fix: stop leaking connector OAuth session tokens to other origins

The connector OAuth popup posted the session token to window.opener with
a '*' target origin, so any page that opened the popup received it. An
attacker with an account on a multi-user deployment could start a flow for
their own pending session, get a victim to finish the provider consent, and
receive a token backed by the victim's Drive/SharePoint/Confluence tokens.

- Post popup results only to allowed frontend origins: the callback origin,
  OIDC_FRONTEND_URL, the new CONNECTOR_ALLOWED_ORIGINS, and localhost:5173
  when the callback runs on a loopback host.
- Render the success page from the callback itself so the token never
  appears in a URL; callback-status ignores session_token/user_email params.
- ConnectorAuth accepts messages only from the popup it opened, on the
  callback origin reported by /api/connectors/auth.
- /api/connectors/disconnect requires auth and only deletes the caller's
  session.
- /api/connectors/sync and /api/remote reject session tokens the caller
  does not own.

Fixes #2766
This commit is contained in:
Alex committed 2026-09-14 21:46:09 +01:00
1 parent d121eec2f5
commit 41b3afed14
14 files changed
+575 -87

No files matched your search

@@ -52,6 +52,7 @@ VITE_CONFLUENCE_CLIENT_ID=your-atlassian-client-id
| `CONFLUENCE_CLIENT_ID` | Client ID from your Atlassian OAuth app | Yes |
| `CONFLUENCE_CLIENT_SECRET` | Client secret from your Atlassian OAuth app | Yes |
| `VITE_CONFLUENCE_CLIENT_ID` | Same Client ID, used by the frontend to show the Confluence option | Yes |
| `CONNECTOR_ALLOWED_ORIGINS` | Comma-separated frontend origins allowed to receive the sign-in result, e.g. `https://docsgpt.example.com`. Not needed when the frontend shares the API origin or runs on `localhost:5173` | When the frontend is on its own origin |
### Step 5: Restart and Use
@@ -62,6 +63,7 @@ Restart your application, then go to the upload section in DocsGPT and select **
## Troubleshooting
- **Option not appearing** — Verify `VITE_CONFLUENCE_CLIENT_ID` is set in the frontend `.env`, then restart.
- **Sign-in popup closes but the account never connects** — The frontend origin is not allowed to receive the result. Add it to `CONNECTOR_ALLOWED_ORIGINS` in the backend `.env`.
- **Authentication failed** — Check that the callback URL matches exactly, including `?provider=confluence`.
- **No accessible sites** — Ensure the authenticating user has access to at least one Confluence Cloud site.
- **Permission denied** — Verify that the Confluence API scopes are enabled in your Atlassian app settings.
@@ -50,6 +50,7 @@ VITE_GOOGLE_CLIENT_ID=your-google-client-id
| `GOOGLE_CLIENT_ID` | OAuth Client ID from GCP Credentials | Yes |
| `GOOGLE_CLIENT_SECRET` | OAuth Client Secret from GCP Credentials | Yes |
| `VITE_GOOGLE_CLIENT_ID` | Same Client ID, used by the frontend to show the Google Drive option | Yes |
| `CONNECTOR_ALLOWED_ORIGINS` | Comma-separated frontend origins allowed to receive the sign-in result, e.g. `https://docsgpt.example.com`. Not needed when the frontend shares the API origin or runs on `localhost:5173` | When the frontend is on its own origin |
<Callout type="warning" emoji="⚠️">
Make sure to use the same Google Client ID in both backend and frontend configurations.
@@ -65,6 +66,7 @@ Restart your application, then go to the upload section in DocsGPT and select **
- **Option not appearing** — Verify `VITE_GOOGLE_CLIENT_ID` is set in the frontend `.env`, then restart.
- **Authentication failed** — Check that the redirect URI matches exactly, including `?provider=google_drive`. Ensure the Google Drive API is enabled.
- **Sign-in popup closes but the account never connects** — The frontend origin is not allowed to receive the result. Add it to `CONNECTOR_ALLOWED_ORIGINS` in the backend `.env`.
- **Permission denied** — Verify the OAuth consent screen is configured and the user has access to the target files.
- **Files not processing** — Check backend logs and verify that backend environment variables are correctly set.
@@ -45,6 +45,7 @@ MICROSOFT_TENANT_ID=your-azure-ad-tenant-id
| `MICROSOFT_CLIENT_SECRET` | Client secret value | Yes | — |
| `MICROSOFT_TENANT_ID` | Directory (tenant) ID | No | `common` |
| `MICROSOFT_AUTHORITY` | Login endpoint override | No | Auto-constructed |
| `CONNECTOR_ALLOWED_ORIGINS` | Comma-separated frontend origins allowed to receive the sign-in result, e.g. `https://docsgpt.example.com` | When the frontend is on its own origin | API origin and `localhost:5173` |
<Callout type="warning">
`MICROSOFT_TENANT_ID=common` (the default) allows any Microsoft account to authenticate. Set this to your specific tenant ID in production.
@@ -60,4 +61,5 @@ Restart your application, then go to the upload section in DocsGPT and select **
- **Option not appearing** — Verify `MICROSOFT_CLIENT_ID` and `MICROSOFT_CLIENT_SECRET` are set, then restart.
- **Authentication failed** — Check that the redirect URI matches exactly, including `?provider=share_point`.
- **Sign-in popup closes but the account never connects** — The frontend origin is not allowed to receive the result. Add it to `CONNECTOR_ALLOWED_ORIGINS` in the backend `.env`.
- **Permission denied** — Ensure admin consent is granted and the user has access to the target files.
+149 -80
View File
@@ -2,7 +2,8 @@ import base64
import html
import json
import uuid
from urllib.parse import urlencode
from typing import Optional
from urllib.parse import urlencode, urlsplit
from flask import (
@@ -19,6 +20,7 @@ from docsgpt.api import api
from docsgpt.api.user.tasks import (
ingest_connector_task,
)
from docsgpt.core.settings import settings
from docsgpt.parser.connectors.connector_creator import ConnectorCreator
from docsgpt.storage.db.repositories.connector_sessions import (
ConnectorSessionsRepository,
@@ -44,6 +46,125 @@ def build_callback_redirect(params: dict) -> str:
return f"{CALLBACK_STATUS_PATH}?{urlencode(params)}"
_LOOPBACK_HOSTS = frozenset({"localhost", "127.0.0.1", "::1"})
_DEV_FRONTEND_PORT = 5173
def _origin_of(url: Optional[str]) -> Optional[str]:
"""Normalized ``scheme://host[:port]`` origin of an http(s) URL, or None."""
if not url:
return None
try:
parts = urlsplit(url.strip())
port = parts.port
except ValueError:
return None
host = parts.hostname
if parts.scheme not in ("http", "https") or not host:
return None
if ":" in host:
host = f"[{host}]"
if port is None or port == {"http": 80, "https": 443}[parts.scheme]:
return f"{parts.scheme}://{host}"
return f"{parts.scheme}://{host}:{port}"
def connector_allowed_origins(request_host_url: str) -> list[str]:
"""Frontend origins the OAuth popup may hand a connector session token to."""
candidates = [
request_host_url,
settings.CONNECTOR_REDIRECT_BASE_URI,
settings.OIDC_FRONTEND_URL,
*(settings.CONNECTOR_ALLOWED_ORIGINS or "").split(","),
]
callback_origin = _origin_of(settings.CONNECTOR_REDIRECT_BASE_URI)
if callback_origin:
callback = urlsplit(callback_origin)
if callback.hostname in _LOOPBACK_HOSTS:
for port in (_DEV_FRONTEND_PORT, callback.port or 80):
candidates += [f"http://localhost:{port}", f"http://127.0.0.1:{port}"]
origins: list[str] = []
for candidate in candidates:
origin = _origin_of(candidate)
if origin and origin not in origins:
origins.append(origin)
return origins
def _js_literal(value) -> str:
"""Encode a value as a JavaScript literal that is safe inside an inline script."""
return json.dumps(value).replace("</", "<\\/").replace("<!--", "<\\!--")
def _render_callback_page(
status: str, message: str, provider_raw: str, session_token: str = "", user_email: str = "",
):
"""Popup page that reports an OAuth result to the opener on allowed origins only."""
status = status if status in ("success", "error", "cancelled") else "error"
provider = html.escape(provider_raw.replace("_", " ").title())
connected_as = (
f"<p>Connected as: {html.escape(user_email)}</p>" if status == "success" and user_email else ""
)
closing_note = (
f"Your {provider} is now connected and ready to use." if status == "success"
else "Feel free to close this window."
)
html_content = f"""
<!DOCTYPE html>
<html>
<head>
<title>{provider} Authentication</title>
<style>
body {{ font-family: Arial, sans-serif; text-align: center; padding: 40px; }}
.container {{ max-width: 600px; margin: 0 auto; }}
.success {{ color: #4CAF50; }}
.error {{ color: #F44336; }}
.cancelled {{ color: #FF9800; }}
</style>
<script>
window.onload = function() {{
const status = {_js_literal(status)};
const sessionToken = {_js_literal(session_token)};
const userEmail = {_js_literal(user_email)};
const providerType = {_js_literal(provider_raw)};
const targetOrigins = {_js_literal(connector_allowed_origins(request.host_url))};
if (status === "success" && window.opener) {{
const payload = {{
type: providerType + '_auth_success',
session_token: sessionToken,
user_email: userEmail
}};
targetOrigins.forEach(function(origin) {{
window.opener.postMessage(payload, origin);
}});
setTimeout(() => window.close(), 3000);
}} else if (status === "cancelled" || status === "error") {{
setTimeout(() => window.close(), 3000);
}}
}};
</script>
</head>
<body>
<div class="container">
<h2>{provider} Authentication</h2>
<div class="{status}">
<p>{html.escape(message)}</p>
{connected_as}
</div>
<p><small>You can close this window. {closing_note}</small></p>
</div>
</body>
</html>
"""
return make_response(
html_content,
200,
{"Content-Type": "text/html", "Cache-Control": "no-store", "Referrer-Policy": "no-referrer"},
)
@connectors_ns.route("/api/connectors/auth")
class ConnectorAuth(Resource):
@@ -78,7 +199,8 @@ class ConnectorAuth(Resource):
return make_response(jsonify({
"success": True,
"authorization_url": authorization_url,
"state": state
"state": state,
"callback_origin": _origin_of(settings.CONNECTOR_REDIRECT_BASE_URI),
}), 200)
except Exception as e:
current_app.logger.error(f"Error generating connector auth URL: {e}", exc_info=True)
@@ -172,14 +294,12 @@ class ConnectorsCallback(Resource):
if not updated:
repo.update_by_legacy_id(value, patch)
# Redirect to success page with session token and user email
return redirect(build_callback_redirect({
"status": "success",
"message": "Authentication successful",
"provider": provider,
"session_token": session_token,
"user_email": user_email
}))
# Render instead of redirecting so the session token never
# lands in a URL (browser history, access logs, Referer).
return _render_callback_page(
"success", "Authentication successful", provider,
session_token=session_token, user_email=user_email,
)
except Exception as e:
current_app.logger.error(f"Error exchanging code for tokens: {str(e)}", exc_info=True)
@@ -345,6 +465,9 @@ class ConnectorDisconnect(Resource):
@api.expect(api.model("ConnectorDisconnectModel", {"provider": fields.String(required=True), "session_token": fields.String(required=False)}))
@api.doc(description="Disconnect a connector session")
def post(self):
decoded_token = request.decoded_token
if not decoded_token:
return make_response(jsonify({"success": False, "error": "Unauthorized"}), 401)
try:
data = request.get_json()
provider = data.get('provider')
@@ -352,11 +475,10 @@ class ConnectorDisconnect(Resource):
if not provider:
return make_response(jsonify({"success": False, "error": "provider is required"}), 400)
if session_token:
with db_session() as conn:
ConnectorSessionsRepository(conn).delete_by_session_token(
session_token,
session_token, decoded_token.get('sub'),
)
return make_response(jsonify({"success": True}), 200)
@@ -428,6 +550,14 @@ class ConnectorSync(Resource):
400
)
with db_readonly() as conn:
session = ConnectorSessionsRepository(conn).get_by_session_token(session_token)
if not session or session.get("user_id") != user_id:
return make_response(
jsonify({"success": False, "error": "Invalid or unauthorized session"}),
401,
)
# Extract configuration from remote_data
file_ids = remote_data.get('file_ids', [])
folder_ids = remote_data.get('folder_ids', [])
@@ -476,74 +606,13 @@ class ConnectorCallbackStatus(Resource):
def get(self):
"""Return HTML page with connector authentication status"""
try:
# Validate and sanitize status to a known value
status_raw = request.args.get('status', 'error')
status = status_raw if status_raw in ('success', 'error', 'cancelled') else 'error'
# Escape all user-controlled values for HTML context
message = html.escape(request.args.get('message', ''))
provider_raw = request.args.get('provider', 'connector')
provider = html.escape(provider_raw.replace('_', ' ').title())
session_token = request.args.get('session_token', '')
user_email = html.escape(request.args.get('user_email', ''))
def safe_js_string(value: str) -> str:
"""Safely encode a string for embedding in inline JavaScript."""
js_encoded = json.dumps(value)
return js_encoded.replace('</', '<\\/').replace('<!--', '<\\!--')
js_status = safe_js_string(status)
js_session_token = safe_js_string(session_token)
js_user_email = safe_js_string(user_email)
js_provider_type = safe_js_string(provider_raw)
html_content = f"""
<!DOCTYPE html>
<html>
<head>
<title>{provider} Authentication</title>
<style>
body {{ font-family: Arial, sans-serif; text-align: center; padding: 40px; }}
.container {{ max-width: 600px; margin: 0 auto; }}
.success {{ color: #4CAF50; }}
.error {{ color: #F44336; }}
.cancelled {{ color: #FF9800; }}
</style>
<script>
window.onload = function() {{
const status = {js_status};
const sessionToken = {js_session_token};
const userEmail = {js_user_email};
const providerType = {js_provider_type};
if (status === "success" && window.opener) {{
window.opener.postMessage({{
type: providerType + '_auth_success',
session_token: sessionToken,
user_email: userEmail
}}, '*');
setTimeout(() => window.close(), 3000);
}} else if (status === "cancelled" || status === "error") {{
setTimeout(() => window.close(), 3000);
}}
}};
</script>
</head>
<body>
<div class="container">
<h2>{provider} Authentication</h2>
<div class="{status}">
<p>{message}</p>
{f'<p>Connected as: {user_email}</p>' if status == 'success' else ''}
</div>
<p><small>You can close this window. {f"Your {provider} is now connected and ready to use." if status == 'success' else "Feel free to close this window."}</small></p>
</div>
</body>
</html>
"""
return make_response(html_content, 200, {'Content-Type': 'text/html'})
# Query params are attacker-controllable, so this page never
# carries a session token; the OAuth callback renders that itself.
return _render_callback_page(
request.args.get('status', 'error'),
request.args.get('message', ''),
request.args.get('provider', 'connector'),
)
except Exception as e:
current_app.logger.error(f"Error rendering callback status page: {e}")
return make_response("Authentication error occurred", 500, {'Content-Type': 'text/html'})
+10
View File
@@ -24,6 +24,7 @@ from docsgpt.security.zip_archive import (
ZipExtractionError,
ZipExtractionLimits,
)
from docsgpt.storage.db.repositories.connector_sessions import ConnectorSessionsRepository
from docsgpt.storage.db.repositories.idempotency import IdempotencyRepository
from docsgpt.storage.db.repositories.sources import SourcesRepository
from docsgpt.storage.db.source_config import SourceConfig
@@ -496,6 +497,15 @@ class UploadRemote(Resource):
),
400,
)
with db_readonly() as conn:
connector_session = ConnectorSessionsRepository(conn).get_by_session_token(session_token)
if not connector_session or connector_session.get("user_id") != user:
if scoped_key:
_release_claim(scoped_key)
return make_response(
jsonify({"success": False, "error": "Invalid or unauthorized session"}),
401,
)
# Process file_ids
file_ids = config.get("file_ids", [])
+3
View File
@@ -251,6 +251,9 @@ class Settings(BaseSettings):
CONNECTOR_REDIRECT_BASE_URI: Optional[str] = (
"http://127.0.0.1:7091/api/connectors/callback" ##add redirect url as it is to your provider's console(gcp)
)
# Comma-separated frontend origins allowed to receive connector OAuth results, e.g. https://docsgpt.example.com.
# The callback origin and OIDC_FRONTEND_URL are always allowed; a loopback callback also allows localhost:5173.
CONNECTOR_ALLOWED_ORIGINS: Optional[str] = None
# Microsoft Entra ID (Azure AD) integration
MICROSOFT_CLIENT_ID: Optional[str] = None # Azure AD Application (client) ID
@@ -307,11 +307,13 @@ class ConnectorSessionsRepository:
result = self._conn.execute(text(sql), params)
return result.rowcount > 0
def delete_by_session_token(self, session_token: str) -> bool:
def delete_by_session_token(self, session_token: str, user_id: str) -> bool:
"""Delete the session behind ``session_token`` only if ``user_id`` owns it."""
result = self._conn.execute(
text(
"DELETE FROM connector_sessions WHERE session_token = :token"
"DELETE FROM connector_sessions "
"WHERE session_token = :token AND user_id = :user_id"
),
{"token": session_token},
{"token": session_token, "user_id": user_id},
)
return result.rowcount > 0
+14
View File
@@ -5,6 +5,7 @@ import { useSelector } from 'react-redux';
import userService from '../api/services/userService';
import { useDarkTheme } from '../hooks';
import { selectToken } from '../preferences/preferenceSlice';
import { isTrustedConnectorMessage } from '../utils/connectorAuthUtils';
import { Button } from './ui/button';
interface ConnectorAuthProps {
@@ -34,6 +35,8 @@ const ConnectorAuth: React.FC<ConnectorAuthProps> = ({
const completedRef = useRef(false);
const intervalRef = useRef<number | null>(null);
const authWindowRef = useRef<Window | null>(null);
// Origin the OAuth callback page is served from, as reported by the backend.
const callbackOriginRef = useRef<string | null>(null);
// Hold the exact listener identity so unmount cleanup removes the same fn.
const messageHandlerRef = useRef<((event: MessageEvent) => void) | null>(
null,
@@ -54,6 +57,16 @@ const ConnectorAuth: React.FC<ConnectorAuthProps> = ({
};
const handleAuthMessage = (event: MessageEvent) => {
// Only the popup we opened, on the callback origin, may report a result.
if (
!isTrustedConnectorMessage(
event,
authWindowRef.current,
callbackOriginRef.current,
)
) {
return;
}
const successGeneric = event.data?.type === 'connector_auth_success';
const successProvider = event.data?.type === `${provider}_auth_success`;
const errorProvider = event.data?.type === `${provider}_auth_error`;
@@ -139,6 +152,7 @@ const ConnectorAuth: React.FC<ConnectorAuthProps> = ({
onError(t('modals.uploadDoc.connectors.auth.authCancelled'));
return;
}
callbackOriginRef.current = authData.callback_origin || null;
authWindow.location.href = authData.authorization_url;
messageHandlerRef.current = handleAuthMessage;
@@ -0,0 +1,58 @@
import { describe, expect, it } from 'vitest';
import { isTrustedConnectorMessage } from './connectorAuthUtils';
const popup = {} as Window;
const otherWindow = {} as Window;
describe('isTrustedConnectorMessage', () => {
it('rejects messages when no auth popup is open', () => {
expect(
isTrustedConnectorMessage(
{ source: popup, origin: 'https://api.example.com' },
null,
'https://api.example.com',
),
).toBe(false);
});
it('rejects messages that do not come from the auth popup', () => {
expect(
isTrustedConnectorMessage(
{ source: otherWindow, origin: 'https://api.example.com' },
popup,
'https://api.example.com',
),
).toBe(false);
});
it('rejects messages from the popup when its origin is unexpected', () => {
expect(
isTrustedConnectorMessage(
{ source: popup, origin: 'https://evil.example.com' },
popup,
'https://api.example.com',
),
).toBe(false);
});
it('accepts messages from the popup on the callback origin', () => {
expect(
isTrustedConnectorMessage(
{ source: popup, origin: 'https://api.example.com' },
popup,
'https://api.example.com',
),
).toBe(true);
});
it('falls back to the popup identity check when no origin is known', () => {
expect(
isTrustedConnectorMessage(
{ source: popup, origin: 'https://api.example.com' },
popup,
null,
),
).toBe(true);
});
});
+12
View File
@@ -0,0 +1,12 @@
/**
* Whether a window message is the OAuth result posted by the connector popup we opened.
* `expectedOrigin` is the callback origin reported by the backend, when known.
*/
export const isTrustedConnectorMessage = (
event: Pick<MessageEvent, 'origin' | 'source'>,
authWindow: Window | null,
expectedOrigin?: string | null,
): boolean => {
if (!authWindow || event.source !== authWindow) return false;
return !expectedOrigin || event.origin === expectedOrigin;
};
+282
View File
@@ -0,0 +1,282 @@
"""Connector OAuth hardening: popup message origin, token exposure, session ownership."""
import base64
import json
from contextlib import contextmanager
from unittest.mock import MagicMock, patch
import pytest
from flask import Flask
@pytest.fixture
def app():
return Flask(__name__)
@contextmanager
def _patch_db(conn, module="docsgpt.api.connector.routes"):
@contextmanager
def _yield():
yield conn
with patch(f"{module}.db_session", _yield), patch(f"{module}.db_readonly", _yield):
yield
def _encode_state(payload):
return base64.urlsafe_b64encode(json.dumps(payload).encode()).decode()
def _seed_session(pg_conn, user, token, provider="google_drive"):
from docsgpt.storage.db.repositories.connector_sessions import ConnectorSessionsRepository
repo = ConnectorSessionsRepository(pg_conn)
row = repo.upsert(user, provider, status="authorized")
repo.update(str(row["id"]), {"session_token": token})
return repo
class TestConnectorAllowedOrigins:
def test_collects_configured_origins(self):
from docsgpt.api.connector.routes import connector_allowed_origins
from docsgpt.core.settings import settings
with patch.object(settings, "CONNECTOR_ALLOWED_ORIGINS", "https://app.example.com/, https://b.example.com/x"), \
patch.object(settings, "OIDC_FRONTEND_URL", "https://sso.example.com/home"), \
patch.object(settings, "CONNECTOR_REDIRECT_BASE_URI", "https://api.example.com/api/connectors/callback"):
origins = connector_allowed_origins("https://api.example.com/")
assert set(origins) == {
"https://app.example.com",
"https://b.example.com",
"https://sso.example.com",
"https://api.example.com",
}
def test_rejects_wildcards_and_non_http_values(self):
from docsgpt.api.connector.routes import connector_allowed_origins
from docsgpt.core.settings import settings
with patch.object(settings, "CONNECTOR_ALLOWED_ORIGINS", "*, javascript:alert(1), null, not a url"), \
patch.object(settings, "OIDC_FRONTEND_URL", None), \
patch.object(settings, "CONNECTOR_REDIRECT_BASE_URI", "https://api.example.com/api/connectors/callback"):
origins = connector_allowed_origins("https://api.example.com/")
assert origins == ["https://api.example.com"]
def test_loopback_dev_frontend_allowed_only_for_loopback_callback(self):
from docsgpt.api.connector.routes import connector_allowed_origins
from docsgpt.core.settings import settings
with patch.object(settings, "CONNECTOR_ALLOWED_ORIGINS", None), \
patch.object(settings, "OIDC_FRONTEND_URL", None), \
patch.object(settings, "CONNECTOR_REDIRECT_BASE_URI", "http://127.0.0.1:7091/api/connectors/callback"):
local = connector_allowed_origins("http://127.0.0.1:7091/")
with patch.object(settings, "CONNECTOR_ALLOWED_ORIGINS", None), \
patch.object(settings, "OIDC_FRONTEND_URL", None), \
patch.object(settings, "CONNECTOR_REDIRECT_BASE_URI", "https://api.example.com/api/connectors/callback"):
public = connector_allowed_origins("http://localhost:7091/")
assert "http://localhost:5173" in local
assert "http://127.0.0.1:5173" in local
assert "http://localhost:5173" not in public
class TestCallbackStatusPage:
def test_never_posts_to_wildcard_origin(self, app):
from docsgpt.api.connector.routes import ConnectorCallbackStatus
with app.test_request_context(
"/api/connectors/callback-status?status=success&provider=google_drive"
):
r = ConnectorCallbackStatus().get()
body = r.get_data(as_text=True)
assert r.status_code == 200
assert "'*'" not in body
assert '"*"' not in body
def test_ignores_session_token_query_param(self, app):
from docsgpt.api.connector.routes import ConnectorCallbackStatus
with app.test_request_context(
"/api/connectors/callback-status?status=success&provider=google_drive"
"&session_token=attacker-supplied&user_email=evil@example.com"
):
r = ConnectorCallbackStatus().get()
body = r.get_data(as_text=True)
assert "attacker-supplied" not in body
assert "evil@example.com" not in body
class TestCallbackDeliversTokenSafely:
def test_success_renders_page_without_redirecting_token(self, app, pg_conn):
from docsgpt.api.connector.routes import ConnectorsCallback
from docsgpt.core.settings import settings
from docsgpt.storage.db.repositories.connector_sessions import ConnectorSessionsRepository
user = "u-cb-origin"
repo = ConnectorSessionsRepository(pg_conn)
pending = repo.upsert(user, "google_drive", status="pending")
state = _encode_state({"provider": "google_drive", "object_id": str(pending["id"])})
fake_auth = MagicMock()
fake_auth.exchange_code_for_tokens.return_value = {"access_token": "at"}
fake_auth.sanitize_token_info.return_value = {"access_token": "at"}
fake_auth.create_credentials_from_token_info.side_effect = RuntimeError("no creds")
with _patch_db(pg_conn), patch(
"docsgpt.api.connector.routes.ConnectorCreator.is_supported", return_value=True,
), patch(
"docsgpt.api.connector.routes.ConnectorCreator.create_auth", return_value=fake_auth,
), patch.object(
settings, "CONNECTOR_ALLOWED_ORIGINS", "https://app.example.com",
), app.test_request_context(
f"/api/connectors/callback?state={state}&code=auth-code",
base_url="https://api.example.com",
):
r = ConnectorsCallback().get()
token = repo.get_by_user_provider(user, "google_drive")["session_token"]
body = r.get_data(as_text=True)
assert token
assert r.status_code == 200
assert "Location" not in r.headers
assert token in body
assert '"https://app.example.com"' in body
assert "'*'" not in body
assert r.headers["Cache-Control"] == "no-store"
assert r.headers["Referrer-Policy"] == "no-referrer"
class TestAuthUrlReportsCallbackOrigin:
def test_includes_callback_origin(self, app, pg_conn):
from docsgpt.api.connector.routes import ConnectorAuth
from docsgpt.core.settings import settings
fake_auth = MagicMock()
fake_auth.get_authorization_url.return_value = "https://ex/auth?state=x"
with _patch_db(pg_conn), patch(
"docsgpt.api.connector.routes.ConnectorCreator.is_supported", return_value=True,
), patch(
"docsgpt.api.connector.routes.ConnectorCreator.create_auth", return_value=fake_auth,
), patch.object(
settings, "CONNECTOR_REDIRECT_BASE_URI", "https://api.example.com/api/connectors/callback",
), app.test_request_context("/api/connectors/auth?provider=google_drive"):
from flask import request
request.decoded_token = {"sub": "u-auth-origin"}
r = ConnectorAuth().get()
assert r.status_code == 200
assert r.json["callback_origin"] == "https://api.example.com"
class TestDisconnectOwnership:
def test_requires_authentication(self, app, pg_conn):
from docsgpt.api.connector.routes import ConnectorDisconnect
repo = _seed_session(pg_conn, "u-owner", "st-noauth")
with _patch_db(pg_conn), app.test_request_context(
"/api/connectors/disconnect", method="POST",
json={"provider": "google_drive", "session_token": "st-noauth"},
):
from flask import request
request.decoded_token = None
r = ConnectorDisconnect().post()
assert r.status_code == 401
assert repo.get_by_session_token("st-noauth") is not None
def test_cannot_delete_another_users_session(self, app, pg_conn):
from docsgpt.api.connector.routes import ConnectorDisconnect
repo = _seed_session(pg_conn, "u-victim", "st-victim")
with _patch_db(pg_conn), app.test_request_context(
"/api/connectors/disconnect", method="POST",
json={"provider": "google_drive", "session_token": "st-victim"},
):
from flask import request
request.decoded_token = {"sub": "u-attacker"}
ConnectorDisconnect().post()
assert repo.get_by_session_token("st-victim") is not None
def test_owner_can_delete_own_session(self, app, pg_conn):
from docsgpt.api.connector.routes import ConnectorDisconnect
repo = _seed_session(pg_conn, "u-self", "st-self")
with _patch_db(pg_conn), app.test_request_context(
"/api/connectors/disconnect", method="POST",
json={"provider": "google_drive", "session_token": "st-self"},
):
from flask import request
request.decoded_token = {"sub": "u-self"}
r = ConnectorDisconnect().post()
assert r.status_code == 200
assert repo.get_by_session_token("st-self") is None
class TestSyncSessionOwnership:
def test_rejects_foreign_session_token(self, app, pg_conn):
from docsgpt.api.connector.routes import ConnectorSync
from docsgpt.storage.db.repositories.sources import SourcesRepository
_seed_session(pg_conn, "u-victim-sync", "st-victim-sync")
attacker = "u-attacker-sync"
src = SourcesRepository(pg_conn).create(
"drive-src", user_id=attacker,
remote_data={"provider": "google_drive", "file_ids": ["f"], "folder_ids": []},
)
delay = MagicMock()
with _patch_db(pg_conn), patch(
"docsgpt.api.connector.routes.ingest_connector_task.delay", delay,
), app.test_request_context(
"/api/connectors/sync", method="POST",
json={"source_id": str(src["id"]), "session_token": "st-victim-sync"},
):
from flask import request
request.decoded_token = {"sub": attacker}
r = ConnectorSync().post()
assert r.status_code == 401
delay.assert_not_called()
class TestRemoteUploadSessionOwnership:
def _post(self, app, pg_conn, user, token, apply_mock):
from docsgpt.api.user.sources.upload import UploadRemote
with _patch_db(pg_conn, "docsgpt.api.user.sources.upload"), patch(
"docsgpt.api.user.sources.upload.ingest_connector_task.apply_async", apply_mock,
), app.test_request_context(
"/api/remote", method="POST",
data={
"user": user, "source": "google_drive", "name": "g",
"data": json.dumps({"session_token": token, "file_ids": ["f1"]}),
},
content_type="multipart/form-data",
):
from flask import request
request.decoded_token = {"sub": user}
return UploadRemote().post()
def test_rejects_foreign_session_token(self, app, pg_conn):
_seed_session(pg_conn, "u-victim-up", "st-victim-up")
apply_mock = MagicMock(return_value=MagicMock(id="t"))
r = self._post(app, pg_conn, "u-attacker-up", "st-victim-up", apply_mock)
assert r.status_code == 401
apply_mock.assert_not_called()
def test_accepts_own_session_token(self, app, pg_conn):
_seed_session(pg_conn, "u-owner-up", "st-owner-up")
apply_mock = MagicMock(return_value=MagicMock(id="t"))
r = self._post(app, pg_conn, "u-owner-up", "st-owner-up", apply_mock)
assert r.status_code == 200
apply_mock.assert_called_once()
+20 -2
View File
@@ -199,8 +199,11 @@ class TestConnectorsCallback:
f"/api/connectors/callback?state={state}&code=auth-code"
):
r = ConnectorsCallback().get()
assert r.status_code == 302
assert "status=success" in r.location
token = ConnectorSessionsRepository(pg_conn).get_by_user_provider(
"u-callback", "google_drive",
)["session_token"]
assert r.status_code == 200
assert token and token in r.get_data(as_text=True)
def test_token_exchange_failure_redirects_error(self, app, pg_conn):
from docsgpt.api.connector.routes import ConnectorsCallback
@@ -425,6 +428,8 @@ class TestConnectorDisconnect:
with app.test_request_context(
"/api/connectors/disconnect", method="POST", json={}
):
from flask import request
request.decoded_token = {"sub": "u"}
r = ConnectorDisconnect().post()
assert r.status_code == 400
@@ -444,9 +449,12 @@ class TestConnectorDisconnect:
method="POST",
json={"provider": "google_drive", "session_token": "st-disc"},
):
from flask import request
request.decoded_token = {"sub": user}
r = ConnectorDisconnect().post()
assert r.status_code == 200
assert r.json["success"] is True
assert repo.get_by_session_token("st-disc") is None
def test_disconnect_without_session_token_succeeds(self, app):
from docsgpt.api.connector.routes import ConnectorDisconnect
@@ -456,6 +464,8 @@ class TestConnectorDisconnect:
method="POST",
json={"provider": "google_drive"},
):
from flask import request
request.decoded_token = {"sub": "u"}
r = ConnectorDisconnect().post()
assert r.status_code == 200
@@ -530,6 +540,14 @@ class TestConnectorSync:
remote_data={"provider": "github", "file_ids": [], "folder_ids": []},
)
from docsgpt.storage.db.repositories.connector_sessions import (
ConnectorSessionsRepository,
)
repo = ConnectorSessionsRepository(pg_conn)
session = repo.upsert(user, "github", status="authorized")
repo.update(str(session["id"]), {"session_token": "st"})
fake_task = MagicMock(id="task-abc")
with _patch_db(pg_conn), patch(
"docsgpt.api.connector.routes.ingest_connector_task.delay",
@@ -597,6 +597,13 @@ class TestRemoteIdempotency:
pytest.skip("no connectors registered in this build")
connector_source = next(iter(supported))
from docsgpt.storage.db.repositories.connector_sessions import (
ConnectorSessionsRepository,
)
repo = ConnectorSessionsRepository(pg_conn)
session = repo.upsert("u", connector_source, status="authorized")
repo.update(str(session["id"]), {"session_token": "tok"})
with _patch_db(pg_conn), patch(
"docsgpt.api.user.sources.upload.ingest_connector_task.apply_async",
apply_mock,
+9 -2
View File
@@ -158,12 +158,17 @@ class TestSourcesUploadExtra:
response = UploadRemote().post()
assert response.status_code == 400
def test_remote_connector_triggers_task(self, app):
def test_remote_connector_triggers_task(self, app, pg_conn):
from docsgpt.api.user.sources.upload import UploadRemote
from docsgpt.storage.db.repositories.connector_sessions import ConnectorSessionsRepository
import json as _json
repo = ConnectorSessionsRepository(pg_conn)
session = repo.upsert("u", "google_drive", status="authorized")
repo.update(str(session["id"]), {"session_token": "st"})
fake_task = MagicMock(id="conn-t")
with patch(
with _patch_upload_db(pg_conn), patch(
"docsgpt.parser.connectors.connector_creator.ConnectorCreator"
".get_supported_connectors",
return_value={"google_drive"},
@@ -210,6 +215,8 @@ class TestConnectorExtra:
"/api/connectors/disconnect", method="POST",
json={"provider": "x", "session_token": "y"},
):
from flask import request
request.decoded_token = {"sub": "u"}
response = ConnectorDisconnect().post()
assert response.status_code == 500