Pin S3 and SMB connector peers

This commit is contained in:
2026-08-04 10:40:46 +02:00
parent 0c68e904cf
commit 92e649477f
15 changed files with 840 additions and 112 deletions
+50 -40
View File
@@ -5,7 +5,7 @@ from datetime import UTC, datetime
from unittest.mock import patch
from govoplan_files.backend.storage.connector_browse import ConnectorBrowseError, _smb_location, browse_connector_profile
from govoplan_files.backend.storage.connector_imports import ConnectorImportError, read_connector_file
from govoplan_files.backend.storage.connector_imports import read_connector_file
from govoplan_files.backend.storage.connector_profiles import ConnectorProfile, connector_profiles_from_payload
from govoplan_files.backend.storage.connector_providers import connector_provider_descriptors
@@ -23,6 +23,10 @@ class FakeS3Client:
self.list_objects_request: dict[str, object] | None = None
self.head_request: dict[str, object] | None = None
self.get_request: dict[str, object] | None = None
self.closed = False
def close(self) -> None:
self.closed = True
def list_objects_v2(self, **kwargs: object) -> dict[str, object]:
self.list_objects_request = dict(kwargs)
@@ -75,42 +79,39 @@ def s3_profile(**overrides: object) -> ConnectorProfile:
class ConnectorProviderTests(unittest.TestCase):
def test_smb_sdk_transport_fails_closed_before_client_creation_in_all_modes(self) -> None:
def test_smb_browse_uses_the_files_owned_pinned_connection_cache(self) -> None:
profile = ConnectorProfile(
id="smb",
label="SMB",
provider="smb",
endpoint_url="smb://files.example.test/share",
)
for allow_private, address in ((False, "93.184.216.34"), (True, "10.0.0.5")):
with self.subTest(allow_private=allow_private), patch.dict(
"os.environ",
{
"APP_ENV": "production",
"GOVOPLAN_CONNECTOR_ALLOW_PRIVATE_NETWORKS": str(allow_private).lower(),
},
), patch(
"govoplan_core.security.outbound_http.socket.getaddrinfo",
return_value=[(2, 1, 6, "", (address, 445))],
), patch("govoplan_files.backend.storage.connector_browse._smbclient_module") as sdk, self.assertRaisesRegex(
ConnectorBrowseError,
"redirects/referrals.*DNS/IP pinning",
):
browse_connector_profile(profile, path="")
sdk.assert_not_called()
with patch.dict(
"os.environ",
{"APP_ENV": "production", "GOVOPLAN_CONNECTOR_ALLOW_PRIVATE_NETWORKS": "false"},
), patch(
"govoplan_core.security.outbound_http.socket.getaddrinfo",
return_value=[(2, 1, 6, "", ("93.184.216.34", 445))],
), patch("govoplan_files.backend.storage.connector_browse._smbclient_module") as sdk:
sdk.return_value.scandir.return_value.__enter__.return_value = iter(())
self.assertEqual([], browse_connector_profile(profile, path=""))
def test_smb_explicit_ip_still_fails_closed_because_the_sdk_may_follow_referrals(self) -> None:
kwargs = sdk.return_value.scandir.call_args.kwargs
self.assertIsInstance(kwargs["connection_cache"], dict)
self.assertTrue(kwargs["require_signing"])
def test_smb_endpoint_preflight_applies_private_network_policy(self) -> None:
profile = ConnectorProfile(id="smb", label="SMB", provider="smb", endpoint_url="smb://10.0.0.5/share")
with patch.dict(
"os.environ",
{"APP_ENV": "production", "GOVOPLAN_CONNECTOR_ALLOW_PRIVATE_NETWORKS": "true"},
{"APP_ENV": "production", "GOVOPLAN_CONNECTOR_ALLOW_PRIVATE_NETWORKS": "false"},
), patch(
"govoplan_core.security.outbound_http.socket.getaddrinfo",
return_value=[(2, 1, 6, "", ("10.0.0.5", 445))],
), self.assertRaisesRegex(ConnectorBrowseError, "redirects/referrals.*DNS/IP pinning"):
), self.assertRaisesRegex(ConnectorBrowseError, "non-public network"):
_smb_location(profile)
def test_smb_import_surfaces_fail_closed_policy_as_an_import_error(self) -> None:
def test_smb_import_uses_the_same_pinned_connection_cache(self) -> None:
profile = ConnectorProfile(id="smb", label="SMB", provider="smb", endpoint_url="smb://10.0.0.5/share")
with patch.dict(
"os.environ",
@@ -118,12 +119,13 @@ class ConnectorProviderTests(unittest.TestCase):
), patch(
"govoplan_core.security.outbound_http.socket.getaddrinfo",
return_value=[(2, 1, 6, "", ("10.0.0.5", 445))],
), patch("govoplan_files.backend.storage.connector_imports._smbclient_module") as sdk, self.assertRaisesRegex(
ConnectorImportError,
"redirects/referrals.*DNS/IP pinning",
):
read_connector_file(profile, library_id="", path="notice.txt", max_bytes=1024)
sdk.assert_not_called()
), patch("govoplan_files.backend.storage.connector_imports._smbclient_module") as sdk:
sdk.return_value.stat.return_value.st_size = 4
sdk.return_value.open_file.return_value.__enter__.return_value.read.return_value = b"test"
downloaded = read_connector_file(profile, library_id="", path="notice.txt", max_bytes=1024)
self.assertEqual(b"test", downloaded.data)
self.assertIsInstance(sdk.return_value.stat.call_args.kwargs["connection_cache"], dict)
def test_provider_descriptors_include_s3_and_reserved_microsoft_providers(self) -> None:
descriptors = {descriptor.provider: descriptor for descriptor in connector_provider_descriptors()}
@@ -172,6 +174,7 @@ class ConnectorProviderTests(unittest.TestCase):
self.assertEqual("govoplan:root/report.xlsx", items[1].external_id)
self.assertEqual("root/report.xlsx", items[1].metadata["key"])
self.assertEqual("next-page", items[1].metadata["next_continuation_token"])
self.assertTrue(client.closed)
def test_s3_browse_lists_buckets_when_profile_has_no_bucket(self) -> None:
client = FakeS3Client()
@@ -181,27 +184,33 @@ class ConnectorProviderTests(unittest.TestCase):
self.assertEqual(["archive"], [item.path for item in items])
def test_s3_sdk_transport_fails_closed_before_client_creation_in_private_mode(self) -> None:
def test_s3_client_is_constructed_through_the_pinned_transport(self) -> None:
with patch.dict(
"os.environ",
{"APP_ENV": "production", "GOVOPLAN_CONNECTOR_ALLOW_PRIVATE_NETWORKS": "true"},
), patch(
"govoplan_core.security.outbound_http.socket.getaddrinfo",
return_value=[(2, 1, 6, "", ("127.0.0.1", 9000))],
), patch("govoplan_files.backend.storage.connector_browse.import_module") as importer, self.assertRaisesRegex(
ConnectorBrowseError,
"until that transport supports.*DNS/IP pinning",
):
), patch(
"govoplan_files.backend.storage.connector_browse.create_pinned_s3_client",
return_value=FakeS3Client(),
) as factory:
browse_connector_profile(s3_profile(), path="")
importer.assert_not_called()
def test_s3_sdk_endpoint_discovery_fails_closed(self) -> None:
with patch("govoplan_files.backend.storage.connector_browse.import_module") as importer, self.assertRaisesRegex(
ConnectorBrowseError,
"endpoint discovery.*cannot guarantee.*DNS/IP pinning",
):
kwargs = factory.call_args.kwargs
self.assertEqual("http://127.0.0.1:9000", kwargs["endpoint_url"])
self.assertEqual("access-key", kwargs["aws_access_key_id"])
self.assertEqual("secret-key", kwargs["aws_secret_access_key"])
self.assertEqual({}, kwargs["config"].proxies)
def test_s3_endpoint_discovery_uses_the_same_pinned_transport(self) -> None:
with patch(
"govoplan_files.backend.storage.connector_browse.create_pinned_s3_client",
return_value=FakeS3Client(),
) as factory:
browse_connector_profile(s3_profile(endpoint_url=None), path="")
importer.assert_not_called()
self.assertNotIn("endpoint_url", factory.call_args.kwargs)
def test_s3_import_downloads_object_and_preserves_remote_identity(self) -> None:
client = FakeS3Client()
@@ -217,6 +226,7 @@ class ConnectorProviderTests(unittest.TestCase):
self.assertEqual("govoplan:root/report.txt", downloaded.external_id)
self.assertEqual("s3://govoplan/root/report.txt", downloaded.external_url)
self.assertEqual("checksum", downloaded.metadata["checksum_sha256"])
self.assertTrue(client.closed)
if __name__ == "__main__":
+13 -6
View File
@@ -10,6 +10,9 @@ from govoplan_files.backend.storage.connector_visibility import (
connector_profile_usable_for_import,
visible_connector_profiles_for_actor,
)
from govoplan_files.backend.storage.connector_providers import (
connector_provider_descriptors,
)
def _profile(
@@ -168,12 +171,16 @@ class ConnectorVisibilityTests(unittest.TestCase):
self,
) -> None:
self.assertTrue(connector_profile_usable_for_import(_profile("webdav")))
self.assertFalse(
connector_profile_usable_for_import(_profile("s3", provider="s3"))
)
self.assertFalse(
connector_profile_usable_for_import(_profile("smb", provider="smb"))
)
descriptors = {
item.provider: item for item in connector_provider_descriptors()
}
for provider in ("s3", "smb"):
self.assertEqual(
descriptors[provider].installed,
connector_profile_usable_for_import(
_profile(provider, provider=provider)
),
)
self.assertFalse(
connector_profile_usable_for_import(
_profile("sharepoint", provider="sharepoint")
+1
View File
@@ -142,6 +142,7 @@ class FilesRuntimeDocumentationTests(unittest.TestCase):
base_path="/classified",
credential_mode="basic",
password_value="credential-secret",
secret_ref="runtime/connector-secret",
)
with patch(
"govoplan_files.backend.documentation.visible_connector_profiles_for_actor",
+3 -1
View File
@@ -4,6 +4,7 @@ import unittest
STATIC_TOPIC_IDS = {
"files.search.managed-content",
"files.workflow.organize-managed-files",
"files.workflow.find-and-download-files",
"files.workflow.share-managed-files",
@@ -152,7 +153,7 @@ class FilesManifestDocumentationTests(unittest.TestCase):
"/api/v1/files/connectors/credentials", {link.href for link in topic.links}
)
def test_operator_topic_covers_recovery_and_fail_closed_s3_smb(self) -> None:
def test_operator_topic_covers_recovery_and_pinned_s3_smb(self) -> None:
topic = self.topic(
"files.reference.integrity-recovery-and-fail-closed-transports"
)
@@ -163,6 +164,7 @@ class FilesManifestDocumentationTests(unittest.TestCase):
self.assertIn("SMB", topic.body)
self.assertIn("fail closed", topic.body)
self.assertIn("DFS referrals", topic.body)
self.assertIn("ambient credential discovery", topic.body)
self.assertIn("does not remove backend blob objects", topic.body)
self.assertIn("bounded resumable integrity scan", topic.body)
self.assertIn("quarantined", topic.body)
+294
View File
@@ -0,0 +1,294 @@
from __future__ import annotations
import socket
import sys
import threading
import types
import unittest
from importlib.util import find_spec
from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer
from unittest.mock import patch
from govoplan_core.security.outbound_http import (
OutboundHttpBlocked,
create_outbound_connection as core_create_outbound_connection,
)
from govoplan_files.backend.storage.sdk_peer_pinning import (
SdkPeerPinningError,
_botocore_transport_types,
create_pinned_s3_client,
install_pinned_smb_transport,
pinned_smb_connection_cache,
)
class _Socket:
def __init__(self) -> None:
self.connected_to: object | None = None
def settimeout(self, _value: object) -> None:
return None
def setsockopt(self, *_args: object) -> None:
return None
def connect(self, value: object) -> None:
self.connected_to = value
def close(self) -> None:
return None
class _S3Handler(BaseHTTPRequestHandler):
def do_GET(self) -> None: # noqa: N802 - stdlib handler contract
body = (
b'<?xml version="1.0" encoding="UTF-8"?>'
b'<ListAllMyBucketsResult xmlns="http://s3.amazonaws.com/doc/2006-03-01/">'
b'<Owner><ID>govoplan</ID><DisplayName>GovOPlaN</DisplayName></Owner>'
b'<Buckets><Bucket><Name>evidence</Name><CreationDate>2026-08-04T00:00:00Z</CreationDate>'
b'</Bucket></Buckets></ListAllMyBucketsResult>'
)
self.send_response(200)
self.send_header("Content-Type", "application/xml")
self.send_header("Content-Length", str(len(body)))
self.end_headers()
self.wfile.write(body)
def log_message(self, _format: str, *_args: object) -> None:
return None
@unittest.skipUnless(find_spec("botocore") and find_spec("boto3"), "boto3 extra is not installed")
class S3PeerPinningTests(unittest.TestCase):
def test_client_is_born_with_the_pinned_http_session(self) -> None:
client = create_pinned_s3_client(
endpoint_url="http://127.0.0.1:9000",
region_name="eu-central-1",
aws_access_key_id="access",
aws_secret_access_key="secret",
)
try:
session_type, _, _ = _botocore_transport_types()
self.assertIsInstance(client._endpoint.http_session, session_type)
self.assertEqual({}, client._endpoint.http_session._proxy_config._proxies)
finally:
client.close()
def test_each_connection_attempt_resolves_and_pins_again(self) -> None:
_, http_connection, _ = _botocore_transport_types()
connection = http_connection(host="objects.example.test", port=80)
sockets = [_Socket(), _Socket()]
with patch(
"govoplan_files.backend.storage.sdk_peer_pinning.create_outbound_connection",
side_effect=sockets,
) as connector:
self.assertIs(sockets[0], connection._new_conn())
self.assertIs(sockets[1], connection._new_conn())
self.assertEqual(2, connector.call_count)
self.assertEqual("objects.example.test", connector.call_args.args[0])
def test_real_sdk_request_uses_the_pinned_socket_path(self) -> None:
server = ThreadingHTTPServer(("127.0.0.1", 0), _S3Handler)
thread = threading.Thread(target=server.serve_forever, daemon=True)
thread.start()
endpoint = f"http://127.0.0.1:{server.server_port}"
try:
with patch.dict(
"os.environ",
{"APP_ENV": "production", "GOVOPLAN_CONNECTOR_ALLOW_PRIVATE_NETWORKS": "true"},
), patch(
"govoplan_files.backend.storage.sdk_peer_pinning.create_outbound_connection",
wraps=core_create_outbound_connection,
) as connector:
client = create_pinned_s3_client(
endpoint_url=endpoint,
region_name="eu-central-1",
aws_access_key_id="access",
aws_secret_access_key="secret",
)
try:
response = client.list_buckets()
finally:
client.close()
finally:
server.shutdown()
server.server_close()
thread.join(timeout=2)
self.assertEqual("evidence", response["Buckets"][0]["Name"])
self.assertEqual("127.0.0.1", connector.call_args.args[0])
def test_every_sdk_selected_origin_uses_a_pinned_pool(self) -> None:
session_type, http_connection, https_connection = _botocore_transport_types()
session = session_type(proxies={})
try:
first = session._manager.connection_from_url("http://one.example.test/root")
redirected = session._manager.connection_from_url("https://two.example.test/root")
self.assertIs(first.ConnectionCls, http_connection)
self.assertIs(redirected.ConnectionCls, https_connection)
finally:
session.close()
def test_tls_connection_retains_the_configured_authority_for_sni(self) -> None:
_, _, https_connection = _botocore_transport_types()
connection = https_connection(host="objects.example.test", port=443)
with patch(
"govoplan_files.backend.storage.sdk_peer_pinning.create_outbound_connection",
return_value=_Socket(),
) as connector:
connection._new_conn()
self.assertEqual("objects.example.test", connection.host)
self.assertEqual("objects.example.test", connection._dns_host)
self.assertEqual("objects.example.test", connector.call_args.args[0])
def test_mixed_answer_and_peer_change_fail_closed(self) -> None:
_, http_connection, _ = _botocore_transport_types()
connection = http_connection(host="objects.example.test", port=80)
records = [
[(socket.AF_INET, socket.SOCK_STREAM, 6, "", ("93.184.216.34", 80))],
[
(socket.AF_INET, socket.SOCK_STREAM, 6, "", ("93.184.216.34", 80)),
(socket.AF_INET, socket.SOCK_STREAM, 6, "", ("10.0.0.5", 80)),
],
]
with patch.dict(
"os.environ",
{"APP_ENV": "production", "GOVOPLAN_CONNECTOR_ALLOW_PRIVATE_NETWORKS": "false"},
), patch(
"govoplan_core.security.outbound_http.socket.getaddrinfo",
side_effect=records,
), patch(
"govoplan_core.security.outbound_http.socket.socket",
return_value=_Socket(),
):
connection._new_conn()
with self.assertRaisesRegex(Exception, "non-public network"):
connection._new_conn()
class _FakeTcp:
def __init__(self, server: str, port: int, timeout: float | None = None) -> None:
self.server = server
self.port = port
self.timeout = timeout
self.connected = False
self._sock = None
self._sock_lock = threading.Lock()
class _FakeConnection:
def __init__(self, _guid: object, server: str, port: int, *, require_signing: bool) -> None:
self.server = server
self.port = port
self.require_signing = require_signing
self.transport = _FakeTcp(server, port)
self.session_table: dict[str, object] = {}
def connect(self, *, timeout: float) -> None:
self.transport.timeout = timeout
self.transport.connect()
class _FakeSession:
def __init__(
self,
connection: _FakeConnection,
*,
username: str | None,
password: str | None,
require_encryption: bool,
auth_protocol: str,
) -> None:
self.connection = connection
self.username = username
self.password = password
self.encrypt_data = require_encryption
self.auth_protocol = auth_protocol
self.encrypt = require_encryption
def connect(self) -> None:
self.connection.session_table[self.username or "anonymous"] = self
class SmbPeerPinningTests(unittest.TestCase):
def setUp(self) -> None:
pinned_smb_connection_cache.cache_clear()
def _modules(self) -> tuple[types.ModuleType, dict[str, types.ModuleType]]:
smbclient = types.ModuleType("smbclient")
pool = types.ModuleType("smbclient._pool")
def original_register_session(
server: str,
username: str | None = None,
password: str | None = None,
port: int = 445,
encrypt: bool | None = None,
connection_timeout: float = 60,
connection_cache: dict[str, object] | None = None,
auth_protocol: str = "negotiate",
require_signing: bool = True,
) -> None:
del server, username, password, port, encrypt, connection_timeout
del connection_cache, auth_protocol, require_signing
pool.register_session = original_register_session
pool.ClientConfig = lambda: types.SimpleNamespace(client_guid="guid")
connection = types.ModuleType("smbprotocol.connection")
connection.Connection = _FakeConnection
connection.Tcp = _FakeTcp
session = types.ModuleType("smbprotocol.session")
session.Session = _FakeSession
transport = types.ModuleType("smbprotocol.transport")
transport.Tcp = _FakeTcp
return smbclient, {
"smbclient._pool": pool,
"smbprotocol.connection": connection,
"smbprotocol.session": session,
"smbprotocol.transport": transport,
}
def test_initial_reconnect_and_referral_hosts_are_each_pinned(self) -> None:
smbclient, modules = self._modules()
sockets = [_Socket(), _Socket(), _Socket()]
with patch.dict(sys.modules, modules), patch(
"govoplan_files.backend.storage.sdk_peer_pinning.create_outbound_connection",
side_effect=sockets,
) as connector:
install_pinned_smb_transport(smbclient)
register = modules["smbclient._pool"].register_session
cache: dict[str, object] = {}
first = register("files.example.test", connection_cache=cache)
first.connection.transport.connected = False
register("files.example.test", connection_cache=cache)
register("dfs-target.example.test", connection_cache=cache)
self.assertEqual(
["files.example.test", "files.example.test", "dfs-target.example.test"],
[call.args[0] for call in connector.call_args_list],
)
def test_referral_to_disallowed_peer_fails_before_session_creation(self) -> None:
smbclient, modules = self._modules()
with patch.dict(sys.modules, modules), patch(
"govoplan_files.backend.storage.sdk_peer_pinning.create_outbound_connection",
side_effect=OutboundHttpBlocked("non-public network"),
):
install_pinned_smb_transport(smbclient)
register = modules["smbclient._pool"].register_session
with self.assertRaisesRegex(ValueError, "non-public network"):
register("private-referral.example.test", connection_cache={})
def test_transport_tampering_after_installation_fails_closed(self) -> None:
smbclient, modules = self._modules()
with patch.dict(sys.modules, modules):
install_pinned_smb_transport(smbclient)
modules["smbprotocol.connection"].Tcp = _FakeTcp
with self.assertRaisesRegex(SdkPeerPinningError, "changed after peer pinning"):
install_pinned_smb_transport(smbclient)
if __name__ == "__main__":
unittest.main()
+2
View File
@@ -9,6 +9,7 @@ from sqlalchemy.orm import Session
from govoplan_access.backend.db.models import Account, Group, User
from govoplan_core.auth import ApiPrincipal
from govoplan_core.core.access import PrincipalRef
from govoplan_core.core.change_sequence import ChangeSequenceEntry
from govoplan_core.core.events import EventObjectRef, EventTenantRef, PlatformEvent
from govoplan_core.core.search import (
SearchAuthorizationRequest,
@@ -35,6 +36,7 @@ class FilesSearchSourceTests(unittest.TestCase):
FileAsset.__table__,
FileFolder.__table__,
FileShare.__table__,
ChangeSequenceEntry.__table__,
),
)
self.session = Session(self.engine)