fix: harden fips startup and app port relays

This commit is contained in:
Dorian 2026-07-27 19:18:39 +01:00
parent 0aee010f9c
commit 709922c293
8 changed files with 366 additions and 34 deletions

View File

@ -44,6 +44,7 @@ Start with:
- [Developer Guide](docs/developer-guide.md)
- [App Developer Guide](docs/app-developer-guide.md)
- [App Manifest Spec](docs/app-manifest-spec.md)
- [Nostr Git Source Hosting Plan](docs/nostr-git-source-hosting.md)
- [Operations Runbook](docs/operations-runbook.md)
- [Troubleshooting](docs/troubleshooting.md)
@ -90,6 +91,7 @@ python3 scripts/check-app-catalog-drift.py --release --strict
| [API Reference](docs/api-reference.md) | JSON-RPC API overview |
| [App Developer Guide](docs/app-developer-guide.md) | How to package and test apps |
| [App Manifest Spec](docs/app-manifest-spec.md) | Manifest schema and validation rules |
| [Nostr Git Source Hosting Plan](docs/nostr-git-source-hosting.md) | ngit/NIP-34 contribution workflow and maintainer model |
| [Apps README](apps/README.md) | Packaged app catalog overview |
| [Image Recipe](image-recipe/README.md) | Bootable image build flow |
| [Operations Runbook](docs/operations-runbook.md) | Production operations and recovery |

View File

@ -83,7 +83,8 @@ impl RpcHandler {
pub(super) async fn handle_fips_install(&self) -> Result<serde_json::Value> {
let identity_dir = fips::identity_dir_from(&self.config.data_dir);
fips::config::install(&identity_dir).await?;
fips::service::activate(fips::SERVICE_UNIT).await?;
let unit = fips::service::activation_unit().await;
fips::service::activate(unit).await?;
let status = fips::FipsStatus::query(&self.config.data_dir).await;
Ok(serde_json::to_value(status)?)
}

View File

@ -54,7 +54,8 @@ pub async fn ensure_activated(data_dir: &std::path::Path) {
tracing::warn!("FIPS auto-activate: config install failed: {:#}", e);
return;
}
if let Err(e) = service::activate(SERVICE_UNIT).await {
let unit = service::activation_unit().await;
if let Err(e) = service::activate(unit).await {
tracing::warn!("FIPS auto-activate: service activate failed: {:#}", e);
return;
}

View File

@ -32,6 +32,16 @@ pub async fn unit_state(unit: &str) -> String {
}
}
/// Whether systemd knows about `unit`.
pub async fn unit_exists(unit: &str) -> bool {
Command::new("systemctl")
.args(["cat", unit])
.output()
.await
.map(|out| out.status.success())
.unwrap_or(false)
}
/// Whether the `fips` debian package is installed on the host.
pub async fn package_installed() -> bool {
// dpkg-query -W -f='${Status}' fips → "install ok installed" when present.
@ -131,17 +141,28 @@ done
Ok(())
}
/// Resolve which systemd unit is actually supervising the fips daemon
/// on this host. Nodes installed from the archipelago ISO run
/// `archipelago-fips.service`; nodes that were apt-installed (or had
/// fips running before archipelago took over) may only have the
/// upstream `fips.service`. Restart/Reconnect must operate on whichever
/// one is running, otherwise the UI button is a silent no-op.
/// Resolve which systemd unit should be started when FIPS is inactive.
/// Newer Archipelago images may ship `archipelago-fips.service`; nodes with
/// the upstream Debian package may only have `fips.service`. Activation must
/// choose a unit systemd can actually load, otherwise the dashboard repeatedly
/// offers an "Activate" action that can never succeed.
pub async fn activation_unit() -> &'static str {
if unit_exists(super::SERVICE_UNIT).await {
return super::SERVICE_UNIT;
}
if unit_exists(super::UPSTREAM_SERVICE_UNIT).await {
return super::UPSTREAM_SERVICE_UNIT;
}
super::SERVICE_UNIT
}
/// Resolve which systemd unit is actually supervising the fips daemon on this
/// host. Restart/Reconnect must operate on whichever one is running, otherwise
/// the UI button is a silent no-op.
///
/// Returns the archipelago-managed unit name if it's active,
/// else the upstream unit name if that's active,
/// else the archipelago-managed name as a default (so activate() can
/// bring it up).
/// else a startable activation unit.
pub async fn active_unit() -> &'static str {
if unit_state(super::SERVICE_UNIT).await == "active" {
return super::SERVICE_UNIT;
@ -149,7 +170,7 @@ pub async fn active_unit() -> &'static str {
if unit_state(super::UPSTREAM_SERVICE_UNIT).await == "active" {
return super::UPSTREAM_SERVICE_UNIT;
}
super::SERVICE_UNIT
activation_unit().await
}
pub async fn mask(unit: &str) -> Result<()> {
@ -248,4 +269,10 @@ mod tests {
// Must not panic regardless of host state.
let _ = package_installed().await;
}
#[tokio::test]
async fn test_unit_exists_is_bool() {
// Must not panic regardless of host state.
let _ = unit_exists("archipelago-bogus-test.service").await;
}
}

View File

@ -940,14 +940,16 @@ impl Server {
);
}
}
if let Err(e) = crate::fips::service::activate(crate::fips::SERVICE_UNIT).await {
let unit = crate::fips::service::activation_unit().await;
if let Err(e) = crate::fips::service::activate(unit).await {
tracing::warn!(
"archipelago-fips activate failed on startup: {} — user can retry via fips.install RPC",
"FIPS activate failed on startup via {}: {} — user can retry via fips.install RPC",
unit,
e
);
return;
}
tracing::info!("archipelago-fips auto-activated on startup");
tracing::info!("FIPS auto-activated on startup via {}", unit);
});
}
@ -1000,10 +1002,8 @@ impl Server {
_ => None,
};
let v6_task = if let Some(port) = v4_any_port {
let v6_addr = SocketAddr::new(
std::net::IpAddr::V6(std::net::Ipv6Addr::UNSPECIFIED),
port,
);
let v6_addr =
SocketAddr::new(std::net::IpAddr::V6(std::net::Ipv6Addr::UNSPECIFIED), port);
match bind_v6_only(v6_addr) {
Ok(listener) => {
info!("IPv6 web listener bound {} (mesh ULA reachable)", v6_addr);
@ -1017,7 +1017,10 @@ impl Server {
)))
}
Err(e) => {
warn!("IPv6 web listener bind {} failed: {} — UI stays v4-only", v6_addr, e);
warn!(
"IPv6 web listener bind {} failed: {} — UI stays v4-only",
v6_addr, e
);
None
}
}
@ -1031,9 +1034,9 @@ impl Server {
// IPv4 only for most apps, so a phone dialing [ULA]:8123 got
// nothing even with the firewall open (HA/FileBrowser/Gitea/
// Portainer/Pine all v4-only on 2026-07-26; a few bind [::]
// themselves). Bridge each catalog launch port to its v4 loopback
// listener with a V6ONLY relay — self-selecting: where the app
// already answers on v6 the bind fails and the app wins.
// themselves). Bridge each catalog launch port on the fips0 ULA
// only. Binding wildcard [::]:port reserves the same host ports
// Podman needs and can restart-loop apps that publish those ports.
let relay_task = tokio::spawn(app_port_v6_relay_loop(tx.subscribe()));
let peer_task = tokio::spawn(peer_late_bind_loop(
@ -1087,10 +1090,14 @@ fn bind_v6_only(addr: SocketAddr) -> std::io::Result<tokio::net::TcpListener> {
tokio::net::TcpListener::from_std(socket.into())
}
fn fips_app_relay_addr(ip: std::net::Ipv6Addr, port: u16) -> SocketAddr {
SocketAddr::new(std::net::IpAddr::V6(ip), port)
}
/// IPv6→IPv4 relay for catalog app launch ports (see the spawn site for
/// why). Rescans every 60s so ports of freshly installed apps get bridged
/// without a daemon restart. Each relay is a V6ONLY listener forwarding
/// raw TCP to the same port on IPv4 loopback.
/// without a daemon restart. Each relay binds to the fips0 ULA only and
/// forwards raw TCP to the same port on IPv4 loopback.
async fn app_port_v6_relay_loop(mut shutdown_rx: tokio::sync::watch::Receiver<bool>) {
use std::collections::HashSet;
let mut bridged: HashSet<u16> = HashSet::new();
@ -1099,6 +1106,7 @@ async fn app_port_v6_relay_loop(mut shutdown_rx: tokio::sync::watch::Receiver<bo
loop {
tokio::select! {
_ = interval.tick() => {
let Some(fips_ip) = crate::fips::iface::fips0_ula() else { continue };
for &port in crate::fips::app_ports::APP_LAUNCH_PORTS {
if bridged.contains(&port) {
continue;
@ -1123,15 +1131,12 @@ async fn app_port_v6_relay_loop(mut shutdown_rx: tokio::sync::watch::Receiver<bo
if !v4_up {
continue;
}
let addr = SocketAddr::new(
std::net::IpAddr::V6(std::net::Ipv6Addr::UNSPECIFIED),
port,
);
// EADDRINUSE = the app itself already answers on v6 —
// exactly when we must stay out of the way.
let addr = fips_app_relay_addr(fips_ip, port);
// EADDRINUSE = fipsd or another process already answers
// on this mesh address/port, so stay out of the way.
let Ok(listener) = bind_v6_only(addr) else { continue };
bridged.insert(port);
debug!("v6 relay bridging [::]:{port} -> 127.0.0.1:{port}");
debug!("v6 relay bridging [{fips_ip}]:{port} -> 127.0.0.1:{port}");
let mut rx = shutdown_rx.clone();
tokio::spawn(async move {
loop {
@ -1945,6 +1950,14 @@ mod merge_tests {
assert!(!is_peer_allowed_path("/rpc/v2"));
}
#[test]
fn app_relay_binds_to_fips_ula_not_wildcard() {
let ula = "fd12:3456:789a::1".parse().unwrap();
let addr = fips_app_relay_addr(ula, 8083);
assert_eq!(addr.ip(), std::net::IpAddr::V6(ula));
assert_eq!(addr.port(), 8083);
}
#[test]
fn preserves_transitional_state_on_merge() {
// existing: user initiated a stop, spawn_transitional set Stopping.

View File

@ -45,12 +45,44 @@ Completed locally in this pass:
schema can be audited without a Python `PyYAML` dependency.
- Updated root/community docs, CI, PR template, app developer notes, and
container/deployment docs toward public contributor expectations.
- Fixed native FIPS activation fallback: nodes that have the packaged
`fips.service` but not `archipelago-fips.service` now start the available
unit instead of repeatedly failing activation against a missing unit. The UI
now labels the action as `Start` instead of making native FIPS look like an
installable app.
- Fixed the FIPS app-port relay design so it binds relays to the node's FIPS
ULA instead of wildcard `[::]`, avoiding collisions with Podman-published app
ports such as FileBrowser `8083` and Botfights `9100`.
- Added `docs/nostr-git-source-hosting.md`, a NIP-34/ngit/GRASP source hosting
plan using a Bitcoin Core-style maintainer model: public review and easy
forks, with canonical merge rights held by a small signed maintainer set.
Verified locally:
- `./scripts/audit-secrets.sh` passes.
- Full `apps/*/manifest.yml` repository audit passes with warnings only.
- `bash -n` passes for the edited shell scripts.
- Targeted FIPS dashboard vitest passes.
- Targeted Rust tests for FIPS service unit detection and FIPS app relay
address selection pass.
Verified on a Linux Archipelago verification node:
- Native FIPS was restored by starting the already-installed packaged
`fips.service`; the daemon became active and joined the FIPS tree.
- Correct local lifecycle API endpoint is HTTP, not HTTPS
(`ARCHY_HOST=127.0.0.1 ARCHY_SCHEME=http`).
- Read-only lifecycle run progressed past login and confirmed required
containers, Bitcoin RPC, ElectrumX TCP, and manifest port-drift checks, but
did not complete cleanly: `botfights` and `filebrowser` remained in
`restarting` longer than the matrix window, and the LND `lncli getinfo`
probe hung. Do not run the destructive gate until those live-node issues are
understood.
- After the node updated to `1.7.116-alpha`, `botfights`, `filebrowser`, and
`lnd` were active/running and ports `8083`/`9100` were held by Podman's
`rootlessport` as expected. The packaged `fips.service` remained installed
and enabled but inactive, so the native FIPS service fallback should still
ship before the public launch.
Still required before public publish:
@ -58,6 +90,10 @@ Still required before public publish:
- Finish Phase 1 password/node/token sanitization beyond the two API keys.
- Publish from fresh history after the sanitized tree is final.
- Run full Rust, frontend, Android, and lifecycle gate verification.
- Resolve the live-node lifecycle blockers above, then rerun the read-only
suite followed by the destructive gate only on an approved verification node.
- Decide the canonical Archipelago maintainer npub and merge-maintainer npub
list before publishing the Nostr Git source-hosting workflow.
---

View File

@ -0,0 +1,252 @@
# Nostr Git Source Hosting Plan
This plan describes how Archipelago can publish and accept contributions to its
source code through `ngit`, NIP-34, and GRASP while keeping the developer
experience inside Archipelago.
## Goals
- Publish Archipelago source from a sanitized, fresh-history repository.
- Make the in-app registry the primary onboarding path for contributors.
- Let contributors clone, branch, push PR branches, open PRs, and discuss issues
with a Nostr identity from their Archipelago node.
- Follow the Bitcoin Core development model: broad public review and easy forks,
with canonical merge authority held by a small maintainer set.
- Give contributors full read, fork, and proposal rights, but no direct merge
rights on the canonical repository.
- Keep the official maintainer identity and merge authority separate from user
node identities.
## Current Building Blocks
Archipelago already has most of the primitives needed for this:
- App manifests and the app registry already install developer tooling as
rootless Podman apps.
- The `gitea` app provides a conventional fallback Git UI and package registry.
- The app launcher already exposes a consent-gated NIP-07 bridge for launched
apps using `getPublicKey`, `signEvent`, NIP-04, and NIP-44 requests.
- The backend exposes node and identity Nostr signing RPC methods.
- FIPS gives nodes a stable mesh identity and private transport path, but repo
announcements and PRs should remain NIP-34 compatible on normal Nostr relays.
- DWN protocol registration exists and can be used later for local contribution
metadata/cache, but should not be required for the first public workflow.
## Protocol Basis
Use existing Nostr Git conventions rather than inventing an Archipelago-only
protocol:
- NIP-34 repository announcement events identify repositories with kind `30617`.
- NIP-34 repository state events publish branch/tag refs with kind `30618`.
- NIP-34 patches, pull requests, PR updates, issues, and status events use kinds
`1617`, `1618`, `1619`, `1621`, and `1630`-`1633`.
- `ngit` provides the `git-remote-nostr` helper for `nostr://` clone URLs and PR
branches.
- GRASP servers provide Git Smart HTTP storage while Nostr events remain the
authority for repository identity, refs, PRs, issues, and maintainer state.
Primary references:
- https://nips.nostr.com/34
- https://docs.rs/crate/ngit/latest/source/README.md
- https://ngit.dev/grasp/
## Recommended Architecture
### Apps
Create two first-party apps:
- `ngit`: CLI/runtime package containing `ngit` and `git-remote-nostr`.
- `archipelago-source`: web UI for cloning Archipelago source, viewing NIP-34
issues/PRs, opening branches, and submitting PR events.
The `archipelago-source` app should depend on `ngit`. It can also recommend
Gitea for users who want a conventional local web Git UI, but Gitea should not
be the source of truth for public contribution permissions.
### Contributor Onboarding
When the user installs `archipelago-source` from the registry:
1. Show a modal before first launch: "Contribute to Archipelago".
2. Explain that the app will use their Archipelago Nostr identity to clone and
sign contribution events.
3. Display the maintainer repository announcement, clone URL, maintainer npub,
and relay/GRASP endpoints.
4. Ask for consent to:
- fetch repository metadata from configured relays,
- clone source through `nostr://`,
- create local branches,
- sign NIP-34 issue/PR/comment events,
- push PR branches to approved GRASP servers.
5. Store approval per app origin, identity id, repository id, and relay set.
This should build on the existing NIP-07 app-launcher bridge, but use a more
specific permission scope than the generic sign-event approval.
### Identity And Permissions
Use four identity classes:
- `archipelago-maintainer`: an offline or tightly controlled Nostr key that
signs the canonical kind `30617` repo announcement and status/merge events.
- `archipelago-merge-maintainer`: one of the small set of maintainer npubs
allowed to advance canonical refs and publish valid merged/applied status.
- `archipelago-build`: release automation key for signed release artifacts and
CI status events. It must not have merge authority.
- `contributor`: user node or app-specific identity used for PRs, issues, and
comments.
Contributor rights:
- Clone the repository.
- Open issues.
- Push proposal branches using `pr/<npub>/<short-topic>` or `pr/<event-id>`.
- Publish NIP-34 PR/update/comment events.
- Rebase and update their own PR branch.
- Run local validation and attach status evidence.
Contributor restrictions:
- Cannot update `refs/heads/main` or release branches in canonical state.
- Cannot publish maintainer-valid merge/applied status.
- Cannot alter the canonical repository announcement.
- Cannot publish release catalog signatures.
Maintainer rights:
- Publish/update the canonical repo announcement.
- Publish canonical `refs/heads/main` state.
- Mark PRs merged/closed/draft via NIP-34 status events.
- Sign release tags and catalog updates.
Fork rights:
- Any contributor can create their own NIP-34 kind `30617` repository
announcement for a fork.
- Fork announcements should use the NIP-34 `u` tag to point back to the
canonical `archy` repository.
- The source app should make forking a first-class path: "Fork on Nostr", clone
the fork locally, push branches to the contributor's GRASP list, and open PRs
back to canonical Archipelago when they want review.
- Forks can have their own maintainer npubs, relays, policies, and release
cadence, but the app should clearly label them as forks unless signed by the
canonical maintainer set.
The GRASP server policy should enforce this by accepting pushes to maintainer
refs only when backed by signed maintainer state, while allowing contributor PR
refs from their own npubs.
## Repository Layout
Canonical repo announcement:
- repo id: `archy`
- display name: `Archipelago`
- clone URLs:
- `nostr://<maintainer-npub>/<relay-hint>/archy`
- `https://<grasp-host>/<maintainer-npub>/archy.git`
- relays:
- Archipelago-operated relay
- at least two public Nostr relays that support the event load
- GRASP servers:
- Archipelago-operated GRASP instance
- one public GRASP-compatible mirror
Keep the existing HTTP Git remote as a mirror during launch. The docs can
present `nostr://` as the preferred contribution path once the workflow is
proven.
## UI Requirements
The source app should provide:
- A first-run contribution modal with a real Archipelago source graphic, not a
generic text-only dialog.
- Current clone status and local path.
- Branch list, changed files, commit form, and push/open-PR flow.
- PR inbox, issue list, maintainer status, and relay health.
- Explicit identity indicator showing which npub will sign events.
- A merge rights indicator that clearly says contributors can propose changes
but cannot merge them.
- A fork flow that creates a user-owned NIP-34 repo announcement and remote,
then offers "Open PR to Archipelago" from any fork branch.
- Maintainer badges based only on pinned canonical maintainer npubs, not relay
metadata or server-side account names.
- Links to container docs, deployment docs, manifest spec, and open-source
readiness tasks.
## Backend Work
Add an RPC module for source contribution workflow:
- `source.repo-info`: returns canonical announcement, clone URL, relay set,
maintainer npubs, and local clone state.
- `source.ensure-ngit`: verifies the `ngit` app/runtime is installed.
- `source.clone`: clones or updates the local source checkout.
- `source.status`: returns branch, dirty files, ahead/behind, and PR state.
- `source.commit`: creates a local commit from selected files.
- `source.fork`: creates a contributor-owned NIP-34 fork announcement and local
remote.
- `source.open-pr`: pushes a PR branch and publishes a kind `1618` event.
- `source.update-pr`: updates the branch and publishes kind `1619`.
- `source.issue`: publishes a kind `1621` event.
Backend must shell out through a narrow command wrapper, never arbitrary user
commands. The wrapper should set an isolated working tree under
`/var/lib/archipelago/source/archy`, run as the Archipelago service user, and
deny operations outside that path.
## Security Model
- Never expose maintainer private keys to an Archipelago node.
- Prefer app-specific contributor identities over the node's default identity.
- Require per-action consent for first PR push, issue creation, and signing any
event that tags the canonical repository.
- Pin the canonical maintainer npub in the app manifest and backend config.
- Keep the canonical merge-maintainer allow list signed by the
`archipelago-maintainer` key; never infer merge rights from GRASP server
accounts.
- Verify the canonical kind `30617` event signature before displaying clone
instructions.
- Treat GRASP servers as untrusted storage; verify Git refs against signed
Nostr state.
- Do not use destructive git operations from the UI without an explicit modal.
- Store local clones and generated patches outside app container writable roots
unless the user exports them.
## MVP
1. Package `ngit` as a first-party app.
2. Stand up one Archipelago-operated GRASP server and one Nostr relay.
3. Publish sanitized fresh-history `archy` through `ngit init`.
4. Add a simple `archipelago-source` app that clones source and links out to the
preferred Nostr Git browser.
5. Add app-launcher consent scopes for repository-specific NIP-34 signing.
6. Allow issues and PR branch submission from contributor npubs.
7. Add a one-click fork flow that publishes a contributor-owned fork
announcement referencing canonical Archipelago.
8. Keep maintainer merge/status publication manual.
## Later
- Native PR review UI with file diffs and inline comments.
- CI status events signed by the build identity.
- FIPS-first source sync between trusted Archipelago nodes.
- Private prerelease repositories using NIP-42 allow lists and/or protected
events if the ecosystem support is mature enough.
- Multi-maintainer policy with threshold signatures or explicit maintainer-list
rotation events.
## Open Questions
- Which maintainer npub should become canonical for `archy`?
- Should contributor identities be node-default or app-specific by default?
- Which GRASP implementation should be deployed first: `ngit-grasp` or another
NIP-34/GRASP-compatible relay?
- Should the source app include a full web Git UI in v1, or launch Gitea/ngit
browser links for review while keeping signing/submission native?
- What exact license and contribution certificate should contributors accept
before submitting PR events?

View File

@ -113,7 +113,7 @@
<div v-if="statusMessage" class="mb-3 p-3 rounded-lg text-xs" :class="statusIsError ? 'bg-red-400/10 text-red-300' : 'bg-green-400/10 text-green-300'">{{ statusMessage }}</div>
<div v-if="status.key_present && !status.service_active" class="flex gap-2 mt-auto pt-3 shrink-0">
<button class="flex-1 min-h-[44px] px-4 py-2 glass-button rounded-lg text-sm font-medium transition-colors" :disabled="installing" @click="installAndActivate">{{ installing ? 'Installing' : 'Activate' }}</button>
<button class="flex-1 min-h-[44px] px-4 py-2 glass-button rounded-lg text-sm font-medium transition-colors" :disabled="installing" @click="installAndActivate">{{ installing ? 'Starting' : 'Start' }}</button>
</div>
</div>
</template>
@ -211,10 +211,10 @@ async function installAndActivate() {
installing.value = true
try {
status.value = await rpcClient.call<FipsStatus>({ method: 'fips.install' })
flash('FIPS installed and activated')
flash('FIPS started')
} catch (e: unknown) {
const msg = e instanceof Error ? e.message : String(e)
flash(`Install failed: ${msg}`, true)
flash(`Start failed: ${msg}`, true)
} finally {
installing.value = false
}