1fork(2) System Calls Manual fork(2)
2
6 fork - create a child process
7
9 Standard C library (libc, -lc)
10
12 #include <unistd.h>
13 pid_t fork(void);
15
17 fork() creates a new process by duplicating the calling process. The
18 new process is referred to as the child process. The calling process
19 is referred to as the parent process.
20 The child process and the parent process run in separate memory spaces.
22 At the time of fork() both memory spaces have the same content. Memory
23 writes, file mappings (mmap(2)), and unmappings (munmap(2)) performed
24 by one of the processes do not affect the other.
25 The child process is an exact duplicate of the parent process except
27 for the following points:
28 • The child has its own unique process ID, and this PID does not match
30 the ID of any existing process group (setpgid(2)) or session.
31 • The child's parent process ID is the same as the parent's process
33 ID.
34 • The child does not inherit its parent's memory locks (mlock(2),
36 mlockall(2)).
37 • Process resource utilizations (getrusage(2)) and CPU time counters
39 (times(2)) are reset to zero in the child.
40 • The child's set of pending signals is initially empty (sigpend‐
42 ing(2)).
43 • The child does not inherit semaphore adjustments from its parent
45 (semop(2)).
46 • The child does not inherit process-associated record locks from its
48 parent (fcntl(2)). (On the other hand, it does inherit fcntl(2)
49 open file description locks and flock(2) locks from its parent.)
50 • The child does not inherit timers from its parent (setitimer(2),
52 alarm(2), timer_create(2)).
53 • The child does not inherit outstanding asynchronous I/O operations
55 from its parent (aio_read(3), aio_write(3)), nor does it inherit any
56 asynchronous I/O contexts from its parent (see io_setup(2)).
57 The process attributes in the preceding list are all specified in
59 POSIX.1. The parent and child also differ with respect to the follow‐
60 ing Linux-specific process attributes:
61 • The child does not inherit directory change notifications (dnotify)
63 from its parent (see the description of F_NOTIFY in fcntl(2)).
64 • The prctl(2) PR_SET_PDEATHSIG setting is reset so that the child
66 does not receive a signal when its parent terminates.
67 • The default timer slack value is set to the parent's current timer
69 slack value. See the description of PR_SET_TIMERSLACK in prctl(2).
70 • Memory mappings that have been marked with the madvise(2) MADV_DONT‐
72 FORK flag are not inherited across a fork().
73 • Memory in address ranges that have been marked with the madvise(2)
75 MADV_WIPEONFORK flag is zeroed in the child after a fork(). (The
76 MADV_WIPEONFORK setting remains in place for those address ranges in
77 the child.)
78 • The termination signal of the child is always SIGCHLD (see
80 clone(2)).
81 • The port access permission bits set by ioperm(2) are not inherited
83 by the child; the child must turn on any bits that it requires using
84 ioperm(2).
85 Note the following further points:
87 • The child process is created with a single thread—the one that
89 called fork(). The entire virtual address space of the parent is
90 replicated in the child, including the states of mutexes, condition
91 variables, and other pthreads objects; the use of pthread_atfork(3)
92 may be helpful for dealing with problems that this can cause.
93 • After a fork() in a multithreaded program, the child can safely call
95 only async-signal-safe functions (see signal-safety(7)) until such
96 time as it calls execve(2).
97 • The child inherits copies of the parent's set of open file descrip‐
99 tors. Each file descriptor in the child refers to the same open
100 file description (see open(2)) as the corresponding file descriptor
101 in the parent. This means that the two file descriptors share open
102 file status flags, file offset, and signal-driven I/O attributes
103 (see the description of F_SETOWN and F_SETSIG in fcntl(2)).
104 • The child inherits copies of the parent's set of open message queue
106 descriptors (see mq_overview(7)). Each file descriptor in the child
107 refers to the same open message queue description as the correspond‐
108 ing file descriptor in the parent. This means that the two file de‐
109 scriptors share the same flags (mq_flags).
110 • The child inherits copies of the parent's set of open directory
112 streams (see opendir(3)). POSIX.1 says that the corresponding di‐
113 rectory streams in the parent and child may share the directory
114 stream positioning; on Linux/glibc they do not.
115
117 On success, the PID of the child process is returned in the parent, and
118 0 is returned in the child. On failure, -1 is returned in the parent,
119 no child process is created, and errno is set to indicate the error.
120
122 EAGAIN A system-imposed limit on the number of threads was encountered.
123 There are a number of limits that may trigger this error:
124 • the RLIMIT_NPROC soft resource limit (set via setrlimit(2)),
126 which limits the number of processes and threads for a real
127 user ID, was reached;
128 • the kernel's system-wide limit on the number of processes and
130 threads, /proc/sys/kernel/threads-max, was reached (see
131 proc(5));
132 • the maximum number of PIDs, /proc/sys/kernel/pid_max, was
134 reached (see proc(5)); or
135 • the PID limit (pids.max) imposed by the cgroup "process num‐
137 ber" (PIDs) controller was reached.
138 EAGAIN The caller is operating under the SCHED_DEADLINE scheduling pol‐
140 icy and does not have the reset-on-fork flag set. See sched(7).
141 ENOMEM fork() failed to allocate the necessary kernel structures be‐
143 cause memory is tight.
144 ENOMEM An attempt was made to create a child process in a PID namespace
146 whose "init" process has terminated. See pid_namespaces(7).
147 ENOSYS fork() is not supported on this platform (for example, hardware
149 without a Memory-Management Unit).
150 ERESTARTNOINTR (since Linux 2.6.17)
152 System call was interrupted by a signal and will be restarted.
153 (This can be seen only during a trace.)
154
156 C library/kernel differences
157 Since glibc 2.3.3, rather than invoking the kernel's fork() system
158 call, the glibc fork() wrapper that is provided as part of the NPTL
159 threading implementation invokes clone(2) with flags that provide the
160 same effect as the traditional system call. (A call to fork() is
161 equivalent to a call to clone(2) specifying flags as just SIGCHLD.)
162 The glibc wrapper invokes any fork handlers that have been established
163 using pthread_atfork(3).
164
166 POSIX.1-2008.
167
169 POSIX.1-2001, SVr4, 4.3BSD.
170
172 Under Linux, fork() is implemented using copy-on-write pages, so the
173 only penalty that it incurs is the time and memory required to dupli‐
174 cate the parent's page tables, and to create a unique task structure
175 for the child.
176
178 See pipe(2) and wait(2) for more examples.
179 #include <signal.h>
181 #include <stdint.h>
182 #include <stdio.h>
183 #include <stdlib.h>
184 #include <unistd.h>
185 int
187 main(void)
188 {
189 pid_t pid;
190 if (signal(SIGCHLD, SIG_IGN) == SIG_ERR) {
192 perror("signal");
193 exit(EXIT_FAILURE);
194 }
195 pid = fork();
196 switch (pid) {
197 case -1:
198 perror("fork");
199 exit(EXIT_FAILURE);
200 case 0:
201 puts("Child exiting.");
202 exit(EXIT_SUCCESS);
203 default:
204 printf("Child is PID %jd\n", (intmax_t) pid);
205 puts("Parent exiting.");
206 exit(EXIT_SUCCESS);
207 }
208 }
209
211 clone(2), execve(2), exit(2), setrlimit(2), unshare(2), vfork(2),
212 wait(2), daemon(3), pthread_atfork(3), capabilities(7), credentials(7)
213Linux man-pages 6.04 2023-03-30 fork(2)