Watch
0
0
Fork
You've already forked hyperhive
0

Compare commits

...
Author SHA1 Message Date
atlas
0f53e9be6c swarm-logs: authenticate as the agent with its own queue secret
swarm-logs reads `swarm/agents/<agent>/queue` from the swarm secret
store with the agent's own store certificate (the read hive-agent makes
for its queue connection) and presents it as
`Bearer swarm-agent.<agent>.<secret>`, which the gateway's
swarm-logs-auth checks. On a 401 it reads the secret again and retries
once, so a secret re-minted between the read and the check does not
fail the query.

The OIDC mint against the hive's shared `hive-<h>-agent` client is gone
from swarm-logs, along with the unused `HIVE_AGENT_OIDC_CA_FILE` anchor.
logs.nix now gates on the store address instead of the queue's token
endpoint and wraps the binary with the paths queue-identity.nix gives
the harness; the host forwards the query URL wherever it forwards a
store identity.

The shared client is still delivered into containers; removing that is

Part of #4459.
2026-10-13 08:38:05 +02:00
atlas
1f4736e79b swarm-logs-auth: check agent tokens on the log store's query route
A new `auth_request` responder on the log store's host. nginx sends it
every `/select/logsql/` request whose `Authorization` is
`Bearer swarm-agent.<agent>.<secret>`; it answers 200 when the secret is
the one stored at `swarm/agents/<agent>/queue` (`swarm-agent-verify`),
401 otherwise. Every other bearer still goes to authelia, so agents
still presenting the hive's shared OIDC client keep working.

`auth_request` takes a fixed URI, so the split is a `map` on
`$http_authorization` plus an internal location that rewrites to the
map's pick. Checked against nginx 1.30.5 with stub upstreams: an agent
token only the stub validator accepts gets 200, a token only the stub
authelia accepts gets 200, every wrong one 401, and the internal
locations 404 from outside.

The responder serves a unix socket in its own traverse-only runtime
directory and reads the store under its own leaf
(`swarm-logs-auth`), whose policy is `swarm/agents/+/queue`, read-only.
glue-bao-tls mints the leaf where the store runs; elsewhere the
operator places it, and eval refuses a log store host without one.

Part of #4459.
2026-10-13 08:38:05 +02:00
atlas
e7dbb94f28 swarm-agent-verify: lift the agent-token store check out of swarm-nats-auth
`CredentialSource`, `Store` and `verify` move from swarm-nats-auth's
`agent_token` module into a small library crate, so a second verifier
can make the same check: the secret in a `swarm-agent.<agent>.<secret>`
token equals the one stored at `swarm/agents/<agent>/queue`, read under
the verifier's own certificate.

`verify` now takes its timeout as an argument; swarm-nats-auth passes
`INTROSPECTION_TIMEOUT` as before. No behaviour change. The store-denial
tests moved with the code; swarm-nats-auth keeps the ones that go
through its policy and its bound.

Part of #4459.
2026-10-13 08:38:05 +02:00
32 changed files with 1050 additions and 531 deletions

View file

@ -193,6 +193,15 @@ hand-maintained per-file tree drifts out of sync with the code.
`claude --print` per classification (no tools, no session) and replies
with the rows and the overall verdict. Holds no forge credential. Design:
[`docs/swarm/review-classifier.md`](docs/swarm/review-classifier.md).
- **`swarm-agent-verify/`** — the check that a presented
`swarm-agent.<agent>.<secret>` token holds the secret stored at
`swarm/agents/<agent>/queue`, read under the verifier's own certificate.
Shared by `swarm-nats-auth` (the queue's auth callout) and
`swarm-logs-auth`.
- **`swarm-logs-auth/`** — the log store gateway's `auth_request` responder
for agents: nginx sends it `/select/logsql/` requests carrying an agent
token, it answers 200 or 401 via `swarm-agent-verify`. Runs on the log
store's host (`swarm-victorialogs.nix`).
- **`swarmctl/`** — swarm-level operator CLI, installed by the
swarm-controller module on the host that runs the daemon. Runs as
**root and acts directly** — it serves no socket, publishes no HTTP

31
Cargo.lock generated
View file

@ -4789,6 +4789,18 @@ version = "2.6.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "13c2bddecc57b384dee18652358fb23172facb8a2c51ccc10d74c157bdea3292"
[[package]]
name = "swarm-agent-verify"
version = "0.1.0"
dependencies = [
"anyhow",
"subtle",
"swarm-queue-client",
"swarm-secret-client",
"tokio",
"tracing",
]
[[package]]
name = "swarm-authelia-bridge"
version = "0.1.0"
@ -4870,6 +4882,23 @@ dependencies = [
"reqwest",
"serde_json",
"swarm-queue-client",
"swarm-secret-client",
"tokio",
]
[[package]]
name = "swarm-logs-auth"
version = "0.1.0"
dependencies = [
"anyhow",
"axum",
"clap",
"swarm-agent-verify",
"swarm-queue-client",
"swarm-secret-client",
"tokio",
"tracing",
"tracing-subscriber",
]
[[package]]
@ -4910,7 +4939,7 @@ dependencies = [
"serde",
"serde_json",
"sha2 0.11.0",
"subtle",
"swarm-agent-verify",
"swarm-queue-client",
"swarm-secret-client",
"tokio",

View file

@ -25,6 +25,7 @@ members = [
"hive-subagent-mcp",
"hive-types",
"hivectl",
"swarm-agent-verify",
"swarm-authelia-bridge",
"swarm-authelia-bridge-sock",
"swarm-controller",
@ -34,6 +35,7 @@ members = [
"swarm-queue-client",
"swarm-review-classifier",
"swarm-logs",
"swarm-logs-auth",
"swarm-secret-client",
"swarmctl",
]
@ -104,6 +106,7 @@ hive-priv-sock = { path = "hive-priv-sock" }
hive-runtime = { path = "hive-runtime" }
hive-sock-client = { path = "hive-sock-client" }
hive-types = { path = "hive-types" }
swarm-agent-verify = { path = "swarm-agent-verify" }
swarm-authelia-bridge-sock = { path = "swarm-authelia-bridge-sock" }
swarm-matrix-client = { path = "swarm-matrix-client" }
swarm-queue-client = { path = "swarm-queue-client" }
@ -234,7 +237,7 @@ futures-util = "0.3"
hmac = "0.13"
sha2 = "0.11"
# Constant-time comparison of a presented secret against the stored one
# (`swarm-nats-auth::agent_token`). Already in the tree through the TLS stack.
# (`swarm-agent-verify`). Already in the tree through the TLS stack.
subtle = "2.6"
# The NATS protocol client, for the swarm queue's auth-callout responder.
# `default-features = false` because the default set is broad - jetstream, kv,

View file

@ -182,8 +182,9 @@ the credential from the store again for each attempt.
The hive's shared queue client is still delivered into the harness unit's
credentials directory, and `swarm.statusPublish.tokenEndpoint` still reaches
the container, but only for `swarm-logs`: the harness never connects with
them.
the container, but only for a `swarm-logs` built before it read the agent's
own queue secret: the harness never connects with them, and a current
`swarm-logs` presents that secret to the log store instead.
<details><summary>What an agent publishes over the queue</summary>

View file

@ -55,27 +55,27 @@ of the cell says how.
<!-- vale write-good.Passive = NO -->
| store path | minter | reader — pulls at runtime, holds in memory | automatic re-mint | automatic re-pull |
| ----------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `swarm/agents/<agent>/matrix/main` | `swarm-controller`, with the swarm's appservice token, at agent creation and in a five-minute pass | the agent container itself, under the certificate its hive passed in | ✅ the pass re-mints when the stored token is missing, unknown to the homeserver, or someone else's | ✅ `hive-matrix-daemon` exits when the homeserver rejects its token, and a five-minute timer restarts it, which reads the store again |
| `swarm/agents/<agent>/matrix/<account>` | `swarm-controller` | the agent container itself, under the certificate its hive passed in | must be stated | must be stated |
| `swarm/controller/swarm-controller/matrix/appservice-token` | `swarm-matrix-ctl`, inside the `hive-matrix` container, once | `swarm-controller`, under its own certificate | ❌ `swarm-matrix-ctl` mints it once; the container keeps its copy and republishes it when the store's differs | ✅ the controller reads it on every five-minute matrix pass |
| `swarm/controller/swarm-controller/oidc/client` | authelia, at its first boot; `swarm-secret-publish` copies it in | `swarm-controller`, under its own certificate, once at start | ❌ authelia mints it once. A re-mint is republished by `swarm-secret-publish`'s path unit | ❌ read once at start; the controller holds the old value until it restarts |
| `swarm/controller/swarm-controller/forge/token` | the forge, inside the `hive-forge` container, once; `swarm-bao-forge-publisher` on the forge's host copies it in under its own certificate, with a write-only grant | `swarm-controller`, under its own certificate, at start | ❌ the forge mints it once and keeps its copy in the container; the publisher writes that copy each time the forge container starts and each time the mint writes the file | ❌ the controller reads it at start, and in the background with backoff until it first finds it; a replaced token reaches it with the next restart |
| `swarm/agents/<agent>/bao-mtls` | the store's agent PKI mount (`deploy.bao.agentPkiMountPath`), which generates the key, at `swarm-controller`'s request at agent creation | `hive-c0re`, under the hive's own certificate, when it writes the agent's container config; then `hive-agent-bao-identity` in the agent container, under the leaf it holds | ✅ `swarm-controller`'s five-minute pass re-issues a live agent's leaf once it's past half its validity (45 of 90 days, read from the certificate itself) | ✅ `hive-agent-bao-identity` re-reads it every 10 minutes and swaps a new leaf into `/run/hive-agent-bao-identity/`, which every store login in the container presents. It reads under the leaf it holds, so it renews only while that leaf is valid |
| `swarm/agents/<agent>/queue` | `swarm-controller`, at agent creation | `hive-agent` in the agent container, under the agent's own certificate, held in memory — the identity it presents to the swarm queue, naming that one agent rather than its hive | ✅ `swarm-controller`'s five-minute pass re-mints a live agent's secret once it's 45 days old by `minted_at` on the stored object; a secret with no `minted_at` gets one stamped, value unchanged. The pass skips agents declared `Destroyed` — declaring an agent destroyed deletes every version of the path instead, the undo of the mint rather than another one | ✅ `hive-agent` reads the path before its first connect and again on every reconnect attempt, so a reconnect after a re-mint presents the new secret. An open connection keeps the secret it connected with; after a revocation the agent keeps retrying under the queue client's backoff |
| `swarm/agents/<agent>/forge-token` | `swarm-controller`, at agent creation and in a pass every 5 minutes over every agent with a store identity | the agent container itself, under its own certificate, fetched to `/run/hive-agent-forge-token/token` | ✅ the controller re-mints when the stored token is missing or no longer matches the forge (last eight characters and scopes) | ✅ the agent re-fetches on a 10-minute timer |
| `swarm/agents/<agent>/forge/<label>` | `swarm-controller`, when an operator links an external forge account in the swarm UI | `hive-agent-forge-accounts` in the agent container, under the agent's own certificate, into `<state>/forge-<label>-token` and `forge-<label>.json` | ❌ an operator's token; replaced only by deleting it and linking the label again | ✅ the agent re-fetches on a 2-minute timer |
| `swarm/agents/<agent>/github-token` | `swarm-controller`, when an operator links a GitHub account in the swarm UI | `hive-agent-github-token` in the agent container, under the agent's own certificate, into `<state>/github-token` | ❌ an operator's token; replaced only by deleting it and linking it again | ✅ the agent re-fetches on a 2-minute timer, and `hive-github-notify` re-reads the file on every poll |
| `swarm/hives/<hive>/matrix/sender-token` | `swarm-controller`, with the swarm's appservice token, for every hive in its directory in a five-minute pass | `swarm-controller` under its own certificate, before it decides whether to mint, and hive-c0re's `stored_sender_token()`, under the hive's own certificate | ✅ the controller's pass re-mints when the stored token is missing, unknown to the homeserver, or someone else's | ✅ hive-c0re's matrix sweep reads the store every run and overwrites its token file when the store's token differs |
| `swarm/hives/<hive>/forge/token` | `swarm-controller`, with the forge's admin API, for every hive in its directory in a five-minute pass | `swarm-controller` under its own certificate, before it decides whether to mint, and hive-c0re's `forge::reader_token()`, under the hive's own certificate | ✅ the controller's pass re-mints when the stored token is missing or no longer matches the forge (last eight characters and scopes) | ✅ hive-c0re reads the store on every forge read |
| `swarm/hives/<hive>/queue/agent` | authelia | `swarm-bao-queue-agent` on the hive's host, under its own per-hive certificate; no agent's policy reaches it | must be stated | must be stated |
| `swarm/hives/<hive>/oidc/client` | authelia, at its first boot; `swarm-secret-publish` copies it in | hive-c0re, under the hive's own certificate, when it first connects to the queue | ❌ authelia mints it once. A re-mint is republished by `swarm-secret-publish`'s path unit | ❌ read until the first connect succeeds, retried with a backoff of up to five minutes while it's missing; then held until hive-c0re restarts |
| `swarm/services/<clientId>/oidc/client` | authelia | that service's own reader, under its own certificate — the queue responder's under `swarm-nats-auth`'s, the homeserver's under `swarm-matrix-ctl`'s, the forge's under `swarm-forge-publisher`'s, the review classifier's under `swarm-review-classifier`'s; the bao UI's by the store's granter | must be stated | must be stated |
| `swarm/services/ci-runner/registration-token` | `swarm-controller`, from the forge's admin API, at start and in a five-minute pass; its grant on the path is write-only | `swarm-bao-ci-runner` on the runner's host, under the runner's own certificate, into a file only `swarm-ci-runner-register` in the runner's container loads; no other principal's policy reads it | ❌ the forge hands out one token until an admin resets it; the controller writes a reset token within five minutes | ✅ `swarm-bao-ci-runner` reads it at boot, on deploy and every ten minutes, and restarts the runner when the value changed |
| `swarm/services/review-classifier/anthropic-api-key` | the operator, by hand in the store's UI, field `api_key` | `swarm-review-classifier` in its container, under its own certificate, at the start of every run; held for that run and handed to `claude`'s environment only; no other principal's policy reads it | ❌ an operator's key; replaced by writing a new one | ✅ read on every run, so a written or removed key takes effect on the next run |
| `swarm/services/review-classifier/forge/token` | `swarm-controller`, with the forge's admin API, for its `review-classifier` account, at start and in a five-minute pass; the secret publisher's policy denies it | `swarm-controller` under its own certificate, before it decides whether to mint, and on every review it posts as the account; the classifier service's policy doesn't read it | ✅ the controller's pass re-mints when the stored token is missing or no longer matches the forge (last eight characters and scopes) | ✅ the controller reads the store on every review it posts |
| _(not in the store)_ a hive's mTLS leaf | the store's own PKI, or an operator placing it by hand | its own client, off disk — the exception above, because it's what makes every other row's pull possible | must be stated | must be stated |
| store path | minter | reader — pulls at runtime, holds in memory | automatic re-mint | automatic re-pull |
| ----------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `swarm/agents/<agent>/matrix/main` | `swarm-controller`, with the swarm's appservice token, at agent creation and in a five-minute pass | the agent container itself, under the certificate its hive passed in | ✅ the pass re-mints when the stored token is missing, unknown to the homeserver, or someone else's | ✅ `hive-matrix-daemon` exits when the homeserver rejects its token, and a five-minute timer restarts it, which reads the store again |
| `swarm/agents/<agent>/matrix/<account>` | `swarm-controller` | the agent container itself, under the certificate its hive passed in | must be stated | must be stated |
| `swarm/controller/swarm-controller/matrix/appservice-token` | `swarm-matrix-ctl`, inside the `hive-matrix` container, once | `swarm-controller`, under its own certificate | ❌ `swarm-matrix-ctl` mints it once; the container keeps its copy and republishes it when the store's differs | ✅ the controller reads it on every five-minute matrix pass |
| `swarm/controller/swarm-controller/oidc/client` | authelia, at its first boot; `swarm-secret-publish` copies it in | `swarm-controller`, under its own certificate, once at start | ❌ authelia mints it once. A re-mint is republished by `swarm-secret-publish`'s path unit | ❌ read once at start; the controller holds the old value until it restarts |
| `swarm/controller/swarm-controller/forge/token` | the forge, inside the `hive-forge` container, once; `swarm-bao-forge-publisher` on the forge's host copies it in under its own certificate, with a write-only grant | `swarm-controller`, under its own certificate, at start | ❌ the forge mints it once and keeps its copy in the container; the publisher writes that copy each time the forge container starts and each time the mint writes the file | ❌ the controller reads it at start, and in the background with backoff until it first finds it; a replaced token reaches it with the next restart |
| `swarm/agents/<agent>/bao-mtls` | the store's agent PKI mount (`deploy.bao.agentPkiMountPath`), which generates the key, at `swarm-controller`'s request at agent creation | `hive-c0re`, under the hive's own certificate, when it writes the agent's container config; then `hive-agent-bao-identity` in the agent container, under the leaf it holds | ✅ `swarm-controller`'s five-minute pass re-issues a live agent's leaf once it's past half its validity (45 of 90 days, read from the certificate itself) | ✅ `hive-agent-bao-identity` re-reads it every 10 minutes and swaps a new leaf into `/run/hive-agent-bao-identity/`, which every store login in the container presents. It reads under the leaf it holds, so it renews only while that leaf is valid |
| `swarm/agents/<agent>/queue` | `swarm-controller`, at agent creation | `hive-agent` in the agent container, under the agent's own certificate, held in memory — the identity it presents to the swarm queue, naming that one agent rather than its hive; `swarm-logs` in the agent container, under the same certificate, on every query, as the bearer it presents at the log store; `swarm-nats-auth` and `swarm-logs-auth`, each under its own certificate, to check a presented secret | ✅ `swarm-controller`'s five-minute pass re-mints a live agent's secret once it's 45 days old by `minted_at` on the stored object; a secret with no `minted_at` gets one stamped, value unchanged. The pass skips agents declared `Destroyed` — declaring an agent destroyed deletes every version of the path instead, the undo of the mint rather than another one | ✅ `hive-agent` reads the path before its first connect and again on every reconnect attempt, so a reconnect after a re-mint presents the new secret. An open connection keeps the secret it connected with; after a revocation the agent keeps retrying under the queue client's backoff. `swarm-logs` reads it on every query and once more after a 401 |
| `swarm/agents/<agent>/forge-token` | `swarm-controller`, at agent creation and in a pass every 5 minutes over every agent with a store identity | the agent container itself, under its own certificate, fetched to `/run/hive-agent-forge-token/token` | ✅ the controller re-mints when the stored token is missing or no longer matches the forge (last eight characters and scopes) | ✅ the agent re-fetches on a 10-minute timer |
| `swarm/agents/<agent>/forge/<label>` | `swarm-controller`, when an operator links an external forge account in the swarm UI | `hive-agent-forge-accounts` in the agent container, under the agent's own certificate, into `<state>/forge-<label>-token` and `forge-<label>.json` | ❌ an operator's token; replaced only by deleting it and linking the label again | ✅ the agent re-fetches on a 2-minute timer |
| `swarm/agents/<agent>/github-token` | `swarm-controller`, when an operator links a GitHub account in the swarm UI | `hive-agent-github-token` in the agent container, under the agent's own certificate, into `<state>/github-token` | ❌ an operator's token; replaced only by deleting it and linking it again | ✅ the agent re-fetches on a 2-minute timer, and `hive-github-notify` re-reads the file on every poll |
| `swarm/hives/<hive>/matrix/sender-token` | `swarm-controller`, with the swarm's appservice token, for every hive in its directory in a five-minute pass | `swarm-controller` under its own certificate, before it decides whether to mint, and hive-c0re's `stored_sender_token()`, under the hive's own certificate | ✅ the controller's pass re-mints when the stored token is missing, unknown to the homeserver, or someone else's | ✅ hive-c0re's matrix sweep reads the store every run and overwrites its token file when the store's token differs |
| `swarm/hives/<hive>/forge/token` | `swarm-controller`, with the forge's admin API, for every hive in its directory in a five-minute pass | `swarm-controller` under its own certificate, before it decides whether to mint, and hive-c0re's `forge::reader_token()`, under the hive's own certificate | ✅ the controller's pass re-mints when the stored token is missing or no longer matches the forge (last eight characters and scopes) | ✅ hive-c0re reads the store on every forge read |
| `swarm/hives/<hive>/queue/agent` | authelia | `swarm-bao-queue-agent` on the hive's host, under its own per-hive certificate; no agent's policy reaches it | must be stated | must be stated |
| `swarm/hives/<hive>/oidc/client` | authelia, at its first boot; `swarm-secret-publish` copies it in | hive-c0re, under the hive's own certificate, when it first connects to the queue | ❌ authelia mints it once. A re-mint is republished by `swarm-secret-publish`'s path unit | ❌ read until the first connect succeeds, retried with a backoff of up to five minutes while it's missing; then held until hive-c0re restarts |
| `swarm/services/<clientId>/oidc/client` | authelia | that service's own reader, under its own certificate — the queue responder's under `swarm-nats-auth`'s, the homeserver's under `swarm-matrix-ctl`'s, the forge's under `swarm-forge-publisher`'s, the review classifier's under `swarm-review-classifier`'s; the bao UI's by the store's granter | must be stated | must be stated |
| `swarm/services/ci-runner/registration-token` | `swarm-controller`, from the forge's admin API, at start and in a five-minute pass; its grant on the path is write-only | `swarm-bao-ci-runner` on the runner's host, under the runner's own certificate, into a file only `swarm-ci-runner-register` in the runner's container loads; no other principal's policy reads it | ❌ the forge hands out one token until an admin resets it; the controller writes a reset token within five minutes | ✅ `swarm-bao-ci-runner` reads it at boot, on deploy and every ten minutes, and restarts the runner when the value changed |
| `swarm/services/review-classifier/anthropic-api-key` | the operator, by hand in the store's UI, field `api_key` | `swarm-review-classifier` in its container, under its own certificate, at the start of every run; held for that run and handed to `claude`'s environment only; no other principal's policy reads it | ❌ an operator's key; replaced by writing a new one | ✅ read on every run, so a written or removed key takes effect on the next run |
| `swarm/services/review-classifier/forge/token` | `swarm-controller`, with the forge's admin API, for its `review-classifier` account, at start and in a five-minute pass; the secret publisher's policy denies it | `swarm-controller` under its own certificate, before it decides whether to mint, and on every review it posts as the account; the classifier service's policy doesn't read it | ✅ the controller's pass re-mints when the stored token is missing or no longer matches the forge (last eight characters and scopes) | ✅ the controller reads the store on every review it posts |
| _(not in the store)_ a hive's mTLS leaf | the store's own PKI, or an operator placing it by hand | its own client, off disk — the exception above, because it's what makes every other row's pull possible | must be stated | must be stated |
<!-- vale write-good.Passive = YES -->

View file

@ -259,8 +259,9 @@ The store's **first reader** is the agent containers' queue credential:
`glue-queue-agent-credential.nix` lands it as two files, the client secret and
the client id it authenticates, because that's how a queue client reads them —
the secret by path, the id by value. No local fallback exists, and none is
possible, so absent files mean this hive's agents have no `swarm-logs`, which
is what a swarm looks like before the publisher on the authelia host has run.
possible, so absent files mean an agent built before `swarm-logs` read its own
queue secret has no working `swarm-logs`, which is what a swarm looks like
before the publisher on the authelia host has run.
The reader runs before `hive-c0re.service`, which wants it (doesn't require
it), so an agent container never renders ahead of the credential; an
unreachable store delays the daemon's start rather than failing it.
@ -274,8 +275,8 @@ it needs no read access to either — and hands them to
`systemd-nspawn --load-credential` through `hive-priv`, which runs as root.
Inside, `hive-agent.service` names the same two ids in `LoadCredential=`, which
lands them in its own `$CREDENTIALS_DIRECTORY`, owned by the agent user and by
nobody else, where the agent's `swarm-logs` reads them. The harness itself
doesn't: it connects to the queue with the agent's own credential. `hive-c0re`
nobody else, where a `swarm-logs` built before it read the agent's own queue
secret reads them. The harness itself doesn't: it connects to the queue with the agent's own credential. `hive-c0re`
never reads the bytes at any point: it runs as `hive-core`, which is the reason
a copy wasn't an option either. When the files aren't there the daemon forwards
nothing and says so in its journal, and the container starts without them.

View file

@ -82,6 +82,13 @@ A hive reads no **swarm service's** OIDC client secret. Every reader of
collector and the store forwarder each fetch theirs with a leaf of their own,
and the store's granter fetches the bao UI's.
Two services read **every agent's queue secret**
(`swarm/agents/<agent>/queue`), each under a leaf of its own whose policy
reaches that one leaf per agent and nothing else: the queue's auth callout
(`swarm-nats-auth`) and the log store's agent check (`swarm-logs-auth`, on the
log store's host). Either one compromised can present any agent's queue
secret, at the queue and at the log store's query route alike.
### Per-hive credentials live under the hive kind
`swarm_secret_client::policy` scopes the **hive** stanza narrowly (unlike the

View file

@ -154,6 +154,9 @@
services.hyperhive.deploy.swarm-ui.package =
lib.mkDefault
self.packages.${pkgs.stdenv.hostPlatform.system}.swarm-ui;
services.hyperhive.deploy.victorialogs.authPackage =
lib.mkDefault
self.packages.${pkgs.stdenv.hostPlatform.system}.swarm-logs-auth;
services.hyperhive.deploy.nats.authPackage =
lib.mkDefault
self.packages.${pkgs.stdenv.hostPlatform.system}.swarm-nats-auth;

View file

@ -7,13 +7,11 @@
# and a default inside the binary would be an address we hope points at
# something.
#
# ⚠️ Gated on the queue being configured as well as on the query URL, because
# the identity this CLI presents IS the queue's: one per-hive machine client,
# one credential pair, delivered once. A swarm with a log store and no agent
# credential has nothing for this to authenticate as, and installing a binary
# that can only fail would be worse than leaving it off PATH — an agent would
# read the 401 as "no logs", which is the exact confusion the store's machine
# route was added to end.
# ⚠️ Gated on the secret store's address as well as on the query URL, because
# the identity this CLI presents is the agent's own queue secret, read from the
# store with the certificate ./bao.nix keeps. Without a store there is nothing
# to authenticate as, and a binary that can only fail would read its 401 as
# "no logs".
{
lib,
pkgs,
@ -22,22 +20,22 @@
}:
let
cfg = config.services.hyperhive.agent.logs;
queueCfg = config.services.hyperhive.agent.queue;
configured = cfg.queryUrl != null && queueCfg.tokenEndpoint != null;
baoCfg = config.services.hyperhive.agent.bao;
configured = cfg.queryUrl != null && baoCfg.addr != null;
# The four coordinates `swarm-logs` reads, all-or-none on its side. Bound
# here as one attrset so the wrapper below cannot set three of them: a
# half-set environment is the failure the binary reports as a deployment
# bug, and it should not be reachable from the module that sets it.
# The same store coordinates ./queue-identity.nix gives the harness, except
# the CA: `%d` expands only inside a unit, so it is spelled as the harness
# unit's credentials directory, which the agent user's shells can read.
#
# 🩸 Two of these are credential **paths**. Neither the id nor the secret
# is read here — a value in `--set` lands in the wrapper script, which is
# in the world-readable nix store.
# 🩸 Paths and a name only. A value in `--set` lands in the wrapper script,
# which is in the world-readable nix store.
wrapperEnv = {
HIVE_AGENT_LOGS_QUERY_URL = cfg.queryUrl;
HIVE_AGENT_OIDC_TOKEN_ENDPOINT = queueCfg.tokenEndpoint;
HIVE_AGENT_OIDC_CLIENT_ID_FILE = queueCfg.clientIdFile;
HIVE_AGENT_OIDC_CLIENT_SECRET_FILE = queueCfg.clientSecretFile;
HIVE_AGENT_NAME = config.services.hyperhive.agent.user.name;
BAO_ADDR = baoCfg.addr;
BAO_CLIENT_CERT = baoCfg.clientCertFile;
BAO_CLIENT_KEY = baoCfg.clientKeyFile;
BAO_CACERT = "/run/credentials/hive-agent.service/hive-agent-bao-server-ca";
};
# Same shape `swarm-controller.nix` wraps `swarmctl` in — `symlinkJoin` +
@ -71,12 +69,6 @@ in
Set by the generated meta flake from the host's own
{option}`services.hyperhive.swarm.victorialogs.domain`.
⚠️ The **whole URL**, not a host to build one from, because this same
string is also the audience `swarm-logs` mints its token for. Two
spellings of one address present as a valid token refused at the
store — the rule `swarm-otel.nix` already states over its own push
targets.
`null` means this hive has not been given the log store's address for
its agents, and `swarm-logs` is then not installed at all. Deliberately
not defaulted: a guessed address resolves cleanly to the wrong place.

View file

@ -33,6 +33,7 @@
./glue-forge-oidc-client.nix
./glue-grafana-oidc-client.nix
./glue-hive-tls-bao-identity.nix
./glue-logs-auth-bao-identity.nix
./glue-matrix-ctl-bao-identity.nix
./glue-matrix-oidc-client.nix
./glue-nats-auth-bao-identity.nix

View file

@ -270,6 +270,13 @@ in
[ -s ${pkiDir}/nats-auth.pem ] || ${signLeaf} ${pkiDir} nats-auth \
${lib.escapeShellArg deployCfg.bao.natsAuthCommonName} "" clientAuth
# The log store's agent auth responder, which reads agent queue
# credentials. Minted here for the matrix-ctl leaf's reason: the log
# store is a swarm singleton, so elsewhere this is the file an operator
# copies.
[ -s ${pkiDir}/logs-auth.pem ] || ${signLeaf} ${pkiDir} logs-auth \
${lib.escapeShellArg deployCfg.bao.logsAuthCommonName} "" clientAuth
# The CI runner's token fetch. Minted here for the matrix-ctl leaf's
# reason: the runner is a swarm singleton, so on a runner host without
# the store this is the file an operator copies.

View file

@ -0,0 +1,31 @@
# Glue: point the log store's agent auth responder at the bao leaf minted for it.
#
# ONE PAIRING PER FILE — the swarm-logs-auth principal ← bao, and nothing else.
# Deleting this leaves a log store with no responder identity unless the
# operator names one, which ./swarm-victorialogs.nix refuses at eval.
#
# ⚠️ The minting is NOT here. ./glue-bao-tls.nix holds the CA and signs the
# leaf. What belongs here is the pairing: which paths the responder presents.
#
# ⚠️ Gated on the store running on this host, the only place the leaf is
# minted. A log-store host without the store gets no default here; the
# operator places the responder's leaf and names it.
#
# Everything is `mkDefault`. An operator naming their own paths wins.
{
lib,
config,
...
}:
let
deployCfg = config.services.hyperhive.deploy;
pkiDir = import ./lib/bao-pki-dir.nix;
in
{
config = lib.mkIf (deployCfg.victorialogs.enable && deployCfg.bao.enable) {
services.hyperhive.deploy.victorialogs = {
authBaoClientCertFile = lib.mkDefault "${pkiDir}/logs-auth.pem";
authBaoClientKeyFile = lib.mkDefault "${pkiDir}/logs-auth-key.pem";
};
};
}

View file

@ -271,29 +271,19 @@ in
HIVE_AGENT_OIDC_TOKEN_ENDPOINT = config.services.hyperhive.swarm.statusPublish.tokenEndpoint;
}
//
# Where an agent reads the swarm's logs, forwarded the same way and gated on
# the same credential: `swarm-logs` authenticates as the queue's own per-hive
# machine client, so an address with no credential behind it would put a
# binary on PATH that can only ever answer 401 — which an agent reads as "no
# logs", the exact confusion the store's machine route was added to end.
# Where an agent reads the swarm's logs. Gated on the store identity below,
# because `swarm-logs` authenticates with the agent's own queue secret, read
# from the store: an address with nothing to authenticate as would put a
# binary on PATH that can only ever answer 401, which an agent reads as "no
# logs".
#
# The **machine** route (`^~ /select/logsql/`), not the browser one at `/`:
# that one ends in `error_page 401 =302` and hands a machine caller
# authelia's login page as a 200 with an HTML body. See
# `swarm-victorialogs.nix`'s location for the whole reasoning.
#
# ⚠️ The full URL rather than the domain, because the same string is also
# the audience the token is minted for — `swarm-otel.nix` states that rule
# over its own push targets, and two spellings present as a valid token
# refused at the store.
lib.optionalAttrs
(
config.services.hyperhive.deploy.hive-controller.queue.agentNatsUrl != null
&& config.services.hyperhive.swarm.statusPublish.tokenEndpoint != null
)
{
HIVE_AGENT_LOGS_QUERY_URL = "https://${config.services.hyperhive.swarm.victorialogs.domain}/select/logsql/query";
}
lib.optionalAttrs haveBaoClientIdentity {
HIVE_AGENT_LOGS_QUERY_URL = "https://${config.services.hyperhive.swarm.victorialogs.domain}/select/logsql/query";
}
//
# Where the swarm's secret store is, and the identity this hive presents to
# it (hive-c0re::lifecycle::agent_identity). `swarm_secret_client` reads these

View file

@ -964,6 +964,17 @@ let
}
'';
# The log store's agent auth responder, which checks the secret an agent
# presents against the one stored for it: the same `+` leaf as the queue
# responder's, and nothing else.
logsAuthReaders = [
{
name = "swarm-logs-auth";
cn = baoDeploy.logsAuthCommonName;
policyText = readStanza "${credentialMountPath}/data/swarm/agents/+/queue";
}
];
# The CI runner's token fetch. The one reader of the path, which the
# controller writes and cannot read back.
ciRunnerReaders = [
@ -1797,6 +1808,19 @@ in
'';
};
logsAuthCommonName = lib.mkOption {
type = lib.types.str;
default = "swarm-logs-auth";
description = ''
Subject the store's `swarm-logs-auth` cert-auth role accepts: the
identity the log store's agent auth responder presents to read agent
queue credentials, `swarm/agents/<agent>/queue` and nothing else under
an agent.
⚠️ Reserved as a hive name by ./swarm.nix, like its siblings.
'';
};
ciRunnerCommonName = lib.mkOption {
type = lib.types.str;
default = "swarm-ci-runner";
@ -3171,6 +3195,8 @@ in
systemd.services.swarm-bao-nats-auth-policy = readerPolicyUnit "write the queue responder's agent-credential-reader bao policy and cert-auth role" natsAuthReaders;
systemd.services.swarm-bao-logs-auth-policy = readerPolicyUnit "write the log store responder's agent-credential-reader bao policy and cert-auth role" logsAuthReaders;
systemd.services.swarm-bao-ci-runner-policy = readerPolicyUnit "write the CI runner's registration-token-reader bao policy and cert-auth role" ciRunnerReaders;
systemd.services.swarm-bao-review-classifier-policy = readerPolicyUnit "write the review classifier's API-key-reader bao policy and cert-auth role" reviewClassifierReaders;
@ -3302,6 +3328,7 @@ in
"swarm-bao-nats-tls-policy.service"
"swarm-bao-hive-leaf-policy.service"
"swarm-bao-ci-runner-policy.service"
"swarm-bao-logs-auth-policy.service"
"swarm-bao-forge-publisher-policy.service"
]
++ granterAfter;

View file

@ -52,6 +52,22 @@ let
error_page 401 =302 https://${hyperhiveCfg.swarm.authelia.domain}/?rd=$target_url;
'';
# The agents' `auth_request` responder (`swarm-logs-auth`). Its socket's
# directory holds that socket alone and is traverse-only for others, the
# shape `swarm-controller.nix` gives its own.
logsAuthUnit = "swarm-logs-auth";
logsAuthSocket = "/run/${logsAuthUnit}/auth.sock";
logsAuthCredential = id: "/run/credentials/${logsAuthUnit}.service/${id}";
# Credential id → host path of the responder's store identity.
logsAuthStoreFiles = {
"bao-client.pem" = deployCfg.victorialogs.authBaoClientCertFile;
"bao-client-key.pem" = deployCfg.victorialogs.authBaoClientKeyFile;
}
// lib.optionalAttrs (deployCfg.bao.serverCaFile != null) {
"bao-ca.pem" = deployCfg.bao.serverCaFile;
};
baoCfg = hyperhiveCfg.swarm.bao;
# Shared host netns, like every sibling swarm container: the collector
# and Grafana reach this at 127.0.0.1:<port>.
privateNetwork = false;
@ -88,7 +104,74 @@ in
'';
};
options.services.hyperhive.deploy.victorialogs.authPackage = lib.mkOption {
type = lib.types.package;
defaultText = lib.literalExpression "hyperhive.packages.\${system}.swarm-logs-auth";
description = ''
The `swarm-logs-auth` package: the gateway's `auth_request` responder
that checks an agent's `swarm-agent.<agent>.<secret>` bearer on
`/select/logsql/` against the secret store.
'';
};
options.services.hyperhive.deploy.victorialogs.authBaoClientCertFile = lib.mkOption {
type = lib.types.nullOr lib.types.str;
default = null;
example = "/var/lib/swarm-bao-pki/logs-auth.pem";
description = ''
Client certificate `swarm-logs-auth` presents to the swarm's secret store
to read agents' queue secrets. Its subject must be
{option}`services.hyperhive.deploy.bao.logsAuthCommonName`.
Required when `deploy.victorialogs.enable` is set; unset fails at eval.
./glue-logs-auth-bao-identity.nix points it at the leaf
./glue-bao-tls.nix mints on a host that runs the store; on any other
log-store host it is a file the operator places.
A path, never a value.
'';
};
options.services.hyperhive.deploy.victorialogs.authBaoClientKeyFile = lib.mkOption {
type = lib.types.nullOr lib.types.str;
default = null;
example = "/var/lib/swarm-bao-pki/logs-auth-key.pem";
description = ''
Private key for {option}`services.hyperhive.deploy.victorialogs.authBaoClientCertFile`.
A path, never a value.
'';
};
config = lib.mkIf deployCfg.victorialogs.enable {
assertions = [
{
assertion =
deployCfg.victorialogs.authBaoClientCertFile != null
&& deployCfg.victorialogs.authBaoClientKeyFile != null;
message = ''
services.hyperhive.deploy.victorialogs.authBaoClientCertFile and
authBaoClientKeyFile must both be set.
swarm-logs-auth reads agents' queue secrets from the swarm's secret
store with this certificate. They default to the leaf the store
mints where services.hyperhive.deploy.bao.enable is set on this
host; elsewhere, issue a leaf whose CN is
services.hyperhive.deploy.bao.logsAuthCommonName and name it in
both options.
'';
}
];
# Which check a `/select/logsql/` request gets. `auth_request` takes a
# fixed URI, so it names `/__swarm_logs_auth`, which rewrites to the
# target this map picks from the caller's `Authorization`.
services.nginx.appendHttpConfig = ''
map $http_authorization $swarm_logs_auth_target {
"~^Bearer swarm-agent\." /__swarm_logs_agent_auth;
default /__hive_authelia;
}
'';
# This store publishes its own health as prometheus metrics on the same
# listener it serves queries on, so the swarm's collector scrapes it with
# no exporter and no extra port — same arrangement as the metrics store.
@ -174,12 +257,35 @@ in
# logs. That is the rule in force, not an omission: read
# permissions are a later thing, and this location is where one
# attaches when it exists.
#
# Two checks: an agent's own `swarm-agent.` bearer goes to
# `swarm-logs-auth`, anything else to authelia (`appendHttpConfig`
# above picks).
"^~ /select/logsql/" = {
# No URI part, so the request path and its query string reach
# the store as the caller sent them.
proxyPass = "http://127.0.0.1:${toString cfg.port}";
extraConfig = ''
auth_request /__hive_authelia;
auth_request /__swarm_logs_auth;
'';
};
# `rewrite … last` re-matches locations, so the subrequest lands in
# one of the two internal locations named by the map.
"= /__swarm_logs_auth" = {
extraConfig = ''
internal;
rewrite ^ $swarm_logs_auth_target last;
'';
};
# The `Authorization` header rides the subrequest unchanged.
"= /__swarm_logs_agent_auth" = {
proxyPass = "http://unix:${logsAuthSocket}:/";
extraConfig = ''
internal;
proxy_pass_request_body off;
proxy_set_header Content-Length "";
'';
};
@ -207,6 +313,53 @@ in
};
};
# The agents' check on `/select/logsql/`. On the host rather than in the
# store's container: it needs nothing of the store, only nginx's reach to
# its socket and its own store identity.
systemd.services.${logsAuthUnit} = {
description = "swarm log store auth_request responder for agent credentials";
wantedBy = [ "multi-user.target" ];
# Where the store's PKI is minted on this host, the leaf is one of its
# files. Ordering only: elsewhere the unit does not exist.
after = [
"network.target"
"swarm-bao-pki.service"
];
serviceConfig = {
ExecStart = lib.concatStringsSep " " [
"${deployCfg.victorialogs.authPackage}/bin/swarm-logs-auth"
"--socket ${logsAuthSocket}"
# The role ./swarm-bao.nix writes on the store's `cert` mount.
"--store-cert-role swarm-logs-auth"
];
# Paths only: `argv` is world-readable via /proc/<pid>/cmdline.
LoadCredential = lib.mapAttrsToList (id: path: "${id}:${path}") logsAuthStoreFiles;
DynamicUser = true;
RuntimeDirectory = logsAuthUnit;
# Traverse-only for others, so nginx reaches the socket path.
RuntimeDirectoryMode = "0751";
# The path never vanishes from under a running nginx; the responder
# unlinks a stale socket as it starts.
RuntimeDirectoryPreserve = "yes";
Restart = "on-failure";
RestartSec = "5s";
SyslogIdentifier = logsAuthUnit;
ProtectSystem = "strict";
ProtectHome = true;
PrivateTmp = true;
PrivateDevices = true;
NoNewPrivileges = true;
};
environment = {
BAO_ADDR = "https://${baoCfg.domain}:${toString baoCfg.port}";
BAO_CLIENT_CERT = logsAuthCredential "bao-client.pem";
BAO_CLIENT_KEY = logsAuthCredential "bao-client-key.pem";
}
// lib.optionalAttrs (logsAuthStoreFiles ? "bao-ca.pem") {
BAO_CACERT = logsAuthCredential "bao-ca.pem";
};
};
containers.${cfg.machine} = {
autoStart = true;
ephemeral = false;

View file

@ -55,6 +55,7 @@ let
deployCfg.bao.granterCommonName
deployCfg.bao.natsAuthCommonName
deployCfg.bao.ciRunnerCommonName
deployCfg.bao.logsAuthCommonName
deployCfg.bao.reviewClassifierCommonName
deployCfg.bao.forgePublisherCommonName
]

View file

@ -40,8 +40,45 @@ let
agentNoBao = shared.agentBare;
# `swarm-logs` authenticates with the agent's queue secret from the store, so
# the store address gates it beside the query URL.
logsUrl = "https://logs.t.local/select/logsql/query";
agentLogs = agentWith {
services.hyperhive.agent.bao.addr = "https://bao.t.local:8200";
services.hyperhive.agent.logs.queryUrl = logsUrl;
};
agentLogsNoBao = agentWith { services.hyperhive.agent.logs.queryUrl = logsUrl; };
swarmLogsIn =
machine:
lib.findFirst (
p: (p.name or "") == "swarm-logs-configured"
) null machine.environment.systemPackages;
agentBaoIdentity = machine: machine.systemd.services.hive-agent-bao-identity;
cases = [
{
# The wrapper names the store identity ./queue-identity.nix gives the
# harness, by path, and nothing of the hive's shared OIDC client.
name = "swarm-logs is wrapped with the agent's own store identity";
ok =
let
w = (swarmLogsIn agentLogs).buildCommand;
a = agentLogs.services.hyperhive.agent;
set = k: v: lib.hasInfix "--set ${k} ${lib.escapeShellArg v}" w;
in
set "HIVE_AGENT_LOGS_QUERY_URL" logsUrl
&& set "HIVE_AGENT_NAME" a.user.name
&& set "BAO_ADDR" "https://bao.t.local:8200"
&& set "BAO_CLIENT_CERT" a.bao.clientCertFile
&& set "BAO_CLIENT_KEY" a.bao.clientKeyFile
&& set "BAO_CACERT" "/run/credentials/hive-agent.service/hive-agent-bao-server-ca"
&& !(lib.hasInfix "OIDC" w);
}
{
# A query URL with nothing to authenticate as installs nothing.
name = "an agent with a log store but no secret store gets no swarm-logs";
ok = swarmLogsIn agentLogsNoBao == null && swarmLogsIn agentLogs != null;
}
{
# Both ids or neither: the secret authenticates nobody without the id it
# belongs to, and `swarm-logs` reads both.

View file

@ -169,7 +169,7 @@ let
_: u: (u.environment.BAO_CLIENT_CERT or null) == granterCertFile
) baoGrantWithConsumers.systemd.services;
# The sixteen units that write a `swarm-*` grant, by name, for the discovery
# The seventeen units that write a `swarm-*` grant, by name, for the discovery
# control below.
grantingUnitNames = [
"swarm-bao-controller-policy"
@ -186,6 +186,7 @@ let
"swarm-bao-operator-viewer-policy"
"swarm-bao-hive-leaf-policy"
"swarm-bao-ci-runner-policy"
"swarm-bao-logs-auth-policy"
"swarm-bao-forge-publisher-policy"
"swarm-bao-review-classifier-policy"
];
@ -704,6 +705,31 @@ let
in
lib.hasInfix "/nats-auth.pem ]" s && lib.hasInfix "swarm-nats-auth \"\" clientAuth" s;
}
{
# The same `+` leaf as the queue responder, read-only, and nothing else:
# this principal checks a presented secret and does nothing with it.
name = "the log store responder's grant is every agent's queue secret, read-only, and nothing else";
ok =
let
s = baoGrantHere.systemd.services.swarm-bao-logs-auth-policy.script;
in
lib.hasInfix "path \"secret/data/swarm/agents/+/queue\" {\n capabilities = [\"read\"]\n}" s
&& lib.length (lib.filter lib.isList (builtins.split "path \"" s)) == 1
&& !(lib.hasInfix "secret/data/swarm/agents/*" s)
&& !(lib.hasInfix "secret/data/swarm/hives" s)
&& !(lib.hasInfix "secret/data/swarm/services" s)
&& lib.hasInfix "auth/cert/certs/swarm-logs-auth" s
&& lib.hasInfix "allowed_common_names=swarm-logs-auth" s
&& lib.hasInfix "token_policies=swarm-logs-auth" s;
}
{
name = "the PKI unit signs the log store responder's leaf under its own subject";
ok =
let
s = baoGrantHere.systemd.services.swarm-bao-pki.script;
in
lib.hasInfix "/logs-auth.pem ]" s && lib.hasInfix "swarm-logs-auth \"\" clientAuth" s;
}
{
# 🩸 The half that makes the policies above bind: a policy grants only
# through a token that carries it, and a token is minted by a cert-auth
@ -915,24 +941,24 @@ let
}
{
# A store host without the granter's pair writes its grants some other
# way, so none of the sixteen units may exist. Without this arm
# way, so none of the seventeen units may exist. Without this arm
# `lib.mkIf haveGranter` could be dropped from any of them and every other
# case here would still pass.
name = "without the granter's pair none of the sixteen granting units render";
name = "without the granter's pair none of the seventeen granting units render";
ok =
let
s = baoGranterOptOut.systemd.services;
in
lib.all (unit: !(s ? ${unit})) (grantingUnitNames ++ [ "swarm-bao-granter-role" ])
# The control: the same store with the pair renders all sixteen.
# The control: the same store with the pair renders all seventeen.
&& lib.all (unit: baoGrantHere.systemd.services ? ${unit}) grantingUnitNames;
}
{
# 🩸 What replaced the silent skip. With no bootstrap token the sixteen still
# 🩸 What replaced the silent skip. With no bootstrap token the seventeen still
# render, and a refused granter fails them with the step that fixes it.
# A store host that never named a token is told to name one, since the
# unit that sets the granter up renders only where it has.
name = "a store host without a bootstrap token renders the sixteen, each failing loudly with the one-time step";
name = "a store host without a bootstrap token renders the seventeen, each failing loudly with the one-time step";
ok =
let
s = baoGranterNoToken.systemd.services;
@ -1695,8 +1721,8 @@ let
}
{
# What makes the case above mean something: discovery by the granter's
# certificate reaches all sixteen units, and each yields calls.
name = "the granter-policy check sees all sixteen granting units, and parses calls from each";
# certificate reaches all seventeen units, and each yields calls.
name = "the granter-policy check sees all seventeen granting units, and parses calls from each";
ok =
lib.sort lib.lessThan (lib.attrNames granterUnits) == lib.sort lib.lessThan grantingUnitNames
&& lib.all (u: baoCalls u.script != [ ]) (lib.attrValues granterUnits)

View file

@ -504,6 +504,44 @@ let
&& !(lib.hasInfix "$args" q.extraConfig)
&& !(lib.hasInfix "?" q.proxyPass);
}
{
# `auth_request` takes a fixed URI, so the split between the agents'
# responder and authelia is a map plus a rewrite in an internal
# location. Every link of that chain is pinned: the map keys on the
# agent-token prefix and falls back to authelia, the query location
# names the switch, the switch rewrites to the map's pick, and the
# agent target reaches the responder's socket, which the unit binds.
name = "the machine query location sends agent bearers to swarm-logs-auth and the rest to authelia";
ok =
let
l = allLocal.services.nginx.virtualHosts."logs.t.local".locations;
http = allLocal.services.nginx.appendHttpConfig;
unit = allLocal.systemd.services.swarm-logs-auth;
in
lib.hasInfix "map \$http_authorization \$swarm_logs_auth_target" http
&& lib.hasInfix "\"~^Bearer swarm-agent\\.\" /__swarm_logs_agent_auth;" http
&& lib.hasInfix "default /__hive_authelia;" http
&& lib.hasInfix "auth_request /__swarm_logs_auth;" l."^~ /select/logsql/".extraConfig
&& lib.hasInfix "rewrite ^ \$swarm_logs_auth_target last;" l."= /__swarm_logs_auth".extraConfig
&& lib.hasInfix "internal;" l."= /__swarm_logs_agent_auth".extraConfig
&& l."= /__swarm_logs_agent_auth".proxyPass == "http://unix:/run/swarm-logs-auth/auth.sock:/"
&& lib.hasInfix "--socket /run/swarm-logs-auth/auth.sock" unit.serviceConfig.ExecStart
&& l ? "= /__hive_authelia";
}
{
# The responder reads the store as itself, with the leaf the store's
# host mints for it.
name = "swarm-logs-auth presents its own store leaf";
ok =
let
v = allLocal.services.hyperhive.deploy.victorialogs;
lc = allLocal.systemd.services.swarm-logs-auth.serviceConfig.LoadCredential;
in
lib.hasSuffix "/logs-auth.pem" v.authBaoClientCertFile
&& lib.hasSuffix "/logs-auth-key.pem" v.authBaoClientKeyFile
&& builtins.elem "bao-client.pem:${v.authBaoClientCertFile}" lc
&& builtins.elem "bao-client-key.pem:${v.authBaoClientKeyFile}" lc;
}
{
# The absence arm for the case above, and the option's own rule — a
# service declares its entry under its own `enable` — made checkable.

View file

@ -163,6 +163,10 @@ in
# rather than every hive's.
swarm-nats-auth = mkBinPackage "swarm-nats-auth" "hyperhive swarm queue auth-callout responder";
# The log store gateway's `auth_request` responder for agents. Out of
# `daemonBins` like the responder above: it runs on the log store's host only.
swarm-logs-auth = mkBinPackage "swarm-logs-auth" "hyperhive swarm log store agent auth responder";
# The matrix container's control binary; `mint` publishes the sender token. Out of `daemonBins` for
# "runs *inside* a container, not on the host" reason as the responder
# above, and with a second one: putting it in the core bundle would place

View file

@ -0,0 +1,18 @@
[package]
name = "swarm-agent-verify"
version.workspace = true
edition.workspace = true
[dependencies]
anyhow.workspace = true
# Comparing an agent's presented secret with the stored one.
subtle.workspace = true
# The agent-token spelling the agent also uses.
swarm-queue-client.workspace = true
# The read of the stored credential the token is checked against.
swarm-secret-client.workspace = true
tokio.workspace = true
tracing.workspace = true
[lints]
workspace = true

View file

@ -0,0 +1,217 @@
//! Checking an agent's own credential against the swarm secret store.
//!
//! The token is `swarm_queue_client::agent_token`'s spelling: a prefix, the
//! agent's name, and its secret. The name is only a claim. It becomes an
//! identity when the secret equals the one stored at
//! `swarm/agents/<agent>/queue`, and [`verify`] grants nothing on the name
//! alone: every path that does not reach that comparison, and every lookup
//! that fails, is a denial.
//!
//! Two verifiers use this: the queue's auth callout (`swarm-nats-auth`) and
//! the log store's `auth_request` responder (`swarm-logs-auth`). Each reads
//! the store under its own certificate.
use std::future::Future;
use std::time::Duration;
use anyhow::Context;
use subtle::ConstantTimeEq;
use swarm_queue_client::agent_token::AgentToken;
use swarm_secret_client::client::{DEFAULT_CERT_MOUNT, Settings};
use swarm_secret_client::queue::{self, AgentCredential};
/// Where the stored credential for an agent comes from.
pub trait CredentialSource {
/// The object at `agent`'s queue path, `None` when nothing is stored
/// there, or an error when the store could not answer.
fn lookup(
&self,
agent: &str,
) -> impl Future<Output = anyhow::Result<Option<AgentCredential>>> + Send;
}
/// The swarm secret store, reached with the verifier's own certificate.
pub struct Store {
settings: Settings,
cert_role: String,
}
impl Store {
/// The store at `settings`, logged in to as `cert_role`.
#[must_use]
pub fn new(settings: Settings, cert_role: String) -> Self {
Self {
settings,
cert_role,
}
}
}
impl CredentialSource for Store {
/// Logs in per lookup, so no token is left to expire inside a long-lived
/// process.
async fn lookup(&self, agent: &str) -> anyhow::Result<Option<AgentCredential>> {
let path = queue::agent_queue_path(agent)?;
let store = swarm_secret_client::SecretStore::connect(
&self.settings,
&self.cert_role,
DEFAULT_CERT_MOUNT,
)
.await
.context("logging in to the secret store")?;
store
.read_optional(&path)
.await
.with_context(|| format!("reading {path}"))
}
}
/// Whether `token`'s secret is the one stored for the agent it names.
///
/// The lookup is bounded by `timeout`, after which this denies.
pub async fn verify(
source: &impl CredentialSource,
token: &AgentToken<'_>,
timeout: Duration,
) -> bool {
let agent = token.agent;
let stored = match tokio::time::timeout(timeout, source.lookup(agent)).await {
Ok(Ok(Some(stored))) => stored,
Ok(Ok(None)) => {
tracing::warn!(
agent,
"no queue credential is stored for this agent; denying"
);
return false;
}
Ok(Err(e)) => {
tracing::warn!(
agent,
error = format!("{e:#}"),
"credential lookup failed; denying"
);
return false;
}
Err(_) => {
tracing::warn!(agent, "credential lookup timed out; denying");
return false;
}
};
if stored.agent != agent {
tracing::warn!(
agent,
stored_agent = %stored.agent,
"the stored credential names a different agent than its path; denying"
);
return false;
}
// Constant time, so the time to refuse says nothing about how much of the
// secret was right. A length mismatch returns early; the length is not
// secret, every minted secret has the same one.
stored
.value
.as_bytes()
.ct_eq(token.secret.as_bytes())
.into()
}
#[cfg(test)]
mod tests {
use super::*;
use swarm_queue_client::agent_token::parse_agent_token;
/// A store holding at most one credential, or failing outright.
enum Fake {
/// `credential` is stored at `path_agent`'s queue path.
Holds {
path_agent: &'static str,
credential: AgentCredential,
},
Fails,
/// A store that accepts the request and never answers.
Hangs,
}
impl CredentialSource for Fake {
async fn lookup(&self, agent: &str) -> anyhow::Result<Option<AgentCredential>> {
match self {
Self::Holds {
path_agent,
credential,
} => Ok((agent == *path_agent).then(|| credential.clone())),
Self::Fails => anyhow::bail!("store unreachable"),
Self::Hangs => std::future::pending().await,
}
}
}
fn stored(path_agent: &'static str, named: &str) -> Fake {
Fake::Holds {
path_agent,
credential: AgentCredential {
value: "Ab9_-zSECRET".to_owned(),
agent: named.to_owned(),
minted_at: None,
},
}
}
fn atlas() -> Fake {
stored("atlas", "atlas")
}
const BOUND: Duration = Duration::from_millis(200);
async fn verified(source: &Fake, token: &str) -> bool {
let token = parse_agent_token(token)
.expect("an agent token")
.expect("well-formed");
verify(source, &token, BOUND).await
}
#[tokio::test]
async fn the_stored_secret_is_accepted() {
assert!(verified(&atlas(), "swarm-agent.atlas.Ab9_-zSECRET").await);
}
#[tokio::test]
async fn a_wrong_secret_is_denied() {
assert!(!verified(&atlas(), "swarm-agent.atlas.Ab9_-zSECREX").await);
assert!(!verified(&atlas(), "swarm-agent.atlas.Ab9_-zSECRE").await);
}
/// The secret check is what stops one agent claiming another's name: the
/// right secret under someone else's name finds that agent's credential,
/// or none.
#[tokio::test]
async fn another_agents_name_with_this_agents_secret_is_denied() {
assert!(!verified(&atlas(), "swarm-agent.argus.Ab9_-zSECRET").await);
}
#[tokio::test]
async fn an_agent_with_nothing_stored_is_denied() {
assert!(!verified(&stored("argus", "argus"), "swarm-agent.atlas.Ab9_-zSECRET").await);
}
#[tokio::test]
async fn a_failed_lookup_is_denied() {
assert!(!verified(&Fake::Fails, "swarm-agent.atlas.Ab9_-zSECRET").await);
}
#[tokio::test]
async fn a_store_that_never_answers_is_denied_within_the_bound() {
let started = std::time::Instant::now();
assert!(!verified(&Fake::Hangs, "swarm-agent.atlas.Ab9_-zSECRET").await);
let took = started.elapsed();
assert!(took >= BOUND, "denied before the bound: {took:?}");
assert!(
took < BOUND + Duration::from_millis(250),
"denied well after the bound: {took:?}"
);
}
#[tokio::test]
async fn a_stored_object_naming_another_agent_is_denied() {
assert!(!verified(&stored("argus", "atlas"), "swarm-agent.argus.Ab9_-zSECRET").await);
}
}

View file

@ -0,0 +1,23 @@
[package]
name = "swarm-logs-auth"
version.workspace = true
edition.workspace = true
[[bin]]
name = "swarm-logs-auth"
path = "src/main.rs"
[dependencies]
anyhow.workspace = true
axum.workspace = true
clap.workspace = true
swarm-agent-verify.workspace = true
# The agent-token spelling, and the rustls provider the store client needs.
swarm-queue-client.workspace = true
swarm-secret-client.workspace = true
tokio.workspace = true
tracing.workspace = true
tracing-subscriber.workspace = true
[lints]
workspace = true

160
swarm-logs-auth/src/main.rs Normal file
View file

@ -0,0 +1,160 @@
//! The log store gateway's `auth_request` responder for agents.
//!
//! nginx sends it every `/select/logsql/` request whose `Authorization` is
//! `Bearer swarm-agent.<agent>.<secret>` (`swarm-victorialogs.nix`). It answers
//! 200 when the secret is the one stored at `swarm/agents/<agent>/queue`, and
//! 401 otherwise, store failures included: `auth_request` turns any other
//! status into a 500 for the caller.
//!
//! It reads the store under its own certificate, whose policy reaches every
//! agent's `queue` secret and nothing else. It serves a unix socket whose
//! directory only the gateway's nginx may enter.
use std::os::unix::fs::PermissionsExt as _;
use std::path::PathBuf;
use std::sync::Arc;
use std::time::Duration;
use anyhow::Context as _;
use axum::extract::State;
use axum::http::{HeaderMap, StatusCode, header};
use clap::Parser;
use swarm_agent_verify::CredentialSource;
use swarm_queue_client::agent_token::parse_agent_token;
/// How long one store lookup may take before the request is denied. Well
/// under nginx's 60s `proxy_read_timeout`, so a slow store reads as a 401
/// rather than a gateway timeout.
const LOOKUP_TIMEOUT: Duration = Duration::from_secs(5);
#[derive(Parser)]
#[command(about = "auth_request responder: checks an agent's queue secret against the store")]
struct Args {
/// Unix socket to serve on.
#[arg(long)]
socket: PathBuf,
/// Cert-auth role this responder logs in to the store as.
#[arg(long, default_value = "swarm-logs-auth")]
store_cert_role: String,
}
/// The answer for one subrequest's `Authorization` header.
async fn decide(source: &impl CredentialSource, authorization: Option<&str>) -> StatusCode {
let Some(token) = authorization.and_then(|v| v.strip_prefix("Bearer ")) else {
tracing::warn!("no bearer token; denying");
return StatusCode::UNAUTHORIZED;
};
let token = match parse_agent_token(token) {
Some(Ok(token)) => token,
Some(Err(e)) => {
tracing::warn!(error = %e, "malformed agent token; denying");
return StatusCode::UNAUTHORIZED;
}
None => {
tracing::warn!("not an agent token; denying");
return StatusCode::UNAUTHORIZED;
}
};
if swarm_agent_verify::verify(source, &token, LOOKUP_TIMEOUT).await {
tracing::info!(agent = token.agent, "verified");
StatusCode::OK
} else {
StatusCode::UNAUTHORIZED
}
}
async fn handle(
State(store): State<Arc<swarm_agent_verify::Store>>,
headers: HeaderMap,
) -> StatusCode {
let authorization = headers
.get(header::AUTHORIZATION)
.and_then(|v| v.to_str().ok());
decide(store.as_ref(), authorization).await
}
#[tokio::main]
async fn main() -> anyhow::Result<()> {
swarm_queue_client::install_crypto_provider();
tracing_subscriber::fmt()
.with_env_filter(
tracing_subscriber::EnvFilter::try_from_default_env()
.unwrap_or_else(|_| tracing_subscriber::EnvFilter::new("info")),
)
// stdout is journald, which keeps ANSI escapes as raw bytes.
.with_ansi(false)
.init();
let args = Args::parse();
let settings = swarm_secret_client::client::Settings::from_env()
.context("reading the secret store's address and this responder's identity")?;
let store = Arc::new(swarm_agent_verify::Store::new(
settings,
args.store_cert_role,
));
// `RuntimeDirectoryPreserve=yes` can leave the previous run's socket; the
// directory is this unit's alone, so unlinking it is safe.
if let Err(e) = std::fs::remove_file(&args.socket)
&& e.kind() != std::io::ErrorKind::NotFound
{
return Err(e)
.with_context(|| format!("clearing stale socket at {}", args.socket.display()));
}
let listener = tokio::net::UnixListener::bind(&args.socket)
.with_context(|| format!("binding {}", args.socket.display()))?;
// nginx is another user and `connect(2)` needs write. The directory is
// the access control, not this mode.
std::fs::set_permissions(&args.socket, std::fs::Permissions::from_mode(0o666))
.with_context(|| format!("chmod {}", args.socket.display()))?;
tracing::info!(socket = %args.socket.display(), "swarm-logs-auth listening");
let app = axum::Router::new().fallback(handle).with_state(store);
axum::serve(listener, app).await.context("serving")
}
#[cfg(test)]
mod tests {
use super::*;
use swarm_secret_client::queue::AgentCredential;
/// A store holding atlas's credential.
struct Atlas;
impl CredentialSource for Atlas {
async fn lookup(&self, agent: &str) -> anyhow::Result<Option<AgentCredential>> {
Ok((agent == "atlas").then(|| AgentCredential {
value: "Ab9_-zSECRET".to_owned(),
agent: "atlas".to_owned(),
minted_at: None,
}))
}
}
#[tokio::test]
async fn the_stored_secret_is_200() {
assert_eq!(
decide(&Atlas, Some("Bearer swarm-agent.atlas.Ab9_-zSECRET")).await,
StatusCode::OK
);
}
#[tokio::test]
async fn everything_else_is_401() {
for authorization in [
None,
Some("swarm-agent.atlas.Ab9_-zSECRET"),
Some("Basic swarm-agent.atlas.Ab9_-zSECRET"),
Some("Bearer swarm-agent.atlas.Ab9_-zSECREX"),
Some("Bearer swarm-agent.argus.Ab9_-zSECRET"),
Some("Bearer swarm-agent.atlas"),
Some("Bearer authelia_at_abc.def"),
] {
assert_eq!(
decide(&Atlas, authorization).await,
StatusCode::UNAUTHORIZED,
"{authorization:?}"
);
}
}
}

View file

@ -12,13 +12,15 @@ path = "src/main.rs"
anyhow.workspace = true
clap.workspace = true
clap-markdown = "0.1"
# The whole point of the crate's dependency list: minting the access token
# is `swarm_queue_client::mint_token_for_blocking`, not a second copy of
# the `client_credentials` request. See `src/auth.rs`.
# The agent-token spelling the gateway's responder parses, and the rustls
# provider the store client needs.
swarm-queue-client.workspace = true
# Blocking, not async: this binary makes exactly two requests and exits, so
# a reactor would exist only to be torn down. It also matches the token
# minter reached for above, which is `swarm-queue-client`'s blocking one.
# The read of this agent's own queue secret.
swarm-secret-client.workspace = true
# Only to drive that one async read; the query itself is blocking.
tokio.workspace = true
# Blocking, not async: this binary makes one request (two on a 401) and
# exits, so a reactor would exist only to be torn down.
reqwest = { workspace = true, features = ["blocking"] }
serde_json.workspace = true

View file

@ -20,40 +20,19 @@ attach at the gateway location, not in this binary.
## How it authenticates
The agent container already holds one identity, the per-hive machine client
`hive-<name>-agent`, delivered as a systemd credential pair. This binary
presents a `client_credentials` access token minted from it as
`Authorization: Bearer`.
As the agent itself. On each query it reads the agent's own queue secret,
`swarm/agents/<agent>/queue`, from the swarm secret store with the agent's
store certificate (`nix/agent-modules/bao.nix`), the same read `hive-agent`
makes for its queue connection. It sends the secret as
`Authorization: Bearer swarm-agent.<agent>.<secret>`, spelled by
`swarm_queue_client::agent_token`.
Minting is **not** implemented here — `swarm_queue_client::mint_token_for_blocking`
already owns the request, its error type and its CA handling, and a
token-endpoint fix has to be findable in one place. What this crate adds is
the agent-shaped half: the client id arrives as a _file_ beside the secret, so
nothing outside `nix/agent-modules/queue.nix` spells `hive-<name>-agent`
twice. That is the same problem `hive-agent/src/swarm_queue.rs` solves, and
`src/auth.rs` is its `decide` restated over this binary's inputs.
At the gateway, nginx hands a bearer of that shape to `swarm-logs-auth`, which
compares it with the stored secret (`swarm-agent-verify`, the check the queue's
auth callout also uses). A 401 makes `swarm-logs` read the secret again and
retry once, which covers a secret re-minted between the read and the check.
🩸 Every credential here is a **path**. The secret's contents are read at the
moment of the request, inside `swarm-queue-client`, and are never bound to a
name in this crate, logged, or rendered into an error.
### The audience is the URL
`swarm-logs` mints its token with the query URL as the requested audience,
because that is what authelia checks at the authz endpoint the gateway's
`auth_request` calls. The same rule `swarm-otel.nix` states over its own push
targets: one binding for the address and the audience, since two spellings of
one address present as a valid token refused at the store.
⚠️ That requires the agent client to be _registered_ for that audience and to
hold `authelia.bearer.authz`. Both are set in `swarm-authelia.nix`'s
`agentClients`; without them authelia answers `invalid_target` at the token
endpoint, or the gateway answers 401 with no explanation.
⚠️ Registration is not issuance, so the token request _asks_ for both: the
audience and the scope are named in the `client_credentials` form, because a
client that is registered for a scope it does not request is handed a token
carrying none, and the gateway refuses that with the same bare 401.
🩸 The secret is never logged or rendered into an error.
## Configuration
@ -62,19 +41,17 @@ shape `swarmctl` is configured in, and for the same reason: every value is
derived from an option that module owns, so a default here would be an address
we _hope_ points at something.
| variable | what |
| ------------------------------------ | ---------------------------------------------------- |
| `HIVE_AGENT_LOGS_QUERY_URL` | the LogsQL query endpoint; also the token's audience |
| `HIVE_AGENT_OIDC_TOKEN_ENDPOINT` | authelia's token endpoint |
| `HIVE_AGENT_OIDC_CLIENT_ID_FILE` | path of the delivered client-id credential |
| `HIVE_AGENT_OIDC_CLIENT_SECRET_FILE` | path of the delivered secret credential |
| `HIVE_AGENT_OIDC_CA_FILE` | optional extra trust anchor; unset in this tree |
| variable | what |
| --------------------------- | ------------------------------------------------ |
| `HIVE_AGENT_LOGS_QUERY_URL` | the LogsQL query endpoint |
| `HIVE_AGENT_NAME` | this agent's name; the store path and role |
| `BAO_ADDR` | the secret store's address |
| `BAO_CLIENT_CERT` | path of the agent's live store certificate |
| `BAO_CLIENT_KEY` | path of its key |
| `BAO_CACERT` | path of the store's CA; missing or empty is fine |
`HIVE_AGENT`, the same prefix the harness reads, because it is the same
identity — an agent authenticates as its hive's agent client whether the
caller is the harness or a CLI the agent typed. The first four are all-or-none:
a half-set environment is a deployment bug, and this binary says so by name
rather than behaving like an unconfigured one.
With no `HIVE_AGENT_LOGS_QUERY_URL` the binary says there is no log store;
with the URL and no store address it reports a deployment bug by name.
## Output

View file

@ -1,187 +1,149 @@
//! The access token this CLI presents to the log store's gateway route.
//! The credential this CLI presents to the log store's gateway route: the
//! agent's own queue secret, read from the swarm secret store with the agent's
//! store certificate.
//!
//! Minting it is **not** implemented here. `swarm-queue-client` already owns
//! the `client_credentials` request against authelia — the basic-auth form,
//! the per-call secret read, the CA-trust builder and the error type that
//! surfaces authelia's own `error_description` — and this crate calls
//! [`swarm_queue_client::mint_token_for_blocking`] rather than writing a
//! second copy of it. A token-endpoint fix has to be findable in one place.
//! The read is `hive-agent`'s `swarm_queue` one: the same variables, the same
//! path (`swarm/agents/<agent>/queue`) and the same cert-auth role. The secret
//! goes out as `Bearer swarm-agent.<agent>.<secret>`, spelled by
//! `swarm_queue_client::agent_token`; the gateway's `swarm-logs-auth`
//! responder checks it against the store.
//!
//! What *is* here is the shape of an agent's identity, which differs from the
//! daemons that crate was written for in exactly one way: **the client id
//! arrives as a file, not as a value.** It rides in the same systemd
//! credential pair as the secret so that nothing outside `queue.nix` ever
//! spells `hive-<name>-agent` again. That is the same problem
//! `hive-agent/src/swarm_queue.rs` solves, and [`Config::from_env`] below is
//! that module's `decide` restated over this binary's inputs — including its
//! rule that a *half*-set environment is a deployment bug rather than an
//! absent integration.
//!
//! 🩸 Every credential in here is a **path**. The secret's contents are read
//! by `swarm-queue-client` at the moment of the request and are never bound to
//! a name in this crate, never logged and never rendered into an error.
use std::path::{Path, PathBuf};
//! 🩸 The secret is never logged and never rendered into an error.
use anyhow::{Context as _, Result, bail};
use swarm_queue_client::{ClientSecret, QueueConfig};
use swarm_secret_client::client::{DEFAULT_CERT_MOUNT, Settings};
use swarm_secret_client::queue::{self, AgentCredential};
use swarm_secret_client::{SecretStore, policy};
/// Variable prefix for this binary's coordinates.
///
/// `HIVE_AGENT`, deliberately the **same** prefix the harness reads, because
/// it is the same identity: an agent authenticates as its hive's agent client
/// whether the caller is the harness or a CLI the agent typed. A prefix of its
/// own would be a second set of variables carrying one fact, which is the
/// disagreement `queue.nix`'s credential pair exists to prevent.
const ENV_PREFIX: &str = "HIVE_AGENT";
/// The full URL of the LogsQL query endpoint, gateway side.
const ENV_QUERY_URL: &str = "HIVE_AGENT_LOGS_QUERY_URL";
/// This agent's name, which the store path and the role are built from.
const ENV_AGENT_NAME: &str = "HIVE_AGENT_NAME";
/// The store's address; the rest of its coordinates are `Settings`'s.
const ENV_ADDR: &str = "BAO_ADDR";
/// The store's CA bundle, which may name an undelivered credential.
const ENV_CACERT: &str = "BAO_CACERT";
/// Where this CLI queries, and who it queries as.
pub struct Config {
/// Full URL of the LogsQL query endpoint, gateway side.
///
/// The whole URL rather than a host to build one from, because this same
/// string is **also the audience** the token is minted for — the rule
/// `swarm-otel.nix` states over its own push targets ("one binding
/// because the same string is both the address requested and the audience
/// the token is minted for — two spellings present as a valid token
/// refused at the store"). Splitting it would reintroduce exactly that.
pub query_url: String,
/// Everything needed to mint a token, in the form the minter wants it.
pub queue: QueueConfig,
agent: String,
settings: Settings,
role: String,
path: String,
}
impl Config {
/// Read the configuration out of the environment the wrapper sets.
///
/// All four coordinates or none, for the reason `swarm-queue-client`'s
/// README gives and `hive-agent`'s `swarm_queue` repeats: a half-set
/// environment produces a process that looks configured and never
/// authenticates. Unlike the harness, this binary *errors* on the absent
/// case rather than degrading — a CLI whose only
/// job is the query has nothing left to do without it, and an operator
/// running it wants to be told why rather than handed an empty result
/// they will read as "no logs matched".
/// Errors rather than degrading: a CLI whose only job is the query has
/// nothing to do without it, and an empty result would read as "no logs
/// matched".
pub fn from_env() -> Result<Self> {
let var = |suffix: &str| std::env::var(format!("{ENV_PREFIX}_{suffix}")).ok();
Self::from_lookup(|k| std::env::var(k).ok())
}
let query_url = var("LOGS_QUERY_URL");
let token_endpoint = var("OIDC_TOKEN_ENDPOINT");
let client_id_file = var("OIDC_CLIENT_ID_FILE");
let client_secret_file = var("OIDC_CLIENT_SECRET_FILE");
// Outside the all-or-none group on purpose: a swarm behind a
// publicly-trusted certificate needs no extra anchor, so a CA path
// with no endpoint is meaningless rather than half-configured. Nothing in this tree sets
// it for an agent — a container already trusts the store's root, which
// `hive_c0re::meta` embeds at build time.
let ca_file = var("OIDC_CA_FILE").map(PathBuf::from);
let (query_url, token_endpoint, client_id_file, client_secret_file) = match (
query_url,
token_endpoint,
client_id_file,
client_secret_file,
) {
(Some(u), Some(t), Some(i), Some(s)) => (u, t, i, s),
(None, None, None, None) => bail!(
"this agent has no swarm log store configured: \
{ENV_PREFIX}_LOGS_QUERY_URL, {ENV_PREFIX}_OIDC_TOKEN_ENDPOINT, \
{ENV_PREFIX}_OIDC_CLIENT_ID_FILE and \
{ENV_PREFIX}_OIDC_CLIENT_SECRET_FILE are all unset"
),
_ => bail!(
"swarm log store half-configured: {ENV_PREFIX}_LOGS_QUERY_URL, \
{ENV_PREFIX}_OIDC_TOKEN_ENDPOINT, {ENV_PREFIX}_OIDC_CLIENT_ID_FILE \
and {ENV_PREFIX}_OIDC_CLIENT_SECRET_FILE are set together or not \
at all — this is a deployment bug, not a missing feature"
),
fn from_lookup(get: impl Fn(&str) -> Option<String>) -> Result<Self> {
let set = |k: &str| get(k).filter(|v| !v.is_empty());
let Some(query_url) = set(ENV_QUERY_URL) else {
bail!("this agent has no swarm log store configured: {ENV_QUERY_URL} is unset");
};
let client_id = read_client_id(Path::new(&client_id_file))?;
if set(ENV_ADDR).is_none() {
bail!(
"{ENV_QUERY_URL} is set but {ENV_ADDR} is not, so there is no secret store \
to read this agent's credential from — this is a deployment bug"
);
}
let agent =
set(ENV_AGENT_NAME).with_context(|| format!("{ENV_AGENT_NAME} is unset or empty"))?;
let settings =
Settings::from_lookup(|k| set(k).filter(|v| k != ENV_CACERT || ca_is_usable(v)))
.context("reading the swarm secret store's coordinates from the environment")?;
Ok(Self {
query_url,
queue: QueueConfig {
// 🩸 Unused by the token mint and deliberately not read from
// the environment. `QueueConfig` is the queue's connect
// config as well as its token config, and this binary needs
// only the second half — taking `HIVE_AGENT_NATS_URL` for it
// would make a broker address a prerequisite for reading
// logs, which is a coupling with no cause. Named so a reader
// of a stack trace is not left wondering.
url: String::from("unused: swarm-logs mints a token and never connects"),
token_endpoint,
client_id,
client_secret: ClientSecret::File(PathBuf::from(client_secret_file)),
ca_file,
},
role: policy::agent_object_name(&agent)?,
path: queue::agent_queue_path(&agent)?,
agent,
settings,
})
}
/// Read this agent's secret from the store and spell it as the bearer.
///
/// Read on every call, so a caller that was refused can call again and
/// present a secret re-minted since.
pub fn bearer(&self) -> Result<String> {
let runtime = tokio::runtime::Builder::new_current_thread()
.enable_all()
.build()
.context("starting the runtime for the secret store read")?;
let stored: Option<AgentCredential> = runtime.block_on(async {
SecretStore::connect(&self.settings, &self.role, DEFAULT_CERT_MOUNT)
.await
.context("logging in to the swarm secret store with this agent's certificate")?
.read_optional(&self.path)
.await
.with_context(|| format!("reading {}", self.path))
})?;
let stored = stored.with_context(|| format!("nothing is stored at {}", self.path))?;
swarm_queue_client::agent_token::format_agent_token(&self.agent, &stored.value)
.with_context(|| format!("spelling the credential stored at {}", self.path))
}
}
/// Read the OIDC client id out of the file the systemd credential landed at.
///
/// Trailing newline stripped and an empty file refused, both copied from
/// `hive-agent`'s `read_client_id`: the writing unit ends the file with a
/// newline, and an id carrying one authenticates as nobody. Where the harness
/// treats absence as the ordinary state of a hive whose credential has not
/// been published yet and carries on, this errors — see [`Config::from_env`]
/// for why a CLI has nothing to carry on with.
fn read_client_id(path: &Path) -> Result<String> {
let raw = std::fs::read_to_string(path)
.with_context(|| format!("reading the OIDC client id from {}", path.display()))?;
let id = raw.trim();
if id.is_empty() {
bail!(
"the OIDC client id at {} is empty — the queue credential has not \
been published to this hive yet",
path.display()
);
}
Ok(id.to_owned())
/// Whether the CA bundle at `path` is a file with bytes in it. `BAO_CACERT`
/// names a systemd credential that may not have been delivered; absent means
/// the container's own trust store.
fn ca_is_usable(path: &str) -> bool {
std::fs::metadata(path).is_ok_and(|m| m.len() > 0)
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn client_id_loses_its_trailing_newline() {
let dir = std::env::temp_dir().join(format!("swarm-logs-id-{}", std::process::id()));
std::fs::create_dir_all(&dir).expect("temp dir");
let path = dir.join("client-id");
std::fs::write(&path, "hive-alpha-agent\n").expect("write");
assert_eq!(read_client_id(&path).expect("read"), "hive-alpha-agent");
std::fs::remove_dir_all(&dir).expect("cleanup");
fn env(pairs: &[(&str, &str)]) -> impl Fn(&str) -> Option<String> {
let pairs: Vec<(String, String)> = pairs
.iter()
.map(|(k, v)| ((*k).to_owned(), (*v).to_owned()))
.collect();
move |k| pairs.iter().find(|(n, _)| n == k).map(|(_, v)| v.clone())
}
/// An empty credential file is the shape a hive has before its secret
/// store has published anything, and it must not become a token request
/// that authenticates as the empty string.
const FULL: &[(&str, &str)] = &[
(ENV_QUERY_URL, "https://logs.example/select/logsql/query"),
(ENV_AGENT_NAME, "atlas"),
(ENV_ADDR, "https://bao.example:8200"),
("BAO_CLIENT_CERT", "/nonexistent/cert"),
("BAO_CLIENT_KEY", "/nonexistent/key"),
];
#[test]
fn an_empty_client_id_is_refused_by_name() {
let dir = std::env::temp_dir().join(format!("swarm-logs-empty-{}", std::process::id()));
std::fs::create_dir_all(&dir).expect("temp dir");
let path = dir.join("client-id");
std::fs::write(&path, "\n").expect("write");
let err = read_client_id(&path).expect_err("empty id must not be accepted");
assert!(
err.to_string().contains("is empty"),
"error should say the id is empty, got: {err}"
);
std::fs::remove_dir_all(&dir).expect("cleanup");
fn no_query_url_says_there_is_no_log_store() {
let err = Config::from_lookup(env(&[])).err().expect("refused");
assert!(err.to_string().contains("no swarm log store"), "{err}");
}
/// A missing file names the path, because the operator's next move is to
/// look at whether the credential was delivered at all.
#[test]
fn a_missing_client_id_file_names_its_path() {
let path = Path::new("/nonexistent/swarm-logs/client-id");
let err = read_client_id(path).expect_err("missing file must not be accepted");
assert!(
err.to_string()
.contains("/nonexistent/swarm-logs/client-id"),
"error should name the path, got: {err}"
);
fn a_query_url_without_a_store_is_a_deployment_bug() {
let err = Config::from_lookup(env(&FULL[..2])).err().expect("refused");
assert!(err.to_string().contains("deployment bug"), "{err}");
}
#[test]
fn a_missing_agent_name_is_named() {
let mut pairs = FULL.to_vec();
pairs.retain(|(k, _)| *k != ENV_AGENT_NAME);
let err = Config::from_lookup(env(&pairs)).err().expect("refused");
assert!(err.to_string().contains(ENV_AGENT_NAME), "{err}");
}
/// The path and role are the agent's own, the ones its policy grants.
#[test]
fn the_read_is_the_agents_own_queue_secret() {
let cfg = Config::from_lookup(env(FULL)).expect("configured");
assert_eq!(cfg.path, "swarm/agents/atlas/queue");
assert_eq!(cfg.role, policy::agent_object_name("atlas").expect("valid"));
}
}

View file

@ -85,6 +85,7 @@ fn main() -> Result<()> {
// keep working on a machine that has no log store wired up at
// all — `markdown-docs` in particular runs in a nix check, where
// no credential exists and none should.
swarm_queue_client::install_crypto_provider();
let cfg = auth::Config::from_env()?;
let stdout = std::io::stdout();
query::run(&cfg, &logsql, limit, format, &mut stdout.lock())

View file

@ -46,41 +46,29 @@ pub fn run(
format: Format,
out: &mut impl Write,
) -> Result<()> {
// Minted per invocation, with the query URL as the audience and the
// `authelia.bearer.authz` scope. All three are load-bearing: authelia
// refuses a token carrying no audience at the authz endpoint the
// gateway's `auth_request` calls, and the audience it checks is the URL
// being requested — so the string sent here and the string requested
// below must be one binding, which is why `Config` holds exactly one.
//
// The scope has to be ASKED for, not merely registered: the agent client
// is granted `authelia.bearer.authz` by the `agentClients` entry in
// `swarm-authelia.nix`, but an OAuth2 server issues no scope the client
// never requested, and a scopeless token is refused at the authz endpoint
// exactly as an unauthenticated one is — a bare nginx 401 with nothing in
// it that names the scope. `swarm-otel.nix` records the same failure
// against the collector's client, on the same string.
let token = swarm_queue_client::mint_token_for_blocking(
&cfg.queue,
Some(&cfg.query_url),
Some("authelia.bearer.authz"),
)
.map_err(|e| anyhow::anyhow!("{}", swarm_queue_client::chain(&e)))
.context("minting an access token for the swarm log store")?;
let http = build_http_client(cfg)?;
let http = build_http_client()?;
let mut form = vec![("query", logsql.to_owned())];
if let Some(limit) = limit {
form.push(("limit", limit.to_string()));
}
let send = || -> Result<reqwest::blocking::Response> {
let bearer = cfg
.bearer()
.context("reading this agent's credential for the swarm log store")?;
http.post(&cfg.query_url)
.bearer_auth(bearer)
.form(&form)
.send()
.with_context(|| format!("querying the swarm log store at {}", cfg.query_url))
};
let response = http
.post(&cfg.query_url)
.bearer_auth(&token)
.form(&form)
.send()
.with_context(|| format!("querying the swarm log store at {}", cfg.query_url))?;
let mut response = send()?;
// A 401 can be a secret re-minted between the store read and the check,
// so it is read again and the query sent once more.
if response.status() == reqwest::StatusCode::UNAUTHORIZED {
response = send()?;
}
let status = response.status();
if !status.is_success() {
@ -140,37 +128,13 @@ fn render(line: &str, format: Format) -> String {
}
}
/// The HTTP client for the query itself.
/// The HTTP client for the query itself, trusting the container's own roots.
///
/// Separate from the one `swarm-queue-client` builds for the token endpoint
/// because they are two hosts — `auth.<swarm>` and `logs.<swarm>` — and the
/// CA-trust knob is declared per endpoint. Same optional anchor, same reason:
/// without it a swarm using its own CA fails at TLS with
/// `invalid peer certificate: UnknownIssuer`, having never looked at the
/// anchor sitting on the host.
///
/// No timeout, deliberately, where the token client has one: a query over a
/// long retention window legitimately takes longer than any number that would
/// be right for a token request, and a client-side cut would truncate a
/// correct answer into a shorter one with no marker saying so.
fn build_http_client(cfg: &Config) -> Result<reqwest::blocking::Client> {
let mut builder = reqwest::blocking::Client::builder();
if let Some(path) = &cfg.queue.ca_file {
let pem = std::fs::read(path).with_context(|| {
format!(
"reading the log-store CA certificate from {}",
path.display()
)
})?;
let cert = reqwest::Certificate::from_pem(&pem).with_context(|| {
format!(
"parsing the log-store CA certificate from {} as PEM",
path.display()
)
})?;
builder = builder.add_root_certificate(cert);
}
builder
/// No timeout, deliberately: a query over a long retention window legitimately
/// takes long, and a client-side cut would truncate a correct answer into a
/// shorter one with no marker saying so.
fn build_http_client() -> Result<reqwest::blocking::Client> {
reqwest::blocking::Client::builder()
.build()
.context("building the log-store HTTP client")
}

View file

@ -23,17 +23,16 @@ serde.workspace = true
serde_json.workspace = true
# The jti digest: base32hex(sha256(claims)) over every JWT this crate signs.
sha2.workspace = true
# Comparing an agent's presented secret with the stored one.
subtle.workspace = true
# For `status::BUCKET` and `notices::STREAM` - the subjects a hive may
# publish to are derived from these names, and every end that touches them
# must agree on the same one. Deliberately WITHOUT the `kv` feature: this
# crate derives subject strings, it never opens the bucket. `notices`
# is name-only too (no `jetstream`/`kv` surface), same reason.
swarm-queue-client = { workspace = true, features = ["notices"] }
# The agent-token spelling the agent also uses, and the read of the stored
# credential it is checked against.
# The store reads: the callout keys, its own OIDC secret, and the agent
# credentials `swarm-agent-verify` checks a presented token against.
swarm-secret-client.workspace = true
swarm-agent-verify.workspace = true
tokio.workspace = true
tracing.workspace = true
tracing-subscriber.workspace = true

View file

@ -1,22 +1,15 @@
//! Verifying an agent's own credential, presented as `auth_token`.
//! Classifying a presented `auth_token`, and the grant for an agent's own.
//!
//! The spelling is `swarm_queue_client::agent_token`'s, shared with the agent
//! that presents it: a prefix, the agent's name, and its secret. The name is only a
//! claim. It becomes an identity when the secret equals the one stored at
//! `swarm/agents/<agent>/queue`, and nothing in this module grants on the name
//! alone: every path that does not reach that comparison, and every lookup
//! that fails, is a denial.
//! that presents it. Whether its secret is the stored one is
//! [`swarm_agent_verify::verify`]'s answer; nothing here grants on the name
//! alone.
//!
//! A token without the prefix is not this module's: it goes to introspection
//! unchanged.
use std::future::Future;
use anyhow::Context;
use subtle::ConstantTimeEq;
use swarm_agent_verify::CredentialSource;
use swarm_queue_client::agent_token::{AgentToken, Malformed, parse_agent_token};
use swarm_secret_client::client::{DEFAULT_CERT_MOUNT, Settings};
use swarm_secret_client::queue::{self, AgentCredential};
use crate::policy::{Permissions, Policy};
@ -41,112 +34,18 @@ pub fn classify(token: &str) -> Presented<'_> {
}
}
/// Where the stored credential for an agent comes from.
pub trait CredentialSource {
/// The object at `agent`'s queue path, `None` when nothing is stored
/// there, or an error when the store could not answer.
fn lookup(
&self,
agent: &str,
) -> impl Future<Output = anyhow::Result<Option<AgentCredential>>> + Send;
}
/// The swarm secret store, reached with this responder's own certificate.
pub struct Store {
settings: Settings,
cert_role: String,
}
impl Store {
/// The store at `settings`, logged in to as `cert_role`.
pub fn new(settings: Settings, cert_role: String) -> Self {
Self {
settings,
cert_role,
}
}
}
impl CredentialSource for Store {
/// Logs in per lookup. An agent connects once per boot and per reconnect,
/// so a login each time costs little, and it leaves no token to expire
/// inside a long-lived process.
async fn lookup(&self, agent: &str) -> anyhow::Result<Option<AgentCredential>> {
let path = queue::agent_queue_path(agent)?;
let store = swarm_secret_client::SecretStore::connect(
&self.settings,
&self.cert_role,
DEFAULT_CERT_MOUNT,
)
.await
.context("logging in to the secret store")?;
store
.read_optional(&path)
.await
.with_context(|| format!("reading {path}"))
}
}
/// Whether `token`'s secret is the one stored for the agent it names.
/// The grant for an agent token, or `None` for a denial.
///
/// The lookup is bounded by introspection's budget, under the server's
/// The store lookup is bounded by introspection's budget, under the server's
/// `authorization.timeout`. The callout loop answers one request at a time, so
/// a store that does not answer holds every request behind it, OIDC ones
/// included, for at most that long, and this then denies.
async fn verify(source: &impl CredentialSource, token: &AgentToken<'_>) -> bool {
let agent = token.agent;
let stored = match tokio::time::timeout(
crate::introspect::INTROSPECTION_TIMEOUT,
source.lookup(agent),
)
.await
{
Ok(Ok(Some(stored))) => stored,
Ok(Ok(None)) => {
tracing::warn!(
agent,
"no queue credential is stored for this agent; denying"
);
return false;
}
Ok(Err(e)) => {
tracing::warn!(
agent,
error = format!("{e:#}"),
"credential lookup failed; denying"
);
return false;
}
Err(_) => {
tracing::warn!(agent, "credential lookup timed out; denying");
return false;
}
};
if stored.agent != agent {
tracing::warn!(
agent,
stored_agent = %stored.agent,
"the stored credential names a different agent than its path; denying"
);
return false;
}
// Constant time, so the time to refuse says nothing about how much of the
// secret was right. A length mismatch returns early; the length is not
// secret, every minted secret has the same one.
stored
.value
.as_bytes()
.ct_eq(token.secret.as_bytes())
.into()
}
/// The grant for an agent token, or `None` for a denial.
pub async fn authorize(
policy: &Policy,
source: &impl CredentialSource,
token: &AgentToken<'_>,
) -> Option<Permissions> {
if !verify(source, token).await {
if !swarm_agent_verify::verify(source, token, crate::introspect::INTROSPECTION_TIMEOUT).await {
return None;
}
let permissions = policy.agent_token_permissions(token.agent);
@ -162,47 +61,28 @@ pub async fn authorize(
#[cfg(test)]
mod tests {
use super::*;
use swarm_secret_client::queue::AgentCredential;
/// A store holding at most one credential, or failing outright.
/// A store holding atlas's credential, or one that never answers.
/// The store's own denials are `swarm_agent_verify`'s tests.
enum Fake {
/// `credential` is stored at `path_agent`'s queue path.
Holds {
path_agent: &'static str,
credential: AgentCredential,
},
Fails,
/// A store that accepts the request and never answers.
Atlas,
Hangs,
}
impl CredentialSource for Fake {
async fn lookup(&self, agent: &str) -> anyhow::Result<Option<AgentCredential>> {
match self {
Self::Holds {
path_agent,
credential,
} => Ok((agent == *path_agent).then(|| credential.clone())),
Self::Fails => anyhow::bail!("store unreachable"),
Self::Atlas => Ok((agent == "atlas").then(|| AgentCredential {
value: "Ab9_-zSECRET".to_owned(),
agent: "atlas".to_owned(),
minted_at: None,
})),
Self::Hangs => std::future::pending().await,
}
}
}
fn stored(path_agent: &'static str, named: &str) -> Fake {
Fake::Holds {
path_agent,
credential: AgentCredential {
value: "Ab9_-zSECRET".to_owned(),
agent: named.to_owned(),
minted_at: None,
},
}
}
fn atlas() -> Fake {
stored("atlas", "atlas")
}
fn policy() -> Policy {
Policy::new(
"hive-".to_owned(),
@ -229,7 +109,7 @@ mod tests {
#[tokio::test]
async fn a_valid_agent_token_is_granted_exactly_its_own_subjects() {
let g = grant(&atlas(), "swarm-agent.atlas.Ab9_-zSECRET")
let g = grant(&Fake::Atlas, "swarm-agent.atlas.Ab9_-zSECRET")
.await
.expect("granted");
assert_eq!(
@ -245,42 +125,7 @@ mod tests {
#[tokio::test]
async fn a_wrong_secret_is_denied() {
assert!(
grant(&atlas(), "swarm-agent.atlas.Ab9_-zSECREX")
.await
.is_none()
);
assert!(
grant(&atlas(), "swarm-agent.atlas.Ab9_-zSECRE")
.await
.is_none()
);
}
/// The secret check is what stops one agent claiming another's name: the
/// right secret under someone else's name finds that agent's credential,
/// or none.
#[tokio::test]
async fn another_agents_name_with_this_agents_secret_is_denied() {
assert!(
grant(&atlas(), "swarm-agent.argus.Ab9_-zSECRET")
.await
.is_none()
);
}
#[tokio::test]
async fn an_agent_with_nothing_stored_is_denied() {
assert!(
grant(&stored("argus", "argus"), "swarm-agent.atlas.Ab9_-zSECRET")
.await
.is_none()
);
}
#[tokio::test]
async fn a_failed_lookup_is_denied() {
assert!(
grant(&Fake::Fails, "swarm-agent.atlas.Ab9_-zSECRET")
grant(&Fake::Atlas, "swarm-agent.atlas.Ab9_-zSECREX")
.await
.is_none()
);
@ -305,15 +150,6 @@ mod tests {
);
}
#[tokio::test]
async fn a_stored_object_naming_another_agent_is_denied() {
assert!(
grant(&stored("argus", "atlas"), "swarm-agent.argus.Ab9_-zSECRET")
.await
.is_none()
);
}
#[tokio::test]
async fn a_verified_agent_with_no_subjects_configured_is_denied() {
let bare = Policy::new(
@ -327,7 +163,7 @@ mod tests {
let Presented::Agent(token) = classify("swarm-agent.atlas.Ab9_-zSECRET") else {
panic!("an agent token");
};
assert!(authorize(&bare, &atlas(), &token).await.is_none());
assert!(authorize(&bare, &Fake::Atlas, &token).await.is_none());
}
/// Everything without the prefix reaches introspection as it arrived.

View file

@ -242,7 +242,7 @@ async fn serve(args: &Args) -> anyhow::Result<()> {
.context("logging in to the secret store for the callout keys and the OIDC client secret")?;
let (callout_xkey, issuer, user_seed) = load_callout_keys(&keys).await?;
let client_secret = load_client_secret(&keys, args).await?;
let store = agent_token::Store::new(settings, args.store_cert_role.clone());
let store = swarm_agent_verify::Store::new(settings, args.store_cert_role.clone());
let client = async_nats::ConnectOptions::with_nkey(user_seed)
.name("swarm-nats-auth")
.connect(&args.nats_url)