Support --cgroup-parent, bridge name option, enable_icc=false and user namespace remap for hosting multi-tenant apps Context

Open
#213 1 comment 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
35/100
Issue type
Feature
Clarity
Mostly clear
Activity status
Active
Tech stack
docker, linux, rust

Research direction

Start by locating how Kiri reads Docker API HostConfig and network options, and how it configures cgroups, Linux bridges, and user namespaces. Check which of the four controls are currently unsupported and whether ignored options can be rejected with a warning or 400 response. Done means the supported behavior and isolation boundary are clearly defined, with tests or documentation covering each requested control.

Written by the indexing model from the issue text.

Description

enhancement
What you're trying to do

I am evaluating Kiri as a runtime for a shared-hosting panel that runs untrusted user binaries, one project per network, in containers from gcr.io/distroless/static with --read-only, --cap-drop=ALL, --security-opt no-new-privileges, and --user uid:uid. Kiri already handles memory (--memory), CPU (--cpus), --pids-limit, read-only rootfs, capabilities, and no-new-privileges, and it forces memory.swap.max=0, which is what we want. Thanks for that.
The following flags are accepted by the Docker API but silently ignored. The container starts, so nothing reports an error, but the protection is missing:
1. HostConfig.CgroupParent is not read. Containers do not go into the given slice, so we cannot set one aggregate CPU/memory limit per project across its containers.
2. com.docker.network.bridge.name network option is not read. Bridges are always named kiriN. This breaks host firewall rules that match bridge names (iifname "gsh-p-*").
3. com.docker.network.bridge.enable_icc=false is not implemented. Containers on the same user-defined network can likely still reach each other. Please confirm and tell us how to get this isolation.
4. User namespace remap (userns-remap or per-container UsernsMode) is not available. Root inside the container is root on the host kernel, with no uid mapping. We would like at least an option to map container root to an unprivileged host uid, or a documented statement on the isolation boundary.

What you'd like Kiri to do

Why it matters
These are defence-in-depth controls for running code from different tenants on one host. Without 1 and 4, a single project cannot be capped as a group, and a container compromise has less separation. Without 2 and 3, a firewall rule written against the bridge name silently does nothing.
Requested
• Either implement the four items above, or state clearly which are out of scope, so users know what the isolation boundary is.
• For any flag that is ignored, return a warning or a 400 instead of silently accepting it.
Environment: Kiri Linux (host, no VM), commit 1befa94 / release , Debian 13 amd64.

Would you like to build it?

No, just suggesting it

Dominant language
Rust
Stars
7.5k
Forks
329
Avg merge
7h 48m
Merged PRs (30d)
105

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 libracoder/kiri

All issues in libracoder/kiri

Similar issues

More Rust issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.