$30
CS 350 Assignment 2: Simple Process Creation and Control
Due: Always consult Blackboard for due dates.
Overview and Objectives
In Assignment 2, you will investigate Linux facilities for process creation, management, and control, both from
within C programs and from the command shell. You will build toward writing a program that creates an N-level
process tree. Each process in the tree, other than leaf processes, will have M children. Tree parameters N and M
will be specified by the user.
Interface, Options, and Parameters
Your program should support the following interface:
prog2tree [-u] [-N <num-levels] [-M <num-children] [-p] [-s <sleep-time]
-u
Print a usage string for your program on stderr and then exit. If –u appears anywhere on the command
line, the program may ignore all other arguments.
-N <num-levels
Create a process tree with <num-levels levels. You may assume that the user will not enter a negative
number. Specifying 0 or 1 for <num-levels should cause your program not to create any child
processes (and therefore no additional levels). Your program should not accept values larger than a
maximum of 4. The default value for <num-levels is 0.
-M <num-children
Create a process tree wherein every process (other than leaf processes) creates <num-children child
processes. You may assume that the user will not enter a negative number or 0. Your program should not
accept values larger than 3. The default value for <num-children is 1.
-p
Leaf processes should pause by calling pause(), immediately upon realizing that they are leaf processes.
When this option is specified, the user will clean up processes from the terminal or shell.
-s <sleep-time
Leaf processes should sleep for <sleep-time seconds, immediately upon realizing that they are leaf
processes. Do this by calling sleep(). Non-leaf processes should not sleep; instead they should wait for
child processes to complete, by calling wait().
If the user specifies neither –p nor –s, then –s is assumed (i.e. the leaf processes should call sleep(), not
pause()), and the default value for <sleep-time is 1.
If the user specifies both –p and –s then notify the user with an error message, print the usage string, and
exit.
If the user specifies an option that your program does not support (-z for example), your program should print the
usage string to standard error (stderr), and exit. If the user does not specify an argument where one is expected,
the program should likewise report the problem, print the usage string to stderr, and then exit.
Output
Before creating any children and before sleeping or pausing, each process in the tree (including the top-level
process) should print to stdout its level, its own Linux process id, and the process id of its parent, in the following
format:
ALIVE: Level 2 process with pid=3452, child of ppid=3450.
Immediately before exiting, each process should print to standard output a message in the following format:
EXITING: Level 2 process with pid=3452, child of ppid=3450.
Levels of the tree are numbered from the top down. In other words, the lone root process at the top of a 3-way tree
with 3 levels is said to be at Level 2. The 3 children of that root process are said to be at Level 1, and the 9 leaf
processes beneath the Level 1 nodes are said to be at Level 0. This is the level numbering that should be reflected
in your output statements. Note that this is backwards from the usual level numbering; we usually say that the root
node resides at level 0 of the tree. I have done this intentionally to make things slightly more straightforward for you
(I hope).
Additional Requirements
Important characteristics for your program:
• The user should be able to run your program such that all processes are running concurrently (by choosing
to have leaf processes pause, or by having them sleep for long enough for the entire tree to be created).
• All processes must use the same executable, namely prog2tree. You may not have a different executable
for different levels of the tree, including leaf processes.
• Every child process must call one of the exec() functions. That is, you may not simply bury the sleep()
or pause() inside the loop structure of your program, after returning as a child from fork(). That leaf
child should instead execute the prog2tree executable.
Initial Steps – Building Toward a Process Tree Program
Some Useful Linux Commands
Begin by running your Program 1 (or some other simple C program) through strace, a program that intercepts
and reports all system calls that the program makes.
$ strace prog1generator –g
Some of the many system calls that strace reports are made by library (e.g. libc) functions that you call directly
from within your program. Others do not correspond as directly to components of your program. As you use
strace to investigate the behavior of your programs, you will learn to identify the calls that correspond directly
to your code. Keep strace in mind, and as needed, please use it to trace your programs throughout the semester
(and beyond).
Next, run the ps command to list processes.
$ ps
(If commands do not work exactly as specified in the assignment, it could be that your version of a command is
“aliased” to some other command. Type “which ps” to find the full path to the ps program that your shell will use.
You can always specify the full path name to programs; there is a ps in /bin, so you may use /bin/ps if
necessary. In general, if something works for a neighbor, but not for you, figure out which program you are really
running, with which. )
Peruse the man page for ps to investigate ps as a utility for listing processes. Also try
$ /bin/ps --help all
Notice that the operating system maintains lots of information about relationships between processes, much of
which you can get at using ps.
Try running top, which also shows running processes.
$ top
Typing ‘q’ gets you back to a shell prompt.
Write a program that prints a message, calls sleep() or pause() (try both), then prints another message.
Run your program and practice “suspending” the program and “resuming” the program with Ctrl-Z and with the
linux command fg (for “foreground”), respectively.
While the program is sleeping, paused, or suspended, “look at it” using ps. You may also use jobs.
Now check out the “kill” command, for killing processes. The kill command can take a pid as a parameter, and
the pid of any running process can be discovered using ps.
Create a paused process, figure out its pid, then kill it and verify that it is gone.
The pid of a process can also be discovered from within a C program, which can also learn the pid of its parent
process. Investigate the functions getpid() and getppid(), and include calls to them inside your simple
sleep/pause program. Verify that the value that your program reports matches the one that the Linux command
utilities show.
Writing Programs that Create Processes
Enter the following program into a file called simplefork.c.
#include <sys/types.h
#include <stdio.h
#include <unistd.h
int main() {
pid_t pid; /* for a child process */
pid = fork();
if (pid < 0) { /* error */
fprintf(stderr, “Fork failed.”);
return 1;
}
if (pid == 0) { /* child process */
execlp(“/bin/ls”, “ls”, NULL);
}
else { /* parent process */
/* parent will wait for the child to complete */
wait(NULL);
printf(“Child Complete.”);
}
return 0;
}
Compile and run it.
From this point forward, keep a copy of each working program after every step (just in case).
Alter your program to exec() your pause/sleep program that reports the pid and parent pid.
Alter the program to create 4 child processes instead of one.
Alter the program to have the first child create another child. Use two different executables for the two children.
Here’s where it gets dangerous....
Use the same executable for the child process and the “grandchild” process. THINK first. Program carefully. Do NOT
create a “fork bomb”!
Use a command line argument, passed through to the child via exec(), to distinguish the first child from the grand
child. Do not do any forking in your code until you are sure that parameter is passed through properly.
Using this code as a building block, you are now ready to tackle the problem of creating an M-way process tree with
N levels.
Submission
Will be the same as Assignment 1, once we get those details worked out. Stay tuned.