sysprobe

Element take-home exercise

Overview

Write a program called sysprobe in your preferred language (C++, Python, or Bash). The program should demonstrate your ability to use Linux system features that are common in embedded development.

You will implement several subcommands that inspect or interact with the system. You can work on any Linux platform with any distribution (we recommend using Ubuntu on a VM/WSL).

No additional hardware or setup is required. Submit your source code, a README with installation and usage instructions, and any necessary Linux configuration files.

Requirements

Program name: sysprobe
Interface: command-line tool with subcommands
Usage: sysprobe [options]

The subcommands will be chosen from the tasks list below.
You must pick any 3 tasks to implement (doing more is optional but not required).

Constraints

• Use only standard libraries (no third-party packages).
• Be robust: your program should run on any Linux embedded board/server/computer without modifications, so handle missing files/devices/permissions gracefully (don’t crash).
• Because you will work with advanced internal features, we recommend working as the root user (or with sudo su).
• Keep it small, readable, and comment where necessary.

Tasks (Choose 3 out of 4)

Task A: sysfs

Sysfs is a pseudo-filesystem in Linux that provides a structured, virtual interface to view and modify kernel data, particularly for devices and drivers. Mounted at /sys, it exposes information about the kernel's device model through a file-based tree, allowing user-space programs to read device attributes and control kernel parameters.

Provide the user of sysprobe with information about the hardware using sysfs.

Find useful data about temperature and CPU, and display it as valid JSON (no need to match a specific format).
Example

# Command
sysprobe get-sys-data

# Possible output
{
 "thermal": [{"name": "zone0", "temp_c": 47.5}],
  "cpufreq_hz": [2400000, 2400000]
}

If files are missing, return an empty array instead of failing.

Task B: Filesystem watch

From man: The inotify API provides a mechanism for monitoring filesystem events. Inotify can be used to monitor individual files or directories. When a directory is monitored, inotify will return events for the directory itself and for files inside the directory.

Monitor a given directory for create/modify/move/delete events for N seconds. Print one line per event. Match the following format:
%Y-%m-%dT%H:%M:%SZ <event> <dir> (see example below). If the directory does not exist, create it first.

# Command
sysprobe watch --dir /tmp/sysprobe --secs 10

# Possible output
2025-08-08T10:00:00Z CREATE /tmp/sysprobe/a
2025-08-08T10:00:01Z MODIFY /tmp/sysprobe/a
2025-08-08T10:00:02Z DELETE /tmp/sysprobe/a

Notice: Use inotify (or similar). No busy loops.

Task C: Tracing (debugfs)

Read about kernel debugfs

Mount the kernel debugfs if it is not already mounted: mount -t debugfs none /sys/kernel/debug.
Inside tracing/events, choose a few events you would like to trace and enable them via the filesystem (only read/write to files are needed for enabling).
Then, read and print the trace pipe to the user for N seconds.

# Command
sysprobe trace --secs 2

# Possible output
sudo-3501    [003] ...1. 23908.815071: sys_write(fd: 7, buf: 5c8099305920, count: 64)
cat-4110    [006] ...1. 23908.814982: sys_write(fd: 1, buf: 7a6c7ccd9000, count: 62)

Task D: Init system

Add a monitoring script to your system that will run on init.

• Choose one of the tasks you implemented, and create an init script using it.
• The script should automatically run after you restart your computer.
• Its content should probe anything you choose from the previous tasks, and write the output to a file with a timestamp. This way, you will have a log file you can track through several boots.
• The script can run for a few seconds and exit; it does not need to be a daemon.
• For example - on every boot, the command sysprobe get-sys-data will run, and these lines will be added to the file sysprobe.log:

System data for Wed May 18 06:45:59 PM EEST 2025:
{
 "thermal": [{"name": "zone0", "temp_c": 47.5}],
  "cpufreq_hz": [2400000, 2400000]
}

You may use any Linux method for init scripts, as long as it works :) Add an explanation for why you chose the init manager you used.

To submit this task, provide a commands file explaining how you made the script run as an init script, along with the script itself.

Deliverables

• Source code

• README with:

  • How to build/run
  • Which tasks you implemented, and any notes about them
  • "Demo" (text/screenshot) of the usage on your machine.

Good Luck!