preferable to losing signals, it was not considered an optimal solu‐
193 tion. Although it is not recommended for this purpose, such an error
194 would be an allowable extension for an implementation.
195 The [ENOMEM] error value is reserved for those implementations that
197 detect and distinguish such a condition. This condition occurs when an
198 implementation detects that there is not enough memory to create the
199 process. This is intended to be returned when [EAGAIN] is inappropriate
200 because there can never be enough memory (either primary or secondary
201 storage) to perform the operation. Since fork() duplicates an existing
202 process, this must be a condition where there is sufficient memory for
203 one such process, but not for two. Many historical implementations
204 actually return [ENOMEM] due to temporary lack of memory, a case that
205 is not generally distinct from [EAGAIN] from the perspective of a con‐
206 forming application.
207 Part of the reason for including the optional error [ENOMEM] is because
209 the SVID specifies it and it should be reserved for the error condition
210 specified there. The condition is not applicable on many implementa‐
211 tions.
212 IEEE Std 1003.1-1988 neglected to require concurrent execution of the
214 parent and child of fork(). A system that single-threads processes was
215 clearly not intended and is considered an unacceptable "toy implementa‐
216 tion" of this volume of IEEE Std 1003.1-2001. The only objection antic‐
217 ipated to the phrase "executing independently" is testability, but this
218 assertion should be testable. Such tests require that both the parent
219 and child can block on a detectable action of the other, such as a
220 write to a pipe or a signal. An interactive exchange of such actions
221 should be possible for the system to conform to the intent of this vol‐
222 ume of IEEE Std 1003.1-2001.
223 The [EAGAIN] error exists to warn applications that such a condition
225 might occur. Whether it occurs or not is not in any practical sense
226 under the control of the application because the condition is usually a
227 consequence of the user's use of the system, not of the application's
228 code. Thus, no application can or should rely upon its occurrence under
229 any circumstances, nor should the exact semantics of what concept of
230 "user" is used be of concern to the application writer. Validation
231 writers should be cognizant of this limitation.
232 There are two reasons why POSIX programmers call fork(). One reason is
234 to create a new thread of control within the same program (which was
235 originally only possible in POSIX by creating a new process); the other
236 is to create a new process running a different program. In the latter
237 case, the call to fork() is soon followed by a call to one of the exec
238 functions.
239 The general problem with making fork() work in a multi-threaded world
241 is what to do with all of the threads. There are two alternatives. One
242 is to copy all of the threads into the new process. This causes the
243 programmer or implementation to deal with threads that are suspended on
244 system calls or that might be about to execute system calls that should
245 not be executed in the new process. The other alternative is to copy
246 only the thread that calls fork(). This creates the difficulty that the
247 state of process-local resources is usually held in process memory. If
248 a thread that is not calling fork() holds a resource, that resource is
249 never released in the child process because the thread whose job it is
250 to release the resource does not exist in the child process.
251 When a programmer is writing a multi-threaded program, the first
253 described use of fork(), creating new threads in the same program, is
254 provided by the pthread_create() function. The fork() function is thus
255 used only to run new programs, and the effects of calling functions
256 that require certain resources between the call to fork() and the call
257 to an exec function are undefined.
258 The addition of the forkall() function to the standard was considered
260 and rejected. The forkall() function lets all the threads in the parent
261 be duplicated in the child. This essentially duplicates the state of
262 the parent in the child. This allows threads in the child to continue
263 processing and allows locks and the state to be preserved without
264 explicit pthread_atfork() code. The calling process has to ensure that
265 the threads processing state that is shared between the parent and
266 child (that is, file descriptors or MAP_SHARED memory) behaves properly
267 after forkall(). For example, if a thread is reading a file descriptor
268 in the parent when forkall() is called, then two threads (one in the
269 parent and one in the child) are reading the file descriptor after the
270 forkall(). If this is not desired behavior, the parent process has to
271 synchronize with such threads before calling forkall().
272 While the fork() function is async-signal-safe, there is no way for an
274 implementation to determine whether the fork handlers established by
275 pthread_atfork() are async-signal-safe. The fork handlers may attempt
276 to execute portions of the implementation that are not async-signal-
277 safe, such as those that are protected by mutexes, leading to a dead‐
278 lock condition. It is therefore undefined for the fork handlers to exe‐
279 cute functions that are not async-signal-safe when fork() is called
280 from a signal handler.
281 When forkall() is called, threads, other than the calling thread, that
283 are in functions that can return with an [EINTR] error may have those
284 functions return [EINTR] if the implementation cannot ensure that the
285 function behaves correctly in the parent and child. In particular,
286 pthread_cond_wait() and pthread_cond_timedwait() need to return in
287 order to ensure that the condition has not changed. These functions can
288 be awakened by a spurious condition wakeup rather than returning
289 [EINTR].
290
292 None.
293
295 alarm() , exec() , fcntl() , posix_trace_attr_getinherited() ,
296 posix_trace_trid_eventid_open() , pthread_atfork() , semop() , signal()
297 , times() , the Base Definitions volume of IEEE Std 1003.1-2001,
298 <sys/types.h>, <unistd.h>
299
301 Portions of this text are reprinted and reproduced in electronic form
302 from IEEE Std 1003.1, 2003 Edition, Standard for Information Technology
303 -- Portable Operating System Interface (POSIX), The Open Group Base
304 Specifications Issue 6, Copyright (C) 2001-2003 by the Institute of
305 Electrical and Electronics Engineers, Inc and The Open Group. In the
306 event of any discrepancy between this version and the original IEEE and
307 The Open Group Standard, the original IEEE and The Open Group Standard
308 is the referee document. The original Standard can be obtained online
309 at http://www.opengroup.org/unix/online.html .
310IEEE/The Open Group 2003 FORK(P)