https://pages.cs.wisc.edu/~remzi/OSTEP/cpu-api.pdf

Process API

fork()

• Duplicates the process
• New process begins from fork() call, not main
• Processes have a unique identifier called pid
  • fork() returns child pid to parent and 0 to child.
• Show p1.c

wait()

• Allows you to wait for a child.
  • Can be used to synchronize
  • Can be used to clean up.
• There are different versions of wait (e.g., can wait on a specific process instead of waiting for any child to terminate)
• The return value in the return from main (or as a parameter to exit) is made available to the parent indirectly through wait
• Show p2.c

exec()

• Runs a new program
  • Specifically transforms the current program into a new program by replacing (most of) the memory and resetting the state.
• To launch a new program is two steps: fork(), then exec()
• Why
• Why do we need fork() at all? Why not just exec()?
• Why do we need to do a fork() then exec? Why not just jump right to exec when desired?
  • It provides the ability to alter the new program’s environment
  • Environment variables
    • (Processes can set using setenv(). Children inherit environment and pass it through exec)
  • stdin / stdout
  • This is what makes file redirection work.
• Several versions of exec. Three basic modifiers:
  • l: Provide arguments as a list. Uses varargs like printf does.
  • v: Provide a pointer to an array (instead of a separate parameter for each argument)
  • p: Use the path. (Otherwise you must specify an absolute path name)
  • e: Provide a custom environment (rather than passing the current environment)