CS 480 M01 Lecture Notes 5 - Sep 7, 2022

Last Time

Perl Warm-Up

Ch 4. Process Control

Software Architecture of the Linux OS
SW Architecture

Components

Address space (memory pages allocated to the process - executed code & libraries, data, stack, heap, ...)
+ set of data structures within the kernel (process control block = task control block):
    addr space map, current state (sleeping, stopped, runnable, ..), execution priority, resources used (CPU, memory, a...) , opened files (& network ports), signal mask (blocked signals), owner,...


Thread: execution context within a process; at least one; each thread has its own stack and CPU context; all share the same address space
States

• Runnable - can be executed
• Sleeping - waiting for some resource
• Zombie - trying to die / not reaped by its parent
• Stopped - suspended / not allowed to execute

Attributes / Parameters (associated with process) affect its execution: amount of processor time it gets, the files it can access, ... :
PID, PPID, UID -> identity (copy of parent's UID) , EUID -> permissions, GID, EGID, Niceness, Control terminal (default linkages for STDIN,STDOUT,STDERR for nondaemon processes) - pseudoterminals these days (window from which you started the program)
    linux: saved UID (EUID when pgm started), FSUID (=> filesystem permissions - rarely used & not portable)
Process Descriptor

changing scheduling priority: nice

Life Cycle

fork()	creates an identical copy of the original process (clone really) with distinct PID and its own accounting information
	=>	0 (child) =>	exec() / execve() / execv() / ... program text replaced data and stack reset new execution starts
	=>	child's PID wait() / waitpid() / waitid() - will receive child's exit code
_exit() - called when process finishes its execution, notifies the kernel that it is ready to die, provides exit code

exception: autonomous processes ( init/systemd:PID=1, ...)

Signals (~ 30 different kinds in Linux) : process-level interrupt requests

• communicating among processes
• sent by terminal driver to kill, intercept or suspend process (^C, ^Z)
• owner or root controlling process
• kernel notifying a process about sth (division by 0)
• Main Signals
  
• When received ...
  1. if a handler was designated => the handler is called = catching - when completed the execution restarts from the point at which the signal received
     
      or (no handler designated for the signal)
  2. kernel takes default action (varies among signals)
  • some signals cannot be caught : KILL, STOP
  • blocking/ignoring - preventing signal from arriving to the process (some cannot be blocked: KILL, STOP, CONT)
    • blocking - signal queued for delivery, delivered after the signal gets unblocked & handler called just once no matter how many times the signal arrived
    • ignoring - signal is discarded
• sending a signal kill, killall, pkill