GCP: pinning only RTMR1/RTMR2 with an endorsed MRTD (and is flashbots-images' measure-gcp still used?)

オープン
#246 コメント 0 件 リアクション 0 件 担当者 0 名 GitHub で見る

メンテナーはふだん 1 日以内に返信

まだ誰も着手していません。

評価

難易度
5/5
見積もり時間
1週間以上
初心者へのやさしさ
30/100
issue の種類
機能追加
明瞭さ
おおむね明確
活発さ
活発
技術スタック
github, shell
領域
devops, security

調査の方向性

Start by tracing make measure-gcp, scripts/gcp_measurements_to_dcap.sh, and the CI and documentation references to determine whether the GCP measurement path is still used. Then review the linked attested-tls proposal and how GCP TDX evidence verifies MRTD; the issue is done when the maintainers decide whether to upstream that safety change and how to handle direct attest measure gcp output.

索引モデルが issue の本文から書いたものです。

説明

At Seismic we admit GCP TDX guests by pinning RTMR1 and RTMR2, while still requiring MRTD to be a firmware build Google has endorsed (like you guys do in your portable measurement policies). We'd like to compute those values with attest measure gcp instead of maintaining our own predictor. Two questions:

1. Is anyone still using make measure-gcp in flashbots-images?

It runs dstack-mr (flashbots/dstack-mr-gcp), whose only consumer is scripts/gcp_measurements_to_dcap.sh. CI and the docs only use measure-portable / measure-portable-gcp. dstack-mr models RTMR2 as the command line plus initrd only, so it probably doesn't match UKIs built with systemd-stub ≥ 256, which also measures each section. If nobody uses it, happy to open a PR removing measure-gcp, the reshape script and the dstack-mr-gcp flake input.

2. Would you take the changes that make RTMR1/RTMR2-only GCP policies safe in attested-tls?

kind: dcap output is rejected (UnsafeAttestMeasureDcapOutput) because MRTD would go unchecked, which makes sense. But an explicit gcp-tdx record that pins only rtmr1/rtmr2 is accepted today and has the same gap. What we'd propose:

  1. Verify Google's endorsement of MRTD for every gcp-tdx evidence, not only inside the portable comparison, and record it with the evidence so archived verification can re-check it. We run this in our fork (https://github.com/SeismicSystems/attested-tls/pull/7), happy to clean it up and upstream it if you guys agree with this approach.
  2. With (1) in place, document gcp-tdx records pinning only rtmr1/rtmr2 as safe. They already parse.
  3. Optionally, accept attest measure gcp output directly. That needs attest to say which platform the output is for, since measure gcp and measure self-hosted both emit kind: dcap (an attestation_type field would do). attested-tls could then map it to a gcp-tdx record.

Would you take (1)? Any preference on (3)?

主要言語
Shell
スター
26
フォーク
18
平均マージ
3日 10時間
マージ済み PR(30日)
13

環境構築

このプロジェクトには開発コンテナ、Dockerfile、コントリビューションガイドがありません。まず README を読み、一般的な手順ははじめてのコントリビューションガイドを参照してください。

はじめの一歩

  1. issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
  2. 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
  3. リポジトリをフォークし、ブランチを切って変更します。
  4. issue 番号を参照したプルリクエストを送ります。

flashbots/flashbots-images のほかの issue

flashbots/flashbots-images の issue をすべて見る

似ている issue

Shell/Bash の issue をもっと見る

新しい issue をメールで受け取る

初心者向けの GitHub issue を短くまとめたダイジェスト。