Writing init program using Python on Linux

As you may know, the first process in Linux is the init (PID 1) — it runs while the system is up. Init is normally a compiled program (statically or dynamically linked). Can we write init in Python? Yes. Two simple ways, with focus on cross-compiling Python and writing the init script.

1) PyInstaller (quick prototype)
Bundle the Python script into a single executable with PyInstaller and place it as /sbin/init. Fast to try, but the result is large and opaque — not recommended for production.

2) Cross-compile / install Python on the target (main focus)

1. Prepare a cross-toolchain and sysroot for the target architecture (e.g. arm64).
2. Configure the Python build for the target (./configure with proper --host/--build/--prefix). Decide static vs dynamic:
   • Static: bigger but fewer runtime deps on target.
   • Dynamic: smaller, but you must ship matching shared libraries (libc, libssl, libpython, …).
3. make and make install into a staged rootfs so /usr/bin/python3 and stdlib are present.
4. Copy required shared .so files into the rootfs if Python is dynamic and fix ld.so paths as needed.
5. Write your init script with a shebang (e.g. #!/usr/bin/python3), put it at /sbin/init inside the rootfs and chmod +x.
6. Test the image under QEMU and on hardware; capture early logs on serial/tty.

Tip: Ensure /dev exists — the kernel usually mounts devtmpfs on /dev. If devtmpfs is disabled or you use an initramfs, create /dev/urandom (device node) in the rootfs before starting Python; otherwise Python may block or fail. Also keep a small fallback init (e.g., BusyBox) for recovery if Python cannot start.