Guix Days: Bootstrapping ARM

During the Guix Days before FOSDEM, some of us discussed bootstrapping on ARM architectures. We focused on how to port Mes to ARM. This post consists of notes from that discussion.

Recap: i686/x86_64 Reduced Binary Seed

We started our discussion by reviewing the current status for i686-linux and x86_64-linux. Jan (janneke) Nieuwenhuizen gave a similar summary a few days later at FOSDEM, and you can read the slides online.

Previously, the size of Guix's binary seed totaled about 250 MB. Now, on the core-updates branch, it's been reduced to about 130 MB. This is nearly a 50% reduction in size, which is great progress! Using this 130 MB reduced binary seed, it's currently possible to bootstrap Guix for both i686-linux and x86_64-linux.

To bootstrap x86_64-linux, we actually "cheat" and bootstrap from the i686-linux bootstrap binaries. This is possible because an x86_64-linux system can natively run i686-linux executables, and also because Guix automatically cross-compiles for the host platform during the bootstrap process. In other words, on an x86_64-linux system, Guix uses the i686-linux bootstrap binaries to build a cross-compilation toolchain, which it then uses to build a final, normal x86_64-linux toolchain. Guix then uses this final toolchain to build everything else for the x86_64-linux system.

That's great news for owners of i686 and x86_64 machines! But what about ARM? Although we could cross-compile the bootstrap binaries for ARM from an x86_64 machine, this isn't great because it would increase the number of things a person or organization would have to verify in order to audit the system. Perhaps more importantly, it would force owners of ARM machines to implicitly trust an x86_64 machine. The dominant vendors of CPUs implementing the x86_64 architecture, Intel and AMD, both include a management engine in many of their products, which represents a serious risk to user freedom.

In the short term, cross-compilation is better than nothing, but in the long term, we'd prefer to bootstrap ARM without cross-compiling from another architecture. Concretely, we'll need to complete at least the following tasks.

TODO: Implement a Mes backend for ARM

We need to implement a new Mes backend for an ARM architecture. We should choose an ARM instruction set that can work on a variety of ARM platforms with minimal fuss. The following candidates were suggested:

After the session concluded, Danny Milosavljevic committed some changes to the wip-arm branch of mes which enabled many of the tests to pass - but some tests still fail, and you can help finish the work!

TODO: Port mescc-tools to ARM, also

The mes project depends upon the mescc-tools project, which also must be ported. The mescc-tools project contains an M1 macro assembler, which would need to be extended to support ARM branches. Currently, ARM branches are very broken.

TODO: Improve Guix integration and merge core-updates

Even if we had a new Mes backend and mescc-tools for ARM, there would still be more to do. The Guix integration is not quite complete - the core-updates branch still needs to be merged with master, and we'll need to fix any problems that arise. Even on i686-linux, the bottom of the bootstrap path is incomplete. Preliminary Guix code exists on the wip-bootstrap branch to achieve a scheme-only bootstrap, but help would be welcome!

You can help!

In summary, you can help the Mes project by doing any of the following things:

There's still plenty of meaty work left to be done! If you're interested, get in touch and we'll help you get started.

Think Big: Bootstrapping without an OS

In addition to the immediate tasks necessary for porting Mes to ARM, we also took some time to think about the long term hopes and dreams of the bootstrappable project.

We discussed how in the long term, in parallel with the aforementioned tasks, it should be possible to investigate how to bootstrap an entire system without relying on a OS or even a kernel running on the machine. For example, one can imagine loading the transitive closure of source (including a tiny, human-readable machine code program to kick off the entire process) into a computer as a kind of "firmware image". When the computer runs, it would execute this "firmware image" and eventually produce a fully bootstrapped system.

Think Bigger: Bootstrapping Hardware

We also briefly talked about how even after we achieve full source software bootstrap, we will still need to tackle the problem of "hardware bootstrap". It isn't clear what form this will eventually take, but surely free hardware design will play an important role in ensuring that we can trust our hardware, too.

About Bootstrappable Builds and Mes

Software is bootstrappable when it does not depend on a binary seed that cannot be built from source. Software that is not bootstrappable - even if it is free software - is a serious security risk for a variety of reasons. The Bootstrappable Builds project aims to reduce the number and size of binary seeds to a bare minimum.

GNU Mes is closely related to the Bootstrappable Builds project. Mes aims to create an entirely source-based bootstrapping path for the Guix System and other interested GNU/Linux distributions. The goal is to start from a minimal, easily inspectable binary (which should be readable as source) and bootstrap into something close to R6RS Scheme.

Currently, Mes consists of a mutual self-hosting scheme interpreter and C compiler. It also implements a C library. Mes, the scheme interpreter, is written in about 5,000 lines of code of simple C. MesCC, the C compiler, is written in scheme. Together, Mes and MesCC can compile a lightly patched TinyCC that is self-hosting. Using this TinyCC and the Mes C library, it is possible to bootstrap the entire Guix System for i686-linux and x86_64-linux.

About GNU Guix

GNU Guix is a transactional package manager and an advanced distribution of the GNU system that respects user freedom. Guix can be used on top of any system running the kernel Linux, or it can be used as a standalone operating system distribution for i686, x86_64, ARMv7, and AArch64 machines.

In addition to standard package management features, Guix supports transactional upgrades and roll-backs, unprivileged package management, per-user profiles, and garbage collection. When used as a standalone GNU/Linux distribution, Guix offers a declarative, stateless approach to operating system configuration management. Guix is highly customizable and hackable through Guile programming interfaces and extensions to the Scheme language.

موضوع‌های مرتبط:

ARM Bootstrapping FOSDEM Guix Days

Unless otherwise stated, blog posts on this site are copyrighted by their respective authors and published under the terms of the CC-BY-SA 4.0 license and those of the GNU Free Documentation License (version 1.3 or later, with no Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts).