The first and last exercises are problem set-esque questions that could easily appear on a midterm or final exam. In fact, all of the questions asked under Problem 4 were on previous midterms and finals. The middle two problems are experiments that'll require you fire up your laptop and run some programs and development tools.

The lab checkoff sheet for all students—both on-campus and off—can be found right here.

This lab was designed by Jerry Cain.

Problem 1: Direct, Singly Indirect, and Doubly Indirect Block Numbers

Assume blocks are 512 bytes in size, block numbers are four-byte ints, and that inodes include space for 6 block numbers. The first three contain direct block numbers, the next two contain singly indirect block numbers, and the final one contains a doubly indirect block number.

• What's the maximum file size?
• How large does a file need to be before the relevant inode requires the first singly indirect block number be used?
• How large does a file need to be before the relevant inode requires the first doubly indirect block number be used?
• Draw as detailed an inode as you can if it's to represent a regular file that's 2049 bytes in size.

Problem 2: Experimenting with the stat utility

This problem is more about exploration and experimentation, and not so much about generating a correct answer. The file system reachable from each myth machine consists of the local file system (restated, it's mounted on the physical machine) and networked drives that are grafted onto the fringe of the local file system so that all of AFS—-which consists of many, many independent file systems from around the globe—all contribute to one virtual file system reachable from your local / directory.

Log into myth52 and use the stat command line utility (which is a user program that makes calls to the stat system call as part of its execution) and prints out oodles of information about a file. Type in the following commands and analyze the output:

• stat /
• stat /tmp
• stat /usr
• stat /usr/bin
• stat /usr/bin/g++
• stat /usr/bin/g++-5

The output for each of the five commands above all produce the same device ID but different inode numbers. Read through this to gain insight on what the Device values are.

For each of the above commands, replace stat with stat -f to get information about the file system on which the file resides (block size, inode table size, number of free blocks, number of free inodes, etc).

Now log into myth55 and run the same commands. Why are the outputs of stat and stat -f the same in some cases and different in others?

Now analyze the output of the stat utility when levied against AFS mounts where the master copies of all /usr/class and /usr/class/cs110 files reside. Do this from both myth52 and myth55.

• stat /usr/class
• stat -f /usr/class
• stat /afs/ir.stanford.edu/class
• stat -f /afs/ir.stanford.edu/class
• stat /usr/class/cs110
• stat /afs/ir.stanford.edu/class/cs110
• stat -f /usr/class/cs110

Why are most of the outputs the same for myth52 compared to myth55? Which ones are symbolic links? Why are the device numbers for remotely hosted file systems so small? What about these commands?

• stat /afs/northstar.dartmouth.edu
• stat -f /afs/northstar.dartmouth.edu
• stat /afs/asu.edu
• stat -f /afs/asu.edu

What files can you see within the dartmouth.edu and asu.edu mounts?

Problem 3: valgrind and orphaned file descriptors

Here's a very short exercise to enhance your understanding of valgrind and what it can do for you. To get started, type the following in to create a local copy of the repository you'll play with for this problem:

~$ git clone /usr/class/cs110/repos/lab1/shared lab1
~$ cd lab1
~$ make

Now open the file and trace through the code to keep tabs on what file descriptors are created, properly closed, and orphaned.

What are dup and dup2? These are system calls that let us copy file descriptors. Here's more information about each of them:

• dup: creates a copy of the entry at the passed-in file descriptor location in the lowest-numbered unused file descriptor position and returns this second file descriptor. What this means is that both these file descriptor entries point to the same open file table entry. This means you can use either to access the file, and reading/writing with one impacts the other. You must also close both the original and the copy. For example, if we say int fd2 = dup(fd1) where fd1 = 3 (and assume no file descriptors beyond 3 are used), this means that fd2 = 4 and thus the index 3 and 4 file descriptor entries both point to the same open file table entry.

• dup2: creates a copy of the entry at the first passed-in file descriptor location in the file descriptor location specified by the second passed-in file descriptor. This is the same as dup, except that instead of the copy being the next available file descriptor, it is specified as the second parameter. If the second parameter is an already-open file descriptor, it is closed before being used. For example, if we say dup2(fd1, fd2) where fd1 = 3 and fd2 = 5, this means that the index 3 and 5 file descriptor entries both point to the same open file table entry.

With this information, try tracing through the program to better understand the file descriptors that are created, closed, and left open. In particular, with the knowledge that open will use the lowest unused file descriptor, as will dup, calculate the actual file descriptor numbers. Then run valgrind ./nonsense to confirm that there aren't any memory leaks or errors (how could there be?), but then run valgrind --track-fds=yes ./nonsense to get information about the file descriptors that were (intentionally) left open. Without changing the logic of the program, insert as many close statements as necessary so that all file descriptors (including 0, 1, and 2) are properly donated back. (In general, you do not have to close file descriptors 0, 1, and 2, but for this problem you should.)

Problem 4: Short Answer Questions

Provide clear answers and/or illustrations for each of the short answer questions below. Each of these questions is either drawn from old exams or based on old exam questions. Questions like this will certainly appear on the midterm.

1. The dup system call accepts a valid file descriptor, claims a new, previously unused file descriptor, configures that new descriptor to alias the same file session as the incoming one, and then returns it. Briefly outline what happens to the relevant file entry table and vnode table entries as a result of dup being called. (Read man dup if you'd like, though don't worry about error scenarios).

2. Now consider the prototype for the link system call (peruse man link). A successful call to link updates the file system so the file identified by oldpath is also identified by newpath. Once link returns, it’s impossible to tell which name was created first. (To be clear, newpath isn’t just a symbolic link, since it could eventually be the only name for the file.) In the context of the file system discussed in lecture and/or the file system discussed in Section 2.5 of the secondary textbook, explain how link might be implemented.

3. Explain what happens when you type cd .././../. at the shell prompt. Frame your explanation in terms of the file system described in Section 2.5 of the secondary textbook, and the fact that the inode number of the current working directory is the only relevant global variable maintained by your shell.

4. All modern file systems allow symbolic links to exist as shortcuts for longer absolute and relative paths (e.g. search_soln might be a symbolic link for /usr/class/cs110/samples/assign1/search_soln, and tests.txt might be a symbolic link for ./mytests/tests.txt). Explain how the absolute pathname resolution process we discussed in lecture would need to change to resolve absolute pathnames to inode numbers when some of the pathname components might be symbolic links.