Porting mimalloc to Solo5

Open
#1,234 0 comments 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 1 day

Nobody has claimed this yet.

Assessment

Difficulty
5/5
Estimated time
Over a week
Newbie friendliness
18/100
Issue type
Feature
Clarity
Needs clarification
Activity status
Stale
Tech stack
c

Research direction

Start by reading Solo5's include/solo5.h memory-range definition and toolchain/cc.in, then compare the linked OCaml sysdeps_solo5.c sbrk implementation. The open questions are stack and heap management, freestanding standard-library responsibilities, and platform detection. Done would be an agreed design and a proper porting pull request, but this issue does not define those decisions.

Written by the indexing model from the issue text.

Description

Solo5 is a sandboxed execution environment primarily intended for, but not limited to, running applications built using various unikernels (a.k.a. library operating systems).

I have a local branch (against dev) with a (mostly) complete port, but there are a few things I'd like to discuss before making a proper PR.

  • Solo5 is a "bare" target. At a program's entry point, it provides a flat memory range in which both the call stack and the heap will live. For now I require the user to reserve memory manually using a static stack size (i.e. mi_manage_os_memory(info->heap_start, info->heap_size - STACK_SIZE, [...])) before doing any allocation, making _mi_prim_alloc always return ENOMEM. However, in OCaml's Solo5 toolchain (where they use dlmalloc), they implement a "dummy" sbrk which dynamically checks if the heap gets too close to the call stack. My opinion is that a static stack size should be fine since processes running on an ordinary OS usually also get a limited stack. It also makes the code simpler.

  • Solo5 doesn't provide a standard C library. The headers (errno.h, string.h, ...) and functions (abort, mem*, ...) mimalloc expects must be come from somewhere. I think the user (not mimalloc) should be responsible for providing such a library. For example, I'm interested in writing unikernels in Koka which requires a lot more from the standard library, and splitting a standard library between different projects depending on their particular needs doesn't make much sense to me.

  • The Solo5 toolchain's cc is just a wrapper around the host system's cc. For now I detect that we're building for Solo5 with defined(__STDC_HOSTED__) && __STDC_HOSTED__ == 0, which is set that way because of the -ffreestanding flag. This is not very robust, we have to check for Solo5 before any other platform (since cc will still define platform-specific constants for the host, like __OpenBSD__) and maybe we'll want to add other freestanding platforms in the future.

I'd be happy to read some opinions. It's fine if this isn't wanted in the main repo, I just wanted to put it out there if anyone's interested.

Dominant language
C
Stars
13.6k
Forks
1.2k
Avg merge
18h 12m
Merged PRs (30d)
20

Getting set up

This project ships no dev container, Dockerfile or contributing guide, so setting up is up to you: start from its README, and see our first-contribution guide for the general steps.

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from microsoft/mimalloc

All issues in microsoft/mimalloc

Similar issues

More C issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.