fork(S)

fork, forkall, fork1 -- create a new process

Synopsis

   #include <sys/types.h>


   #include <unistd.h>
   
   pid_t fork(void);
   

   pid_t fork1(void);
   

   pid_t forkall(void);

Description

fork causes creation of a new process. The new process (child process) is an exact copy of the calling process (parent process). This means the child process inherits the following attributes from the parent process:

real user ID, real group ID, effective user ID, effective group ID
environment
close-on-exec flag [see exec(S)]
signal handling settings (that is, SIG_DFL, SIG_IGN, SIG_HOLD, function address)
supplementary group IDs
set-user-ID mode bit
set-group-ID mode bit
profiling on/off status
nice value [see nice(S)]
scheduler class [see priocntl(S)]
all attached shared memory segments [see shmop]
process group ID
session ID [see exit]
current working directory
root directory
file mode creation mask [see umask(S)]
resource limits [see getrlimit]
controlling terminal
working and maximum privilege sets

Scheduling priority and any per-process scheduling parameters that are specific to a given scheduling class may or may not be inherited according to the policy of that particular class (see priocntl(S)).

Multithreaded processes should call:

fork: to create a single-thread child process. (This is the standard call for X/Open- and POSIX-conformant applications.)
fork1: if they will exec immediately.
forkall: to duplicate all threads and LWPs in the child.

By default, for multi-threaded applications, a call to fork is equivalent to a call to forkall. A new process created by forkall duplicates the calling process, including the set of threads (and underlying LWPs) that exist at the time of the call. (See ``Compatibility and standards conformance,'' below.)

The fork1 system call will create a new process with a single LWP. The parent process is not changed. The fork1 system call should be used only by multithreaded processes that intend to have the new process immediately call the exec system call. Since exec will terminate all but one thread, there is no need to duplicate all threads with forkall.

If the creation of the new process would use the last process table slot, the calling process must have the P_SYSOPS privilege.

The child process differs from the parent process in the following ways:

The child process has a unique process ID which does not match any active process group ID.
The child process has a different parent process ID (that is, the process ID of the parent process).
The child process has its own copy of the parent's file descriptors and directory streams. Each of the child's file descriptors shares a common file pointer with the corresponding file descriptor of the parent.
All semadj values are cleared [see semop].
Process locks, text locks and data locks are not inherited by the child [see plock].
The child process's tms structure is cleared: tms_utime, stime, cutime, and cstime are set to 0 [see times].
The time left until an alarm clock signal is reset to 0.
The set of signals pending for the child process is initialized to the empty set.
Record locks set by the parent process are not inherited by the child process [see fcntl(S)].
The following statement applies only to forkall: LWPs waiting on a synchronization queue may not be in the same order in the child process, but the order of release of LWPs from a synchronization object is scheduling policy-specific. If FIFO ordering on a synchronization object is required, it is called out in the specification of that object.

Return values

On success, fork, fork1, and forkall return 0 to the child process and return the process ID of the child process to the parent process. On failure, fork, fork1, and forkall return a value of (pid_t)-1 to the parent process, set errno to identify the error, and do not create a child process.

Errors

In the following conditions, fork fails and sets errno to:

EAGAIN: The system limit on number of LWPs per real user-id would be exceeded if the call succeeds, and the calling process does not have the P_SYSOPS privilege. The system lacked the necessary resources to create another process.
EAGAIN: Total amount of system memory available when reading via raw I/O is temporarily insufficient.
EINTR: A forkall has been interrupted by a forkall executed by another thread in the process.

References

alarm(S), exec(S), fcntl(S), getrlimit(S), nice(S), plock(S), priocntl(S), ptrace(S), semop(S), shmop(S), signal(S), system(S), times(S), umask(S), wait(S)

Compatibility and standards conformance

By default, fork is synonymous with forkall.

The X/Open and POSIX specifications, however, equate a call to fork with a call to fork1. To enable the POSIX behavior for an application, the application must include the following global definition in its source code:

   int	__thread_sys_behavior = 1;

When a multi-threaded process calls fork, and the process defines __thread_sys_behavior as above, the new process contains a replica of the calling thread and its entire address space, possibly including the states of mutexes and other resources. Consequently, to avoid errors, the child process may only execute async-signal safe operations until such time as one of the exec functions is called. Fork handlers may be established by means of the pthread_atfork function in order to maintain application invariants across fork calls.

Note that the method an application uses to specify the X/Open behavior is expected to change in a future release.

Notices

Considerations for threads programming

Threads in the creating process are unaffected by these system calls except possibly for the error indication EINTR in the creating process.

It is expected that processes created via fork1 (only the calling thread replicated) will soon call exec(S) to redefine the address space. The single remaining thread should take care when requesting resources (for example, locks) that were duplicated from the parent. The fork1 system call does duplicate the lock (part of the address space) but not the thread that may have been holding that lock at the time of the system call. The surviving thread would block forever since there is no thread to release the resource.