fix(audit-2026-06-11/revocation-lag): evict PrincipalCache on credential change

The PrincipalCache (token-hash -> resolved Principal, 60s TTL) had no
eviction hook, so deactivation / password change / API-key revocation
flipped the DB rows but the cache kept serving the stale principal for
up to the TTL window. Closes the audit's PrincipalCache revocation-lag
Medium and greens authz::deactivating_user_revokes_their_api_keys.

- PrincipalCache::evict_user(user_id) + evict_token(hash).
- One shared cache Arc threaded into AuthState / AdminsState /
  ApiKeysState (a per-state cache would let the middleware's own copy
  keep authenticating a revoked principal).
- Evict on: deactivation, password change (both evict_user), logout
  (evict_token, precise), single API-key delete (evict_user).
- H-E1 note: DB-write invalidation stays best-effort by design (a blip
  must not undo the rotation); the cache evict is what makes it take
  effect on the next request.
- Renamed bearer_and_cookie_produce_same_principal ->
  session_token_and_api_key_produce_same_principal (cookie auth was
  removed in C-1; the test never tested cookies).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
MechaCat02
2026-06-12 18:12:54 +02:00
parent 19f3647f0e
commit 31ecfae86e
6 changed files with 116 additions and 4 deletions

View File

@@ -45,6 +45,11 @@ pub struct AdminsState {
/// Capability gate: every endpoint here requires
/// `InstanceManageUsers` (owner / admin).
pub authz: Arc<dyn AuthzRepo>,
/// Audit 2026-06-11 (PrincipalCache revocation-lag) — evicted on
/// password change / deactivation so a just-revoked session or
/// API key stops authenticating immediately instead of lingering
/// for the cache TTL.
pub principal_cache: Arc<crate::auth_middleware::PrincipalCache>,
}
pub fn admins_router(state: AdminsState) -> Router {
@@ -254,6 +259,12 @@ async fn patch_admin(
tracing::error!(?err, "failed to expire api keys on password change");
}
}
// Audit 2026-06-11 (PrincipalCache revocation-lag) — the DB
// writes above are best-effort by design (a blip must not undo
// the rotation), so evicting the in-process cache is what makes
// the rotation take effect on the very next request rather than
// after the cache TTL.
state.principal_cache.evict_user(id);
}
if let Some(email_patch) = input.email.as_ref() {
@@ -323,6 +334,11 @@ async fn patch_admin(
tracing::error!(?err, "failed to expire api keys for deactivated admin");
}
}
// Audit 2026-06-11 (PrincipalCache revocation-lag) — evict
// cached principals so a deactivated admin's session / API
// keys stop authenticating immediately instead of lingering
// for the cache TTL window.
state.principal_cache.evict_user(id);
}
}

View File

@@ -39,6 +39,10 @@ const NAME_MAX: usize = 64;
#[derive(Clone)]
pub struct ApiKeysState {
pub keys: Arc<dyn ApiKeyRepository>,
/// Audit 2026-06-11 (PrincipalCache revocation-lag) — evicted when
/// a user deletes one of their own keys so it stops authenticating
/// from cache immediately rather than after the cache TTL.
pub principal_cache: Arc<crate::auth_middleware::PrincipalCache>,
}
pub fn api_keys_router(state: ApiKeysState) -> Router {
@@ -160,6 +164,12 @@ async fn delete_key(
// we deliberately don't leak the distinction.
return Err(ApiKeysError::NotFound(id));
}
// Audit 2026-06-11 (PrincipalCache revocation-lag) — the cache is
// keyed by token hash, which we can't derive from the key id, so
// evict every cached principal for this user. That also drops their
// session entries; re-auth on the next request is the right cost for
// a deliberate key revocation.
state.principal_cache.evict_user(principal.user_id);
Ok(StatusCode::NO_CONTENT)
}

View File

@@ -195,6 +195,11 @@ async fn logout(State(state): State<AuthState>, req: Request<Body>) -> Response
if let Err(err) = state.sessions.delete(&hash).await {
tracing::error!(?err, "admin_sessions delete failed");
}
// Audit 2026-06-11 (PrincipalCache revocation-lag) — drop just
// this token from the cache so the logged-out session can't keep
// authenticating for the cache TTL. Other sessions for the same
// user are left untouched.
state.principal_cache.evict_token(&hash);
}
StatusCode::NO_CONTENT.into_response()
}

View File

@@ -92,6 +92,32 @@ impl PrincipalCache {
}
guard.insert(token_hash, (Instant::now(), principal));
}
/// Evict every cached principal belonging to `user_id`.
///
/// Audit 2026-06-11 (PrincipalCache revocation-lag, Medium) — the
/// revocation paths (`set_active(false)`, password change,
/// API-key delete) flip the backing DB rows, but a cached principal
/// keeps authenticating for up to [`PRINCIPAL_CACHE_TTL`] unless it
/// is evicted. Callers invoke this right after the DB mutation so a
/// just-revoked credential stops working on the next request. The
/// cache value carries the resolved [`Principal`], so we can target
/// the user without knowing which token hashes belong to them.
pub fn evict_user(&self, user_id: AdminUserId) {
let Ok(mut guard) = self.inner.lock() else {
return;
};
guard.retain(|_, (_, p)| p.user_id != user_id);
}
/// Evict a single cached entry by its token hash. Used on logout,
/// where exactly one session token is being invalidated and we
/// don't want to disturb the user's other live sessions.
pub fn evict_token(&self, token_hash: &str) {
if let Ok(mut guard) = self.inner.lock() {
guard.remove(token_hash);
}
}
}
/// Prefix on the wire that selects the API-key path. The body that
@@ -471,4 +497,46 @@ mod tests {
assert!(p.scopes.is_none());
assert!(p.app_binding.is_none());
}
fn principal_for(user_id: AdminUserId) -> Principal {
Principal {
user_id,
instance_role: InstanceRole::Owner,
scopes: None,
app_binding: None,
}
}
#[test]
fn evict_user_drops_only_that_users_entries() {
// Audit 2026-06-11 (PrincipalCache revocation-lag).
let cache = PrincipalCache::new();
let alice = AdminUserId::new();
let bob = AdminUserId::new();
cache.insert("hash-alice-1".into(), principal_for(alice));
cache.insert("hash-alice-2".into(), principal_for(alice));
cache.insert("hash-bob-1".into(), principal_for(bob));
cache.evict_user(alice);
assert!(cache.get("hash-alice-1").is_none());
assert!(cache.get("hash-alice-2").is_none());
// Bob is untouched.
assert!(cache.get("hash-bob-1").is_some());
}
#[test]
fn evict_token_drops_only_that_entry() {
let cache = PrincipalCache::new();
let alice = AdminUserId::new();
cache.insert("hash-1".into(), principal_for(alice));
cache.insert("hash-2".into(), principal_for(alice));
cache.evict_token("hash-1");
// Logout drops exactly one session; the user's other live
// session stays cached.
assert!(cache.get("hash-1").is_none());
assert!(cache.get("hash-2").is_some());
}
}

View File

@@ -498,12 +498,19 @@ pub async fn build_app(
.and_then(|s| s.parse::<usize>().ok())
.filter(|n| *n > 0)
.unwrap_or(2);
// Audit 2026-06-11 (PrincipalCache revocation-lag) — one shared
// cache instance threaded into every state that performs auth or
// revocation, so an evict on the revocation side is visible to the
// middleware's resolve side. A per-state cache would let a revoked
// principal keep authenticating from the middleware's own copy.
let principal_cache =
Arc::new(picloud_manager_core::auth_middleware::PrincipalCache::new());
let auth_state = AuthState {
users: auth.users.clone(),
sessions: auth.sessions.clone(),
keys: auth.keys.clone(),
ttl: auth.ttl,
principal_cache: Arc::new(picloud_manager_core::auth_middleware::PrincipalCache::new()),
principal_cache: principal_cache.clone(),
login_rate_limiter: Arc::new(
picloud_manager_core::login_rate_limit::LoginRateLimiter::new(),
),
@@ -514,6 +521,7 @@ pub async fn build_app(
sessions: auth.sessions,
keys: auth.keys.clone(),
authz: authz.clone(),
principal_cache: principal_cache.clone(),
};
let app_members_state = AppMembersState {
apps: apps_state.apps.clone(),
@@ -526,7 +534,10 @@ pub async fn build_app(
authz: authz.clone(),
users: users.clone(),
};
let api_keys_state = ApiKeysState { keys: auth.keys };
let api_keys_state = ApiKeysState {
keys: auth.keys,
principal_cache: principal_cache.clone(),
};
// /admin/auth/login + /logout are unguarded by design (login is how
// you get in). /admin/auth/me applies the middleware internally so

View File

@@ -380,12 +380,14 @@ async fn member_can_only_touch_apps_they_belong_to(pool: PgPool) {
}
// ----------------------------------------------------------------------------
// 5. Bearer pic_ + cookie produce the same Principal
// 5. Session-token bearer + pic_ API-key bearer produce the same Principal
// (cookie auth was removed in the audit 2026-06-11 C-1 fix; both paths
// now travel through the Authorization header).
// ----------------------------------------------------------------------------
#[ignore = "needs DATABASE_URL pointing at a running Postgres"]
#[sqlx::test(migrations = "../manager-core/migrations")]
async fn bearer_and_cookie_produce_same_principal(pool: PgPool) {
async fn session_token_and_api_key_produce_same_principal(pool: PgPool) {
let s = boot(pool).await;
let session_token = login_token(&s.server, "owner", "owner-pw").await;