hyperlight-dev / hyperlight-dev/hyperlight
Consider creating a wrapper for hyperlight-host that locates sandboxes in remote processes
- 主要言語
- Rust
- スター
- 4.7k
- フォーク
- 208
- 平均マージ
- 1日 7時間
- マージ済み PR(30日)
- 47
説明
It should be possible to create a drop-in replacement for the hyperlight-host crate that exposes the same public API, but wraps the `UninitializedSandbox` and `MultiUseSandbox` types with wrappers that implement its calls via (optionally?) making RPCs to different processes. This might be useful for people who wish to run sandboxes in remote, low-privilege processes for even-more-defense-in-depth (e.g. against hypervisor kernel vulnerabilities), and might allow working around per-process resource limits.
This should be possible to do via a new crate which is a drop-in wrapper around the hyperlight-host crate, avoiding the need for any changes in Hyperlight itself.
Semantically, it also shouldn't require any changes to other crates that depend on hyperlight-host, like hyperlight-{wasm,js,unikraft}. Unfortunately, due to rust-lang/cargo#9227, there may be distribution challenges due to the inability to for a root crate to override `hyperlight-host` with some hypothetical `hyperlight-host-remote-process-wrapper` throughout the transitive dependency graph. The wrapper crate would either need to use `name = "hyperlight-host"` and be used as a git/path dependency, or hyperlight-{wasm,js,unikraft} would need to be made aware of its existence/configured to optionally depend on it.
コントリビューションガイド
調査の方向性
Start by reviewing the public API of the hyperlight-host crate, especially UninitializedSandbox and MultiUseSandbox, and how hyperlight-wasm, hyperlight-js, and hyperlight-unikraft depend on it. Investigate the Cargo dependency and override constraints described in the issue. Done would mean a separate drop-in crate can preserve the API while locating sandboxes in remote processes without requiring changes to dependent crates.
索引モデルが issue の本文から書いたものです。
評価
- 技術スタック
- rust
- 領域
- operating-systems, security
- issue の種類
- 機能追加
- 難易度
- 5/5
- 見積もり時間
- 1週間以上
- 活発さ
- 活発
- 明瞭さ
- 説明が足りない
- 初心者へのやさしさ
- 35/100