3.1 The Source Directory

All references to the “Source Directory” below are to the CLS directory ~idallen/cst8177/14f/assignment02/ and that name starts with a tilde character ~ followed by a userid with no intervening slash. The leading tilde indicates to the shell that the pathname starts with the HOME directory of the account idallen (seven letters).

4.1 Set Up – The Base Directory on the CLS

1. Do a Remote Login to the Course Linux Server (CLS) from any existing computer, using the host name appropriate for whether you are on-campus or off-campus. All work in this assignment must be done on the CLS.

2. Create the following directory structure in your CLS HOME directory and record (for study purposes) the series of Unix commands you used to create it. Spelling and capitalization must be exactly as shown:

   CST8177-14F
   `-- Assignments
       `-- assignment02

This assignment02 directory is the Base Directory for most pathnames in this assignment. Store your files and answers in this Base Directory.

3. There is a Checking Program named assignment02check in the Source Directory on the CLS. Follow the instructions in the first two steps at the start of Checking Program to create a working symbolic link to this program.

Check your work so far using the checking program symlink.

4.2 Exploring the search PATH mechanism

This task explores the shell Search Path.

NOTE: In this task, you will change the PATH variable temporarily for the current shell session only. If at any time you want to undo these temporary changes, you can simply exit the shell that has the changes and then start a new shell, or re-login if you exited your login shell. (Below, we suggest you create a subshell so that you don’t log out when you exit!)

Since this task involves temporary changes to shell variables, you cannot resume this task in the middle if you log out or exit the shell. You must always start this task over from the beginning.

Do NOT edit your .bashrc to make permanent changes that affect future login sessions, unless explicitly told to do so.

1. Start a second copy of bash (a nested shell or subshell). (Watch the MP4 video on Subshells. Some sysadmin like to put the nesting level of the shell in their prompt using the $SHLVL variable.) You can use the ps command to verify that you have two copies of bash running. Re-read the NOTE, above.

2. Display the value of the shell environment variable PATH on your screen and verify that the standard system directory /bin is in the list of directories in your shell’s search PATH. (If the /bin directory is not listed in your PATH, something is very wrong; contact your instructor.)

   Hints: There are several ways to see the value of an environment variable (or local variable) on your screen. You could use a command that shows single lines of text on your screen. Review Shell Variables.

3. Use a command to discover which of your PATH directories contains the ls program. Use redirection to put the output of this command (the absolute path to the ls command) in file lspath.txt in your Base Directory. You will need to know this pathname later.

4. Store the current value of your PATH variable in a new shell variable named oldpath by entering a shell assignment statement oldpath=$PATH (no blanks). Make sure that both variables PATH and oldpath have the same values.

   Hints: Some commands that can display the value of environment variable PATH cannot be used to display the value of the new local non-environment variable oldpath. Choose correctly. Since setting the value of a shell variable doesn’t save it on disk, the saved value in oldpath is temporary, only valid for this shell. The value will be lost if you exit this shell and start a new shell. If you exit this shell before completing this task, you will lose the value of oldpath and must start the task over from the beginning.

5. Put the value of the shell variable oldpath into a new file named savepath.txt in your Base Directory. Hint: Display the value of the variable and use output redirection to put it into the file. It’s easy and accurate.

6. Remove all write permissions from the savepath.txt file, so that you don’t accidentally overwrite it.

7. Set your PATH variable to have no value. One method is to type nothing after the equals sign – just push Enter after the equals: PATH= 
   Use echo to display the new (empty) value of PATH on your screen. (Most other commands will not be found; echo is built-in to the shell and doesn’t need a valid search PATH.)

8. Try to run any non-builtin command (for example, ls, who, date, etc.) and notice that the attempt fails because the shell cannot find a program with that name in your current (empty) search PATH. Type the ls command name and use the correct syntax to redirect just the shell error message into a file called path_error.txt in your Base Directory.

   Hints: Recall how to redirect errors using shell Shell I/O Redirection. Display the contents of the file to make sure the error message is in the file: If you use the right absolute command pathname, you can run a file display command even if no commands can be found because of the empty PATH variable of the current shell. See the next question for an example of using an absolute pathname as a command name.

9. Run the ls program using its absolute pathname (the one you discovered and saved earlier). You do not need any directories in your PATH to run a command using a pathname containing slashes. The PATH is not used to look for command names containing slashes.

   Issue a shell built-in command (for example, echo, pwd, shopt) and verify that built-in commands are part of the shell process itself and also do not need to be found in PATH directories.

10. Redirect the output of shell built-in shopt into a file called shopt.txt in your Base Directory. Because this is a shell built-in command, the search PATH is not used.

11. Restore your PATH variable to its original value by restoring its value from the oldpath variable.

    Hints: This is the reverse shell assignment statement compared to what you did earlier to save the value: simply exchange the names. If you make a mistake, re-read the NOTE, above.

12. Display your PATH variable to verify it is restored (has the same value as in the variable oldpath).

13. Try a few non-builtin commands to verify they are found with the restored PATH in place.

14. Exit the nested sub-shell and return to your original login shell.

Check your work so far using the checking program symlink.

4.3 Creating your first shell script

1. In the Base Directory create the file firstscript.sh, containing the following two lines:
   1. The first line of the file should be exactly these 12 characters: #!/bin/sh -u (plus one more character for the newline at the end of the line).
   2. The second line of the file should be a command that displays the exact text: *** It's not a "Micro$oft" World *** on the screen. Make sure you count the asterisks and get the capitalization, quotes, and dollar sign correct. You will test that you got this right in the next few steps.

2. Make the firstscript.sh file executable for only the owner (you). Group and others can read the file but not execute it. Hint: Review Permissions from last term.

3. Execute the script and edit it to fix any errors. When the output of the script is correct on your screen, redirect the output into file firstscriptout in the Base Directory.

   Hints: Execute the script by typing its name. You will note that the shell cannot find firstscript.sh as an executable command name with no slashes, because the command is not on your search PATH. You need to avoid the search PATH by using a relative pathname containing a slash to execute this file, just as you do when executing the check symlink. Review Executing a program in the current directory.)

You have just finished creating your own very simple Linux command – a script that prints some text to its standard output. Linux sysadmin create dozens of command scripts that do various system administration tasks. The scripts eliminate typing errors and make repeating tasks much easier.

4. Create the directory bin in your own HOME directory as a place to keep your own personal commands. Hint: Make sure you create this directory in your own HOME directory.

5. We will give our new command a more command-like name, without the file extension. We will at the same time link the new name into our private bin directory that we just created. To do this, create a hard link from your firstscript.sh script to the name firstscript (no extension) inside your bin directory. You now have one file inode with two names. Hint: Review Links and Inodes.

6. Make the Base Directory your current directory. Now, without changing directories again, use a command to display the inode numbers and names of both script files to verify that they are hard links to the same inode. Use one command with two relative pathnames. When the two output lines are correct (showing the relative pathnames and the identical inode numbers), save the output into file sameinodes.txt in the current (base) directory.

7. Change to your HOME directory. Without changing directories again, execute your new firstscript command that is inside your bin directory using the shortest relative pathname (from your HOME). Put the relative pathname you used into file binrel.txt in your Base Directory.

   Hint: A fast way to create binrel.txt is to echo the pathname that you just used and use redirection to create the file in the base directory.

8. Start a subshell. Append the absolute pathname of your new bin directory to the right end (not the start) of the PATH variable of your current shell. Remember how you did this – you will need to do it again later in your .bashrc file. Hint: Review Appending to PATH.

9. Display your PATH to confirm that the absolute path of your bin directory is on the right end (not at the start). do not add the directory more than once – it should only appear once in your PATH variable. If you need to start over, exit this subshell and then start another one to get a fresh, good copy of PATH.

10. if you have done the previous steps correctly, you can now type the command name firstscript (no slashes, no file extension) and the shell will find your firstscript script in your bin directory and execute it (because your bin directory is now in the PATH for this shell).

    Hint: Now that your bin directory is in your search PATH, you can use the TAB key to do command name completion on the firstscript command name, so you don’t have to type the whole name. Let the shell do your typing for you!

11. When you have your own firstscript command working using your modified PATH variable, save a copy of your modified PATH variable in file binpath.txt in your Base Directory. Make sure there are no duplicate directories in PATH before you save it. Your HOME bin directory must be at the end.

The change we just made to this shell’s PATH variable will be lost when the shell exits. We need to save this change so that it happens every time we log in. We do this using the .bashrc file.

12. Create a backup copy of your .bashrc file, in case you need to start over.

13. Make the necessary changes to your .bashrc file to add the absolute path of your bin directory to the right end of your PATH variable. The syntax to do this is exactly as you did earlier with your interactive shell. Save the file.

14. Log out and log in to have your .bashrc execute. Check your PATH variable to make sure the absolute path of your bin directory is at the right end of your PATH. Verify that your firstscript command from your own bin directory still works (no slashes). The shell will do command completion on this name if you use the TAB key.

You will note that every time you start a new nested subshell, your .bashrc appends another copy of your HOME bin to your shell’s search PATH variable. We will learn in later assignments how to avoid this duplication.

Check your work so far using the checking program symlink.

4.4 Putting the current directory in PATH is a bad idea

1. Make your HOME directory your current directory.

2. Create a fake ls command using these steps:
   1. Copy the firstscript script you created above to a new file named ls in the current (your HOME) directory.
   2. Edit the new file, changing the text message on the second line to read exactly this (one long line): You are running fake ls with all your privileges - this script could remove all your files!
   3. Make the new file executable if it is not already.
   4. Execute the ls script in the current directory and make sure it prints your text message on the screen. Hint: To execute the ls script, you need a relative command pathname that includes a slash.

You have just finished setting a trap for anyone who has the current directory . near the beginning their PATH and runs ls while in your HOME directory.

3. Still in your HOME directory, type the ordinary ls command name and note that you get the real ls command, because your PATH does not include the current directory. You do not have . at the start of your PATH so you didn’t run the fake ls in the current (HOME) directory.

4. Start a subshell (a second copy of the shell).

5. Make yourself a victim: temporarily put the current directory . at the front (start) of your current shell’s PATH variable. Hint: Review Appending to PATH, but make sure you add to the front (left) end of the PATH this time.

6. Display the new value of your PATH variable and verify that the current directory name . is the first directory in the list, followed by a colon character (:) and the rest of the search PATH. Redirect and save this new PATH into file badpath.txt in the Base Directory.

7. Still in your HOME directory (where the fake ls command resides), type the ls command name as you normally would. Notice that your PATH setting means you’ve just run the fake ls command, not the real ls command!

   This fake ls script executes with your permissions, and it has privileges to do anything to your account that your own userid can do.

Imagine you are running as the super-user root account (not a good idea) and you have made the mistake of putting the current directory in your PATH as we have done here. (DON’T DO THIS!) Imagine further that you notice a new directory on the system that shouldn’t be there. As root you change into this directory and type ls to see what it contains. A user has placed a fake ls command in this directory. Because you’re root, that fake ls command in the current directory runs with root privileges – and it can destroy your system.

NEVER PUT THE CURRENT DIRECTORY IN YOUR PATH!

8. Exit the subshell that has the bad PATH.

   Make sure your current shell’s PATH does not contain the current directory before you continue! Log out and back in, if needed.

Check your work so far using the checking program symlink.

4.5 Creating a script

This task explores Processes and Jobs.

In this task you will create a script that prints out the number of processes each user is currently running on the system. You will create a series of scripts, each being an enhancement of the previous one, until you have the final product.

NOTE: The command that gives you a “snapshot of the current processes” has several kinds of options, some with dashes (UNIX style) and some without (BSD style), because it is really two or three commands merged into one program. You sometimes find that mixing options from the different types (with and without dashes) gives error messages. It’s best not to mix option types. See the first few paragraphs of the man page.

1. Run a command that prints out a “full-format listing” and “select all processes” running on the system. Search in the man page for the terms “Select all processes” and “full-format” to discover the correct two UNIX-style options. (UNIX-style means the option letters must be preceded by a dash, unlike BSD-style options.) When you run the command with these two UNIX-style options, the output will be several hundred lines long, with every line starting with the userid that owns that process. The first few lines will look similar to this:

   UID        PID  PPID  C STIME TTY          TIME CMD
   root         1     0  0 Sep12 ?        00:00:11 /sbin/init
   root         2     0  0 Sep12 ?        00:00:01 [kthreadd]

   (Hint: Pipe the huge output into a command that displays only the first few lines, so you can confirm the above format.)

2. Create a new two-line script file procs.sh in your Base Directory by copying your existing two-line firstscript.sh file and replacing the second line of the new file to use the process listing command you used from the previous step. Do NOT change the first line.

3. Run the new script procs.sh to make sure it works. It should generate the full-format, all-processes listing for you, just as if you had typed it yourself on the command line.

As described in the section on awk in Data Mining, the command awk '{print $1}' reads lines from standard input and prints just the first space-delimited field of each line from that input. The output of the process listing command from your new script is conveniently in space-delimited columns.

4. Run a command pipeline that pipes the output from your procs.sh script into an awk command line that selects just the first leading (userid) column of each line. You should end up with a list of several hundred userids on your screen. The first three lines will be the same as the first column of the sample output given above.

5. Copy procs.sh to procs_users.sh in the same directory.

6. Change the second line in the new procs_users.sh file to be a command pipeline prints just the first column of the process listing. (Add the previously-used awk command to the end of the line to make it a command pipeline, as you did in a previous step.) Do not change the first line.

7. Run the new script procs_users.sh to make sure it works. It should generate a list of several hundred userids, just as if you had typed the two commands yourself on the command line.

8. Pipe the output of your new procs_users.sh script into the sort command. The output will be the same list of several hundred userids, now in sorted order.

9. Copy procs_users.sh to procs_sorted.sh in the same directory.

10. Enhance the command pipeline in the new procs_sorted.sh file, adding a further pipeline into the sort command so that the one column of userids comes out sorted.

11. Run the new script procs_sorted.sh to make sure it works. It should generate a list of several hundred userids in sorted order, just as if you had typed the three commands yourself on the command line.

12. Pipe the output of your new procs_sorted.sh script into the command that counts adjacent identical lines. Instead of hundreds of lines of output, the output will be counts of the number of processes being run by each unique userid. You will see a few dozen lines, depending on how many users are logged in. The lines will all have this output format of number followed by userid, though the actual numbers and accounts may differ:

     1 102
     1 UID
     2 avahi
     1 colord
     2 cst8207a
     8 cst8207c
     ...etc...

    Hint: There is a command that can count the number of occurrences of adjacent input lines, displaying each unique line preceded by the count of the number of times that line appeared. To recall this unique command name and its option to count adjacent lines, review the Command List from last term and the many counting Examples of pipes in the pages Shell I/O Redirection and Data Mining.

13. Copy procs_sorted.sh to procs_counted.sh.

14. Enhance the command pipeline in the new procs_counted.sh script file to show the counts of processes run by each unique userid.

15. Run the new script procs_counted.sh to make sure it works. It should generate a list of more than a dozen unique userids, each preceded by a count, just as if you had typed the four commands yourself on the command line.

We can now see that there are two odd things in the output that are not account userids:

• 102 – a number, not a userid. This appears to be a bug in the process listing program, since the passwd file does have an account entry for uid 102: messagebus
• UID – the column heading from the first line of output of the process listing. We don’t want to see this in our output.

16. Copy procs_counted.sh to procs_counted2.sh.

17. Enhance the command pipeline in the new procs_counted2.sh file to eliminate the bogus UID column heading in the output. There are at least two ways to do this; pick either way and make the change:
    • One way is to insert a pipeline filter that removes any line containing the string UID. Hint: the command that searches inside files for lines matching a pattern has an option to invert the match and only show non-matching lines. Search the man page.
    • Another way is to see if the process listing command has an option to print no header line. Hint: It does! Searching the man page for no header will find it. You will need to read about the problematic use of the single-letter option and how they solved it using long options.
    • DO NOT use any removal method that relies on the number of lines output or the order of the lines output, since the number of lines and the user names in those lines will change from day to day.

18. Run the new script procs_counted2.sh to make sure it works. It should generate a list of more than a dozen userids in sorted order, with no bogus UID in the output.
    • Re-read the note above about how NOT to remove the UID line!

19. Copy procs_counted2.sh to procs_counted_header.sh.

20. Enhance the new procs_counted_header.sh script so that it prints out its own header (title) line before it generates the output. The header line should be the exact text “NumProc Account” (15 characters, including the single space, with upper-case for the N, P, and A) and you can optionally output some dashed underlining under it as well, to make the heading stand out. Line up the blank in the title with the blank in the two columns of output. The output will look similar to this, though the actual numbers and accounts may differ:

    NumProc Account
    ------- -------
          1 102
          2 avahi
          1 colord
          2 cst8207a
         10 cst8207c
    ... etc ...

21. Run the new script procs_counted_header.sh to make sure it works. It should generate its own header line, some optional underlining, followed by a list of more than a dozen userids in sorted order, with no bogus UID in the output.
    • Re-read the note above about how NOT to remove the UID line!

22. The long script name procs_counted_header.sh is helpful for knowing what the script does, but it’s much too long for a command name. Make a hard link from procs_counted_header.sh to the name pch in your HOME bin directory.

23. Because your bin directory is in your PATH, you should now be able to run pch as a command name with no slashes. You have written your second Linux command script.

Check your work so far using the checking program symlink.

4.6 When you are done

That is all the tasks you need to do.

Check your work a final time using the Checking Program and save the output as described below. Submit your mark following the directions below.