fork and exec

Quick Reference

The fork-replace-wait pattern:

pid_t pid = fork();
if (pid == 0) {
    // CHILD: replace with a different program
    execlp("ls", "ls", "-l", NULL);
    perror("launch failed");
    _exit(1);
} else {
    // PARENT: wait for child
    wait(NULL);
}

fork() return values:

Pipe recipe for cmd1 | cmd2:

1. pipe(fd)           — create the pipe
2. fork() child 1     — dup2(fd[1], STDOUT), close both, execlp(cmd1)
3. fork() child 2     — dup2(fd[0], STDIN), close both, execlp(cmd2)
4. parent             — close both fd ends, wait() twice

Common signals:

Common Pitfalls

• Forgetting wait() — Child becomes a zombie. Parent must collect exit status.
• Code after execlp — Only runs if the call fails. Always follow with perror + _exit(1).
• Not closing pipe ends — Reader blocks forever if a write end stays open.
• Using exit() in child — Use _exit(1) to avoid flushing parent’s stdio buffers.
• Fork bomb — fork() in a loop without an exit condition creates exponentially many processes. Always have child exit or replace itself.