1FORK(2) Linux Programmer's Manual FORK(2)
2
6 fork - create a child process
7
9 #include <unistd.h>
10 pid_t fork(void);
12
14 fork() creates a new process by duplicating the calling process. The
15 new process, referred to as the child, is an exact duplicate of the
16 calling process, referred to as the parent, except for the following
17 points:
18 * The child has its own unique process ID, and this PID does not match
20 the ID of any existing process group (setpgid(2)).
21 * The child's parent process ID is the same as the parent's process
23 ID.
24 * The child does not inherit its parent's memory locks (mlock(2),
26 mlockall(2)).
27 * Process resource utilizations (getrusage(2)) and CPU time counters
29 (times(2)) are reset to zero in the child.
30 * The child's set of pending signals is initially empty (sigpend‐
32 ing(2)).
33 * The child does not inherit semaphore adjustments from its parent
35 (semop(2)).
36 * The child does not inherit record locks from its parent (fcntl(2)).
38 * The child does not inherit timers from its parent (setitimer(2),
40 alarm(2), timer_create(2)).
41 * The child does not inherit outstanding asynchronous I/O operations
43 from its parent (aio_read(3), aio_write(3)), nor does it inherit any
44 asynchronous I/O contexts from its parent (see io_setup(2)).
45 The process attributes in the preceding list are all specified in
47 POSIX.1-2001. The parent and child also differ with respect to the
48 following Linux-specific process attributes:
49 * The child does not inherit directory change notifications (dnotify)
51 from its parent (see the description of F_NOTIFY in fcntl(2)).
52 * The prctl(2) PR_SET_PDEATHSIG setting is reset so that the child
54 does not receive a signal when its parent terminates.
55 * The default timer slack value is set to the parent's current timer
57 slack value. See the description of PR_SET_TIMERSLACK in prctl(2).
58 * Memory mappings that have been marked with the madvise(2) MADV_DONT‐
60 FORK flag are not inherited across a fork().
61 * The termination signal of the child is always SIGCHLD (see
63 clone(2)).
64 * The port access permission bits set by ioperm(2) are not inherited
66 by the child; the child must turn on any bits that it requires using
67 ioperm(2).
68 Note the following further points:
70 * The child process is created with a single thread—the one that
72 called fork(). The entire virtual address space of the parent is
73 replicated in the child, including the states of mutexes, condition
74 variables, and other pthreads objects; the use of pthread_atfork(3)
75 may be helpful for dealing with problems that this can cause.
76 * The child inherits copies of the parent's set of open file descrip‐
78 tors. Each file descriptor in the child refers to the same open
79 file description (see open(2)) as the corresponding file descriptor
80 in the parent. This means that the two descriptors share open file
81 status flags, current file offset, and signal-driven I/O attributes
82 (see the description of F_SETOWN and F_SETSIG in fcntl(2)).
83 * The child inherits copies of the parent's set of open message queue
85 descriptors (see mq_overview(7)). Each descriptor in the child
86 refers to the same open message queue description as the correspond‐
87 ing descriptor in the parent. This means that the two descriptors
88 share the same flags (mq_flags).
89 * The child inherits copies of the parent's set of open directory
91 streams (see opendir(3)). POSIX.1-2001 says that the corresponding
92 directory streams in the parent and child may share the directory
93 stream positioning; on Linux/glibc they do not.
94
96 On success, the PID of the child process is returned in the parent, and
97 0 is returned in the child. On failure, -1 is returned in the parent,
98 no child process is created, and errno is set appropriately.
99
101 EAGAIN fork() cannot allocate sufficient memory to copy the parent's
102 page tables and allocate a task structure for the child.
103 EAGAIN It was not possible to create a new process because the caller's
105 RLIMIT_NPROC resource limit was encountered. To exceed this
106 limit, the process must have either the CAP_SYS_ADMIN or the
107 CAP_SYS_RESOURCE capability.
108 ENOMEM fork() failed to allocate the necessary kernel structures
110 because memory is tight.
111 ENOSYS fork() is not supported on this platform (for example, hardware
113 without a Memory-Management Unit).
114
116 SVr4, 4.3BSD, POSIX.1-2001.
117
119 Under Linux, fork() is implemented using copy-on-write pages, so the
120 only penalty that it incurs is the time and memory required to dupli‐
121 cate the parent's page tables, and to create a unique task structure
122 for the child.
123 Since version 2.3.3, rather than invoking the kernel's fork() system
125 call, the glibc fork() wrapper that is provided as part of the NPTL
126 threading implementation invokes clone(2) with flags that provide the
127 same effect as the traditional system call. (A call to fork() is
128 equivalent to a call to clone(2) specifying flags as just SIGCHLD.)
129 The glibc wrapper invokes any fork handlers that have been established
130 using pthread_atfork(3).
131
133 See pipe(2) and wait(2).
134
136 clone(2), execve(2), exit(2), setrlimit(2), unshare(2), vfork(2),
137 wait(2), daemon(3), capabilities(7), credentials(7)
138
140 This page is part of release 3.53 of the Linux man-pages project. A
141 description of the project, and information about reporting bugs, can
142 be found at http://www.kernel.org/doc/man-pages/.
143Linux 2013-03-12 FORK(2)