title: Project 1 Design Document author:
- Craig Collins
- Mat Fukano
- Ryan Gliever
- Leland Miller date: April 2014 abstract: This document describes our teams design for a simple MINIX shell written in C. Our design was worked on together, but each member coded a separate shell. ...
At the highest level our shell is a loop that takes user input, then runs commands that the user typed in. For each command in the loop our shell first scans through the arguments to find special feature symbols such as & for running a command in the background or > for redirection, and then copies the arguments besides the special symbols into a new array. This new array is passed into execvp() for execution in the child process.
This loop can be contained in a function separate from the code that executes the commands. It should be able to take an array of arguments, which is provided to us in the template, and call the function that executes commands. The function that executes the commands should take the command, a file to redirect output into, a file to redirect input into, and a whether or not the process should be run in the background or not. This leaves us with two major functions:
void strparse (char **args);
void executecmd (char **args, char* output, char* input, bool bg);If redirection is not desired, output or input should be set to null. strparse() will also look for special cases such as exit and deal with them accordingly. In other words, executecmd() will only be responsible for forking and executing system commands.
When the “exit” command is executed, the shell closes. This is done by using strcmp() to compare the first argument to string “exit” and using system call exit() to quit the program. This checking will be done in the strparse() function.
When a command with no arguments is read, we will fork() into a new child process. In the new process we will run execvp() with the command. execvp() transfers control of the new process to the new program while the main process is in a loop wait()ing for the child to finish.
Commands with arguments will be handled similarly to commands without arguments. We will fork() into a new child process, and run execvp() with the command, and execvp() handles the command arguments and passes them to the new program to be executed. The parent process still wait()s for the child process to finish.
Output:
Output of a command that is to be redirected to a specified file will be done with freopen(). We will pass in the specified file name, mode “w”, and stdout. Then, we will run execvp() on the command, and the output will be redirected to the specified file.
Input:
If input to a command is redirected from a file, we would use freopen(). We will use the "r" mode which will read in the contents of the file specified. stdin will be the stream
The only real difference in handling background processes is that the main thread should not wait for the child to finish, and instead the parent process (the shell) should just continue accepting input. The only special consideration is the killing of zombie processes.
We decided that dealing with the zombie processes will be done while we wait for the foreground processes to finish. This seemed more efficient and cleaner than using signal handlers. This way of handling zombie processes allows us to decide when our program will "clean" up the zombie processes (by waiting for them), instead of the processes interrupting our shell program. In order to implement this, when a foreground process is started we will have a loop that waits for any process, it will only break this loop is the process id returned is that of our current foreground process.
All in all this project has a relatively simple structure. The following is a general overview of our project in pseudo-code (note that args and new_args are arrays of parsed strings):
executecmd (new_args, output, input, bg)
fork process
if parent
if not bg
loop until wait returns foreground pid
if child
set redirection (using freopen)
execvp
strparse (args)
new_args, output, input, bg
for arg in args
if arg is a shell command
set needed flag
else
push arg to new_args
executecmd new_arg output input bg
main()
while true
get input
args = parse input
strparse (args)
Error handling will have to be done after the fork() and the execvp() calls. Other information regarding specific projects and how to build them will be included with the individual projects.