f778122f5a
A swarm runs one homeserver and a homeserver has one appservice sender account, so "mint it once" is a property of the thing being minted rather than something a lock has to enforce. That is what makes this account the one to move first: no trigger route, no controller change and no agent list — a boot-time oneshot beside tuwunel is the whole mechanism. `swarm-matrix-minter` runs inside `containers.hive-matrix`, which already holds the appservice token: the rendered registration is bound in read-only because that is how tuwunel is handed it. What the container lacked was an identity of its own, so this adds one — a leaf from the store's CA with a grant of exactly one path, not the hive's leaf, which reads every secret in the store. Both ends of the credential ship here. The minter reads the path it publishes to before it touches the homeserver, and returning on a non-empty read IS the "only once"; `hive-c0re`'s `ensure_hive_user` reads the same path, authenticating with the hive name already in `HYPERHIVE_HIVE_NAME`. The existing mint-then-`M_USER_IN_USE`-login ladder stays as the fallback for a store that is empty, unconfigured or unreachable, which is every swarm deployed before this — so nothing needs backfilling and nothing breaks if the rest of the sequence never lands. The credential is not an admin credential, and is not named like one. It is the access token of the appservice registration's own `sender_localpart` — `@hive:<server_name>`, an account the homeserver creates for itself when it loads the registration. The store path is `swarm/services/matrix/sender-token`, the host path is `matrix/access-token`, and the homeserver no longer runs an `admin_execute` promotion for that account at boot. Everything the hive provisions with it — the Space, the chat room, their hierarchy and join rules, the invites — rides on being the creator of those rooms at power level 100, not on homeserver admin; there is no Synapse admin API here to need, tuwunel has none. Two operations do need an admin *sender* and therefore stop working: `hivectl matrix promote-user` and `hivectl matrix reset-password`, both `!admin …` messages into `#admins:<server>`, plus the password-reset recovery path that an agent with a lost password file falls back to. They are swarm-level operations and are left failing loudly rather than served by an over-privileged token every other call site would also carry. The sweep's own admin-rights check and self-repair go with them: an account that is deliberately not an admin has nothing to check. `ephemeral = false` stays, and hive root can still read the container's filesystem. Accepted: what this buys is identity separation — no hive *process* holds or reads the appservice token — not physical isolation. Refs #4345
44 lines
2 KiB
Markdown
44 lines
2 KiB
Markdown
# swarm-matrix-minter
|
|
|
|
A boot-time oneshot that runs **inside `containers.hive-matrix`**, beside the
|
|
homeserver, and puts the `@hive:` account's access token into the swarm's secret
|
|
store under an identity of its own.
|
|
|
|
## Why it lives in the matrix container
|
|
|
|
The credential it mints is authorised by the appservice `as_token`, and the
|
|
container already holds that: `nix/host-modules/hive-matrix.nix` bind-mounts the
|
|
rendered appservice registration into it read-only, because that is how tuwunel
|
|
itself is handed the registration. Minting anywhere else would mean copying the
|
|
`as_token` to a second holder — and the point of this component is that the hive
|
|
stops being one.
|
|
|
|
It is not the swarm controller for the same reason, plus a structural one: a
|
|
homeserver has exactly **one** `@hive:` account and a swarm runs one
|
|
homeserver, so "mint it once" needs no lock, no lease and no trigger surface — it
|
|
is a property of the thing being minted.
|
|
|
|
## Idempotency
|
|
|
|
The **store** is the key, not the homeserver. A run reads
|
|
`swarm/services/matrix/hive-access-token` first and returns without touching the
|
|
homeserver when something is already there. Only an empty path reaches the mint
|
|
ladder:
|
|
|
|
1. `POST /_matrix/client/v3/register` with `"type": "m.login.application_service"`
|
|
— one round trip, no UIAA.
|
|
2. `M_USER_IN_USE` (the expected arm on a homeserver that has already loaded the
|
|
registration, since the account is the appservice's own `sender_localpart`) →
|
|
`POST /_matrix/client/v3/login` as the appservice, same pinned `device_id`, so
|
|
the old device is replaced rather than duplicated.
|
|
3. Write the result to the store.
|
|
|
|
A crash between the homeserver call and the store write is recoverable: the next
|
|
run takes arm 2.
|
|
|
|
## 🩸 A secret is a path, never a value
|
|
|
|
Nothing here logs, prints or interpolates a token. The mint ladder's errors are
|
|
built from the homeserver's _status_ and its `errcode`, never its body, because a
|
|
`/login` response body is an access token. The one identifier this binary logs is
|
|
the store path it wrote.
|