bao: mint the controller's leaf, and point the controller at it

glue-bao-tls.nix signs a third leaf. It is minted whether or not a
controller runs here, because the case it serves is the one where it
does not: a controller elsewhere needs a leaf from this CA and cannot
sign one, so issuing it here turns "obtain a certificate out of band"
into "copy this file".

glue-controller-bao-identity.nix holds the pairing and nothing else --
which paths this host's controller reads. Gated on the leaf existing
rather than on deploy.bao.enable, so a controller on the store's host
and one three networks away with an out-of-band leaf get the same
wiring; gating on the store would have made the co-located case the
only supported shape. The directory comes from deploy.bao.clientCertFile
rather than repeating glue-bao-tls.nix's literal, so moving the PKI
moves both.

module-eval gains three cases and two fixtures, because nothing asserted
the PKI script before: an earlier commit added a leaf to that rendered
unit and left the derivation unchanged. The fixture's CN is deliberately
a value no default could supply, so "the role and the leaf both carry
it" says they read one option rather than that both happen to say
swarm-controller.

Gates: 62 module properties hold (59 before, plus these three), on a
derivation hash that actually moved -- this suite is a cache hit when
only fixtures change, so an unchanged hash would have meant the cases
never ran. nix fmt clean, all three scripts/check-*.sh exit 0.
This commit is contained in:
atlas 2026-09-07 22:18:32 +02:00
commit c18aee8f74
4 changed files with 112 additions and 2 deletions

View file

@ -24,6 +24,7 @@
./hive-tls.nix
./otel.nix
./glue-bao-tls.nix
./glue-controller-bao-identity.nix
./glue-matrix-bao-token.nix
./swarm-authelia.nix
./swarm-bao.nix

View file

@ -16,8 +16,8 @@
# ⚠️ Not the hive CA and not the swarm CA. The store will eventually
# distribute both, and an authority you must already hold a certificate from
# cannot be one the store hands out — reach the store to get the CA material,
# need a cert from that CA to reach the store. This CA signs exactly two
# things and distributes nothing, so it cannot enter that cycle.
# need a cert from that CA to reach the store. This CA signs a fixed, short
# list of leaves and distributes nothing, so it cannot enter that cycle.
#
# ⚠️ Files like this are the only place a `deploy.<foo>` value may derive from
# a `deploy.<bar>.enable`. Everywhere else that is forbidden. The exception
@ -126,6 +126,17 @@ in
${lib.escapeShellArg cfg.domain} ${lib.escapeShellArg "DNS:${cfg.domain}"} serverAuth
[ -s ${pkiDir}/client.pem ] || ${signLeaf} ${pkiDir} client \
${lib.escapeShellArg clientCn} "" clientAuth
# Minted whether or not a controller runs here, because the case it
# serves is the one where it does not: a controller elsewhere needs a
# leaf from this CA and has no way to sign one. Issuing it here turns
# "obtain a certificate out of band" into "copy this file".
#
# Its own CN rather than the reader's above: the controller's policy
# lets it create roles for every hive, and the reader's leaf carries
# this hive's name.
[ -s ${pkiDir}/controller.pem ] || ${signLeaf} ${pkiDir} controller \
${lib.escapeShellArg deployCfg.bao.controllerCommonName} "" clientAuth
'';
};
};

View file

@ -0,0 +1,41 @@
# Glue: point the swarm controller at the bao leaf minted for it.
#
# ONE PAIRING PER FILE — controller ← bao, and nothing else. Deleting this
# leaves a controller that takes operator-provided certificate paths, which is
# what every deployment that does not mint its own already does.
#
# ⚠️ The minting is NOT here. ./glue-bao-tls.nix holds the CA and signs the
# leaf, because the thing that owns a private key owns issuing from it; a
# second signer would duplicate that file's `openssl` helper to no benefit.
# What belongs here is the pairing: which paths this host's controller reads.
#
# ⚠️ Gated on the leaf existing, not on the store being enabled. A controller
# on the store's own host is one deployment; a controller three networks away
# holding a leaf issued out of band is another, and both want the same wiring.
# `deploy.bao.enable` would have made the first one the only supported shape.
#
# Everything is `mkDefault`. An operator naming their own paths wins.
{
lib,
config,
...
}:
let
hyperhiveCfg = config.services.hyperhive;
deployCfg = hyperhiveCfg.deploy;
baoDeploy = deployCfg.bao;
# Where ./glue-bao-tls.nix puts the leaves. Derived from the reader's own
# path rather than repeating that file's directory literal: an operator who
# moves the PKI moves both, and the two cannot drift apart.
haveMintedPki = baoDeploy.clientCertFile != null;
pkiDir = if haveMintedPki then builtins.dirOf baoDeploy.clientCertFile else null;
in
{
config = lib.mkIf (hyperhiveCfg.enable && deployCfg.swarm-controller.enable && haveMintedPki) {
services.hyperhive.deploy.swarm-controller = {
baoClientCertFile = lib.mkDefault "${pkiDir}/controller.pem";
baoClientKeyFile = lib.mkDefault "${pkiDir}/controller-key.pem";
};
};
}

View file

@ -254,6 +254,20 @@ let
baoGrantNoStore = hive {
deploy.bao.bootstrapTokenFile = "/run/secrets/bao-bootstrap.token";
};
# Store and controller on one machine, with a CN no default could supply.
# The odd value is what lets the case below tell "both ends read the same
# option" from "both ends happen to say swarm-controller".
baoControllerHere = hive {
deploy.bao.enable = true;
deploy.bao.bootstrapTokenFile = "/run/secrets/bao-bootstrap.token";
deploy.bao.controllerCommonName = "cn-marker-not-a-default";
deploy.swarm-controller.enable = true;
};
# The controller with no store, which is every spread deployment. Nothing
# mints here, so the pairing must leave the paths unset rather than name
# files this host will never have.
controllerNoStore = hive { deploy.swarm-controller.enable = true; };
# The store and the token, with no CA to trust. `mkForce` because the PKI
# glue supplies one by default here — this is the deployment that brings its
# own certificates and has not named the authority yet, and it separates
@ -527,6 +541,49 @@ let
&& lib.hasInfix "auth/cert/certs/swarm-controller" s
&& lib.hasInfix "/var/lib/swarm-bao-tls/client-ca.pem" s;
}
{
# Nothing asserted the PKI script before this, so a third leaf could be
# added to it and every case still passed — measured, not assumed: the
# commit that added one left `module-eval`'s derivation unchanged.
name = "the store mints a leaf for the controller, and the controller is pointed at it";
ok =
let
m = baoControllerHere;
pki = m.systemd.services.swarm-bao-pki.script;
in
lib.hasInfix "controller.pem" pki
&&
m.services.hyperhive.deploy.swarm-controller.baoClientCertFile
== "/var/lib/swarm-bao-pki/controller.pem"
&&
m.services.hyperhive.deploy.swarm-controller.baoClientKeyFile
== "/var/lib/swarm-bao-pki/controller-key.pem";
}
{
# What makes the one above mean something: a controller with no store
# has nothing to be pointed at. Naming a path here would be a file this
# host never gets, which fails at a TLS handshake rather than at eval.
name = "a controller on a host with no store is left without certificate paths";
ok =
let
c = controllerNoStore.services.hyperhive.deploy.swarm-controller;
in
c.baoClientCertFile == null && c.baoClientKeyFile == null;
}
{
# The CN is an interface between two files: the store writes a role that
# matches it, the PKI mints a leaf that carries it. They read one option,
# and this is what says so — the fixture's value cannot come from a
# default, so matching it in both places is not a coincidence.
name = "the cert-auth role and the minted leaf take their subject from one option";
ok =
let
m = baoControllerHere;
role = m.containers.swarm-bao.config.systemd.services.swarm-bao-controller-policy.script;
pki = m.systemd.services.swarm-bao-pki.script;
in
lib.hasInfix "cn-marker-not-a-default" role && lib.hasInfix "cn-marker-not-a-default" pki;
}
{
# The arm that makes the one above mean something. A role's trust anchor
# is the CA, so with none named there is nothing to write — and the