PA2: Implementing a Linux Shell solution


PA2: Implementing a Linux Shell

In this programming assignment, you are going to implement your very own linux shell. Your
shell should have the ability to function almost as much as the linux/ubuntu shell in your OS,
which lets a user navigate through the file system and perform a wide variety of tasks using a
series of easy to remember and simple commands. Please take a look at this website for some
interesting and commonly used linux commands.
Features of Shell



5/5 - (3 votes)

PA2: Implementing a Linux Shell

In this programming assignment, you are going to implement your very own linux shell. Your
shell should have the ability to function almost as much as the linux/ubuntu shell in your OS,
which lets a user navigate through the file system and perform a wide variety of tasks using a
series of easy to remember and simple commands. Please take a look at this website for some
interesting and commonly used linux commands.
Features of Shell
The shell maintains many variables which allow the user to maintain some settings and
information visible throughout the system. This is analogous to a collection of global variables in
a C program. The difference is that all programs running from the shell can use environment
variables. For instance, the current working directory and the PATH are two of many important
variables. As its name implies, the current working directory variable keeps track of the user’s
current directory. The PATH variable on the other hand consists of a colon separated directory
list that is searched wherever you type a command in the terminal. If an executable file by the
same name does not exist in any of these directories or the current directory, the shell says
“command not found”, which is something we experience when trying to run a program before
installing it. One may modify the PATH at any time to add and remove directories to search for
executables. The following shows the commands for printing the current directory and the
content of the PATH variable:
shell pwd
shell echo $PATH
Command Pipelining
While the individual linux commands are useful for doing specific tasks (e.g., grep for
searching, ls for listing files, echo for printing), sometimes the problems at hand are more
complicated as they require multiple commands together and feeding the output of one to the
input of the next. The linux shell lets you combine commands by putting the character | between
then. A pipe in between two commands causes the standard output of one to be redirected into
the standard input of another. An example of this is provided below, using the pipe operation to
search for all processes with the name “bash”.
shell ps -elf | grep bash | awk ‘{print $10}’ | sort
Input/Output Redirection
Many times, the output of a program is not intended for immediate human consumption (if at
all). Even if someone isn’t intending to look at the output of your program, it is still immensely
helpful to have it print out status/logging messages during execution. If something goes wrong,
those messages can be reviewed to help pinpoint bugs. Since it is impractical to have all
messages from all system programs print out to a screen to be reviewed at a later date, sending
that data to a file as it is printed is desired.
Other times, a program might require an extensive list of input commands. It would be an
unnecessary waste of programmer time to have to sit and type them out individually. Instead,
pre-written text in a file can be redirected to serve as the input of the program as if it were
entered in the terminal window.
In short, the shell implements input redirection by redirecting the standard input of a program to
an file opened for reading. Similarly, output redirection is implemented by changing the
standard output (and sometimes also standard error) to point to a file opened for writing.
shell echo “This text will go to a file” temp.txt
shell cat temp.text
This text will go to a file
shell cat < temp1.txt
This text came from a file
Background Processes
When you run a command in the shell, it suspends until the command finishes. We often do not
notice this effect because many commonly used commands finish soon after they start.
However, if the command takes a while to finish, the shell stays inactive for that duration and
you cannot use it. For instance, typing sleep 5 in the shell causes the shell to suspend for 5
seconds. Only after that the prompt comes back and you can type the next command. You can
change this behavior by sending the program to the background and continue using the shell. If
you type sleep 5 & in the shell for example, it will return the shell immediately because the
corresponding process for the sleep runs in the background instead of in the foreground where
regular programs run.
Implementation Hint: First of all, remove the ‘&’ symbol from the command before passing it to
exec(). But use the symbol to set a boolean that tells you that it is supposed to run in the
background. After that, from the parent process do not call waitpid(), which you have been
doing for the regular processes. Rather put the pid into a list/vector of background processes
that are currently running and periodically check on the list to make sure that they do not
become Zombies or do not stay in that state for too long. A good frequency of check is before
scanning the next input from the user inside the main loop. Keep in mind that waitpid() function
suspends when called on a running process. Therefore, calling it as it is on background
processes may cause your whole program to get suspended. However, there is an option in
waitpid() that makes it non-blocking, which is the desired way of calling it on background
processes. You can find this option from the man pages.
Use of Single/Double Quotes
White-spaces are usually treated as argument separators except when they are used inside
quotes. For example, notice the difference between the following two commands:
shell echo -e “cat\ta.txt”
cat a.txt
shell echo “-e cat\ta.txt”
-e cat\ta.txt
Note that the “-e” option for the echo command prints the string with interpretation of some
symbols. Now, in the first command, the string is put inside quotes to make sure that it is
interpreted as a single string. As a result of using the -e option, the string is printed after
interpreting the “\t” as a tab. In the second example, “-e” is part of the string which masks ‘\t’
interpretation and prints a multi-word sentence which is impossible without putting quotes
around. Also note the following example:
shell echo -e `<<<<< This message contains a |||line feed \n’
<<<<< This message contains a |||line feed
Which does not consider the above command to have redirections or pipes because the
corresponding symbols are inside quotation marks.
Implementation Hint: In your command parser that looks for certain tokens, have a counter of
how many single or double quotes you have seen so far. If the number is odd, it means you are
inside a quote and you should just ignore the token. On the other hand, if you are outside
(indicated by an even count), you should consider the token by its face value.
Your Task
In this assignment you will design a simple shell which implements a subset of the functionality
of the Bourne Again Shell (bash). The requirements are below:
1. Continually prompt the user for the next input
2. Parse the user input to extract the command(s) in it and execute it (them). The rules for
parsing are described below
3. For executing a command from the shell, you must use the fork()+exec() function
pair. You cannot use the system() function to do it because that creates a child
process internally without giving us explicit control. In addition, your shell must wait for
the executed command to finish which is achieved by using the wait()/waitpid()
4. Support input redirection from a file and output redirection to a file. Note that a single
command can have both
5. Allow piping multiple commands together connected by “|” symbols in between them.
Every process preceding the symbol must redirect its standard output to the standard
input of the following process. This is done using an Interprocess Communication
mechanism called pipe that is initiated by calling the pipe() system call
6. Run the user command in the background if the command contains a “&” symbol. Note
that you must avoid creating Zombie processes in this case
7. Allow directory handling commands (e.g., pwd, cd). Note that some of these commands
are not recognized by the exec() functions because there are no executables by the
same name. These are some additional shell features that must be implemented using
system calls instead of forwarding to exec().
8. Print a custom prompt to be shown before taking each command. This should include
your user name and current date-time
9. Bonus: Allow $-sign expansion. See the last command in the grading instructions
command list in the following
10. Write a report describing your unique (you can skip things that are common to the entire
class) design choices, algorithms (e.g., how you implemented single/double quotes,
$-sign expansions) and any implemented bonus features with implementation technique
11. Make a youtube video of the demonstration following the grading instructions and
include a link to the video in the report
Grading Instructions for Shell
1. Checking single and double quotes: 5 points (2.5 points each)
A. echo “Hello world | Life is Good Great $”
B. echo ‘Hello world | Life is Good Great $’
2. Simple commands with arguments: 20 points
If commands only w/o arguments work deduct 15 points.
A. ls
B. ls -l /sbin # should list from the /sbin directory that has many
C. ls -l -a
D. ls -la #this should be same as c
E. ps -aux
F. ls -l .. # lists the parent directory
G. To check if they wait for the child process, run the following:
H. sleep 5
This should block the shell for 5 seconds. If the shell comes back immediately, deduct 5 points
because they are not using wait()/waitpid().
3. Input/Output redirection (20 points – 10 points each):
A. ps aux a
B. grep /init < a
C. grep /init < a b # grep output should go to file b
If they do not have the last part (i.e., input output redirection in the same line), deduct 5 points
4. Single pipe (10 pts):
Any single pipe command
5. Two or more pipes (20 points):
A. ps aux | awk ‘/init/{print $1}’ | sort -r
B. ps aux | awk ‘/init/{print $1}’ | sort -r | awk ‘/ro/’ | grep ro
C. ps aux | awk ‘{print $1$11}’ | sort -r | grep root
6. Two or more pipes with input and output redirection (5 points):
A. ps aux test.txt
B. awk ‘{print $1$11}'<test.txt | head -10 | head -8 | head -7 |
sort output.txt
C. cat output.txt
7. Background processes: 10 points
A. sleep 1 &
B. sleep 2 &
C. sleep 20 &
The prompt should come back immediately, otherwise deduct 5 points. Also, look for the code
where they handled background processes. They should get the pid of the child process, put
that in a vector or list, and before getting every command they should cycle through each pid in
the list with waitpid (pid, 0, WNOHANG). This should not take time at all and they cannot get
stuck in a long running process because WNOHANG makes the function non-blocking.
Give some partial points for effort. Otherwise, deduct full 10 points if they do not have anything
significant about bg processes.
8. Directory processing (5):
A. cd ../../
B. cd .
C. cd /home/
D. cd – # goes back to the last directory you were in before this
one. It is similar to the back button in the browser
9. Own prompt 5 points and misc: Deduct 5 points if they do not print any prompt
10. Bonus (10 pts): $ sign expansion
A. cat /proc/$(ps|grep bash|head -1|awk ‘{print $1}’)/status
B. mkdir a
C. cd $(ls -l | grep ‘^d’|head -1|awk ‘{print $9}’) # goes inside
the first directory in the current directory
11. Other open-ended bonus options (anything outside the below list will be credited based on
A. [3 pts] Command history (pressing Up/Down button goes to previous/next command)
B. [3 pts] Autocomplete by pressing tab