C on Linux: files and processes with system calls

User space and the kernel

Your program runs in user space and cannot touch hardware, other processes' memory or the disk directly. To do those things it asks the operating-system kernel through a system call. The C library wraps each system call in a function: open, read, write, close, fork, execvp, waitpid and so on (documented in section 2 of the manual: man 2 open). printf, fopen and fgets (section 3) are library functions that add buffering and formatting and eventually call write or read.

File descriptors

#include <fcntl.h>
#include <unistd.h>
#include <stdio.h>

int main(void) {
    int fd = open("log.txt", O_WRONLY | O_CREAT | O_APPEND, 0644);   /* returns a small int, or -1 */
    if (fd == -1) { perror("open log.txt"); return 1; }
    const char msg[] = "order 1042 placed\n";
    ssize_t n = write(fd, msg, sizeof msg - 1);                       /* bytes written, or -1 */
    if (n == -1) perror("write");
    close(fd);

    write(1, "done\n", 5);        /* fd 1 is stdout */
    return 0;
}

read(fd, buf, n) may return FEWER bytes than requested (a "short read"), for example from a pipe or network socket, and 0 at end of file. Robust code loops until it has what it needs. The same is true for write on sockets and pipes.

Errors: -1 and errno

System calls report failure by returning -1 and setting the global errno to a code such as ENOENT (no such file) or EACCES (permission denied). perror("context") or strerror(errno) turns it into a readable message. Check errno only right after a call has failed, because other calls may change it.

Processes: fork, exec, wait

#include <sys/wait.h>
#include <unistd.h>
#include <stdio.h>

int main(void) {
    pid_t pid = fork();                        /* now TWO processes run from here */
    if (pid == -1) { perror("fork"); return 1; }
    if (pid == 0) {                            /* child */
        char *args[] = {"ls", "-l", NULL};
        execvp("ls", args);                    /* replace the child's program with ls */
        perror("execvp");                      /* only reached if exec failed */
        _exit(127);
    }
    int status;                                /* parent */
    waitpid(pid, &status, 0);                  /* wait for the child to finish */
    if (WIFEXITED(status)) printf("child exited with %d\n", WEXITSTATUS(status));
    return 0;
}

Where to go next

The same system-call style continues into sockets (network programming), pipes and signals, memory mapping (mmap) and threads (pthread_create). These topics are the core of Linux systems programming and of many interviews for systems roles; the man pages and the book "The Linux Programming Interface" are standard references.