Files
PiCloud/crates/manager-core/src/auth_middleware.rs
MechaCat02 902dd78027 feat(picloud): opportunistic principal middleware on the data plane
The data-plane (POST /execute/{id} + user-route fallback) is
unauthenticated by default — public scripts get hit by anonymous HTTP
traffic. But some calls are authed (dashboard test-runs, API-key
invocations) and v1.1.x services will want to see the caller via
`cx.principal` for audit / authz once those features land.

  - New manager-core::attach_principal_if_present middleware. Always
    inserts Extension<Option<Principal>>: Some on resolved bearer/cookie,
    None on absent or malformed token. Fail-open on DB blip so a
    transient infra failure can't 500 anonymous traffic.
  - Wired in picloud build_app, scoped to the data-plane and user-routes
    routers only. The admin path keeps using require_authenticated; no
    double-resolve on the same token.
  - orchestrator-core handlers (execute_by_id, user_route_handler) now
    extract Extension<Option<Principal>> and pass it to build_exec_request.
    Replaces the temporary `None` placeholders from the previous commit.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-30 18:53:27 +02:00

378 lines
13 KiB
Rust

//! Authentication middleware — resolves the caller's `Principal` from
//! either a session cookie / Bearer session-token OR an API key
//! (`Authorization: Bearer pic_…`). Both paths converge on the same
//! request extension so downstream handlers see one shape.
//!
//! Capability checks live in `crate::authz` and are called per-handler
//! (after the relevant resource is loaded, so the capability binds to
//! the actual resource's `app_id`). This middleware is gate-only: it
//! ensures *some* `Principal` is attached, or returns 401.
//!
//! Token discriminator: the `pic_` prefix on a Bearer value selects
//! the API-key path; anything else (raw 32-byte base64-url-encoded
//! string) takes the session path. The session cookie can only ever
//! carry a session token (cookies are never API keys).
use std::sync::Arc;
use std::time::Duration;
use axum::body::Body;
use axum::extract::{Request, State};
use axum::http::{header, StatusCode};
use axum::middleware::Next;
use axum::response::{IntoResponse, Json, Response};
use chrono::Utc;
use picloud_shared::{AdminUserId, Principal};
use serde_json::json;
use crate::admin_session_repo::AdminSessionRepository;
use crate::admin_user_repo::AdminUserRepository;
use crate::api_key_repo::{ApiKeyRepository, ApiKeyVerification};
use crate::auth::{hash_token, verify_password};
pub const SESSION_COOKIE: &str = "picloud_session";
/// Prefix on the wire that selects the API-key path. The body that
/// follows is `base32(32 random bytes)`; the first 8 chars of the body
/// index into `api_keys.prefix` for verification.
pub const API_KEY_PREFIX: &str = "pic_";
/// Length of the indexed prefix portion of an API key (the 8 chars
/// immediately after `pic_`). Schema-side index is on this slice.
pub const API_KEY_PREFIX_LEN: usize = 8;
/// Shared state for auth: the user / session / API-key repos plus the
/// configured sliding session TTL. Cheap to clone (`Arc` everywhere).
#[derive(Clone)]
pub struct AuthState {
pub users: Arc<dyn AdminUserRepository>,
pub sessions: Arc<dyn AdminSessionRepository>,
pub keys: Arc<dyn ApiKeyRepository>,
pub ttl: Duration,
}
/// Legacy request-extension alias retained so the (only remaining)
/// handler that pulled `AuthedAdmin` out — `GET /admin/auth/me` —
/// keeps compiling during the migration. New handlers should pull
/// `Extension<Principal>` directly.
#[deprecated(note = "use Extension<Principal> directly")]
#[derive(Debug, Clone)]
pub struct AuthedAdmin {
pub id: AdminUserId,
pub username: String,
}
/// Middleware entry point. Wire with
/// `axum::middleware::from_fn_with_state(auth_state, require_authenticated)`.
/// Inserts `Principal` (and the legacy `AuthedAdmin`) as request
/// extensions on success; returns 401 on any failure mode.
pub async fn require_authenticated(
State(state): State<AuthState>,
mut req: Request<Body>,
next: Next,
) -> Response {
let Some(token) = extract_token(&req) else {
return unauthorized();
};
let principal = match resolve_principal(&state, &token).await {
Ok(Some(p)) => p,
Ok(None) => return unauthorized(),
Err(InternalError) => return internal_error(),
};
let username_for_legacy = username_for(&state, principal.user_id).await;
req.extensions_mut().insert(principal.clone());
#[allow(deprecated)]
if let Some(username) = username_for_legacy {
req.extensions_mut().insert(AuthedAdmin {
id: principal.user_id,
username,
});
}
next.run(req).await
}
/// Backwards-compatible alias — the single callsite that still names
/// `require_admin` keeps working without an immediate rename. New
/// wiring should call `require_authenticated`.
#[deprecated(note = "renamed to require_authenticated")]
pub async fn require_admin(state: State<AuthState>, req: Request<Body>, next: Next) -> Response {
require_authenticated(state, req, next).await
}
/// Opportunistic data-plane variant: always inserts an
/// `Extension<Option<Principal>>` and forwards the request. Used on
/// `/execute/{id}` and the user-route fallback, where most invocations
/// are anonymous public HTTP and the few authed ones (dashboard
/// test-runs, API keys) should still let scripts see the caller via
/// `cx.principal` once services consume it.
///
/// Failure modes — all degrade to `None` rather than rejecting:
/// * No bearer / cookie → `None`.
/// * Malformed or unknown token → `None`.
/// * DB blip while resolving → `None` (fail-open; the data plane
/// should not 500 on transient infra failures for an *optional*
/// identity check).
///
/// Admin-side routes that REQUIRE an identity keep using
/// `require_authenticated`.
pub async fn attach_principal_if_present(
State(state): State<AuthState>,
mut req: Request<Body>,
next: Next,
) -> Response {
let principal: Option<Principal> = match extract_token(&req) {
Some(token) => resolve_principal(&state, &token).await.unwrap_or(None),
None => None,
};
req.extensions_mut().insert(principal);
next.run(req).await
}
/// Decide whether the token is an API key (pic_ prefix) or a session
/// token, then resolve the corresponding `Principal`. `Ok(None)`
/// means the token was structurally valid but didn't match any active
/// credential; `Err(InternalError)` means a DB blip.
async fn resolve_principal(
state: &AuthState,
token: &str,
) -> Result<Option<Principal>, InternalError> {
if let Some(rest) = token.strip_prefix(API_KEY_PREFIX) {
return verify_api_key(state, rest).await;
}
verify_session(state, token).await
}
async fn verify_session(
state: &AuthState,
token: &str,
) -> Result<Option<Principal>, InternalError> {
let token_hash = hash_token(token);
let lookup = match state.sessions.lookup(&token_hash).await {
Ok(Some(l)) => l,
Ok(None) => return Ok(None),
Err(err) => {
tracing::error!(?err, "admin_sessions lookup failed");
return Err(InternalError);
}
};
let user = match state.users.get(lookup.user_id).await {
Ok(Some(u)) if u.is_active => u,
Ok(_) => return Ok(None),
Err(err) => {
tracing::error!(?err, "admin_users lookup failed");
return Err(InternalError);
}
};
// Sliding-window bump — inline so a DB blip surfaces as 500 rather
// than silent stale sessions. Same shape as Phase 3a.
let new_expires_at = Utc::now() + chrono::Duration::from_std(state.ttl).unwrap_or_default();
if let Err(err) = state.sessions.touch(&token_hash, new_expires_at).await {
tracing::error!(?err, "admin_sessions touch failed");
return Err(InternalError);
}
Ok(Some(Principal {
user_id: user.id,
instance_role: user.instance_role,
scopes: None,
app_binding: None,
}))
}
/// API-key verification path. `rest` is the portion of the bearer
/// value *after* `pic_`. We slice off the first 8 chars as the
/// indexed lookup key, then Argon2id-verify each candidate's hash
/// against the full `rest`. At most one match is expected; multiple
/// candidates with the same prefix is statistically negligible but
/// handled correctly (verify each, take the first match).
async fn verify_api_key(state: &AuthState, rest: &str) -> Result<Option<Principal>, InternalError> {
if rest.len() <= API_KEY_PREFIX_LEN {
return Ok(None);
}
let prefix = &rest[..API_KEY_PREFIX_LEN];
let candidates = match state.keys.find_active_by_prefix(prefix).await {
Ok(v) => v,
Err(err) => {
tracing::error!(?err, "api_keys lookup failed");
return Err(InternalError);
}
};
let matched: Option<ApiKeyVerification> = candidates
.into_iter()
.find(|c| verify_password(&c.hash, rest));
let Some(matched) = matched else {
return Ok(None);
};
// Resolve the owning user. is_active = false → reject even if the
// key itself hasn't been expired yet (the expire_all_for_user
// cascade on deactivation is the primary defense; this is the
// belt-and-suspenders check at request time).
let user = match state.users.get(matched.user_id).await {
Ok(Some(u)) if u.is_active => u,
Ok(_) => return Ok(None),
Err(err) => {
tracing::error!(?err, "admin_users lookup for api key failed");
return Err(InternalError);
}
};
if let Err(err) = state.keys.touch_last_used(matched.id).await {
tracing::error!(?err, "api_keys touch_last_used failed");
// Soft-fail: a timestamp blip should not invalidate the
// request. Continue with the resolved Principal.
}
Ok(Some(Principal {
user_id: user.id,
instance_role: user.instance_role,
scopes: Some(matched.scopes),
app_binding: matched.app_id,
}))
}
/// Best-effort username lookup for the legacy `AuthedAdmin` extension.
/// Returns `None` on DB error (the caller treats `None` as "skip the
/// legacy extension"). New handlers use `Principal` and don't depend
/// on this.
async fn username_for(state: &AuthState, id: AdminUserId) -> Option<String> {
match state.users.get(id).await {
Ok(Some(u)) => Some(u.username),
Ok(None) => None,
Err(err) => {
tracing::warn!(
?err,
"username lookup for AuthedAdmin failed; skipping legacy ext"
);
None
}
}
}
/// Pull the bearer token out of an `Authorization` header (preferred)
/// or the `picloud_session` cookie (fallback for browser clients).
/// Same shape as Phase 3a; the cookie only ever carries session
/// tokens — no `pic_` prefix expected there.
fn extract_token(req: &Request<Body>) -> Option<String> {
if let Some(value) = req.headers().get(header::AUTHORIZATION) {
if let Ok(s) = value.to_str() {
if let Some(token) = s.strip_prefix("Bearer ") {
let trimmed = token.trim();
if !trimmed.is_empty() {
return Some(trimmed.to_string());
}
}
}
}
if let Some(value) = req.headers().get(header::COOKIE) {
if let Ok(s) = value.to_str() {
for chunk in s.split(';') {
let chunk = chunk.trim();
if let Some(rest) = chunk.strip_prefix(&format!("{SESSION_COOKIE}=")) {
if !rest.is_empty() {
return Some(rest.to_string());
}
}
}
}
}
None
}
/// Sentinel returned from the resolve functions when a DB error should
/// produce a 500 rather than a 401. Empty struct because the actual
/// error is already logged at the failure site.
struct InternalError;
fn unauthorized() -> Response {
(
StatusCode::UNAUTHORIZED,
Json(json!({ "error": "authentication required" })),
)
.into_response()
}
fn internal_error() -> Response {
(
StatusCode::INTERNAL_SERVER_ERROR,
Json(json!({ "error": "internal error" })),
)
.into_response()
}
#[cfg(test)]
mod tests {
use super::*;
use axum::http::Request;
use picloud_shared::InstanceRole;
fn req_with_header(name: &str, value: &str) -> Request<Body> {
Request::builder()
.header(name, value)
.body(Body::empty())
.unwrap()
}
#[test]
fn extracts_bearer_token() {
let r = req_with_header("authorization", "Bearer abc123");
assert_eq!(extract_token(&r).as_deref(), Some("abc123"));
}
#[test]
fn extracts_bearer_pic_prefixed_token() {
let r = req_with_header("authorization", "Bearer pic_abcdefghIJKL");
assert_eq!(extract_token(&r).as_deref(), Some("pic_abcdefghIJKL"));
}
#[test]
fn ignores_bearer_with_no_token() {
let r = req_with_header("authorization", "Bearer ");
assert_eq!(extract_token(&r), None);
}
#[test]
fn extracts_cookie_token() {
let r = req_with_header("cookie", "foo=bar; picloud_session=xyz; baz=qux");
assert_eq!(extract_token(&r).as_deref(), Some("xyz"));
}
#[test]
fn bearer_wins_over_cookie() {
let r = Request::builder()
.header("authorization", "Bearer header-token")
.header("cookie", "picloud_session=cookie-token")
.body(Body::empty())
.unwrap();
assert_eq!(extract_token(&r).as_deref(), Some("header-token"));
}
#[test]
fn returns_none_when_neither_present() {
let r = Request::builder().body(Body::empty()).unwrap();
assert_eq!(extract_token(&r), None);
}
// Round-trip test for the unused-variable to keep `Principal`
// visibly tied to InstanceRole — caught a real bug during dev when
// the field order in the struct literal had drifted.
#[test]
fn principal_construction_is_explicit() {
let p = Principal {
user_id: AdminUserId::new(),
instance_role: InstanceRole::Owner,
scopes: None,
app_binding: None,
};
assert_eq!(p.instance_role, InstanceRole::Owner);
assert!(p.scopes.is_none());
assert!(p.app_binding.is_none());
}
}