osmk 0.1.3

Write x86 operating systems in Python — one instruction at a time

osmk

Write x86 operating systems in Python — one instruction at a time.

Every function call maps to an x86 assembly instruction. build() assembles it into a bootable 512-byte disk image.

Install

pip install osmk

You also need NASM (the assembler) and optionally QEMU (to run your OS):

# Ubuntu/Debian
sudo apt install nasm qemu-system-x86

# macOS
brew install nasm qemu

# Windows
# NASM: https://nasm.us → download & add to PATH
# QEMU: https://qemu.org/download/#windows

Or test without installing QEMU at copy.sh/v86 — upload your .img as a floppy disk image.

Quick Start

import osmk
from osmk import ax, bx

osmk.start("hello.img")    # begin a new OS

osmk.ax(5)                 # mov ax, 5
osmk.bx(5)                 # mov bx, 5
osmk.add(ax, bx)           # add ax, bx  →  ax = 0x000A
osmk.writehex()             # print AX as hex → "000A"

osmk.build()                # assemble → hello.img (512 bytes)

Run it:

qemu-system-i386 -drive format=raw,file=hello.img

API Reference

Lifecycle

Function	Description
start(filename, bits=16, org=0x7C00)	Begin a new OS program
build()	Assemble and create the bootable image
dump_asm()	Return the generated NASM source (for debugging)
run(memory=128)	Build and launch in QEMU

Register Setters

Every register is callable. osmk.ax(5) → mov ax, 5.

osmk.ax(5)          # mov ax, 5
osmk.bx(ax)         # mov bx, ax
osmk.al(0x41)       # mov al, 0x41
osmk.eax(0x1000)    # mov eax, 0x1000

Available: ax bx cx dx si di sp bp · al ah bl bh cl ch dl dh · eax ebx ecx edx esi edi esp ebp

Instructions

# Arithmetic
osmk.add(ax, bx)       osmk.sub(ax, 1)
osmk.mul(bx)            osmk.div(cx)
osmk.inc(ax)            osmk.dec(cx)
osmk.neg(ax)

# Bitwise
osmk.and_(ax, 0xFF)    osmk.or_(ax, bx)
osmk.xor(ax, ax)       osmk.not_(ax)
osmk.shl(ax, 4)        osmk.shr(ax, 1)

# Compare & Jump
osmk.cmp(ax, 0)
osmk.je("done")        osmk.jne("loop")
osmk.jl("less")        osmk.jg("greater")
osmk.jz("zero")        osmk.jnz("nonzero")
osmk.jmp("start")      osmk.jc("carry")

# Stack
osmk.push(ax)           osmk.pop(bx)

# Subroutines
osmk.call("myfunc")     osmk.ret()

# Interrupts
osmk.interrupt(0x10)    # int 0x10
osmk.cli()              osmk.sti()
osmk.hlt()

# Labels
osmk.label("loop")

Labels & Scoping

# Global labels — must be unique
osmk.label("main")

# Local labels — prefix with '.', scoped to nearest global label above
osmk.label("main")
osmk.label(".loop")        # belongs to 'main'
osmk.jnz(".loop")

osmk.label("other")
osmk.label(".loop")        # different .loop, belongs to 'other' — no collision
osmk.jnz(".loop")

Output Helpers

osmk.writehex()         # print AX as 4-digit hex (e.g. "000A")
osmk.writechar("A")     # print character 'A'
osmk.writechar()        # print whatever's in AL
osmk.prints("Hello!")   # print a string
osmk.newline()          # print CR+LF
osmk.readchar()         # wait for keypress → AL
osmk.halt()             # halt CPU (cli + hlt loop)

Functions

Define reusable subroutines with the function() context manager. ret is appended automatically. Labels inside are auto-scoped as local labels, so multiple functions can use the same label names without collision.

with osmk.function("print_stars") as fn:
    fn.mov(cx, 5)
    fn.label("loop")         # becomes .loop (local to print_stars)
    fn.asm("mov ah, 0x0E")
    fn.asm("mov al, '*'")
    fn.interrupt(0x10)
    fn.dec(cx)
    fn.jnz(".loop")
    # ret is auto-appended here

with osmk.function("print_dashes") as fn:
    fn.mov(cx, 5)
    fn.label("loop")         # different .loop — no collision
    fn.asm("mov ah, 0x0E")
    fn.asm("mov al, '-'")
    fn.interrupt(0x10)
    fn.dec(cx)
    fn.jnz(".loop")

osmk.call("print_stars")
osmk.call("print_dashes")

The function builder supports the full DSL: fn.mov(), fn.add(), fn.sub(), fn.inc(), fn.dec(), fn.cmp(), fn.push(), fn.pop(), fn.call(), fn.interrupt(), fn.label(), fn.jmp(), fn.je(), fn.jne(), fn.jz(), fn.jnz(), fn.jl(), fn.jg(), fn.asm(), fn.comment().

Custom / Escape Hatch

When osmk doesn't cover what you need:

# Inject raw NASM assembly
osmk.asm("mov ah, 0x0E")
osmk.asm("""
    mov ah, 0x00
    mov al, 0x03
    int 0x10
""")

# Inject raw bytes
osmk.raw(b"\x90\x90\x90")   # 3x NOP

# Define data
osmk.data("my_var", "dw 0x1234")

# Add comments
osmk.comment("this sets up video mode")

Explicit MOV

osmk.mov(ax, 0x0E)     # same as osmk.ax(0x0E)
osmk.mov(bx, ax)       # same as osmk.bx(ax)

Examples

Hello World

import osmk

osmk.start("hello.img")
osmk.prints("Hello, World!")
osmk.newline()
osmk.prints("My first OS!")
osmk.halt()
osmk.build()

Keyboard Echo

import osmk
from osmk import al

osmk.start("echo.img")
osmk.prints("Type anything:")
osmk.newline()

osmk.label("loop")
osmk.readchar()        # key → AL
osmk.writechar()       # echo it

osmk.cmp(al, 13)       # Enter key?
osmk.jne("loop")       # no → keep looping
osmk.writechar(10)     # yes → also print line feed

osmk.jmp("loop")
osmk.build()

Counter

import osmk
from osmk import ax, cx

osmk.start("counter.img")

osmk.ax(0)
osmk.cx(16)

osmk.label("count")
osmk.writehex()
osmk.prints(" ")
osmk.inc(ax)
osmk.dec(cx)
osmk.jnz("count")

osmk.halt()
osmk.build()

How It Works

1. Each osmk.* call appends an x86 assembly instruction to an internal list
2. build() wraps them in a bootloader template (16-bit real mode, MBR)
3. NASM assembles it into a flat binary
4. The binary is exactly 512 bytes with the 0xAA55 boot signature

The generated code runs in 16-bit real mode using BIOS interrupts for I/O. Use dump_asm() to see exactly what assembly your Python generates.

Resources

• OSDev Wiki — OS development reference
• NASM Documentation — assembler docs
• x86 Instruction Reference — complete instruction set
• Writing a Simple OS — Nick Blundell's guide
• BIOS Interrupt List — int 0x10, 0x13, 0x16, etc.

License

MIT