vinzenz/hyperhive
vinzenz/hyperhive
Watch
0
0
Fork
You've already forked hyperhive
 0
Code Activity

drop the hive's push of the legacy core/meta repo

Browse source 
Nothing reads core/meta: builds come from internal/meta. Remove push_meta, ensure_meta_remote, meta_read_access, ensure_repo, the per-agent forge sync that only served them, and the git commit path in meta.rs; remove_permission_files now just deletes the seed files.
This commit is contained in:
atlas 2026-10-25 16:35:46 +01:00 • committed by mara
commit b6b568e97e
15 changed files with 81 additions and 518 deletions
Download patch file Download diff file Expand all files Collapse all files

4
docs/agent-lifecycle/config-changes.md
View file

@ -289,8 +289,8 @@ mounts via `set_nspawn_flags`:
manage any agent", so the mounted set is every agent.
- `/var/lib/hyperhive/applied/` → `/applied/` (RO) — every agent's
authoritative applied repo, including `.git`.
- `/var/lib/hyperhive/meta/` → `/meta/` (RO) — the swarm-wide
deploy flake.
- `/var/lib/hyperhive/meta/` → `/meta/` (RO) — the hive's plain state
(topology, swarm limits).
An agent without the permission only has its direct children's config
dirs.

17
docs/agent-lifecycle/persistence.md
View file

@ -387,17 +387,12 @@ here from merged config PRs) + any other files those PRs commit.
`.git/` carries the proposal / approved / building / deployed / failed /
denied tag history.
Under `/var/lib/hyperhive/meta/` — the swarm-wide deploy flake plus
system-level config files. Single git repo for the whole host; hive-c0re
commits every mutation that should survive a restart here.
Contents:
Under `/var/lib/hyperhive/meta/` — system-level config files, plain state
with no git history that anything reads. The flake agents build from is the
swarm's `internal/meta`, not here. Contents:
- `flake.nix` — declares one `nixpkgs` input per agent + one
`nixosConfigurations.<n>` output per agent. `flake.lock` is the
canonical "what's deployed where." The git log is the deploy
audit trail (one commit per successful deploy or hyperhive bump).
- `topology.json` — the agent roster (`["alice", "bob", "ruth"]`).
Written by `topology::reconcile` on every meta sync; read by
Written by `topology::reconcile` from `agent_set::refresh`; read by
`topology::all_agents`, which is the set the `manage_root_agent`
permission grants mounts over. The reader also accepts a map-shaped
file carrying a `parent` per agent, and keeps its keys.
@ -419,13 +414,13 @@ Contents:
hive-wide only and come straight off `HiveEnv`, so this file has no field
for them.
- `resource-limits.json` — retired. Per-agent overrides set on the hive used
to live here; nothing reads or writes it now. An old copy in the meta repo
to live here; nothing reads or writes it now. An old copy in this directory
is harmless and you can delete it.
An agent holding `manage_root_agent`, normally ruth, has the meta dir
RO-mounted at `/meta/`.
Outside the meta repo, `/var/lib/hyperhive/swarm-agent-permissions.json`
Outside this directory, `/var/lib/hyperhive/swarm-agent-permissions.json`
holds the permissions the swarm grants this hive's agents, as
`workers::swarm_permissions` last read them from the swarm's
`agent-permissions` bucket (every 30 seconds): a copy, not the

4
docs/integrations/forge.md
View file

@ -122,9 +122,7 @@ Two things live in the `agent-configs` Forgejo organization:
[config-changes.md § Config changes](../agent-lifecycle/config-changes.md#config-changes)).
hive-c0re pushes nothing to this org; it only reads from it.
Repos stay private, so an agent can't read another
agent's config. (Agents remain read-only collaborators on the legacy
`core/meta`, the hive's old permission-seed repo.)
The agent's **persistent meta flake input** is this repo's `main` head,
agent's config. The agent's **persistent meta flake input** is this repo's `main` head,
pinned by sha in `internal/meta` (`agent-<n>.url =
git+<forge>/agent-configs/<n>.git?rev=<sha>`; see
[config-changes.md § Meta flake](../agent-lifecycle/config-changes.md#meta-flake)).

8
docs/web-ui/shape.md
View file

@ -233,10 +233,10 @@ enough — wiping and rebuilding still destroys the state even if
`JSON.stringify({...visible fields...})`. On each render, the pattern
reuses cache-hit rows verbatim (preserving textarea draft, checkbox
state, and event listeners) and rebuilds only cache-miss rows. Used for: containers (`containerRowCache`),
rebuild-queue entries (`rebuildQueueRowCache`). The spawn-form input+focus and meta-input
checkboxes use a lighter snapshot-then-restore pattern (snapshot
before `replaceChildren`, restore after) since they're single
values rather than per-row caches.
rebuild-queue entries (`rebuildQueueRowCache`). The spawn-form input+focus
uses a lighter snapshot-then-restore pattern (snapshot
before `replaceChildren`, restore after) since it's a single
value rather than a per-row cache.
<!-- vale write-good.Passive = NO -->

2
frontend/packages/dashboard/src/tabs.js
View file

@ -172,7 +172,7 @@ import { renderUserMenu } from "./user-menu.js";
// broker traffic flows through the SSE channel. The only
// /api/state fetches are the initial cold load and the
// post-submit refetch on forms without `data-no-refresh`
// (tombstones, meta-input updates).
// (tombstones).
if (pollTimer) {
clearTimeout(pollTimer);
pollTimer = null;

119
hive-c0re/src/forge/mod.rs
View file

@ -1,6 +1,5 @@
//! Optional Forgejo wiring — the legacy `core/meta` repo (the hive's old
//! permission-seed repo, which `meta::remove_permission_files` still commits
//! to) and its read-access grants. Those writes, as `core`, run only beside a local `hive-forge`
//! Optional Forgejo wiring — the `core` admin user and token and its avatar.
//! Those writes, as `core`, run only beside a local `hive-forge`
//! container; reads of the forge go through [`reader_token`] on every hive.
//! The swarm-wide orgs, teams and repos, and each agent's read grant on
//! `internal/docs`, are swarm-controller's. Full design:
@ -15,10 +14,7 @@ mod users;
pub(crate) use internal_meta::internal_meta_source;
pub use reader::reader_token;
pub use reconcile::{fetch_forge_main, reconcile_config_apply, reconcile_config_status};
pub use repos::{
clone_config_into_proposed, ensure_meta_remote, ensure_repo, fetch_config_main_into_applied,
meta_read_access, push_meta,
};
pub use repos::{clone_config_into_proposed, fetch_config_main_into_applied};
pub use users::core_token;
use std::future::IntoFuture;
@ -74,13 +70,13 @@ pub(crate) fn validate_env() -> Result<()> {
Ok(())
}
/// Git remote for `repo` (e.g. `"core/meta"`) — **credential-free**.
/// Git remote for `repo` (e.g. `"hyperhive/hyperhive"`) — **credential-free**.
///
/// The token does not go here. A URL is a process argument, and `argv` is
/// world-readable through `/proc/<pid>/cmdline` for as long as the git child
/// lives, so a credentialed remote publishes the core admin token to every
/// local user on the host. Credentials travel in the environment instead, via
/// [`core_auth_header`] and [`crate::lifecycle::git_command_authed`] —
/// [`basic_auth_header`] and [`crate::lifecycle::git_command_authed`] —
/// `/proc/<pid>/environ` is owner-only.
pub(crate) fn forge_git_url(repo: &str) -> String {
git_url_with_base(forge_http_base(), repo)
@ -92,11 +88,6 @@ fn git_url_with_base(base: &str, repo: &str) -> String {
format!("{base}/{repo}.git")
}
/// The `http.extraHeader` value authenticating as the forge core user.
pub(crate) fn core_auth_header(token: &str) -> String {
basic_auth_header("core", token)
}
/// The `http.extraHeader` value carrying `token`. Forgejo takes the account
/// from the token, so `login` only labels the request.
///
@ -120,16 +111,6 @@ fn basic_auth_header(login: &str, token: &str) -> String {
/// still can't reach *another* agent's config. hive-c0re only reads from it.
pub(crate) const CONFIG_ORG: &str = "agent-configs";
/// Leak `s` to get a `&'static str` warning `kind` for the small, bounded
/// set of boot warnings in [`ensure_all`] keyed by a runtime name (at
/// most a handful per process). [`crate::warnings::set_boot_warning`]
/// requires a `'static` kind so distinct orgs/repos don't clobber each
/// other's warning entry; leaking a few short strings once per boot is
/// cheap and bounded, unlike a per-request or per-loop-iteration leak.
fn static_kind(s: String) -> &'static str {
Box::leak(s.into_boxed_str())
}
/// Probe whether `hive-forge` exists as a nixos-container. Cheap —
/// `nixos-container list` is just a directory scan in /etc. Routed
/// through hive-priv: `nixos-container` needs root, and hive-c0re runs
@ -216,42 +197,7 @@ where
.map_err(|_| anyhow::anyhow!("{op}: no answer from the forge within {API_TIMEOUT:?}"))
}
/// Per-agent forge sync: grant read access to the legacy `core/meta`, and wire
/// the `meta` remote in its proposed repo. All operations are idempotent;
/// failures are logged as warnings but don't abort the caller.
///
/// `core_token` is `core_token()` — passed in so callers that already
/// fetched it don't re-read the file. Pass `None` to skip the
/// `meta_read_access` step (safe: the access grant is best-effort).
///
/// Called by both `ensure_all()` (startup sweep) and `rebuild_agent`
/// (per-rebuild) so the two paths stay equivalent.
/// Returns `true` if all steps succeeded, `false` if any step failed. The
/// caller can use the return value to aggregate per-agent failures into a
/// dashboard warning (see [`ensure_all`]); the rebuild path ignores it and
/// relies on the journal `warn!` lines alone (a rebuild is its own retry).
pub async fn sync_agent(name: &str, core_token: Option<&str>) -> bool {
// The agent's forge account, token and `agent-configs/<name>` repo are
// swarm-controller's; this hive creates none of them and pushes nothing
// to the repo.
let mut ok = true;
// Grant read-only access to the legacy core/meta and wire the `meta`
// remote into the proposed repo. Agents build from `internal/meta`, not
// from this repo.
if let Some(token) = core_token
&& let Err(e) = meta_read_access(name, token).await
{
tracing::warn!(%name, error = ?e, "forge: ensure_meta_read_access failed");
ok = false;
}
if let Err(e) = ensure_meta_remote(name).await {
tracing::warn!(%name, error = ?e, "forge: ensure_meta_remote failed");
ok = false;
}
ok
}
/// The `core_token.is_some()` half of [`ensure_all`]: the legacy meta repo, the
/// The `core_token.is_some()` half of [`ensure_all`]: the
/// core avatar, and removing this hive's own config-PR hook —
/// every step that needs an authenticated forge client. The swarm-wide
/// objects (orgs, the `operators` team, mirrors, `internal/docs`,
@ -260,17 +206,6 @@ pub async fn sync_agent(name: &str, core_token: Option<&str>) -> bool {
/// `ensure_all` under clippy's function-length limit; not meant to be called
/// from anywhere else.
async fn ensure_all_orgs_and_repos(token: &str) {
// The legacy meta repo lives at core/meta. The only push to it is
// `meta::remove_permission_files`' commit; make sure the repo exists
// before that push hits a 404.
if let Err(e) = ensure_repo("meta", token).await {
tracing::warn!(error = ?e, "forge: ensure_repo core/meta failed");
crate::warnings::set_boot_warning(
"forge_ensure_meta_repo",
"crit",
format!("forge: core/meta repo provisioning failed: {e}"),
);
}
if let Err(e) = ensure_core_avatar(token).await {
tracing::warn!(error = ?e, "forge: ensure_core_avatar failed");
crate::warnings::set_boot_warning(
@ -382,11 +317,10 @@ async fn wait_until_ready() -> bool {
}
/// Sweep the forge at startup: seed the `core` admin user (hive-c0re's own
/// identity for pushing the legacy meta repo + driving the API) and its token, the
/// `core/meta` repo and the local knowledge clone, then run [`sync_agent`]
/// over every existing container (manager + sub-agents). Agent accounts,
/// identity for driving the API, and the one nix's git reads as) and its token,
/// then its avatar. Agent accounts,
/// tokens and `agent-configs/<name>` repos are the swarm-controller's, as is
/// the `agent-configs` org; this sweep pushes nothing to them.
/// the `agent-configs` org; this sweep touches none of them.
/// Called once at hive-c0re startup. Per-step failures are logged
/// but don't abort the sweep.
pub async fn ensure_all() {
@ -413,37 +347,6 @@ pub async fn ensure_all() {
if let Some(token) = core_token.as_deref() {
ensure_all_orgs_and_repos(token).await;
}
let Ok(containers) = crate::lifecycle::list().await else {
tracing::warn!("forge: nixos-container list failed; skipping user sweep");
crate::warnings::set_boot_warning(
"forge_container_list",
"crit",
"forge: nixos-container list failed; per-agent forge sync skipped this boot",
);
return;
};
let mut sync_failed: Vec<String> = Vec::new();
for c in containers {
let Some(name) = c.strip_prefix(crate::lifecycle::AGENT_PREFIX) else {
continue;
};
if !sync_agent(name, core_token.as_deref()).await {
sync_failed.push(name.to_owned());
}
}
if !sync_failed.is_empty() {
// Use static_kind to mint a `&'static str` key from the failed-agent
// list (bounded leak: one per hive-c0re boot, not per request).
let key = static_kind(format!("forge_sync_agent_{}", sync_failed.join("_")));
crate::warnings::set_boot_warning(
key,
"warn",
format!(
"forge: per-agent sync failed for: {} (see journal for per-step detail)",
sync_failed.join(", ")
),
);
}
}
#[cfg(test)]
@ -526,9 +429,9 @@ mod tests {
/// back to `core:<token>`, so the swap is genuinely equivalent auth and
/// not a silent downgrade to anonymous.
#[test]
fn core_auth_header_is_basic_core_token() {
fn basic_auth_header_is_basic_core_token() {
use base64::Engine as _;
let header = super::core_auth_header("s3cret");
let header = super::basic_auth_header("core", "s3cret");
let b64 = header
.strip_prefix("Authorization: Basic ")
.expect("basic auth header");

240
hive-c0re/src/forge/repos.rs
View file

@ -1,207 +1,10 @@
//! Repo + org plumbing on the local Forgejo: the meta + knowledge repos,
//! meta mirroring (`push_meta`), collaborator grants, pull-mirrors, and
//! branch-protection rules. The typed API-client constructor + org-name
//! constants live in the module root (`super`).
//! Repo plumbing on the local Forgejo: reseeding an agent's `applied` and
//! `proposed` repos from `agent-configs/<name>`. The typed API-client
//! constructor + org-name constants live in the module root (`super`).
use std::path::Path;
use std::sync::Mutex;
use anyhow::{Context, Result};
use forgejo_api::structs::{
AddCollaboratorOption, AddCollaboratorOptionPermission, CreateRepoOption, Repository,
};
use forgejo_api::{ApiErrorKind, ForgejoError};
use reqwest::StatusCode;
use crate::coordinator::Coordinator;
use super::{
CONFIG_ORG, api, bounded, core_auth_header, core_token, forge_git_url, forge_http_base,
is_present, reader_token,
};
/// Creation options for an empty repo defaulting to `main`.
fn repo_option(name: &str, private: bool) -> CreateRepoOption {
CreateRepoOption {
auto_init: Some(false),
default_branch: Some("main".to_owned()),
description: None,
gitignores: None,
issue_labels: None,
license: None,
name: name.to_owned(),
object_format_name: None,
private: Some(private),
readme: None,
template: None,
trust_model: None,
}
}
/// Whether a create-style call failed because the object already
/// exists. Forgejo signals this as HTTP 409 (conflict) or 422
/// (validation). The typed client surfaces those as
/// `ApiErrorKind::Other(409)` / `ValidationFailed` when the endpoint
/// spec lists the status, or as a bare `UnexpectedStatusCode`
/// otherwise — match all shapes defensively.
fn is_already_exists(e: &ForgejoError) -> bool {
match e {
ForgejoError::ApiError(api) => match api.error_kind() {
ApiErrorKind::ValidationFailed => true,
ApiErrorKind::Other(s) => *s == StatusCode::CONFLICT,
_ => false,
},
ForgejoError::UnexpectedStatusCode(s) => {
*s == StatusCode::CONFLICT || *s == StatusCode::UNPROCESSABLE_ENTITY
}
_ => false,
}
}
/// Fold a repo-creation result's "already exists" (409 / 422) into
/// success. `label` is `<owner>/<name>` — purely for log + error
/// context.
fn created_or_exists(res: Result<Repository, ForgejoError>, label: &str) -> Result<()> {
match res {
Ok(_) => {
tracing::info!(%label, "forge: created repo");
Ok(())
}
Err(e) if is_already_exists(&e) => {
tracing::debug!(%label, "forge: repo already exists");
Ok(())
}
Err(e) => Err(e).with_context(|| format!("create repo {label}")),
}
}
/// Create a repo in the token-owner's own namespace. `token` belongs
/// to the user we want the repo owned by (we use `core`'s token for
/// `core/meta`). Idempotent.
pub async fn ensure_repo(name: &str, token: &str) -> Result<()> {
let label = format!("core/{name}");
let res = bounded(
&format!("create repo {label}"),
api(token)?.create_current_user_repo(repo_option(name, true)),
)
.await?;
created_or_exists(res, &label)
}
/// Dashboard-warning guard for the meta non-fast-forward condition.
/// Held while the `core/meta` remote is ahead of our local mirror; cleared
/// automatically when the next push succeeds.
static META_NON_FF_GUARD: Mutex<Option<crate::warnings::WarningGuard>> = Mutex::new(None);
/// Push `dir` (the legacy hive meta repo) to `core/meta` on the local forge.
/// Its one caller is `meta::remove_permission_files`' commit.
/// Best-effort: returns Err which callers log + ignore. No-op when
/// the core token isn't present yet (forge container not provisioned).
pub async fn push_meta(dir: &Path) -> Result<()> {
let Some(token) = core_token() else {
return Ok(());
};
// Token-in-URL push. Forgejo accepts `oauth2:<token>` or just
// any-username:<token>; using `core` matches the owner so the
// remote name is self-describing.
//
// No --force: `meta.rs` uses regular `git commit` (append-only),
// so the push is always fast-forward in normal operation. Per
// operator directive, hive-c0re must not force-push anywhere. If
// the remote is somehow ahead (e.g. split-brain or manual push) we
// raise a dashboard warning and leave the remote intact rather than
// silently erasing its history.
let url = forge_git_url("core/meta");
let out = crate::lifecycle::run_with_deadline(
crate::lifecycle::git_command_authed(&core_auth_header(&token))
.current_dir(dir)
.args(["push", &url, "HEAD:main"]),
crate::lifecycle::FETCH_DEADLINE,
"git push core/meta",
)
.await?;
if !out.status.success() {
let stderr = String::from_utf8_lossy(&out.stderr);
if stderr.contains("non-fast-forward") || stderr.contains("fetch first") {
let msg = "forge/core/meta push rejected (non-fast-forward): \
remote is ahead of local; leaving remote history intact";
tracing::warn!("{msg}");
// Raise a persistent hive warning; cleared on next successful push.
if let Ok(mut g) = META_NON_FF_GUARD.lock() {
*g = Some(crate::warnings::set_warning(
"forge-meta-non-ff",
"warn",
msg,
));
}
return Ok(());
}
anyhow::bail!(
"git push core/meta failed ({}): {}",
out.status,
stderr.trim()
);
}
// Successful push: clear any outstanding non-ff warning.
if let Ok(mut g) = META_NON_FF_GUARD.lock() {
*g = None;
}
tracing::info!("forge: pushed meta to core/meta");
Ok(())
}
/// Grant agent `name` read-only collaborator access to `core/meta` on
/// the forge so the agent can clone/fetch that legacy repo. Idempotent:
/// re-adding an existing collaborator succeeds (Forgejo answers 204
/// either way).
pub async fn meta_read_access(name: &str, core_token: &str) -> Result<()> {
add_collaborator(
"core",
"meta",
name,
AddCollaboratorOptionPermission::Read,
core_token,
)
.await?;
tracing::info!(%name, "forge: granted meta read access");
Ok(())
}
/// Add the forge `core/meta.git` URL as the `meta` remote in the
/// agent's proposed config repo so the agent (and the manager) can
/// fetch that legacy repo from the forge. Idempotent: no-op when the
/// remote already points at the right URL, or when the proposed repo
/// does not exist yet. No-op when the forge is not running.
pub async fn ensure_meta_remote(name: &str) -> Result<()> {
if !is_present().await {
return Ok(());
}
// A malformed name has no proposed config repo (repos are only created
// under a validated Ident), so there's nothing to wire — no-op.
let Ok(agent) = hive_types::Ident::parse(name) else {
return Ok(());
};
let proposed_dir = Coordinator::agent_proposed_dir(&agent);
if !proposed_dir.join(".git").exists() {
return Ok(());
}
let want = format!("{}/core/meta.git", forge_http_base());
let existing = crate::lifecycle::git_command()
.current_dir(&proposed_dir)
.args(["remote", "get-url", "meta"])
.output()
.await
.context("git remote get-url meta")?;
if existing.status.success() {
let current = String::from_utf8_lossy(&existing.stdout).trim().to_owned();
if current == want {
return Ok(());
}
crate::lifecycle::git(&proposed_dir, &["remote", "set-url", "meta", &want]).await
} else {
crate::lifecycle::git(&proposed_dir, &["remote", "add", "meta", &want]).await
}
}
use super::{CONFIG_ORG, forge_git_url, reader_token};
/// Reseed a revived agent's `applied` repo — missing its `.git` entirely,
/// e.g. after `destroy --purge` — by fetching `agent-configs/<name>`'s
@ -334,38 +137,3 @@ pub async fn clone_config_into_proposed(dir: &Path, name: &str) -> bool {
}
}
}
/// Add `user` as a collaborator on `owner/repo` at `permission`.
/// Idempotent: adding an existing collaborator just updates its
/// permission (Forgejo answers 204 either way; a 201 from older
/// versions is tolerated defensively).
async fn add_collaborator(
owner: &str,
repo: &str,
user: &str,
permission: AddCollaboratorOptionPermission,
token: &str,
) -> Result<()> {
let op = format!("add collaborator {user} to {owner}/{repo}");
let res = bounded(
&op,
api(token)?.repo_add_collaborator(
owner,
repo,
user,
AddCollaboratorOption {
permission: Some(permission),
},
),
)
.await?;
match res {
Ok(()) => {}
Err(ForgejoError::UnexpectedStatusCode(s)) if s == StatusCode::CREATED => {}
Err(e) => {
return Err(e).context(op);
}
}
tracing::debug!(%owner, %repo, %user, ?permission, "forge: collaborator set");
Ok(())
}

2
hive-c0re/src/forge/users.rs
View file

@ -14,7 +14,7 @@ use super::{api, bounded, forge_admin};
const TOKEN_NAME_PREFIX: &str = "hyperhive";
// Where the host-side `core` admin token lives. Used by hive-c0re itself
// to push the meta repo + drive admin API calls. Root-only. Literal lives
// to drive admin API calls and as the credential nix's git reads. Root-only. Literal lives
// in `crate::paths`; aliased here under the long-standing name.
use crate::paths::FORGE_CORE_TOKEN as CORE_TOKEN_PATH;
// The `forge/core-avatar-set` marker lives in `crate::paths` — a one-shot

6
hive-c0re/src/job_queue/exec.rs
View file

@ -446,11 +446,7 @@ async fn run_rebuild_bookkeeping(coord: &Arc<Coordinator>, name: &str) -> Result
// The `Rebuilt` manager event is emitted exactly once per agent by the DAG's
// `EmitRebuilt` tail — emitting ok here and letting a failed tail `Reconcile`
// add a contradictory !ok would double-report the same rebuild.
// Full forge sync on every successful rebuild so the rebuild path is
// equivalent to the startup sweep: meta access, the `meta` remote and the
// `internal/docs` grant all recover without a hive-c0re restart. (No
// matrix step: the swarm mints an agent's matrix account, not this hive.)
crate::forge::sync_agent(name, crate::forge::core_token().as_deref()).await;
// No `kick_agent` here any more — reporting from `hive-agent`'s own
// boot instead avoids the duplicate-restart-message problem this used
// to cause: the container is still stopped at this point (`swap_update` requires it stopped and

2
hive-c0re/src/lifecycle/git.rs
View file

@ -125,7 +125,7 @@ pub fn git_command() -> Command {
/// [`git_command`] carrying HTTP credentials **in its environment**.
///
/// `auth_header` is an `Authorization:` line (see
/// [`crate::forge::core_auth_header`]) handed to git as `http.extraHeader`
/// `forge::basic_auth_header`) handed to git as `http.extraHeader`
/// through `GIT_CONFIG_*`, which git reads exactly like a config file. The
/// alternative — userinfo in the remote URL — puts the secret in `argv`, and
/// `/proc/<pid>/cmdline` is world-readable while `/proc/<pid>/environ` is

9
hive-c0re/src/lifecycle/mod.rs
View file

@ -9,10 +9,11 @@ mod tests;
#[cfg(test)]
pub use git::git_update_ref;
#[cfg(test)]
pub use git::run_with_deadline;
pub use git::{
FETCH_DEADLINE, GIT_LOCAL_DEADLINE, git, git_authed, git_command, git_command_authed,
git_is_ancestor, git_read_tree_reset, git_rev_parse, git_tag, git_update_ref_cas,
run_with_deadline,
git, git_authed, git_command, git_command_authed, git_is_ancestor, git_read_tree_reset,
git_rev_parse, git_tag, git_update_ref_cas,
};
pub use host_config::write_dropins;
pub use setup::{
@ -372,7 +373,7 @@ fn stop_failed_guards()
/// Leak a per-agent warning `kind`. Agent names are a small, bounded set —
/// leaking one short string per distinct agent that ever hits a stop
/// failure is cheap, same reasoning as `forge::static_kind`.
/// failure is cheap.
fn static_stop_kind(name: &str) -> &'static str {
Box::leak(format!("agent_stop_failed_{name}").into_boxed_str())
}

164
hive-c0re/src/meta.rs
View file

@ -1,149 +1,69 @@
//! The hive's `meta/` directory: plain state beside the hive's other files
//! (`topology.json`, the swarm limits, the permission seed files) that the
//! manager container sees read-only. Agents are built from the swarm's
//! `internal/meta` ([`crate::meta_rev`]), not from anything here; a hive that
//! predates it also has a git repo in this directory, which
//! [`remove_permission_files`] still commits to.
//! (`topology.json`, the swarm limits) that the manager container sees
//! read-only. Agents are built from the swarm's `internal/meta`
//! ([`crate::meta_rev`]), not from anything here.
use std::path::Path;
use anyhow::{Context, Result, bail};
use crate::lifecycle;
const GIT_NAME: &str = "c0re";
const GIT_EMAIL: &str = "c0re@hyperhive.local";
use anyhow::{Context, Result};
/// Where the manager sees this directory inside its container (RO bind).
pub const CONTAINER_MANAGER_META_MOUNT: &str = "/meta";
/// Delete the meta repo's permission files and commit the removal. Their
/// grants live in the swarm's `agent-permissions` bucket once
/// `workers::permissions_import` has sent them.
/// Delete the permission seed files a hive that predates the swarm's
/// permissions left in the meta directory. Their grants live in the swarm's
/// `agent-permissions` bucket once `workers::permissions_import` has sent
/// them. A file already gone is fine.
///
/// # Errors
///
/// Removing a file or a git stage/commit step fails.
pub async fn remove_permission_files() -> Result<()> {
let dir = crate::paths::meta_root();
let mut removed = Vec::new();
for file in crate::workers::permissions_import::META_FILES {
/// Removing a file fails.
pub fn remove_permission_files() -> Result<()> {
remove_in(
&crate::paths::meta_root(),
crate::workers::permissions_import::META_FILES,
)
}
fn remove_in(dir: &std::path::Path, files: &[&str]) -> Result<()> {
for file in files {
match std::fs::remove_file(dir.join(file)) {
Ok(()) => {}
Err(e) if e.kind() == std::io::ErrorKind::NotFound => {}
Err(e) => return Err(e).with_context(|| format!("removing {file}")),
}
git(&dir, &["rm", "--cached", "--ignore-unmatch", "-q", file]).await?;
// Only a tracked file has a removal to commit; naming an untracked
// one makes `git commit -- <path>` fail.
if paths_dirty(&dir, &[file]).await? {
removed.push(*file);
}
}
if !removed.is_empty() {
git_commit_paths(
&dir,
"permissions moved to the swarm's agent-permissions",
&removed,
)
.await?;
}
Ok(())
}
/// Run `git <args>` in `dir` within [`lifecycle::GIT_LOCAL_DEADLINE`]; the caller
/// reads the exit status.
async fn git_output(dir: &Path, args: &[&str]) -> Result<std::process::Output> {
lifecycle::run_with_deadline(
lifecycle::git_command().current_dir(dir).args(args),
lifecycle::GIT_LOCAL_DEADLINE,
&format!("git {} in {}", args.join(" "), dir.display()),
)
.await
}
async fn git(dir: &Path, args: &[&str]) -> Result<()> {
let out = git_output(dir, args).await?;
if !out.status.success() {
bail!(
"git {} failed ({}): {}",
args.join(" "),
out.status,
String::from_utf8_lossy(&out.stderr).trim()
);
}
Ok(())
}
/// Path-limited commit: commits ONLY the given paths, so unrelated staged
/// content can never be swept into the commit.
async fn git_commit_paths(dir: &Path, message: &str, paths: &[&str]) -> Result<()> {
let name = format!("user.name={GIT_NAME}");
let email = format!("user.email={GIT_EMAIL}");
let mut args = vec!["-c", &name, "-c", &email, "commit", "-m", message, "--"];
args.extend_from_slice(paths);
git(dir, &args).await?;
if let Err(e) = crate::forge::push_meta(dir).await {
tracing::warn!(error = ?e, "forge: meta push after commit failed (non-fatal)");
}
Ok(())
}
/// True when any of `paths` differs between HEAD and the index or working
/// tree.
async fn paths_dirty(dir: &Path, paths: &[&str]) -> Result<bool> {
let mut args = vec!["diff", "--quiet", "HEAD", "--"];
args.extend_from_slice(paths);
let out = git_output(dir, &args).await?;
// Exit 0 = no differences; 1 = differences; anything else (e.g.
// no HEAD yet on a fresh repo) → treat as dirty so the commit
// path runs and surfaces real errors loudly.
Ok(!out.status.success())
}
#[cfg(test)]
mod tests {
use super::*;
use super::remove_in;
/// A path-limited commit must leave an unrelated staged file untouched.
#[tokio::test]
async fn path_limited_commit_leaves_unrelated_staged_file_alone() {
let tmp = tempfile::tempdir().expect("tempdir");
let dir = tmp.path();
git(dir, &["init", "--initial-branch=main"])
.await
.expect("git init");
std::fs::write(dir.join("resource-limits.json"), "{}").expect("write");
std::fs::write(dir.join("other.json"), "v1").expect("write");
git(dir, &["add", "-A"]).await.expect("add");
git_commit_paths(dir, "seed", &["resource-limits.json", "other.json"])
.await
.expect("seed commit");
// Something stages a change (uncommitted)…
std::fs::write(dir.join("other.json"), "v2-staged").expect("write");
git(dir, &["add", "other.json"]).await.expect("stage");
// …and a limits change commits, path-limited.
std::fs::write(dir.join("resource-limits.json"), r#"{"alice":[]}"#).expect("write");
git(dir, &["add", "resource-limits.json"])
.await
.expect("add");
git_commit_paths(
dir,
"set resource limits for alice",
&["resource-limits.json"],
)
.await
.expect("path-limited commit");
// The limits file is committed; the staged one is not.
#[test]
fn the_listed_files_go_and_nothing_else() {
let dir = tempfile::tempdir().unwrap();
for name in ["capabilities.json", "tool-groups.json", "topology.json"] {
std::fs::write(dir.path().join(name), "{}").unwrap();
}
remove_in(dir.path(), &["capabilities.json", "tool-groups.json"]).unwrap();
assert!(!dir.path().join("capabilities.json").exists());
assert!(!dir.path().join("tool-groups.json").exists());
assert!(
!paths_dirty(dir, &["resource-limits.json"])
.await
.expect("check"),
"limits file must be committed"
dir.path().join("topology.json").exists(),
"other state stays"
);
// A file already gone is fine.
remove_in(dir.path(), &["capabilities.json"]).unwrap();
}
#[test]
fn a_file_that_cannot_be_removed_is_an_error_naming_it() {
let dir = tempfile::tempdir().unwrap();
// A directory where the file belongs fails with a non-NotFound error.
std::fs::create_dir(dir.path().join("capabilities.json")).unwrap();
let err = remove_in(dir.path(), &["capabilities.json"]).unwrap_err();
assert!(
paths_dirty(dir, &["other.json"]).await.expect("check"),
"the staged file must survive the limits commit"
format!("{err:#}").contains("removing capabilities.json"),
"{err:#}"
);
}
}

12
hive-c0re/src/migrate.rs
View file

@ -50,17 +50,7 @@ pub async fn run() -> Result<()> {
tracing::info!("migration: {KILL_SWITCH} set — skipping");
return Ok(());
}
// Stale meta index lock: a previous hive-c0re crash mid-`git add`
// can leave `.git/index.lock` behind, which blocks every
// subsequent meta op until somebody `rm`s it manually. We just
// booted so nothing of ours is holding it; safe to clear.
let meta_lock = crate::paths::meta_git_index_lock();
if meta_lock.exists() {
match std::fs::remove_file(&meta_lock) {
Ok(()) => tracing::warn!("cleared stale meta/.git/index.lock"),
Err(e) => tracing::warn!(error = ?e, "clear stale meta lock failed"),
}
}
let names = match enumerate_agents().await {
Ok(names) => names,
Err(e) => {

7
hive-c0re/src/paths.rs
View file

@ -193,13 +193,6 @@ pub fn meta_root() -> PathBuf {
state_root().join("meta")
}
/// `meta/.git/index.lock` — git index lock; checked before meta git ops
/// so a stale lock from a crashed process can be cleared.
#[must_use]
pub fn meta_git_index_lock() -> PathBuf {
meta_root().join(".git/index.lock")
}
/// `meta-rev/` — one file per agent naming the `internal/meta` commit its
/// rebuild builds from (`crate::meta_rev`).
///

1
hive-c0re/src/workers/permissions_import.rs
View file

@ -222,7 +222,6 @@ async fn send_at(client: &async_nats::Client, hive: &str, meta: &Path) -> Result
"import: the controller took the meta repo's permissions"
);
crate::meta::remove_permission_files()
.await
.context("deleting the imported permission files")?;
Ok(ImportOutcome::Imported)
}