Linux System Programming by Robert Love [Chapters 5-8]

Chapter 5 - Processes

• A binary is a compiled blob of code

• A process is a running program, includes

  • the binary image
  • virtualized memory instance
  • kernel resources like open files
  • security context like user
  • one or more threads
  • a thread has a virtualized processor, stack, registers and instruction pointer
  • in a single threaded process, the process IS the thread
  • all threads share the same memory
• A process has a PID that's guaranteed to be unique at a given moment
• The idle process has PID 0
• The first process is called the init process (pid 1)

• The kernel tries the following init programs:

  • /sbin/init
  • /etc/init
  • /bin/init
  • /bin/sh
• The init process handles the remainder of the boot process, launching services and a login program
• The kernel keeps monotinically increasing the PID till it wraps around
• Every process (child) is spawned from another process (parent)
• Each process has a ppid(parent pid)
• Each process inherits the parent's UID and GID
• Processes belong to process groups. Normally children have the same group as the parent
• when you do a ls | less both ls and less will have the same process group
• to get your own pid:

#include <sys/types.h>
#include <unistd.h>
// returns the process ID of the invoking process
pid_t getpid (void);
// returns the process ID of the invoking process’s parent
pid_t getppid (void);
// Neither call will return an error
printf ("My pid=%jd\n", (intmax_t) getpid ());
printf ("Parent's pid=%jd\n", (intmax_t) getppid ());
// intmax_t is guaranteed to be larger than any int on the system

Running processes

• fork creates a near duplicate of the parent process
• exec family of calls loads a binary into memory, replacing the previous program, and executes it
• to execute a new program, you have to fist fork, then exec

#include <unistd.h>
int ret;
  ret = execl ("/bin/vi", "vi", "/home/kidd/hooks.txt", NULL);
  // Arguments must be NULL terminated
  if (ret == −1)
    perror ("execl");

• normally, execl does not return since there's nothing to return to
• it clears out most attributes of the calling process including signals and locks
• it keeps the open file ids. They should be either manually closed or using fcntl

• In addition, we have execlp, execle, execv, execvp, execve

  • l means list, v mean array (variadic vs array args)
  • e means it takes a new environment
  • p means it seaches the PATH for the filename

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

int main() {
    int pid = fork();
    if(pid == -1) {
        perror("fork"); return 1;
    }
    if(pid == 0) { //child
        int ret;
        ret = execlp("ls", "ls", NULL);
        if (ret == -1) {
            perror("exec"); return 1;
        }
    } else {
        printf("Parent\n");
    }

}