Moss: a Rust Linux-compatible kernel in 26,000 lines of code

Original Article ↗ Hacker News Discussion ↗

Project scope and current capabilities

  • Rust + a bit of assembly, ARM64-only so far; boots on QEMU and several dev boards.
  • Aims for Linux user‑space binary compatibility; currently runs most BusyBox commands and tools like vim.
  • Major gaps: only read‑only FAT32 filesystem, no networking stack yet, limited syscalls and drivers, not a drop‑in Ubuntu/Android kernel.
  • Uses a libkernel that can run in user space for easier debugging of core logic (e.g., memory management), seen as getting some microkernel‑like development benefits in a monolithic design.

Rust async/await in the kernel

  • Several comments explain Rust async/await as compiler‑generated state machines (Futures) plus an executor, not tied to OS‑level threads or a heavyweight runtime.
  • Others argue that, in a Linux‑compatible kernel with real threads and schedulers, integrating async, work‑stealing, and multi‑CPU scheduling is non‑trivial.
  • Embedded Rust frameworks (e.g., single‑threaded async RTOS‑like environments) are cited as evidence that async can work with minimal runtimes.
  • One thread links this to historic coroutine‑style I/O frameworks, noting Rust removes much of the “hand‑rolled coroutine” pain.

Linux compatibility strategy

  • “Linux‑compatible” here primarily means user‑space ABI: same syscalls and behavior where implemented, not internal kernel APIs.
  • Missing syscalls currently cause a panic, prompting the author to implement them as they appear in real workloads.
  • Commenters note this is a pragmatic way to focus kernel work on actually‑used syscalls without immediately recreating the full Linux toolchain.

Licensing debate (MIT vs copyleft)

  • Large subthread: some fear an MIT‑licensed, Linux‑ABI‑compatible kernel will be forked by vendors, filled with proprietary drivers, and shipped on locked‑down hardware.
  • Others counter that:
    • GPL has not prevented widespread binary blobs in Linux, nor guaranteed contributions; enforcement is uneven.
    • FreeBSD and other permissive kernels do get meaningful contributions from some large users.
    • LLVM/Clang and many popular user‑space projects show permissive licenses can still attract corporate investment.
  • Philosophical clash:
    • Copyleft advocates stress protecting end‑user freedom and the “commons.”
    • Permissive‑license advocates emphasize unconditional “gifts,” developer freedom, and reject moral pressure on the author’s choice.

Long‑term goals and questions

  • Author mentions mid‑term goal of being self‑hosting (edit, fetch deps, and build Moss from Moss), and long‑term aspiration as a daily‑driver OS on chosen hardware.
  • Commenters ask about outperforming Linux via different page sizes or interrupt strategies, adding networking (e.g., via smoltcp), reimplementing KVM, and comparisons with other Rust kernels.