Processes

• process = program in execution
• running a program multiple times will result in multiple processes
• concurrent processes = processes that run simultaneously

Zombie processes

• a parent process may be interested in the execution status (exit code) of a child process ⇒ the parent process uses either the wait or waitpid system calls
• if the child process ends before the parent calls wait? the system keeps the child process in a zombie state: it doesn't execute anything, but it appears in the list of processes
• a zombie process stops when the parent calls wait or waitpid
• zombie processes are problematic because if they are not "waited" they will grow in number, occupying PIDs and bringing the system to a point where no other processes can be created because of lack of PIDs
• to avoid zombie processes, always call wait for each child process you create

// create one child process

pid = fork();
if(pid == 0) {
	 // do some stuff
	 exit(0);
}
// do some other stuff
wait(0);

// create multiple child process

for(i=0; i<3; i++) {
	 pid = fork();
	 if(pid == 0) {
		 // do some stuff
		 exit(0);
	 }
}
// do some other stuff
for(i=0; i<3; i++) {
	 wait(0);
}

<unistd.h>

• Functions specific to processes

  fork()

  exit(n)

  wait(p)

  exec*(c, lc)

  system(c)

• Functions for IPC

  pipe(f)

  mkfifo(name, rights)

  FILE *popen(c, “r|w”)

  pclose(FILE *)

  dup2(fo,fn)

fork()

• UNIX way of creating a process is to duplicate the current process, by making a copy of it using the system call fork
• this leads to another unfamiliar aspect: source code that looks simple and appears to start at the beginning and finish at the end, will actually execute along split path
• after calling fork, there will be two processes:
  • the original one (called parent)
  • and the copy of it (called child)