hyperhive/nix/host-modules/swarm-ca.nix
atlas cc74a2c1f6 feat(nix): make the swarm CA hierarchy opt-in
A swarm's services and its hives can live on different hosts, and a host
has no way to tell whether it is the one holding the root — so setting
the swarm CA up is an operator action, not something a host infers.
`swarm.ca.autoConfigure` is off by default and is the only thing that
issues a hive sub-CA, which is also the only case that *can*: signing
one needs the root's private key, and that key is deliberately not on a
hive host otherwise.

With the flag off, a hive self-signs its CA exactly as it always has.
That keeps a plain hive working out of the box; what it gives up is
membership of a swarm's trust hierarchy, which is the right thing to
give up for a hive nobody has federated.

Replaces a default derived from `swarm.peers == { }`. That read "no
peers declared" as "everything is local", which is not the same claim —
a hive can belong to a swarm it has not declared yet, or to one whose
services run elsewhere entirely.
2026-08-05 15:57:50 +02:00

163 lines
6.7 KiB
Nix

# The swarm root CA: the anchor a whole swarm shares, and the issuer of
# each hive's own CA (which is where it gets used — see ./hive-tls.nix).
#
# Why a hierarchy at all: cross-hive trust is hand-pinned today
# (`swarm.peers.<d>.caCert`), so every hive must name every peer — O(n²)
# configuration that a new hive can only join by editing all the others.
# One root makes it O(1): trust the root once and every present *and
# future* peer validates.
#
# Two provisioning modes, ONE structure — what differs is who puts the
# artifacts on disk, never what the artifacts are:
#
# - autoconfigured (the default while the hive declares no peers): the
# unit below generates the root here on first boot.
# - operator-provided (multi-host): the operator installs the root cert
# — and, on a host that does not hold the root key, the hive CA too —
# into the state dirs, and this unit does nothing. Splitting a
# single-host swarm across hosts is then a matter of moving the
# service dirs and turning `autoConfigure` off, not of a second code
# path that has to be kept agreeing with the first.
#
# The root KEY is the reason this is a runtime file and not a nix option:
# the store is world-readable and content-addressed, so a key committed
# to a flake is a key *published* to every consumer of that flake. The
# root CERT has the opposite property, and is already distributed at
# build time via `swarm.peers.<d>.caCert`. That asymmetry is what makes
# the layering work at all.
{
lib,
config,
pkgs,
...
}:
let
cfg = config.services.hyperhive.swarm.ca;
hyperhiveCfg = config.services.hyperhive;
# The subject CN is a label for a human reading a chain, not an
# identity anything authenticates against. Fall through swarm name →
# hive domain → a constant so a hive that has set neither still
# evaluates; a missing `domain` is reported by its own assertion in
# hive-network.nix, and shouldn't also surface here as a null.
swarmLabel =
if hyperhiveCfg.swarm.name != null then
hyperhiveCfg.swarm.name
else if hyperhiveCfg.domain != null then
hyperhiveCfg.domain
else
"hyperhive";
in
{
options.services.hyperhive.swarm.ca = {
autoConfigure = lib.mkOption {
type = lib.types.bool;
default = false;
example = true;
description = ''
Run the whole swarm CA on this one host: generate the swarm
root when it is missing, and issue this hive's CA under it.
**Off by default, deliberately.** A swarm's services and its
hives can live on different hosts, and this host has no way to
tell whether it is the one holding the root so the swarm CA
is something an operator sets up, not something a host decides
it is. Turn this on for an all-on-one-host deployment (dev
boxes, single-hive swarms) and get the hierarchy for free.
With it off, both artifacts are operator-provided: the root
under `stateDir`, and this hive's CA under
`services.hyperhive.tls.stateDir`. A hive given neither keeps
the self-signed CA it has always had it simply isn't part of
a swarm's trust hierarchy, which is the correct outcome for a
hive nobody has federated yet.
'';
};
stateDir = lib.mkOption {
type = lib.types.str;
default = "/var/lib/swarm-ca";
description = ''
Host directory holding the swarm root CA: `root.pem` (the
anchor, safe to distribute this is what a peer hive is
pointed at via `swarm.peers.<d>.caCert`) and `root-key.pem`
(0600, the one file that must never reach the nix store or
another host). The directory itself is 0700: nothing reads
out of it but the hive CA issuance in `hive-tls.nix`.
Moving the swarm CA to its own host is a matter of moving this
directory and setting `autoConfigure = false` here.
'';
};
validityDays = lib.mkOption {
type = lib.types.int;
default = 10950;
description = ''
Validity window of the swarm root CA in days (default ~30y).
Deliberately longer than `services.hyperhive.tls.caValidityDays`:
the root must outlive the hive CAs it issues, or those chains
expire out from under hives that are still perfectly happy with
their own intermediate. Rotating a root is the one operation in
this system with no partial-failure mode it invalidates every
peer at once, paced by the slowest peer's rebuild so it is
never automatic and this window is meant to be uneventful.
'';
};
};
config = lib.mkIf (hyperhiveCfg.enable && cfg.autoConfigure) {
systemd.services.swarm-ca = {
description = "Generate the swarm root CA when absent";
wantedBy = [ "multi-user.target" ];
path = [ pkgs.openssl ];
serviceConfig = {
Type = "oneshot";
RemainAfterExit = true;
UMask = "0077";
# Pin the journal identity (else it's the `script` store-path wrapper).
SyslogIdentifier = "swarm-ca";
};
script = ''
set -euo pipefail
d=${lib.escapeShellArg cfg.stateDir}
install -d -m 0700 "$d"
root="$d/root.pem"
rootk="$d/root-key.pem"
# Note the asymmetry with the hive CA in hive-tls.nix, which
# regenerates itself once expired: a root is never replaced
# automatically, not even an expired one. Consumers hold this
# cert, so replacing it is a swarm-wide flag day that wants an
# operator running it deliberately, with both roots trusted
# across the overlap.
if [ -s "$root" ] && [ -s "$rootk" ]; then
echo "swarm root CA already present at $root leaving it alone"
exit 0
fi
# Half a root is not a root. Generating a fresh key beside an
# already-distributed cert (or the reverse) leaves every
# consumer trusting an anchor that no longer signs anything
# and it would look like it worked.
if [ -e "$root" ] || [ -e "$rootk" ]; then
echo "swarm root CA half-provisioned ($root / $rootk) refusing to generate over it" >&2
exit 1
fi
echo "generating swarm root CA at $root"
# pathlen:1 the root signs hive CAs, which sign leaves. One
# intermediate below the root and no deeper.
openssl req -x509 -newkey rsa:4096 -nodes -sha256 \
-days ${toString cfg.validityDays} \
-keyout "$rootk" -out "$root" \
-subj "/CN=swarm-ca ${swarmLabel}" \
-addext "basicConstraints=critical,CA:TRUE,pathlen:1" \
-addext "keyUsage=critical,keyCertSign,cRLSign"
chmod 0600 "$rootk"
chmod 0644 "$root"
'';
};
};
}