FORK1(9) Kernel Developer's Manual FORK1(9)

NAME

fork1create a new process

SYNOPSIS

#include <sys/types.h>
#include <sys/proc.h>

int
fork1(struct lwp *l1, int flags, int exitsig, void *stack, size_t stacksize, void (*func)(void *), void *arg, register_t *retval, struct proc **rnewprocp);

DESCRIPTION

fork1() creates a new process out of the process behind l1, which is assumed to be the current lwp. This function is used primarily to implement the fork(2) and vfork(2) system calls, but is versatile enough to be used as a backend for e.g. the __clone(2) call.

The flags argument controls the semantics of the fork operation, and is made up of the bitwise-OR of the following values:

FORK_PPWAIT
The parent process will sleep until the child process successfully calls execve(2) or exits (either by a call to _exit(2) or abnormally).
FORK_SHAREVM
The child process will share the parent's virtual address space. If this flag is not specified, the child will get a copy-on-write snapshot of the parent's address space.
FORK_SHARECWD
The child process will share the parent's current directory, root directory, and file creation mask.
FORK_SHAREFILES
The child process will share the parent's file descriptors.
FORK_SHARESIGS
The child process will share the parent's signal actions.
FORK_NOWAIT
The child process will at creation time be inherited by the init process.
FORK_CLEANFILES
The child process will not copy or share the parent's descriptors, but rather will start out with a clean set.

A flags value of 0 indicates a standard fork operation.

The exitsig argument controls the signal sent to the parent on child death. If normal operation desired, SIGCHLD should be supplied.

It is possible to specify the child userspace stack location and size by using the stack and stacksize arguments, respectively. Values NULL and 0, respectively, will give the child the default values for the machine architecture in question.

The arguments func and arg can be used to specify a kernel function to be called when the child process returns instead of child_return(). These are used for example in starting the init process and creating kernel threads.

The retval argument is provided for the use of system call stubs. If retval is not NULL, it will hold the following values after successful completion of the fork operation:

retval[0]
This will contain the pid of the child process.
retval[1]
In the parent process, this will contain the value 0. In the child process, this will contain 1.

User level system call stubs typically subtract 1 from retval[1] and bitwise-AND it with retval[0], thus returning the pid to the parent process and 0 to the child.

If rnewprocp is not NULL, *rnewprocp will point to the newly created process upon successful completion of the fork operation.

RETURN VALUES

Upon successful completion of the fork operation, fork1() returns 0. Otherwise, the following error values are returned:
[EAGAIN]
The limit on the total number of system processes would be exceeded.
[EAGAIN]
The limit RLIMIT_NPROC on the total number of processes under execution by this user id would be exceeded.

SEE ALSO

execve(2), fork(2), vfork(2)
January 4, 2008 NetBSD 6.1