LinuxQuestions.org - Creating Linux Processes using C

you create a new process with fork(). It's usage is like this:

Code:

pid_t pid;

pid = fork();

if (pid < 0)

    err(1, "fork failed!");

else if (pid == 0) {

    printf("hi, I'm the child process\n");

    printf("goodbye\n");

    exit(0);

} else {

    printf("I'm the parent and I just forked a child, pid %d", pid);

}

Except for the return value from fork(), the child is an exact copy of the parent. All memory appears to be copied; they don't write on eachothers variables. (Actually, it's copy-on-write, but the look is identical)

If you want, you can get information about child processes which have died through a SIGCHLD handler. If you don't want, you should probably do this:

Code:

if (signal(SIGCHLD, SIG_IGN) == SIG_ERR)

    err(EXIT_FAILURE, "signal");

That way, you don't have idle, defunct processes ("zombies") sitting around waiting for the parent to exit.

The shared memory API I've used is SysV IPC. The functions you want are shmget(), shmat(), shmdt(), and shmctl(). There are man pages for all of these, as well as ipc(5) which is worth a read as well. The API is very ugly, but there's nothing suprising about it aside from ugliness. If you have questions I can probably|maybe answer :D

You will probably want some sort of locking on your shared memory: I've used SysV semaphores for this. The functions for those are semget(), semop(), and semctl(). Again, they all have man pages and are ugly.

You could also implement this thing with threads -- one library for that is called pthread. I find fork() to be much more intuitive than the threading stuff, and the SysV stuff is pretty reasonable once you write 10 line wrapper functions to take care of error handling and such.