fix: enforce production transport policy

Every PostgreSQL connection this repository opened was plaintext, in every
environment: no connection site set `ssl` at all, so credentials, session
tokens, and row data crossed the network in the clear and nothing in
configuration, startup, or `pnpm doctor` said so. The public identity URL
boundary was also looser than it appeared.

`core/src/database/transport.ts` is now the one typed policy. It has exactly two
modes: `verify-full` (TLS required, certificate chain verified, and server
identity verified against the configured host — the production default) and
`disable` (the development/test default, refused outright in production). There
is deliberately no encryption-only `require` mode, no chain-only `verify-ca`,
and no `rejectUnauthorized: false` escape hatch anywhere in the configuration
surface; every loose or aliased spelling is rejected in both environments, so a
typo never quietly becomes a weaker mode. Production has no plaintext exception
at all, not even for a loopback database — a connection string cannot prove the
database is process-local.

Verification is certificate *and* hostname. The policy supplies an explicit
`checkServerIdentity` pinned to the host parsed from DATABASE_URL, because the
pinned driver sets TLS `servername` only for non-IP hosts and Node performs no
identity check when it is unset — so `rejectUnauthorized: true` alone accepts a
certificate from any host the trusted CA ever signed when connecting by IP.
Both halves are proven against a real TLS-enabled PostgreSQL, including that the
same connection without the pinned callback succeeds. Every CA, certificate, and
key is generated at test runtime and deleted afterwards.

A DATABASE_URL may no longer carry its own TLS parameters. The pinned driver
merges parsed connection-string values over explicitly passed options, so a
`?sslmode=disable` in the URL would silently replace the policy's verified `ssl`
object; that now fails with a stable code, reported by both `pnpm doctor` and
the migration CLIs rather than as "failed unexpectedly".

One shared environment reader keeps `process.env` out of Core's reusable
constructors, and every executable that opens a connection resolves the policy
identically: the API (validated before any pool, Better Auth instance, module
runtime, or listener exists), both migration commands, the bootstrap CLI, and
Doctor's own connection. Doctor reports it as a blocking failure, never a
warning, naming only the effective mode.

Separately, the identity base URL and every trusted origin must now be exactly
one canonical origin — the WHATWG origin serialization must equal the supplied
string — so a path, trailing slash, credentials, explicit default port, or
non-canonical case is rejected rather than normalized into something that no
longer matches what a browser sends. Production's loopback rejection now
classifies the address instead of comparing against three literals, covering
`.localhost`, all of 127.0.0.0/8, 0.0.0.0, `::`, expanded and IPv4-mapped
forms. This is a configuration-shape guarantee only: it does not resolve DNS or
prove reachability, and TLS protects the connection, not the exposure.

Co-Authored-By: Claude Opus 5 <[email protected]>
This commit is contained in:
ekklesia
2026-07-26 17:03:52 -05:00
co-authored by Claude Opus 5
parent 7a0f624fc3
commit 6b6b77c105
34 changed files with 2035 additions and 52 deletions
+60 -3
View File
@@ -49,12 +49,14 @@ No real values are shown here — configure per environment and never commit the
| Variable | Requirement |
| --- | --- |
| `DATABASE_URL` | An absolute `postgres://` or `postgresql://` URL to an already-migrated database. |
| `DATABASE_URL` | An absolute `postgres://` or `postgresql://` URL to an already-migrated database. It must **not** carry a TLS parameter (`sslmode`, `ssl`, `sslrootcert`, …) — see [Database transport (TLS)](#database-transport-tls). |
| `EKKLESIA_DATABASE_TLS_MODE` | Optional. Exactly `"verify-full"` or `"disable"`. Defaults to `verify-full` in production and `disable` in development/test. Production refuses `disable`. |
| `EKKLESIA_DATABASE_CA` | Optional. PEM text of a private certificate authority to trust in addition to the system store, for a database using an internal CA. |
| `EKKLESIA_API_HOST` | Optional. Listen address. Defaults to loopback (`127.0.0.1`). An explicit non-loopback address is permitted for future deployment. |
| `EKKLESIA_API_PORT` | Optional. Listen port, `0`–`65535` (`0` requests an OS-assigned port). Defaults to `3000`. |
| `EKKLESIA_IDENTITY_SECRET` | At least 32 characters, high entropy (for example `openssl rand -base64 32`). Same requirement as Phase 4A. |
| `EKKLESIA_IDENTITY_BASE_URL` | An absolute `http`/`https` URL. `https` is required in production; loopback hosts are rejected in production. |
| `EKKLESIA_IDENTITY_TRUSTED_ORIGINS` | Comma-separated list of absolute origins (no path/query/fragment, no wildcards). Used for both Better Auth's own CSRF/origin checks and the API's CORS policy — there is only one trusted-origin list. |
| `EKKLESIA_IDENTITY_BASE_URL` | Exactly one canonical `http`/`https` origin — scheme, host, and non-default port only, with no path, trailing slash, query, fragment, credentials, or explicit default port. `https` is required in production; loopback and unspecified hosts are rejected in production. |
| `EKKLESIA_IDENTITY_TRUSTED_ORIGINS` | Comma-separated list of canonical origins, validated identically to the base URL (no wildcards). Used for both Better Auth's own CSRF/origin checks and the API's CORS policy — there is only one trusted-origin list. |
| `EKKLESIA_PUBLIC_SIGNUP_ENABLED` | Optional, **development/test only**. Exactly `"true"` enables public sign-up when `NODE_ENV=development`; defaults to disabled, and a production configuration rejects `"true"` outright. Omitted, misspelled, or any other value never enables it. See [Public registration policy](#public-registration-policy). |
| `EKKLESIA_CLIENT_ADDRESS_MODE` | Optional. `"direct"` (default when unset) or `"trusted-proxy"`. See [Client-address resolution](#client-address-resolution). |
| `EKKLESIA_CLIENT_ADDRESS_HEADER` | Required only when `EKKLESIA_CLIENT_ADDRESS_MODE=trusted-proxy`. The one forwarded-header name the reverse proxy is trusted to set (for example `x-forwarded-for`). |
@@ -499,6 +501,61 @@ authorization, and module domains all read from this one pool through independen
views — see [Authorization](authorization.md#shared-database-composition) for why there is exactly
one pool, not two.
## Database transport (TLS)
`core/src/database/transport.ts` is the one typed policy deciding how *every* Ekklesia executable
reaches PostgreSQL. It has exactly two modes:
| Mode | Meaning | Default in |
| --- | --- | --- |
| `verify-full` | TLS required, certificate chain verified, **and** the server's identity verified against the configured host. | production |
| `disable` | No TLS. Refused outright in production. | development/test |
There is deliberately no encryption-only mode (`require`), no chain-only mode (`verify-ca`), and no
`rejectUnauthorized: false` escape hatch: encryption without identity verification is exactly what
an interception attack defeats. Every loose or aliased spelling — `VERIFY-FULL`, `" verify-full"`,
`require`, `prefer`, `no-verify`, `true`, `1`, empty — is rejected with `DATABASE_TLS_MODE_INVALID`
in *both* environments, so a typo can never quietly become a weaker mode.
**Production has no plaintext exception — not even for a loopback database.** `disable` in
production fails with `DATABASE_TLS_REQUIRED_IN_PRODUCTION` whether `DATABASE_URL` points at a
remote host, `127.0.0.1`, `localhost`, or `[::1]`. A connection string cannot prove the database is
actually process-local, so no "same host, therefore safe" carve-out exists.
Verification is certificate *and* hostname, not encryption alone. The policy sets
`rejectUnauthorized: true` and an explicit `checkServerIdentity` pinned to the host parsed from
`DATABASE_URL`. That second part closes a real gap rather than restating a default: the pinned `pg`
driver sets TLS `servername` only for non-IP hosts, and Node performs **no** identity check when it
is unset — so `rejectUnauthorized: true` alone accepts a certificate from any host the trusted CA
ever signed when you connect by IP address. Both halves are proven against a real TLS-enabled
PostgreSQL in `core/tests/database-tls.integration.ts`.
**`DATABASE_URL` may not carry its own TLS parameters.** `sslmode`, `ssl`, `sslcert`, `sslkey`,
`sslrootcert`, `sslpassword`, `sslnegotiation`, and `uselibpqcompat` are all rejected
(case-insensitively) with `DATABASE_TLS_CONNECTION_STRING_CONFLICT`. This is not stylistic: the
pinned driver merges parsed connection-string values *over* explicitly passed options, so a
`?sslmode=disable` in the URL would silently replace the policy's verified `ssl` object. TLS is
decided in exactly one place, and it is never the URL.
For a database behind an internal CA, supply the CA's PEM text in `EKKLESIA_DATABASE_CA`; it is
added to the trusted set for verification and never appears in any error, log, or `pnpm doctor`
output. Loading it from a file path or a secret store is out of scope here, as are certificate
rotation and client-certificate authentication.
The same policy is resolved identically by every executable that opens a connection — `apps/api`,
`pnpm migrations:plan`, `pnpm migrations:apply`, the platform-owner bootstrap CLI, and
`pnpm doctor`'s own connection — through one shared environment reader
(`core/src/cli/database-transport-environment.ts`), so `process.env` stays out of Core's reusable
constructors. `apps/api` validates it in `createApiRuntimeConfig`, *before* any pool, Better Auth
instance, module runtime, or listener exists. `pnpm doctor` reports it as `database.transport` in
its PostgreSQL section: a verified production policy passes, and an unsafe or invalid one is a
blocking `NOT READY` failure — never a warning — naming only the effective mode, never the URL,
host, port, credentials, or CA material.
TLS protects the *connection*, not the exposure: it does not make an internet-reachable database
safe, and the network boundary remains the deployment's own responsibility. See
[ADR 0018](../architecture/decisions/0018-identity-session-resolution-and-database-transport.md).
## Graceful shutdown
`startApi()` (the composition function `apps/api/src/server.ts` exports, used by both the
+8
View File
@@ -416,6 +416,14 @@ unaffected and remain available for standalone/testable use — the CLI bootstra
`createAuthorizationDrizzleDatabase` directly for exactly that reason, since it only ever needs
the authorization schema and has no reason to share a pool with anything else.
Both the shared pool and the per-domain factories accept the same optional `transport` policy —
Ekklesia's single PostgreSQL TLS decision, described in
[API § Database transport (TLS)](api.md#database-transport-tls). Every executable that opens a
connection (the API, both migration commands, the platform-owner bootstrap CLI, and `pnpm doctor`)
resolves it the same way, so production connections are TLS-verified regardless of which factory
built them. Omitting it is a development/test affordance only: no `ssl` option is set and the
connection is plaintext, which a production configuration refuses outright.
`apps/api/tests/lifecycle.test.ts` proves this composition explicitly: exactly one
`new Pool(...)` call for a full `startApi()` lifecycle (covering identity, authorization, and
module database usage), and exactly one `pool.end()` call regardless of how many concurrent
+26 -7
View File
@@ -126,12 +126,28 @@ intentionally contains no real values — configure them per environment, never
| Value | Requirement |
| --- | --- |
| Secret | At least 32 characters, high entropy (for example `openssl rand -base64 32`). No built-in default is ever provided; a missing or too-short secret is rejected explicitly. |
| Base URL | An absolute `http`/`https` URL. In production, `https` is required and loopback hosts are rejected. |
| Trusted origins | One or more absolute `http`/`https` origins with no path, query, or fragment. Wildcards are rejected in this slice. In production, `https` is required and loopback hosts are rejected. |
| Base URL | Exactly one canonical `http`/`https` **origin** — scheme, host, and non-default port only. In production, `https` is required and loopback/unspecified hosts are rejected. |
| Trusted origins | One or more canonical `http`/`https` origins. Wildcards are rejected in this slice. In production, `https` is required and loopback/unspecified hosts are rejected. |
Local development origins (for example `http://localhost:3000`) are only accepted when the caller
explicitly passes `environment: "development"` — never silently. Every configuration failure
raises a sanitized `IdentityConfigError` carrying only a stable code, never the offending value.
**"Canonical origin" is literal**: the value must equal the WHATWG URL parser's own origin
serialization of itself. A path (`https://id.example.test/auth`), a bare trailing slash
(`https://id.example.test/`), a query, a fragment, embedded credentials, an explicitly written
default port (`https://id.example.test:443`), or non-canonical case is rejected with
`BASE_URL_INVALID`/`TRUSTED_ORIGIN_INVALID` rather than normalized behind your back — so the stored
value always matches what a browser actually sends in an `Origin` header.
Production rejects loopback and unspecified hosts by classifying the address, not by comparing it
against a short list: `localhost` and the whole reserved `.localhost` namespace (RFC 6761), all of
`127.0.0.0/8` (not just `127.0.0.1`), `0.0.0.0`, `::1` in both compressed and expanded spellings,
`::`, and IPv4-mapped forms such as `::ffff:127.0.0.1`. Local development origins (for example
`http://localhost:3000`) are only accepted when the caller explicitly passes
`environment: "development"` — never silently. Every configuration failure raises a sanitized
`IdentityConfigError` carrying only a stable code, never the offending value.
This is a guarantee about the *shape* of the configured value and nothing more. It does not resolve
DNS and does not prove reachability: a public hostname that resolves to a loopback or private
address passes this validation, and a firewall, network policy, or reverse proxy — not URL
parsing — is what actually determines who can reach a deployment.
Secret rotation (Better Auth's plural `secrets` list) and cross-subdomain cookies are deferred; see
ADR 0010.
@@ -165,7 +181,9 @@ pinned to `8`/`128` instead of left on Better Auth's fallback default.
`operationTimeoutMs`, applied as both `pg`'s `query_timeout` and `statement_timeout` — bounding a
hung or slow query, which `connectionTimeoutMs` alone cannot do since it only bounds establishing
the connection. `apps/api` sets both timeouts explicitly (see [API](api.md)); an invalid value is
rejected the same sanitized way an invalid connection string already is.
rejected the same sanitized way an invalid connection string already is. It also accepts an optional
`transport` policy — the PostgreSQL TLS decision described in
[API § Database transport](api.md#database-transport-tls); every production executable passes one.
## How a principal is resolved
@@ -194,7 +212,8 @@ internal failures and returns `null` for them exactly as it does for a genuinely
A missing session table, a dropped or drifted column, a revoked `SELECT` privilege, an unreachable
database, and a query timeout therefore all produce a sanitized
`503 {"code":"PRINCIPAL_RESOLUTION_FAILED"}` — never a silent anonymous result. The response body,
logs, and error carry no relation name, column name, SQL, driver message, host, or stack.
logs, and error carry no relation name, column name, SQL, driver message, host, or stack. See
[ADR 0018](../architecture/decisions/0018-identity-session-resolution-and-database-transport.md).
## What is not operational yet
+23 -4
View File
@@ -14,7 +14,9 @@ module migration discovery, identity schema drift detection, formatting, linting
Migration-runner unit tests use a narrow database/session port to prove deterministic planning,
approval, transaction orchestration, cleanup, and sanitized errors. They do not claim to prove
PostgreSQL behavior. Identity unit tests cover branded opaque-ID contracts, principal shapes,
configuration validation (secret, base URL, and trusted-origin rules), bounded/validated
configuration validation (secret rules, canonical-origin-only base URL and trusted origins, and the
full loopback/unspecified classifier production rejects — `.localhost`, all of `127.0.0.0/8`,
`0.0.0.0`, `::1` compressed and expanded, `::`, and IPv4-mapped forms), bounded/validated
connection and operation (`query_timeout`/`statement_timeout`) database timeouts, the Core identity
runtime (explicit construction, exact schema mapping, only email/password and rate limiting
enabled, an explicit `advanced.disableCSRFCheck`/`disableOriginCheck: false` pair, an explicit
@@ -22,9 +24,12 @@ error-only sanitized logger that strips Better Auth's own message and arguments
they contain, pinned 8–128 password-length bounds, sanitized failures, exact whole-name
session-cookie detection that never fires on a prefixed, suffixed, or differently-cased lookalike,
and that a broken session store during principal resolution is never silently reported as anonymous
while an unrelated cookie performs no identity database work at all), and Core migration discovery
(deterministic ordering, the reserved `core` identity, and collision rejection) — all without a
live database. `apps/api`'s own unit tests cover its pure Hono app factory (route shapes, module
while an unrelated cookie performs no identity database work at all), the PostgreSQL transport
policy (`core/tests/database-transport.test.ts`: mode validation including every loose or aliased
spelling, production's plaintext refusal with no loopback exception, connection-string
TLS-parameter rejection, and that no configuration path yields `rejectUnauthorized: false` or leaks
CA material into an error), and Core migration discovery (deterministic ordering, the reserved
`core` identity, and collision rejection) — all without a live database. `apps/api`'s own unit tests cover its pure Hono app factory (route shapes, module
route mounting/enforcement/CORS/response sanitization, CORS allow/deny/preflight, deterministic 404
and sanitized error responses), its runtime configuration, its idempotent shutdown lifecycle
(concurrent callers share one server-close-then-module-runtime-stop-then-pool-close promise; a
@@ -125,6 +130,20 @@ completes and is independently verifiable, through the harness's own administrat
never the API's own pool — before that pool closes. See
[ADR 0016](../architecture/decisions/0016-module-database-access.md).
A fourth dedicated suite (`core/tests/database-tls.integration.ts`, included in the same
`pnpm test:migrations:integration` run) proves the production PostgreSQL transport policy against a
real TLS-enabled PostgreSQL rather than against the shape of an options object: a trusted
certificate whose identity matches the host connects and is confirmed encrypted through
`pg_stat_ssl`; a certificate that does not chain to the configured CA is refused
(`UNABLE_TO_VERIFY_LEAF_SIGNATURE`); the same server reached through an address its certificate does
not cover is refused (`ERR_TLS_CERT_ALTNAME_INVALID`); the identical connection *without* the
policy's pinned `checkServerIdentity` succeeds, which is the concrete gap the policy closes; and
`verify-full` cannot be satisfied by a plaintext-only endpoint. Every CA, certificate, and private
key is generated at test runtime into a temporary directory and deleted afterwards — **no private
key material is ever committed to this repository** — and the disposable container is removed in
`afterAll` even when an assertion fails. See
[ADR 0018](../architecture/decisions/0018-identity-session-resolution-and-database-transport.md).
CI runs the same command in the dedicated `migration-smoke` job. After this change is merged, add
the following pull-request status context to `main` branch protection: