Fix join being denied after being invited over federation

changelog.d/18075.bugfix

@@ -0,0 +1 @@
+Fix join being denied after being invited over federation. Also fixes other out-of-band membership transitions.

synapse/event_auth.py

@@ -566,6 +566,7 @@ def _is_membership_change_allowed(
     logger.debug(
         "_is_membership_change_allowed: %s",
         {
+            "caller_membership": caller.membership if caller else None,
             "caller_in_room": caller_in_room,
             "caller_invited": caller_invited,
             "caller_knocked": caller_knocked,
@@ -677,7 +678,8 @@ def _is_membership_change_allowed(
                 and join_rule == JoinRules.KNOCK_RESTRICTED
             )
         ):
-            if not caller_in_room and not caller_invited:
+            # You can only join the room if you are invited or are already in the room.
+            if not (caller_in_room or caller_invited):
                 raise AuthError(403, "You are not invited to this room.")
         else:
             # TODO (erikj): may_join list

synapse/events/__init__.py

@@ -42,7 +42,7 @@
 from typing_extensions import Literal
 from unpaddedbase64 import encode_base64
 
-from synapse.api.constants import RelationTypes
+from synapse.api.constants import EventTypes, RelationTypes
 from synapse.api.room_versions import EventFormatVersions, RoomVersion, RoomVersions
 from synapse.synapse_rust.events import EventInternalMetadata
 from synapse.types import JsonDict, StrCollection
@@ -325,12 +325,17 @@ def __str__(self) -> str:
     def __repr__(self) -> str:
         rejection = f"REJECTED={self.rejected_reason}, " if self.rejected_reason else ""
 
+        conditional_membership_string = ""
+        if self.get("type") == EventTypes.Member:
+            conditional_membership_string = f"membership={self.membership}, "
+
         return (
             f"<{self.__class__.__name__} "
             f"{rejection}"
             f"event_id={self.event_id}, "
             f"type={self.get('type')}, "
             f"state_key={self.get('state_key')}, "
+            f"{conditional_membership_string}"
             f"outlier={self.internal_metadata.is_outlier()}"
             ">"
         )

synapse/events/builder.py

@@ -24,7 +24,7 @@
 import attr
 from signedjson.types import SigningKey
 
-from synapse.api.constants import MAX_DEPTH
+from synapse.api.constants import MAX_DEPTH, EventTypes
 from synapse.api.room_versions import (
     KNOWN_EVENT_FORMAT_VERSIONS,
     EventFormatVersions,
@@ -109,6 +109,19 @@ def state_key(self) -> str:
     def is_state(self) -> bool:
         return self._state_key is not None
 
+    def is_mine_id(self, string: str) -> bool:
+        """Determines whether a user ID or room alias originates from this homeserver.
+
+        Returns:
+            `True` if the hostname part of the user ID or room alias matches this
+            homeserver.
+            `False` otherwise, or if the user ID or room alias is malformed.
+        """
+        localpart_hostname = string.split(":", 1)
+        if len(localpart_hostname) < 2:
+            return False
+        return localpart_hostname[1] == self._hostname
+
     async def build(
         self,
         prev_event_ids: List[str],
@@ -142,6 +155,46 @@ async def build(
                 self, state_ids
             )
 
+            # Check for out-of-band membership that may that may have been exposed on
+            # `/sync` but the events have not been de-outliered yet so they won't be
+            # part of the room state yet.
+            #
+            # This helps in situations where a remote homeserver invites a local user to
+            # a room that we're already participating in; and we've persisted the invite
+            # as an out-of-band membership (outlier), but it hasn't been pushed to us as
+            # part of a `/send` transaction yet and de-outliered. This also helps for
+            # any of the other out-of-band membership transitions.
+            #
+            # As an optimization, we could check if the room state already includes a
+            # non-`leave` membership event, then we can assume the membership event has
+            # been de-outliered and we don't need to check for an out-of-band
+            # membership. But we don't have the necessary information from a
+            # `StateMap[str]` and we'll just have to take the hit of this extra lookup
+            # for any membership event for now.
+            if self.type == EventTypes.Member and self.is_mine_id(self.state_key):
+                (
+                    _membership,
+                    member_event_id,
+                ) = await self._store.get_local_current_membership_for_user_in_room(
+                    user_id=self.state_key,
+                    room_id=self.room_id,
+                )
+                # There is no need to check if the membership is actually an
+                # out-of-band membership (`outlier`) as we would end up with the
+                # same result either way (adding the member event to the
+                # `auth_event_ids`).
+                if (
+                    member_event_id is not None
+                    # We only need to be careful about duplicating the event in the
+                    # `auth_event_ids` list (duplicate `type`/`state_key` is part of the
+                    # authorization rules)
+                    and member_event_id not in auth_event_ids
+                ):
+                    auth_event_ids.append(member_event_id)
+                    # Also make sure to point to the previous membership event that will
+                    # allow this one to happen so the computed state works out.
+                    prev_event_ids.append(member_event_id)
+
         format_version = self.room_version.event_format
         # The types of auth/prev events changes between event versions.
         prev_events: Union[StrCollection, List[Tuple[str, Dict[str, str]]]]

synapse/federation/federation_server.py

@@ -1256,7 +1256,6 @@ async def _process_incoming_pdus_in_room_inner(
         # has started processing).
         while True:
             async with lock:
-                logger.info("handling received PDU in room %s: %s", room_id, event)
                 try:
                     with nested_logging_context(event.event_id):
                         # We're taking out a lock within a lock, which could

synapse/handlers/federation_event.py

@@ -2272,8 +2272,8 @@ async def persist_events_and_notify(
                 event_and_contexts, backfilled=backfilled
             )
 
-            # After persistence we always need to notify replication there may
-            # be new data.
+            # After persistence, we always need to notify replication there may be new
+            # data (backfilled or not) because TODO.
             self._notifier.notify_replication()
 
             if self._ephemeral_messages_enabled:

tests/federation/test_federation_devices.py

@@ -0,0 +1,161 @@
+#
+# This file is licensed under the Affero General Public License (AGPL) version 3.
+#
+# Copyright (C) 2024 New Vector, Ltd
+#
+# This program is free software: you can redistribute it and/or modify
+# it under the terms of the GNU Affero General Public License as
+# published by the Free Software Foundation, either version 3 of the
+# License, or (at your option) any later version.
+#
+# See the GNU Affero General Public License for more details:
+# <https://www.gnu.org/licenses/agpl-3.0.html>.
+#
+# Originally licensed under the Apache License, Version 2.0:
+# <http://www.apache.org/licenses/LICENSE-2.0>.
+#
+# [This file includes modifications made by New Vector Limited]
+#
+#
+
+import logging
+from unittest.mock import AsyncMock, Mock
+
+from twisted.test.proto_helpers import MemoryReactor
+
+from synapse.handlers.device import DeviceListUpdater
+from synapse.server import HomeServer
+from synapse.types import JsonDict
+from synapse.util import Clock
+from synapse.util.retryutils import NotRetryingDestination
+
+from tests import unittest
+
+logger = logging.getLogger(__name__)
+
+
+class DeviceListResyncTestCase(unittest.HomeserverTestCase):
+    def prepare(self, reactor: MemoryReactor, clock: Clock, hs: HomeServer) -> None:
+        self.store = self.hs.get_datastores().main
+
+    def test_retry_device_list_resync(self) -> None:
+        """Tests that device lists are marked as stale if they couldn't be synced, and
+        that stale device lists are retried periodically.
+        """
+        remote_user_id = "@john:test_remote"
+        remote_origin = "test_remote"
+
+        # Track the number of attempts to resync the user's device list.
+        self.resync_attempts = 0
+
+        # When this function is called, increment the number of resync attempts (only if
+        # we're querying devices for the right user ID), then raise a
+        # NotRetryingDestination error to fail the resync gracefully.
+        def query_user_devices(
+            destination: str, user_id: str, timeout: int = 30000
+        ) -> JsonDict:
+            if user_id == remote_user_id:
+                self.resync_attempts += 1
+
+            raise NotRetryingDestination(0, 0, destination)
+
+        # Register the mock on the federation client.
+        federation_client = self.hs.get_federation_client()
+        federation_client.query_user_devices = Mock(side_effect=query_user_devices)  # type: ignore[method-assign]
+
+        # Register a mock on the store so that the incoming update doesn't fail because
+        # we don't share a room with the user.
+        self.store.get_rooms_for_user = AsyncMock(return_value=["!someroom:test"])
+
+        # Manually inject a fake device list update. We need this update to include at
+        # least one prev_id so that the user's device list will need to be retried.
+        device_list_updater = self.hs.get_device_handler().device_list_updater
+        assert isinstance(device_list_updater, DeviceListUpdater)
+        self.get_success(
+            device_list_updater.incoming_device_list_update(
+                origin=remote_origin,
+                edu_content={
+                    "deleted": False,
+                    "device_display_name": "Mobile",
+                    "device_id": "QBUAZIFURK",
+                    "prev_id": [5],
+                    "stream_id": 6,
+                    "user_id": remote_user_id,
+                },
+            )
+        )
+
+        # Check that there was one resync attempt.
+        self.assertEqual(self.resync_attempts, 1)
+
+        # Check that the resync attempt failed and caused the user's device list to be
+        # marked as stale.
+        need_resync = self.get_success(
+            self.store.get_user_ids_requiring_device_list_resync()
+        )
+        self.assertIn(remote_user_id, need_resync)
+
+        # Check that waiting for 30 seconds caused Synapse to retry resyncing the device
+        # list.
+        self.reactor.advance(30)
+        self.assertEqual(self.resync_attempts, 2)
+
+    def test_cross_signing_keys_retry(self) -> None:
+        """Tests that resyncing a device list correctly processes cross-signing keys from
+        the remote server.
+        """
+        remote_user_id = "@john:test_remote"
+        remote_master_key = "85T7JXPFBAySB/jwby4S3lBPTqY3+Zg53nYuGmu1ggY"
+        remote_self_signing_key = "QeIiFEjluPBtI7WQdG365QKZcFs9kqmHir6RBD0//nQ"
+
+        # Register mock device list retrieval on the federation client.
+        federation_client = self.hs.get_federation_client()
+        federation_client.query_user_devices = AsyncMock(  # type: ignore[method-assign]
+            return_value={
+                "user_id": remote_user_id,
+                "stream_id": 1,
+                "devices": [],
+                "master_key": {
+                    "user_id": remote_user_id,
+                    "usage": ["master"],
+                    "keys": {"ed25519:" + remote_master_key: remote_master_key},
+                },
+                "self_signing_key": {
+                    "user_id": remote_user_id,
+                    "usage": ["self_signing"],
+                    "keys": {
+                        "ed25519:" + remote_self_signing_key: remote_self_signing_key
+                    },
+                },
+            }
+        )
+
+        # Resync the device list.
+        device_handler = self.hs.get_device_handler()
+        self.get_success(
+            device_handler.device_list_updater.multi_user_device_resync(
+                [remote_user_id]
+            ),
+        )
+
+        # Retrieve the cross-signing keys for this user.
+        keys = self.get_success(
+            self.store.get_e2e_cross_signing_keys_bulk(user_ids=[remote_user_id]),
+        )
+        self.assertIn(remote_user_id, keys)
+        key = keys[remote_user_id]
+        assert key is not None
+
+        # Check that the master key is the one returned by the mock.
+        master_key = key["master"]
+        self.assertEqual(len(master_key["keys"]), 1)
+        self.assertTrue("ed25519:" + remote_master_key in master_key["keys"].keys())
+        self.assertTrue(remote_master_key in master_key["keys"].values())
+
+        # Check that the self-signing key is the one returned by the mock.
+        self_signing_key = key["self_signing"]
+        self.assertEqual(len(self_signing_key["keys"]), 1)
+        self.assertTrue(
+            "ed25519:" + remote_self_signing_key in self_signing_key["keys"].keys(),
+        )
+        self.assertTrue(remote_self_signing_key in self_signing_key["keys"].values())