Project 1: System Inspector (v 1.0)
Starter repository on GitHub: https://classroom.github.com/a/C9U2qWAU
Our journey through the operating system starts in user space, outside the kernel. In this project, we’ll implement a Unix utility that inspects the system it runs on and creates a summarized report for the user. If you’ve ever used the
top command from a shell, our program will be somewhat similar. To give you an idea of how your program will work, here’s a quick example:
Hostname: magical-unicorn | Kernel Version: 5.15.4 | CPUs: 2 Uptime: 16 days, 11:25:42 Load Average (1/5/15 min): 0.09 0.14 0.09 CPU Usage: [######--------------] 30.0% Memory Usage: [#####---------------] 29.3% (421.7 MiB / 1.4 GiB) Tasks: 101 total 2 running, 0 waiting, 98 sleeping, 1 stopped, 0 zombie PID | Task Name | State | User ----------+---------------------------+--------------+----------------- 9783 | sshd | running | root 11642 | sl | stopped | mmalensek 12701 | inspector | running | mmalensek
To get this information, you will use the
proc, the process information pseudo-filesystem. While there are other ways to get the information displayed above, you are restricted to using
proc in this assignment. There are two great resources for finding out what information is available in
cd /procin your shell and then run
lsto view the files. You’ll see process IDs and several other virtual files that are updated dynamically with system information. Each line shown above in the process listing corresponds to a numbered directory in
- Check out the man page:
man procfs. (If this doesn’t work, you probably need to install the man pages:
pacman -Sy man-pages). The manual has a complete description of every file and directory stored under
For a quick example, try running
cat /proc/uptime. You’ll see the number of seconds the system has been running printed to the terminal.
In this assignment, you will get experience working with:
close()system calls for reading file data
- Tokenizing text files
readdirfunctions for listing directory contents
- Argument parsing with
- Load averages, calculating CPU usage, and Linux tasks
- Text-based UIs with ncurses
The program supports a few command line options. We’ll let the program do the talking by printing usage information (-h option):
[magical-unicorn:~/P1-malensek]$ ./inspector -h Usage: ./inspector [-ho] [-i interval] [-p procfs_dir] Options: * -h Display help/usage information * -i interval Set the update interval (default: 1000ms) * -p procfs_dir Set the expected procfs mount point (default: /proc) * -o Operate in one-shot mode (no curses or live updates)
Pay particular attention to the
-p flag. This allows us to change the directory where
proc is mounted (
/proc by default). We will use this option to test your code with our own pre-populated copy of
Populating the Output
Your main responsibility in this project is implementing a variety of procfs-related functions (prefixed
pfs_*) to retrieve the required information. The UI for the program was written using the ncurses library – assuming your
pfs_* functions work properly, you shouldn’t need to worry about modifying the UI (although you are welcome to do so). In ‘one-shot’ mode, the simple UI is disabled and printed directly to the terminal instead.
Here’s some tips to guide your implementation:
- Remember to close the files/directories you open! If too many file descriptors are left open, subsequent open calls will fail.
- Truncate strings that are too long to display (applicable to the process and user names). Process names should be no longer than 25 characters and usernames must be no longer than 15 characters. Otherwise, you can run into memory issues.
- Use the total and available memory when reporting memory usage.
- If you need to store a list of items that you don’t know the size of ahead of time,
realloccan be a good option.
Hostname, Kernel Version, CPUs, Load Average
These should be straightforward to locate and extract the data for with minimal post-processing.
Format uptime as
D day[s], HH:MM:SS. If the machine has been running for 1 day, do not print the ’s' in ‘days.’ If the machine has been running for less than 1 day, only print the hours, minutes, and seconds of uptime.
Overall CPU Usage
CPU usage is calculated by sampling over a period of time, i.e., the CPU was active for 70% of one second. You will record the CPU usage, sleep for one second (this is handled by the UI code), and then get a second reading to determine the usage percentage. The CPU metrics in
/proc/stat will add up to 100% because idle time is included. You’ll need to track idle time separately, so the calculation will look something like:
(usage2 - usage1) / (total2 - total1)
total1 are from the first sample and
total2 are from the second sample. You’re effectively calculating the change in these two samples to get a usage percentage.
If the CPU usage percentage is NaN (not a number), or you encounter errors performing the calculation, report 0%. Since this calculation requires two samples, your function should take the previous sample as one of its arguments.
You should support all of the process states listed in the
proc man pages. Additionally, for our purposes, we will consider ‘idle’ processes equivalent to ‘sleeping.’ To build the “running, waiting, sleeping, …” output, the specific state flags you’ll be interested in are
Idle and sleeping (I and S) processes are not shown in the program output.
You’ll notice that
proc doesn’t contain information about the username associated with running processes, but it does provide their ID numbers (UIDs). To map UIDs to usernames, you will need to parse the contents of
/etc/passwd. While there are functions that will do this for you, such as
getpwuid, you are required to build your own lookup functionality. (Note:
getpwuid can seemingly leak memory in certain situations, so we’re avoiding it here).
Restrictions: you may use any standard C library functionality. External libraries are not allowed unless permission is granted in advance. Your code must compile and run on your VM set up with Arch Linux as described in class – failure to do so will receive a grade of 0.
While there are several ways to retrieve the system information displayed by your project, you must retrieve the data from
One of the major components of this assignment is reading and parsing text files. To read the files, you are required to use the
read system call instead of the fancier C library functions like
scanf, etc. You also shouldn’t use
strtok, and processing the text files using regular expressions is not allowed.
Rationale: we’re using
read here to get familiar with how I/O works at a lower level. You will need to be able to understand
read for subsequent assignments. As for
strtok, it has several pitfalls (including not being reentrant or thread safe) that make it a bad choice. You can either use
strsep or the
next_token implementation provided in class.
Failure to follow these guidelines will result in severe deductions (at least 10 points) or a 0.
Check your code against the provided test cases. You should make sure your code runs on your Arch Linux VM.
Submission: submit via GitHub by checking in your code before the project deadline.
Your grade is based on:
- Passing the test cases (
- Remember to run
make testupdateto pull in the latest test cases.
- When you are satisfied with your project, use
make gradeto test it on our test hardware.
- You should continue to test your code for robustness beyond the provided test cases; i.e., just because you hard-coded a function to pass does not guarantee you will receive the points for its functionality.
- Remember to run
- Code review.
- Evaluation of code quality, stylistic consistency, cleanliness, efficiency, and documentation.
- Github issues will be filed when code quality must be improved. There will be a separate code review deadline to resolve these issues.
- Initial project specification posted (2/9)