Computer Science 330
Operating Systems

Spring 2012, Siena College

Home     Syllabus     Schedule

Quote: UNIX system calls, reading about those can be about as interesting as reading the phone book... - George Williams, 3/12/91

In this lab, we will learn and/or review several aspects of Unix systems programming, focusing on those things you will need for the shell project.

You may work alone or in a group of 2 or 3 on this lab.

Start a plain text file answers.txt in your directory for this lab in which you will answer the questions scattered throughout the lab.

Low-level File Operations

You have used at least some of the C standard file I/O routines defined in stdio.h, such as fopen(), fscanf(), fprintf(), and fclose(). These provide relatively "high-level" access to files in that you deal with data types rather than a low-level stream of bytes.

Underneath the stdio functions, you will find those low-level operations: open(), close(), read(), write().

Error Checking and Reporting

Before we look at the use of all of these, we recall the standard error reporting mechanism.

Most Unix system calls can fail for a variety of reasons. You should always check the return value of system calls that may fail.

Read the intro(2) man page on winterstorm and the errno(3) and perror(3) man pages on a Linux system to learn about or refresh your knowledge of the errno error condition and the system calls perror(3) and strerror(3) that allow you to print out (hopefully) meaningful error messages when you detect a failed system call.

For example, consider a program that uses the low-level open and close system calls:

See Example:
~jteresco/shared/cs330/examples/perror

Copy this program to your directory and compile it (it has a Makefile).

With Unix system calls, there are a lot of good reasons that something can fail. It's worth your trouble to check these return conditions and print meaningful error messages.

A More Complete Example

Whereas fopen returns a value of type FILE *, the open call returns an int. This int has a special meaning - it is a file descriptor. It can subsequently be used in read and write calls, and is later passed to close when we are done.

There are three file descriptors that are automatically created for each process:

Read through the man pages for these four system calls, then consider this example:

See Example:
~jteresco/shared/cs330/examples/everyother

Running a New Program - the exec Calls

Recall that the fork() system call lets you have two copies of a process - each running the same program and executing at the statement immediately following the fork() call.

See Example:
~jteresco/shared/cs330/examples/forking

Sometimes this is what you want, but more likely you will want to start a new process to run some new program.

To create processes that do "something else", the fork() is followed by one of these "exec" calls, in the child process:

execl() - exec a process with list of arguments

execv() - exec a process with args specified in an array

execlp() - list, but search the existing path for the program.

execvp() - array, but search the existing path for the program.

execvP() - array, but specify a search path for the program.

The man pages have details.

The related vfork() system call is often more appropriate when the child process will be doing an exec() immediately. It doesn't duplicate all of the memory for the parent process. Beware: this may cause you trouble in the shell if you use it, since the parent is usually suspended until the child exits or calls an exec.

We consider a series of example programs.

See Example:
~jteresco/shared/cs330/examples/exec

Start by looking at the exec program:

• This one uses execlp. The parameters are the program to run and its arguments.
• This is a "varargs" function call - we can send any number of parameters.

Note that we can specify a program by its name only (like "ls"), in which case the search path is used to try to find a program to run. We can also give a full path to the program (like "/bin/ls") in which case the program must be at the exact path specified.

Next, we look at a program that doesn't use any of the "exec" calls, but which will be useful as we look at further examples: procinfo. This one simply prints the process id and the command-line parameters (including one beyond the last).

Use the execprocinfo program to execute procinfo.

Next, look at exec2, which uses execvp() instead of execlp(). This is the "list" form rather than the "varargs" form. We pass a NULL-terminated array of parameters.

Our last example program is execwithargs, which uses its command-line parameters to determine which program it should become (weird).

Practice With exec

Write a program execlsloop.c that loops forever (well, until you kill it) and every 5 seconds, creates a child process that executes ls -l and waits for that child process to finish. You may use any of the class examples as a starting point if you'd like. (5 points)

Signals

We next consider a form of interprocess communication in a Unix system known as signals.

We can send a signal SIGNAL to a process pid with the command

kill -SIGNAL pid

For example, if we launch a program at our Unix prompt to sleep for 60 seconds and put it into the background:

-> sleep 60 &

you should see output something like:

[1] 96132

where "96132" would be the process id of the sleep process you just created, and [1] is the job number within your Unix shell of the process.

We can then send signals to that process by using its pid or %1 which will refer to job number 1.

For example:

-> kill -TERM %1

will send the SIGTERM signal to try to terminate the process. If you do this, you should see output similar to:

[1]+  Terminated              sleep 60

Now launch another sleep 60 process in the background. Assuming this becomes shell job 1, issue these commands:

-> kill -STOP %1
-> kill -CONT %1

and wait until the sleep command finishes.

Every process has signal handlers that are used to respond to signals sent to the process. Basically, it's a function that gets called asynchronously when a signal is received.

A default signal handler is installed when a process begins.

Two system calls are used to send and catch signals:

signal() replaces default handler. This lets you trap many signals and handle them appropriately.

Be careful not to confuse this signal() with the signal() operation on semaphores!

See Example:
~jteresco/shared/cs330/examples/signals

The sigalrm-example.c example is compute-bound process that "wakes up" every 5 seconds to report on its progress.

The setitimer(2) system call is used to set a "timer" which will cause a SIGALRM signal to be sent to the process at some time in the future (in this case, every 5 seconds).

We can ignore a signal completely by setting its handler to SIG_IGN, and restore the default handler with SIG_DFL.

Consider this enhanced example: sigalrm-example2.c

A process can also send signals with kill(). Don't let the name fool you, you can send any signal with kill(), not just SIGKILL.

Note that SIGTERM's handler sends the process a SIGINT.

Final note about signals: SIGCHLD will be useful for your shell projects. This gets sent to a process's parent when the process terminates.

Pipes

Processes may wish to send data streams to each other. Unix pipes are one way to achieve this. You've almost certainly used Unix pipes at the command line. You can also use them in programs.

An unnamed pipe can be created using the

int pipe(int fd[]);

system call. fd is an array of two int values. These are file descriptors, very similar to the file descriptors used for file I/O using open(), read(), and write().

fd[0] is the "read end" and fd[1] is the "write end". 0 return means success. -1 means failure.

read() and write() again operate only on basic streams of bytes - any structure must be added.

See Example:
~jteresco/shared/cs330/examples/pipes

pipe1.c is an example of communication between two processes, a parent and its child created by fork(), communicating via an unnamed pipe.

This required that the values of fd are shared between the parent and child processes. This is fine when you create your pipe just before a fork(), but what if we have two processes already in existence that wish to communicate through a pipe?

We can create a named pipe with mkfifo (command or system call).

pipe2.c augments our simple example using a named pipe.

pipeprocs.c is an example that's a little more interesting: two independent processes communicate through a pipe.

Duplicating file descriptors

We can use the dup2() system call to "reroute" input or output from one file descriptor to another file descriptor. This is how your I/O redirection and pipes will work in the shell.

Back in the exec example set, see and try execredir.c.

Note that we don't close the file here and in fact are not given an opportunity to do so since we lose control once the execlp call occurs.

We have seen that you can also obtain file descriptors from open(), pipe(). The fd's at the ends of a pipe can be passed to dup2() as well - this will be useful in the shell - set the output of one process to be the input of another through a pipe.

Submission and Evaluation

This lab is graded out of 40 points.

By 4:00 PM, Wednesday, February 29, 2012, submit your answers.txt and execlsloop.c files by email to jteresco AT siena.edu.