Computer Science 432 Operating Systems Williams College Fall 2006

After a few weeks of small labs while being distracted by Reading Period and an exam, it's back to a bigger programming task.

For this lab, you are to write a C program called the Cow Shell (cowsh), a mini command shell interpreter. cowsh is similar to familiar Unix shells such as the Bourne shell (sh) the Bourne-Again shell (bash), and C shell (csh, tcsh). You will learn about process creation, pipes, input/output redirection, background process management, signals, and interrupt handling, and gain extensive experience with C.

The program is designated a "team program" (groups of size 2 or 3 permitted) for honor code purposes. Collaboration within a group is, of course, unrestricted. You may discuss the program with members of other groups, but what you turn in must be your own group's work. Groups must be formed no later than 9:00 AM, Monday, October 23, 2006, and be confirmed by all group members by electronic mail to terescoj@cs.williams.edu. All group members will be assigned the same grade for the lab. There are many subtasks that can be carved off and assigned to group members, so everyone is encouraged to join a group.

Requirements

Like the Unix shells you use every day, cowsh should issue a prompt (perhaps "cowsh#"), at which it reads commands from the user and executes them.

Your shell should interpret the following commands and provide the following functionality:

• exit: exit from the shell
• help: display a message listing usage of all commands
• Execute a command (program on disk). As in the shells you use, the command should be located according to the PATH environment variable Appropriate choice of exec function will help here. The arguments following the command should be passed to the command. For example,

         cowsh# cat cat.c
  

  should execute cat with one argument, cat.c
  • Input and output redirection should be implemented. For example,

    	cowsh# cat < cat.c > myfile.c
    

    should cause the cat program to read from cat.c and write to the file myfile.c.

    	cowsh# ls -l >> dirlistings.txt
    

    should cause the output of ls -l to be appended to the end of the existing file dirlistings.txt. An individual command may only redirect input once and output once, but those redirections may be specified in any order.

            cowsh# > alines grep -i < Makefile a
    

    should be interpreted the same as the more usual

            cowsh# grep -i a < Makefile > alines
    

  • Pipes should be implemented.
    • For example,

             cowsh# cat cat.c | wc > count.txt
      

      should cause the output of cat cat.c to be the input of wc > count.txt.
    • You should allow a sequence of pipes to be specified:

             cowsh# ls -l *.c | grep "Oct 31" | wc -l
      

      The program should be able to handle a sequence of pipes of any length.
    • Only the first command in a pipeline may have input redirection. Only the last command in a pipeline may have output redirection. Redirection of other commands should be reported as an ambiguous command line.
  • Typing <ctrl-c> should abort a command being run, but not cause cowsh to terminate.
• A list of commands separated by semicolons should be executed in sequence.
• As with the familiar Unix shells, appending an & to the end of your command line will execute the command in the background.
  • An & may be anywhere in a command line, but it is treated like a ; in that anything on the command line following an & should be treated as a new command. For example,

           cowsh# sleep 5 & sleep 5
    

    should launch one instance of sleep 5 in the background, then a second in the foreground.
  • Typing <ctrl-c> should not kill the shell or any commands running in the background.
  • When any background command terminates, it should be reported.

     
           cowsh# sleep 55 &        ; invoke sleep in background
           cowsh# cat < hello.c     ; give other commands
           cowsh# ...               ; other commands
           cowsh# ...               ; other commands
           [2] "sleep" terminated   ; sleep command is done
    

  • All backgrounded processes should be maintained in a process table by cowsh, so that the program name is displayed when it terminates, and for use in the jobs and kill commands.
• jobs: Displays all the active background programs that have been started from this shell, along with their ids. (This id need not be that same as the actual process id. For example, it could be the index into your process table).

         cowsh# jobs
         PID     Name
         [0]    mycat < myfile.c > newfile.c
         [1]    sleep 20
         [4]    grep cowsh < doc | wc
  

• kill: Without any arguments, prints a usage statement:

         cowsh# kill
         Usage: kill <pid> [<pid> ...]
  

  Otherwise it kills the processes with the specified ids and displays the processes killed.

         cowsh# kill 4
         [4] "grep" Killed
         cowsh# kill 3 7
         [3] "cat" Killed
         [7] "wc" Killed
  

• Implement a builtin cd command.
• Errors should be reported meaningfully.
• Additional functionality of your group's choosing should also be implemented (see also the Submission and Evaluation section below). Ideas for extra functionality include:
  • command history: history command, !command, !!, ^ modification of previous command
  • control structures in the shell (for, while, if)
  • aliases
  • user-specified prompts (as in bash, tcsh)
  • More advanced redirection and pipes: <<, >&, |&
  • <ctrl-z> trapping and corresponding job control (fg, bg commands)
  • Parenthesis-delimited subshells
  See builtin(1) for more ideas.

Notes

• A similar project is described at the end of Chapter 3 of SG&G. You may find some useful information and examples there.
• The system(3) system call is not to be used.
• The system calls that you should use are fork(2), a variant of exec(3), signal(3), kill(2), open(2), dup2(2), close(2), pipe(2), and chdir(2).
• You may find the readline(3) and the strsep(3) functions to be useful. readline(3) is part of the GNU Readline Library and strsep(3) is in the Standard C Library.
• You may assume the builtin commands (exit, help, jobs, kill) stand alone on a command line (no pipes, I/O redirection).
• Use cowsh scripts to test your shell.
• A shell may run for a long time. Be careful about memory management.
• gcc's -Wall flag will report additional compiler warnings. This can help you find some of the more subtle bugs that are so common in C code before they have a chance to cause their subtle problems. If you aren't absolutely certain that a given warning is harmless, fix it!
• A verbose/debugging mode is essential for, well, debugging.
• Keep your code under source code control. You may wish to use subversion to track your changes and to allow you to go back to previously working code. If you're working in a group, this is even more useful, allowing easier concurrent development. Group members should request a Unix group for safe sharing of files.
• The following might be a good order to tackle the required functionality.
  • exit and help commands
  • run a command (with no argument passing, no redirection, no pipes)
  • run a command with argument passing
  • run a command with input and output redirection
  • run a command in the background
  • jobs command
  • kill command
  • <ctrl-c> trapping
  • trapping termination of background processes
  • pipes
  • ;- or &-separated commands
  • cd command
  • extra functionality
• /usr/cs-local/share/cs432/shell/cowsh contains my version of cowsh.

Submission and Evaluation

All necessary files should be submitted using turnin as a single "tar" file, lab5.tar. Include a Makefile to allow easy compilation of the Cow Shell program.

I will compile and test your shell programs on CSLab FreeBSD systems.

Your program will be graded based on a total of 50 points.

• 6 points for documentation. Your code should be commented appropriately throughout. Please also include a longer comment at the top of your program describing your implementation. This documentation should list your important design decisions, the assumptions that you made (if any), the additional functionality implemented, and any other relevant information.
• 1 point for a functional Makefile.
• 1 point for implementing the exit and help commands.
• 4 points for running a command with no arguments.
• 4 points for running a command with arguments.
• 4 points for running a command with input/output redirection.
• 4 points for launching a command in the background.
• 4 points for launching a sequence of jobs separated by ;'s or &'s.
• 2 points for the jobs command.
• 2 points for the kill command.
• 2 points for correctly trapping the <ctrl-c> keystroke.
• 3 points for reporting termination of backgrounded jobs.
• 4 points for a pipeline of two processes.
• 3 points for an arbitrarily long pipeline of processes.
• 1 point for a builtin cd command.
• Up to 5 points for extra functionality.
  • 1-point enhancements: aliases, more advanced redirection and pipes: <<, >&, |&, user-specified prompts.
  • 2-point enhancements: command history, control structures in the shell (for, while, if), modification of previous command with ^.
  • 3-point enhancements: <ctrl-z> trapping and corresponding job control (fg, bg commands).
  • Please ask about other possible enhancements. Point values will be based on level of difficulty.

Penalties may be applied for poor design choices, poor formatting of code, poor programming style, or if your program compiles with warnings (when using gcc -Wall).

Have fun and good luck!