vxgmichel/arm-linux-from-scratch

An ARM Linux program from scratch

A 32-bit ARM Linux program written in assembly using customasm.

It converts lowercase characters to uppercase, leaving other characters untouched.

Rationale

The idea behind this project was to produce a working Linux binary without relying on any tool that contains knowledge about Linux or any specific CPU. Surprisingly enough, all that's needed to achieve this is:

• An agnostic assembler
• The ELF specification (to produce a working executable)
• The instruction set specification for the target architecture (for the actual program)
• The Linux system call specification (to communicate with the kernel)

Here's the corresponding tools and resources I ended up using:

• First python, then the amazing customasm project
• In-depth: ELF - The Extensible & Linkable Format by stacksmashing
• The THUMB instruction set reference document
• The Linux system call table for ARM 32-bit

I also used some bits of the following ARM documents:

• The ARM instruction set reference document
• The ARM-THUMB procedure call standard
• The quick reference card for THUMB

Tools

First install qemu and customasm:

# Install qemu, unless you're already on a raspberry pi
$ sudo apt install qemu-user
# Install customasm, or download from: https://github.com/hlorenzi/customasm/releases
$ cargo install customasm

Assembly

Then assemble the upper binary:

# Assemble with customasm and give executable access
$ customasm upper.asm -o upper && chmod +x upper
customasm v0.11.4 (x86_64-unknown-linux-gnu)
assembling `upper.asm`...
writing `upper`...
success

Execution

And try it out:

# Run with qemu explicitely
$ qemu-arm upper --help
Usage: upper [OPTION]
Transform lowercase ascii characters to uppercase.

Other ascii characters are left untouched.
Data is read from stdin and written to stdout.

  --help  display this help and exit

# Or implicitely
$ echo "1.. 2.. This is a test.." | ./upper
1.. 2.. THIS IS A TEST..

# The binary is actually quite small, less than half a KB
wc -c upper
475 upper

# customasm can also produce an annotated binary
$ customasm upper.asm -f annotated -p
customasm v0.11.4 (x86_64-unknown-linux-gnu)
assembling `upper.asm`...
success

  outp | addr | data

   0:0 |    0 | 7f          ; #d 0x7f
   1:0 |    1 | 45 4c 46    ; "ELF"
   ...