From e03d7c5fd0577df5b62cd34559925c6cfe3e0360 Mon Sep 17 00:00:00 2001 From: Patrick Cloke Date: Fri, 7 Oct 2022 12:38:46 -0400 Subject: Remove support for the unstable dir flag on relations. (#14106) From MSC3715, this was unused by clients (and there was no way for clients to know it was supported). Matrix 1.4 defines the stable field. --- synapse/handlers/relations.py | 33 ++++++++++++++++----------------- 1 file changed, 16 insertions(+), 17 deletions(-) (limited to 'synapse/handlers') diff --git a/synapse/handlers/relations.py b/synapse/handlers/relations.py index 63bc6a7aa5..cc5e45c241 100644 --- a/synapse/handlers/relations.py +++ b/synapse/handlers/relations.py @@ -21,6 +21,7 @@ from synapse.api.errors import SynapseError from synapse.events import EventBase, relation_from_event from synapse.logging.opentracing import trace from synapse.storage.databases.main.relations import _RelatedEvent +from synapse.streams.config import PaginationConfig from synapse.types import JsonDict, Requester, StreamToken, UserID from synapse.visibility import filter_events_for_client @@ -72,13 +73,10 @@ class RelationsHandler: requester: Requester, event_id: str, room_id: str, + pagin_config: PaginationConfig, + include_original_event: bool, relation_type: Optional[str] = None, event_type: Optional[str] = None, - limit: int = 5, - direction: str = "b", - from_token: Optional[StreamToken] = None, - to_token: Optional[StreamToken] = None, - include_original_event: bool = False, ) -> JsonDict: """Get related events of a event, ordered by topological ordering. @@ -88,14 +86,10 @@ class RelationsHandler: requester: The user requesting the relations. event_id: Fetch events that relate to this event ID. room_id: The room the event belongs to. + pagin_config: The pagination config rules to apply, if any. + include_original_event: Whether to include the parent event. relation_type: Only fetch events with this relation type, if given. event_type: Only fetch events with this event type, if given. - limit: Only fetch the most recent `limit` events. - direction: Whether to fetch the most recent first (`"b"`) or the - oldest first (`"f"`). - from_token: Fetch rows from the given token, or from the start if None. - to_token: Fetch rows up to the given token, or up to the end if None. - include_original_event: Whether to include the parent event. Returns: The pagination chunk. @@ -114,6 +108,9 @@ class RelationsHandler: if event is None: raise SynapseError(404, "Unknown parent event.") + # TODO Update pagination config to not allow None limits. + assert pagin_config.limit is not None + # Note that ignored users are not passed into get_relations_for_event # below. Ignored users are handled in filter_events_for_client (and by # not passing them in here we should get a better cache hit rate). @@ -123,10 +120,10 @@ class RelationsHandler: room_id=room_id, relation_type=relation_type, event_type=event_type, - limit=limit, - direction=direction, - from_token=from_token, - to_token=to_token, + limit=pagin_config.limit, + direction=pagin_config.direction, + from_token=pagin_config.from_token, + to_token=pagin_config.to_token, ) events = await self._main_store.get_events_as_list( @@ -162,8 +159,10 @@ class RelationsHandler: if next_token: return_value["next_batch"] = await next_token.to_string(self._main_store) - if from_token: - return_value["prev_batch"] = await from_token.to_string(self._main_store) + if pagin_config.from_token: + return_value["prev_batch"] = await pagin_config.from_token.to_string( + self._main_store + ) return return_value -- cgit 1.5.1 From a86b2f6837f0a067b0a014fbf5140e8773b8da2e Mon Sep 17 00:00:00 2001 From: Shay Date: Tue, 11 Oct 2022 11:18:45 -0700 Subject: Fix a bug where redactions were not being sent over federation if we did not have the original event. (#13813) --- changelog.d/13813.bugfix | 1 + synapse/federation/sender/__init__.py | 29 +++++++++++++++++-------- synapse/handlers/appservice.py | 9 +++++--- synapse/storage/databases/main/events_worker.py | 15 +++++++++---- synapse/storage/databases/main/stream.py | 28 +++++++++++------------- tests/handlers/test_appservice.py | 18 +++++++++------ 6 files changed, 62 insertions(+), 38 deletions(-) create mode 100644 changelog.d/13813.bugfix (limited to 'synapse/handlers') diff --git a/changelog.d/13813.bugfix b/changelog.d/13813.bugfix new file mode 100644 index 0000000000..23388788ff --- /dev/null +++ b/changelog.d/13813.bugfix @@ -0,0 +1 @@ +Fix a long-standing bug where redactions were not being sent over federation if we did not have the original event. diff --git a/synapse/federation/sender/__init__.py b/synapse/federation/sender/__init__.py index a6cb3ba58f..774ecd81b6 100644 --- a/synapse/federation/sender/__init__.py +++ b/synapse/federation/sender/__init__.py @@ -353,21 +353,25 @@ class FederationSender(AbstractFederationSender): last_token = await self.store.get_federation_out_pos("events") ( next_token, - events, event_to_received_ts, - ) = await self.store.get_all_new_events_stream( + ) = await self.store.get_all_new_event_ids_stream( last_token, self._last_poked_id, limit=100 ) + event_ids = event_to_received_ts.keys() + event_entries = await self.store.get_unredacted_events_from_cache_or_db( + event_ids + ) + logger.debug( "Handling %i -> %i: %i events to send (current id %i)", last_token, next_token, - len(events), + len(event_entries), self._last_poked_id, ) - if not events and next_token >= self._last_poked_id: + if not event_entries and next_token >= self._last_poked_id: logger.debug("All events processed") break @@ -508,8 +512,14 @@ class FederationSender(AbstractFederationSender): await handle_event(event) events_by_room: Dict[str, List[EventBase]] = {} - for event in events: - events_by_room.setdefault(event.room_id, []).append(event) + + for event_id in event_ids: + # `event_entries` is unsorted, so we have to iterate over `event_ids` + # to ensure the events are in the right order + event_cache = event_entries.get(event_id) + if event_cache: + event = event_cache.event + events_by_room.setdefault(event.room_id, []).append(event) await make_deferred_yieldable( defer.gatherResults( @@ -524,9 +534,10 @@ class FederationSender(AbstractFederationSender): logger.debug("Successfully handled up to %i", next_token) await self.store.update_federation_out_pos("events", next_token) - if events: + if event_entries: now = self.clock.time_msec() - ts = event_to_received_ts[events[-1].event_id] + last_id = next(reversed(event_ids)) + ts = event_to_received_ts[last_id] assert ts is not None synapse.metrics.event_processing_lag.labels( @@ -536,7 +547,7 @@ class FederationSender(AbstractFederationSender): "federation_sender" ).set(ts) - events_processed_counter.inc(len(events)) + events_processed_counter.inc(len(event_entries)) event_processing_loop_room_count.labels("federation_sender").inc( len(events_by_room) diff --git a/synapse/handlers/appservice.py b/synapse/handlers/appservice.py index 203b62e015..66f5b8d108 100644 --- a/synapse/handlers/appservice.py +++ b/synapse/handlers/appservice.py @@ -109,10 +109,13 @@ class ApplicationServicesHandler: last_token = await self.store.get_appservice_last_pos() ( upper_bound, - events, event_to_received_ts, - ) = await self.store.get_all_new_events_stream( - last_token, self.current_max, limit=100, get_prev_content=True + ) = await self.store.get_all_new_event_ids_stream( + last_token, self.current_max, limit=100 + ) + + events = await self.store.get_events_as_list( + event_to_received_ts.keys(), get_prev_content=True ) events_by_room: Dict[str, List[EventBase]] = {} diff --git a/synapse/storage/databases/main/events_worker.py b/synapse/storage/databases/main/events_worker.py index 7cdc9fe98f..d4104462b5 100644 --- a/synapse/storage/databases/main/events_worker.py +++ b/synapse/storage/databases/main/events_worker.py @@ -474,7 +474,7 @@ class EventsWorkerStore(SQLBaseStore): return [] # there may be duplicates so we cast the list to a set - event_entry_map = await self._get_events_from_cache_or_db( + event_entry_map = await self.get_unredacted_events_from_cache_or_db( set(event_ids), allow_rejected=allow_rejected ) @@ -509,7 +509,9 @@ class EventsWorkerStore(SQLBaseStore): continue redacted_event_id = entry.event.redacts - event_map = await self._get_events_from_cache_or_db([redacted_event_id]) + event_map = await self.get_unredacted_events_from_cache_or_db( + [redacted_event_id] + ) original_event_entry = event_map.get(redacted_event_id) if not original_event_entry: # we don't have the redacted event (or it was rejected). @@ -588,11 +590,16 @@ class EventsWorkerStore(SQLBaseStore): return events @cancellable - async def _get_events_from_cache_or_db( - self, event_ids: Iterable[str], allow_rejected: bool = False + async def get_unredacted_events_from_cache_or_db( + self, + event_ids: Iterable[str], + allow_rejected: bool = False, ) -> Dict[str, EventCacheEntry]: """Fetch a bunch of events from the cache or the database. + Note that the events pulled by this function will not have any redactions + applied, and no guarantee is made about the ordering of the events returned. + If events are pulled from the database, they will be cached for future lookups. Unknown events are omitted from the response. diff --git a/synapse/storage/databases/main/stream.py b/synapse/storage/databases/main/stream.py index 530f04e149..ffeb2b3683 100644 --- a/synapse/storage/databases/main/stream.py +++ b/synapse/storage/databases/main/stream.py @@ -1024,28 +1024,31 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): "after": {"event_ids": events_after, "token": end_token}, } - async def get_all_new_events_stream( - self, from_id: int, current_id: int, limit: int, get_prev_content: bool = False - ) -> Tuple[int, List[EventBase], Dict[str, Optional[int]]]: + async def get_all_new_event_ids_stream( + self, + from_id: int, + current_id: int, + limit: int, + ) -> Tuple[int, Dict[str, Optional[int]]]: """Get all new events - Returns all events with from_id < stream_ordering <= current_id. + Returns all event ids with from_id < stream_ordering <= current_id. Args: from_id: the stream_ordering of the last event we processed current_id: the stream_ordering of the most recently processed event limit: the maximum number of events to return - get_prev_content: whether to fetch previous event content Returns: - A tuple of (next_id, events, event_to_received_ts), where `next_id` + A tuple of (next_id, event_to_received_ts), where `next_id` is the next value to pass as `from_id` (it will either be the stream_ordering of the last returned event, or, if fewer than `limit` events were found, the `current_id`). The `event_to_received_ts` is - a dictionary mapping event ID to the event `received_ts`. + a dictionary mapping event ID to the event `received_ts`, sorted by ascending + stream_ordering. """ - def get_all_new_events_stream_txn( + def get_all_new_event_ids_stream_txn( txn: LoggingTransaction, ) -> Tuple[int, Dict[str, Optional[int]]]: sql = ( @@ -1070,15 +1073,10 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): return upper_bound, event_to_received_ts upper_bound, event_to_received_ts = await self.db_pool.runInteraction( - "get_all_new_events_stream", get_all_new_events_stream_txn - ) - - events = await self.get_events_as_list( - event_to_received_ts.keys(), - get_prev_content=get_prev_content, + "get_all_new_event_ids_stream", get_all_new_event_ids_stream_txn ) - return upper_bound, events, event_to_received_ts + return upper_bound, event_to_received_ts async def get_federation_out_pos(self, typ: str) -> int: if self._need_to_reset_federation_stream_positions: diff --git a/tests/handlers/test_appservice.py b/tests/handlers/test_appservice.py index af24c4984d..7e4570f990 100644 --- a/tests/handlers/test_appservice.py +++ b/tests/handlers/test_appservice.py @@ -76,9 +76,13 @@ class AppServiceHandlerTestCase(unittest.TestCase): event = Mock( sender="@someone:anywhere", type="m.room.message", room_id="!foo:bar" ) - self.mock_store.get_all_new_events_stream.side_effect = [ - make_awaitable((0, [], {})), - make_awaitable((1, [event], {event.event_id: 0})), + self.mock_store.get_all_new_event_ids_stream.side_effect = [ + make_awaitable((0, {})), + make_awaitable((1, {event.event_id: 0})), + ] + self.mock_store.get_events_as_list.side_effect = [ + make_awaitable([]), + make_awaitable([event]), ] self.handler.notify_interested_services(RoomStreamToken(None, 1)) @@ -95,10 +99,10 @@ class AppServiceHandlerTestCase(unittest.TestCase): event = Mock(sender=user_id, type="m.room.message", room_id="!foo:bar") self.mock_as_api.query_user.return_value = make_awaitable(True) - self.mock_store.get_all_new_events_stream.side_effect = [ - make_awaitable((0, [event], {event.event_id: 0})), + self.mock_store.get_all_new_event_ids_stream.side_effect = [ + make_awaitable((0, {event.event_id: 0})), ] - + self.mock_store.get_events_as_list.side_effect = [make_awaitable([event])] self.handler.notify_interested_services(RoomStreamToken(None, 0)) self.mock_as_api.query_user.assert_called_once_with(services[0], user_id) @@ -112,7 +116,7 @@ class AppServiceHandlerTestCase(unittest.TestCase): event = Mock(sender=user_id, type="m.room.message", room_id="!foo:bar") self.mock_as_api.query_user.return_value = make_awaitable(True) - self.mock_store.get_all_new_events_stream.side_effect = [ + self.mock_store.get_all_new_event_ids_stream.side_effect = [ make_awaitable((0, [event], {event.event_id: 0})), ] -- cgit 1.5.1 From 9c23442ac909afe3d827534b00d52ee182d2f423 Mon Sep 17 00:00:00 2001 From: Andrew Morgan <1342360+anoadragon453@users.noreply.github.com> Date: Wed, 12 Oct 2022 14:37:20 +0100 Subject: Correct field name for stripped state events when knocking. `knock_state_events` -> `knock_room_state` (#14102) --- changelog.d/14102.bugfix | 1 + synapse/federation/federation_client.py | 2 +- synapse/federation/federation_server.py | 9 ++++++++- synapse/handlers/federation.py | 20 ++++++++++++++++---- 4 files changed, 26 insertions(+), 6 deletions(-) create mode 100644 changelog.d/14102.bugfix (limited to 'synapse/handlers') diff --git a/changelog.d/14102.bugfix b/changelog.d/14102.bugfix new file mode 100644 index 0000000000..d71e108f7c --- /dev/null +++ b/changelog.d/14102.bugfix @@ -0,0 +1 @@ +Fix a bug introduced in Synapse v1.37.0 in which an incorrect key name was used for sending and receiving room metadata when knocking on a room. \ No newline at end of file diff --git a/synapse/federation/federation_client.py b/synapse/federation/federation_client.py index 4dca711cd2..b220ab43fc 100644 --- a/synapse/federation/federation_client.py +++ b/synapse/federation/federation_client.py @@ -1294,7 +1294,7 @@ class FederationClient(FederationBase): return resp[1] async def send_knock(self, destinations: List[str], pdu: EventBase) -> JsonDict: - """Attempts to send a knock event to given a list of servers. Iterates + """Attempts to send a knock event to a given list of servers. Iterates through the list until one attempt succeeds. Doing so will cause the remote server to add the event to the graph, diff --git a/synapse/federation/federation_server.py b/synapse/federation/federation_server.py index 907940e19e..28097664b4 100644 --- a/synapse/federation/federation_server.py +++ b/synapse/federation/federation_server.py @@ -824,7 +824,14 @@ class FederationServer(FederationBase): context, self._room_prejoin_state_types ) ) - return {"knock_state_events": stripped_room_state} + return { + "knock_room_state": stripped_room_state, + # Since v1.37, Synapse incorrectly used "knock_state_events" for this field. + # Thus, we also populate a 'knock_state_events' with the same content to + # support old instances. + # See https://github.com/matrix-org/synapse/issues/14088. + "knock_state_events": stripped_room_state, + } async def _on_send_membership_event( self, origin: str, content: JsonDict, membership_type: str, room_id: str diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index 986ffed3d5..44e70c6c3c 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -781,15 +781,27 @@ class FederationHandler: # Send the signed event back to the room, and potentially receive some # further information about the room in the form of partial state events - stripped_room_state = await self.federation_client.send_knock( - target_hosts, event - ) + knock_response = await self.federation_client.send_knock(target_hosts, event) # Store any stripped room state events in the "unsigned" key of the event. # This is a bit of a hack and is cribbing off of invites. Basically we # store the room state here and retrieve it again when this event appears # in the invitee's sync stream. It is stripped out for all other local users. - event.unsigned["knock_room_state"] = stripped_room_state["knock_state_events"] + stripped_room_state = ( + knock_response.get("knock_room_state") + # Since v1.37, Synapse incorrectly used "knock_state_events" for this field. + # Thus, we also check for a 'knock_state_events' to support old instances. + # See https://github.com/matrix-org/synapse/issues/14088. + or knock_response.get("knock_state_events") + ) + + if stripped_room_state is None: + raise KeyError( + "Missing 'knock_room_state' (or legacy 'knock_state_events') field in " + "send_knock response" + ) + + event.unsigned["knock_room_state"] = stripped_room_state context = EventContext.for_outlier(self._storage_controllers) stream_id = await self._federation_event_handler.persist_events_and_notify( -- cgit 1.5.1 From b6baa46db078c3ef9e6c5751bccb8d2e1c5c5402 Mon Sep 17 00:00:00 2001 From: Shay Date: Wed, 12 Oct 2022 11:01:00 -0700 Subject: Fix a bug where the joined hosts for a given event were not being properly cached (#14125) --- changelog.d/14125.bugfix | 1 + synapse/handlers/federation_event.py | 4 +- synapse/handlers/message.py | 91 +++++++++++++++++++----------------- 3 files changed, 51 insertions(+), 45 deletions(-) create mode 100644 changelog.d/14125.bugfix (limited to 'synapse/handlers') diff --git a/changelog.d/14125.bugfix b/changelog.d/14125.bugfix new file mode 100644 index 0000000000..852f00ebb9 --- /dev/null +++ b/changelog.d/14125.bugfix @@ -0,0 +1 @@ +Fix a bug introduced in v1.69.0rc1 where the joined hosts for a given event were not being properly cached. diff --git a/synapse/handlers/federation_event.py b/synapse/handlers/federation_event.py index da319943cc..f382961099 100644 --- a/synapse/handlers/federation_event.py +++ b/synapse/handlers/federation_event.py @@ -414,7 +414,9 @@ class FederationEventHandler: # First, precalculate the joined hosts so that the federation sender doesn't # need to. - await self._event_creation_handler.cache_joined_hosts_for_event(event, context) + await self._event_creation_handler.cache_joined_hosts_for_events( + [(event, context)] + ) await self._check_for_soft_fail(event, context=context, origin=origin) await self._run_push_actions_and_persist_event(event, context) diff --git a/synapse/handlers/message.py b/synapse/handlers/message.py index da1acea275..4e55ebba0b 100644 --- a/synapse/handlers/message.py +++ b/synapse/handlers/message.py @@ -1390,7 +1390,7 @@ class EventCreationHandler: extra_users=extra_users, ), run_in_background( - self.cache_joined_hosts_for_event, event, context + self.cache_joined_hosts_for_events, events_and_context ).addErrback( log_failure, "cache_joined_hosts_for_event failed" ), @@ -1491,62 +1491,65 @@ class EventCreationHandler: await self.store.remove_push_actions_from_staging(event.event_id) raise - async def cache_joined_hosts_for_event( - self, event: EventBase, context: EventContext + async def cache_joined_hosts_for_events( + self, events_and_context: List[Tuple[EventBase, EventContext]] ) -> None: - """Precalculate the joined hosts at the event, when using Redis, so that + """Precalculate the joined hosts at each of the given events, when using Redis, so that external federation senders don't have to recalculate it themselves. """ - if not self._external_cache.is_enabled(): - return - - # If external cache is enabled we should always have this. - assert self._external_cache_joined_hosts_updates is not None + for event, _ in events_and_context: + if not self._external_cache.is_enabled(): + return - # We actually store two mappings, event ID -> prev state group, - # state group -> joined hosts, which is much more space efficient - # than event ID -> joined hosts. - # - # Note: We have to cache event ID -> prev state group, as we don't - # store that in the DB. - # - # Note: We set the state group -> joined hosts cache if it hasn't been - # set for a while, so that the expiry time is reset. + # If external cache is enabled we should always have this. + assert self._external_cache_joined_hosts_updates is not None - state_entry = await self.state.resolve_state_groups_for_events( - event.room_id, event_ids=event.prev_event_ids() - ) + # We actually store two mappings, event ID -> prev state group, + # state group -> joined hosts, which is much more space efficient + # than event ID -> joined hosts. + # + # Note: We have to cache event ID -> prev state group, as we don't + # store that in the DB. + # + # Note: We set the state group -> joined hosts cache if it hasn't been + # set for a while, so that the expiry time is reset. - if state_entry.state_group: - await self._external_cache.set( - "event_to_prev_state_group", - event.event_id, - state_entry.state_group, - expiry_ms=60 * 60 * 1000, + state_entry = await self.state.resolve_state_groups_for_events( + event.room_id, event_ids=event.prev_event_ids() ) - if state_entry.state_group in self._external_cache_joined_hosts_updates: - return + if state_entry.state_group: + await self._external_cache.set( + "event_to_prev_state_group", + event.event_id, + state_entry.state_group, + expiry_ms=60 * 60 * 1000, + ) - state = await state_entry.get_state( - self._storage_controllers.state, StateFilter.all() - ) - with opentracing.start_active_span("get_joined_hosts"): - joined_hosts = await self.store.get_joined_hosts( - event.room_id, state, state_entry + if state_entry.state_group in self._external_cache_joined_hosts_updates: + return + + state = await state_entry.get_state( + self._storage_controllers.state, StateFilter.all() ) + with opentracing.start_active_span("get_joined_hosts"): + joined_hosts = await self.store.get_joined_hosts( + event.room_id, state, state_entry + ) - # Note that the expiry times must be larger than the expiry time in - # _external_cache_joined_hosts_updates. - await self._external_cache.set( - "get_joined_hosts", - str(state_entry.state_group), - list(joined_hosts), - expiry_ms=60 * 60 * 1000, - ) + # Note that the expiry times must be larger than the expiry time in + # _external_cache_joined_hosts_updates. + await self._external_cache.set( + "get_joined_hosts", + str(state_entry.state_group), + list(joined_hosts), + expiry_ms=60 * 60 * 1000, + ) - self._external_cache_joined_hosts_updates[state_entry.state_group] = None + self._external_cache_joined_hosts_updates[ + state_entry.state_group + ] = None async def _validate_canonical_alias( self, -- cgit 1.5.1 From 3bbe532abb7bfc41467597731ac1a18c0331f539 Mon Sep 17 00:00:00 2001 From: Patrick Cloke Date: Thu, 13 Oct 2022 08:02:11 -0400 Subject: Add an API for listing threads in a room. (#13394) Implement the /threads endpoint from MSC3856. This is currently unstable and behind an experimental configuration flag. It includes a background update to backfill data, results from the /threads endpoint will be partial until that finishes. --- changelog.d/13394.feature | 1 + synapse/_scripts/synapse_port_db.py | 2 + synapse/config/experimental.py | 3 + synapse/handlers/relations.py | 86 ++++++++++- synapse/rest/client/relations.py | 50 ++++++- synapse/storage/databases/main/cache.py | 1 + synapse/storage/databases/main/events.py | 38 ++++- synapse/storage/databases/main/relations.py | 166 ++++++++++++++++++++- .../schema/main/delta/73/09threads_table.sql | 30 ++++ tests/rest/client/test_relations.py | 151 +++++++++++++++++++ 10 files changed, 522 insertions(+), 6 deletions(-) create mode 100644 changelog.d/13394.feature create mode 100644 synapse/storage/schema/main/delta/73/09threads_table.sql (limited to 'synapse/handlers') diff --git a/changelog.d/13394.feature b/changelog.d/13394.feature new file mode 100644 index 0000000000..68de079cf3 --- /dev/null +++ b/changelog.d/13394.feature @@ -0,0 +1 @@ +Experimental support for [MSC3856](https://github.com/matrix-org/matrix-spec-proposals/pull/3856): threads list API. diff --git a/synapse/_scripts/synapse_port_db.py b/synapse/_scripts/synapse_port_db.py index 5fa599e70e..d850e54e17 100755 --- a/synapse/_scripts/synapse_port_db.py +++ b/synapse/_scripts/synapse_port_db.py @@ -72,6 +72,7 @@ from synapse.storage.databases.main.registration import ( RegistrationBackgroundUpdateStore, find_max_generated_user_id_localpart, ) +from synapse.storage.databases.main.relations import RelationsWorkerStore from synapse.storage.databases.main.room import RoomBackgroundUpdateStore from synapse.storage.databases.main.roommember import RoomMemberBackgroundUpdateStore from synapse.storage.databases.main.search import SearchBackgroundUpdateStore @@ -206,6 +207,7 @@ class Store( PusherWorkerStore, PresenceBackgroundUpdateStore, ReceiptsBackgroundUpdateStore, + RelationsWorkerStore, ): def execute(self, f: Callable[..., R], *args: Any, **kwargs: Any) -> Awaitable[R]: return self.db_pool.runInteraction(f.__name__, f, *args, **kwargs) diff --git a/synapse/config/experimental.py b/synapse/config/experimental.py index f44655516e..1860006536 100644 --- a/synapse/config/experimental.py +++ b/synapse/config/experimental.py @@ -101,6 +101,9 @@ class ExperimentalConfig(Config): # MSC3848: Introduce errcodes for specific event sending failures self.msc3848_enabled: bool = experimental.get("msc3848_enabled", False) + # MSC3856: Threads list API + self.msc3856_enabled: bool = experimental.get("msc3856_enabled", False) + # MSC3852: Expose last seen user agent field on /_matrix/client/v3/devices. self.msc3852_enabled: bool = experimental.get("msc3852_enabled", False) diff --git a/synapse/handlers/relations.py b/synapse/handlers/relations.py index cc5e45c241..1fdd7a10bc 100644 --- a/synapse/handlers/relations.py +++ b/synapse/handlers/relations.py @@ -11,6 +11,7 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. +import enum import logging from typing import TYPE_CHECKING, Dict, FrozenSet, Iterable, List, Optional, Tuple @@ -20,7 +21,7 @@ from synapse.api.constants import RelationTypes from synapse.api.errors import SynapseError from synapse.events import EventBase, relation_from_event from synapse.logging.opentracing import trace -from synapse.storage.databases.main.relations import _RelatedEvent +from synapse.storage.databases.main.relations import ThreadsNextBatch, _RelatedEvent from synapse.streams.config import PaginationConfig from synapse.types import JsonDict, Requester, StreamToken, UserID from synapse.visibility import filter_events_for_client @@ -32,6 +33,13 @@ if TYPE_CHECKING: logger = logging.getLogger(__name__) +class ThreadsListInclude(str, enum.Enum): + """Valid values for the 'include' flag of /threads.""" + + all = "all" + participated = "participated" + + @attr.s(slots=True, frozen=True, auto_attribs=True) class _ThreadAggregation: # The latest event in the thread. @@ -482,3 +490,79 @@ class RelationsHandler: results.setdefault(event_id, BundledAggregations()).replace = edit return results + + async def get_threads( + self, + requester: Requester, + room_id: str, + include: ThreadsListInclude, + limit: int = 5, + from_token: Optional[ThreadsNextBatch] = None, + ) -> JsonDict: + """Get related events of a event, ordered by topological ordering. + + Args: + requester: The user requesting the relations. + room_id: The room the event belongs to. + include: One of "all" or "participated" to indicate which threads should + be returned. + limit: Only fetch the most recent `limit` events. + from_token: Fetch rows from the given token, or from the start if None. + + Returns: + The pagination chunk. + """ + + user_id = requester.user.to_string() + + # TODO Properly handle a user leaving a room. + (_, member_event_id) = await self._auth.check_user_in_room_or_world_readable( + room_id, requester, allow_departed_users=True + ) + + # Note that ignored users are not passed into get_relations_for_event + # below. Ignored users are handled in filter_events_for_client (and by + # not passing them in here we should get a better cache hit rate). + thread_roots, next_batch = await self._main_store.get_threads( + room_id=room_id, limit=limit, from_token=from_token + ) + + events = await self._main_store.get_events_as_list(thread_roots) + + if include == ThreadsListInclude.participated: + # Pre-seed thread participation with whether the requester sent the event. + participated = {event.event_id: event.sender == user_id for event in events} + # For events the requester did not send, check the database for whether + # the requester sent a threaded reply. + participated.update( + await self._main_store.get_threads_participated( + [eid for eid, p in participated.items() if not p], + user_id, + ) + ) + + # Limit the returned threads to those the user has participated in. + events = [event for event in events if participated[event.event_id]] + + events = await filter_events_for_client( + self._storage_controllers, + user_id, + events, + is_peeking=(member_event_id is None), + ) + + aggregations = await self.get_bundled_aggregations( + events, requester.user.to_string() + ) + + now = self._clock.time_msec() + serialized_events = self._event_serializer.serialize_events( + events, now, bundle_aggregations=aggregations + ) + + return_value: JsonDict = {"chunk": serialized_events} + + if next_batch: + return_value["next_batch"] = str(next_batch) + + return return_value diff --git a/synapse/rest/client/relations.py b/synapse/rest/client/relations.py index b31ce5a0d3..d1aa1947a5 100644 --- a/synapse/rest/client/relations.py +++ b/synapse/rest/client/relations.py @@ -13,12 +13,15 @@ # limitations under the License. import logging +import re from typing import TYPE_CHECKING, Optional, Tuple +from synapse.handlers.relations import ThreadsListInclude from synapse.http.server import HttpServer -from synapse.http.servlet import RestServlet +from synapse.http.servlet import RestServlet, parse_integer, parse_string from synapse.http.site import SynapseRequest from synapse.rest.client._base import client_patterns +from synapse.storage.databases.main.relations import ThreadsNextBatch from synapse.streams.config import PaginationConfig from synapse.types import JsonDict @@ -78,5 +81,50 @@ class RelationPaginationServlet(RestServlet): return 200, result +class ThreadsServlet(RestServlet): + PATTERNS = ( + re.compile( + "^/_matrix/client/unstable/org.matrix.msc3856/rooms/(?P[^/]*)/threads" + ), + ) + + def __init__(self, hs: "HomeServer"): + super().__init__() + self.auth = hs.get_auth() + self.store = hs.get_datastores().main + self._relations_handler = hs.get_relations_handler() + + async def on_GET( + self, request: SynapseRequest, room_id: str + ) -> Tuple[int, JsonDict]: + requester = await self.auth.get_user_by_req(request) + + limit = parse_integer(request, "limit", default=5) + from_token_str = parse_string(request, "from") + include = parse_string( + request, + "include", + default=ThreadsListInclude.all.value, + allowed_values=[v.value for v in ThreadsListInclude], + ) + + # Return the relations + from_token = None + if from_token_str: + from_token = ThreadsNextBatch.from_string(from_token_str) + + result = await self._relations_handler.get_threads( + requester=requester, + room_id=room_id, + include=ThreadsListInclude(include), + limit=limit, + from_token=from_token, + ) + + return 200, result + + def register_servlets(hs: "HomeServer", http_server: HttpServer) -> None: RelationPaginationServlet(hs).register(http_server) + if hs.config.experimental.msc3856_enabled: + ThreadsServlet(hs).register(http_server) diff --git a/synapse/storage/databases/main/cache.py b/synapse/storage/databases/main/cache.py index a9f25a5904..0ce3156c9c 100644 --- a/synapse/storage/databases/main/cache.py +++ b/synapse/storage/databases/main/cache.py @@ -259,6 +259,7 @@ class CacheInvalidationWorkerStore(SQLBaseStore): self._attempt_to_invalidate_cache("get_applicable_edit", (relates_to,)) self._attempt_to_invalidate_cache("get_thread_summary", (relates_to,)) self._attempt_to_invalidate_cache("get_thread_participated", (relates_to,)) + self._attempt_to_invalidate_cache("get_threads", (room_id,)) async def invalidate_cache_and_stream( self, cache_name: str, keys: Tuple[Any, ...] diff --git a/synapse/storage/databases/main/events.py b/synapse/storage/databases/main/events.py index 060fe71454..6698cbf664 100644 --- a/synapse/storage/databases/main/events.py +++ b/synapse/storage/databases/main/events.py @@ -35,7 +35,7 @@ import attr from prometheus_client import Counter import synapse.metrics -from synapse.api.constants import EventContentFields, EventTypes +from synapse.api.constants import EventContentFields, EventTypes, RelationTypes from synapse.api.errors import Codes, SynapseError from synapse.api.room_versions import RoomVersions from synapse.events import EventBase, relation_from_event @@ -1616,7 +1616,7 @@ class PersistEventsStore: ) # Remove from relations table. - self._handle_redact_relations(txn, event.redacts) + self._handle_redact_relations(txn, event.room_id, event.redacts) # Update the event_forward_extremities, event_backward_extremities and # event_edges tables. @@ -1866,6 +1866,34 @@ class PersistEventsStore: }, ) + if relation.rel_type == RelationTypes.THREAD: + # Upsert into the threads table, but only overwrite the value if the + # new event is of a later topological order OR if the topological + # ordering is equal, but the stream ordering is later. + sql = """ + INSERT INTO threads (room_id, thread_id, latest_event_id, topological_ordering, stream_ordering) + VALUES (?, ?, ?, ?, ?) + ON CONFLICT (room_id, thread_id) + DO UPDATE SET + latest_event_id = excluded.latest_event_id, + topological_ordering = excluded.topological_ordering, + stream_ordering = excluded.stream_ordering + WHERE + threads.topological_ordering <= excluded.topological_ordering AND + threads.stream_ordering < excluded.stream_ordering + """ + + txn.execute( + sql, + ( + event.room_id, + relation.parent_id, + event.event_id, + event.depth, + event.internal_metadata.stream_ordering, + ), + ) + def _handle_insertion_event( self, txn: LoggingTransaction, event: EventBase ) -> None: @@ -1989,13 +2017,14 @@ class PersistEventsStore: txn.execute(sql, (batch_id,)) def _handle_redact_relations( - self, txn: LoggingTransaction, redacted_event_id: str + self, txn: LoggingTransaction, room_id: str, redacted_event_id: str ) -> None: """Handles receiving a redaction and checking whether the redacted event has any relations which must be removed from the database. Args: txn + room_id: The room ID of the event that was redacted. redacted_event_id: The event that was redacted. """ @@ -2024,6 +2053,9 @@ class PersistEventsStore: self.store._invalidate_cache_and_stream( txn, self.store.get_thread_participated, (redacted_relates_to,) ) + self.store._invalidate_cache_and_stream( + txn, self.store.get_threads, (room_id,) + ) self.db_pool.simple_delete_txn( txn, table="event_relations", keyvalues={"event_id": redacted_event_id} diff --git a/synapse/storage/databases/main/relations.py b/synapse/storage/databases/main/relations.py index e7fbf950e6..ac9b96ab44 100644 --- a/synapse/storage/databases/main/relations.py +++ b/synapse/storage/databases/main/relations.py @@ -14,6 +14,7 @@ import logging from typing import ( + TYPE_CHECKING, Collection, Dict, FrozenSet, @@ -29,17 +30,46 @@ from typing import ( import attr from synapse.api.constants import MAIN_TIMELINE, RelationTypes +from synapse.api.errors import SynapseError from synapse.events import EventBase from synapse.storage._base import SQLBaseStore -from synapse.storage.database import LoggingTransaction, make_in_list_sql_clause +from synapse.storage.database import ( + DatabasePool, + LoggingDatabaseConnection, + LoggingTransaction, + make_in_list_sql_clause, +) from synapse.storage.databases.main.stream import generate_pagination_where_clause from synapse.storage.engines import PostgresEngine from synapse.types import JsonDict, RoomStreamToken, StreamKeyType, StreamToken from synapse.util.caches.descriptors import cached, cachedList +if TYPE_CHECKING: + from synapse.server import HomeServer + logger = logging.getLogger(__name__) +@attr.s(slots=True, frozen=True, auto_attribs=True) +class ThreadsNextBatch: + topological_ordering: int + stream_ordering: int + + def __str__(self) -> str: + return f"{self.topological_ordering}_{self.stream_ordering}" + + @classmethod + def from_string(cls, string: str) -> "ThreadsNextBatch": + """ + Creates a ThreadsNextBatch from its textual representation. + """ + try: + keys = (int(s) for s in string.split("_")) + return cls(*keys) + except Exception: + raise SynapseError(400, "Invalid threads token") + + @attr.s(slots=True, frozen=True, auto_attribs=True) class _RelatedEvent: """ @@ -56,6 +86,76 @@ class _RelatedEvent: class RelationsWorkerStore(SQLBaseStore): + def __init__( + self, + database: DatabasePool, + db_conn: LoggingDatabaseConnection, + hs: "HomeServer", + ): + super().__init__(database, db_conn, hs) + + self.db_pool.updates.register_background_update_handler( + "threads_backfill", self._backfill_threads + ) + + async def _backfill_threads(self, progress: JsonDict, batch_size: int) -> int: + """Backfill the threads table.""" + + def threads_backfill_txn(txn: LoggingTransaction) -> int: + last_thread_id = progress.get("last_thread_id", "") + + # Get the latest event in each thread by topo ordering / stream ordering. + # + # Note that the MAX(event_id) is needed to abide by the rules of group by, + # but doesn't actually do anything since there should only be a single event + # ID per topo/stream ordering pair. + sql = f""" + SELECT room_id, relates_to_id, MAX(topological_ordering), MAX(stream_ordering), MAX(event_id) + FROM event_relations + INNER JOIN events USING (event_id) + WHERE + relates_to_id > ? AND + relation_type = '{RelationTypes.THREAD}' + GROUP BY room_id, relates_to_id + ORDER BY relates_to_id + LIMIT ? + """ + txn.execute(sql, (last_thread_id, batch_size)) + + # No more rows to process. + rows = txn.fetchall() + if not rows: + return 0 + + # Insert the rows into the threads table. If a matching thread already exists, + # assume it is from a newer event. + sql = """ + INSERT INTO threads (room_id, thread_id, topological_ordering, stream_ordering, latest_event_id) + VALUES %s + ON CONFLICT (room_id, thread_id) + DO NOTHING + """ + if isinstance(txn.database_engine, PostgresEngine): + txn.execute_values(sql % ("?",), rows, fetch=False) + else: + txn.execute_batch(sql % ("?, ?, ?, ?, ?",), rows) + + # Mark the progress. + self.db_pool.updates._background_update_progress_txn( + txn, "threads_backfill", {"last_thread_id": rows[-1][1]} + ) + + return txn.rowcount + + result = await self.db_pool.runInteraction( + "threads_backfill", threads_backfill_txn + ) + + if not result: + await self.db_pool.updates._end_background_update("threads_backfill") + + return result + @cached(uncached_args=("event",), tree=True) async def get_relations_for_event( self, @@ -776,6 +876,70 @@ class RelationsWorkerStore(SQLBaseStore): "get_if_user_has_annotated_event", _get_if_user_has_annotated_event ) + @cached(tree=True) + async def get_threads( + self, + room_id: str, + limit: int = 5, + from_token: Optional[ThreadsNextBatch] = None, + ) -> Tuple[List[str], Optional[ThreadsNextBatch]]: + """Get a list of thread IDs, ordered by topological ordering of their + latest reply. + + Args: + room_id: The room the event belongs to. + limit: Only fetch the most recent `limit` threads. + from_token: Fetch rows from a previous next_batch, or from the start if None. + + Returns: + A tuple of: + A list of thread root event IDs. + + The next_batch, if one exists. + """ + # Generate the pagination clause, if necessary. + # + # Find any threads where the latest reply is equal / before the last + # thread's topo ordering and earlier in stream ordering. + pagination_clause = "" + pagination_args: tuple = () + if from_token: + pagination_clause = "AND topological_ordering <= ? AND stream_ordering < ?" + pagination_args = ( + from_token.topological_ordering, + from_token.stream_ordering, + ) + + sql = f""" + SELECT thread_id, topological_ordering, stream_ordering + FROM threads + WHERE + room_id = ? + {pagination_clause} + ORDER BY topological_ordering DESC, stream_ordering DESC + LIMIT ? + """ + + def _get_threads_txn( + txn: LoggingTransaction, + ) -> Tuple[List[str], Optional[ThreadsNextBatch]]: + txn.execute(sql, (room_id, *pagination_args, limit + 1)) + + rows = cast(List[Tuple[str, int, int]], txn.fetchall()) + thread_ids = [r[0] for r in rows] + + # If there are more events, generate the next pagination key from the + # last thread which will be returned. + next_token = None + if len(thread_ids) > limit: + last_topo_id = rows[-2][1] + last_stream_id = rows[-2][2] + next_token = ThreadsNextBatch(last_topo_id, last_stream_id) + + return thread_ids[:limit], next_token + + return await self.db_pool.runInteraction("get_threads", _get_threads_txn) + @cached() async def get_thread_id(self, event_id: str) -> str: """ diff --git a/synapse/storage/schema/main/delta/73/09threads_table.sql b/synapse/storage/schema/main/delta/73/09threads_table.sql new file mode 100644 index 0000000000..aa7c5e9a2e --- /dev/null +++ b/synapse/storage/schema/main/delta/73/09threads_table.sql @@ -0,0 +1,30 @@ +/* Copyright 2022 The Matrix.org Foundation C.I.C + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +CREATE TABLE threads ( + room_id TEXT NOT NULL, + -- The event ID of the root event in the thread. + thread_id TEXT NOT NULL, + -- The latest event ID and corresponding topo / stream ordering. + latest_event_id TEXT NOT NULL, + topological_ordering BIGINT NOT NULL, + stream_ordering BIGINT NOT NULL, + CONSTRAINT threads_uniqueness UNIQUE (room_id, thread_id) +); + +CREATE INDEX threads_ordering_idx ON threads(room_id, topological_ordering, stream_ordering); + +INSERT INTO background_updates (ordering, update_name, progress_json) VALUES + (7309, 'threads_backfill', '{}'); diff --git a/tests/rest/client/test_relations.py b/tests/rest/client/test_relations.py index 988cdb746d..d595295e2c 100644 --- a/tests/rest/client/test_relations.py +++ b/tests/rest/client/test_relations.py @@ -1707,3 +1707,154 @@ class RelationRedactionTestCase(BaseRelationsTestCase): relations[RelationTypes.THREAD]["latest_event"]["event_id"], related_event_id, ) + + +class ThreadsTestCase(BaseRelationsTestCase): + @unittest.override_config({"experimental_features": {"msc3856_enabled": True}}) + def test_threads(self) -> None: + """Create threads and ensure the ordering is due to their latest event.""" + # Create 2 threads. + thread_1 = self.parent_id + res = self.helper.send(self.room, body="Thread Root!", tok=self.user_token) + thread_2 = res["event_id"] + + self._send_relation(RelationTypes.THREAD, "m.room.test") + self._send_relation(RelationTypes.THREAD, "m.room.test", parent_id=thread_2) + + # Request the threads in the room. + channel = self.make_request( + "GET", + f"/_matrix/client/unstable/org.matrix.msc3856/rooms/{self.room}/threads", + access_token=self.user_token, + ) + self.assertEquals(200, channel.code, channel.json_body) + thread_roots = [ev["event_id"] for ev in channel.json_body["chunk"]] + self.assertEqual(thread_roots, [thread_2, thread_1]) + + # Update the first thread, the ordering should swap. + self._send_relation(RelationTypes.THREAD, "m.room.test") + + channel = self.make_request( + "GET", + f"/_matrix/client/unstable/org.matrix.msc3856/rooms/{self.room}/threads", + access_token=self.user_token, + ) + self.assertEquals(200, channel.code, channel.json_body) + thread_roots = [ev["event_id"] for ev in channel.json_body["chunk"]] + self.assertEqual(thread_roots, [thread_1, thread_2]) + + @unittest.override_config({"experimental_features": {"msc3856_enabled": True}}) + def test_pagination(self) -> None: + """Create threads and paginate through them.""" + # Create 2 threads. + thread_1 = self.parent_id + res = self.helper.send(self.room, body="Thread Root!", tok=self.user_token) + thread_2 = res["event_id"] + + self._send_relation(RelationTypes.THREAD, "m.room.test") + self._send_relation(RelationTypes.THREAD, "m.room.test", parent_id=thread_2) + + # Request the threads in the room. + channel = self.make_request( + "GET", + f"/_matrix/client/unstable/org.matrix.msc3856/rooms/{self.room}/threads?limit=1", + access_token=self.user_token, + ) + self.assertEquals(200, channel.code, channel.json_body) + thread_roots = [ev["event_id"] for ev in channel.json_body["chunk"]] + self.assertEqual(thread_roots, [thread_2]) + + # Make sure next_batch has something in it that looks like it could be a + # valid token. + next_batch = channel.json_body.get("next_batch") + self.assertIsInstance(next_batch, str, channel.json_body) + + channel = self.make_request( + "GET", + f"/_matrix/client/unstable/org.matrix.msc3856/rooms/{self.room}/threads?limit=1&from={next_batch}", + access_token=self.user_token, + ) + self.assertEquals(200, channel.code, channel.json_body) + thread_roots = [ev["event_id"] for ev in channel.json_body["chunk"]] + self.assertEqual(thread_roots, [thread_1], channel.json_body) + + self.assertNotIn("next_batch", channel.json_body, channel.json_body) + + @unittest.override_config({"experimental_features": {"msc3856_enabled": True}}) + def test_include(self) -> None: + """Filtering threads to all or participated in should work.""" + # Thread 1 has the user as the root event. + thread_1 = self.parent_id + self._send_relation( + RelationTypes.THREAD, "m.room.test", access_token=self.user2_token + ) + + # Thread 2 has the user replying. + res = self.helper.send(self.room, body="Thread Root!", tok=self.user2_token) + thread_2 = res["event_id"] + self._send_relation(RelationTypes.THREAD, "m.room.test", parent_id=thread_2) + + # Thread 3 has the user not participating in. + res = self.helper.send(self.room, body="Another thread!", tok=self.user2_token) + thread_3 = res["event_id"] + self._send_relation( + RelationTypes.THREAD, + "m.room.test", + access_token=self.user2_token, + parent_id=thread_3, + ) + + # All threads in the room. + channel = self.make_request( + "GET", + f"/_matrix/client/unstable/org.matrix.msc3856/rooms/{self.room}/threads", + access_token=self.user_token, + ) + self.assertEquals(200, channel.code, channel.json_body) + thread_roots = [ev["event_id"] for ev in channel.json_body["chunk"]] + self.assertEqual( + thread_roots, [thread_3, thread_2, thread_1], channel.json_body + ) + + # Only participated threads. + channel = self.make_request( + "GET", + f"/_matrix/client/unstable/org.matrix.msc3856/rooms/{self.room}/threads?include=participated", + access_token=self.user_token, + ) + self.assertEquals(200, channel.code, channel.json_body) + thread_roots = [ev["event_id"] for ev in channel.json_body["chunk"]] + self.assertEqual(thread_roots, [thread_2, thread_1], channel.json_body) + + @unittest.override_config({"experimental_features": {"msc3856_enabled": True}}) + def test_ignored_user(self) -> None: + """Events from ignored users should be ignored.""" + # Thread 1 has a reply from an ignored user. + thread_1 = self.parent_id + self._send_relation( + RelationTypes.THREAD, "m.room.test", access_token=self.user2_token + ) + + # Thread 2 is created by an ignored user. + res = self.helper.send(self.room, body="Thread Root!", tok=self.user2_token) + thread_2 = res["event_id"] + self._send_relation(RelationTypes.THREAD, "m.room.test", parent_id=thread_2) + + # Ignore user2. + self.get_success( + self.store.add_account_data_for_user( + self.user_id, + AccountDataTypes.IGNORED_USER_LIST, + {"ignored_users": {self.user2_id: {}}}, + ) + ) + + # Only thread 1 is returned. + channel = self.make_request( + "GET", + f"/_matrix/client/unstable/org.matrix.msc3856/rooms/{self.room}/threads", + access_token=self.user_token, + ) + self.assertEquals(200, channel.code, channel.json_body) + thread_roots = [ev["event_id"] for ev in channel.json_body["chunk"]] + self.assertEqual(thread_roots, [thread_1], channel.json_body) -- cgit 1.5.1 From 126a15794c95002560709283640ad412636b29b8 Mon Sep 17 00:00:00 2001 From: Patrick Cloke Date: Fri, 14 Oct 2022 08:30:05 -0400 Subject: Do not allow a None-limit on PaginationConfig. (#14146) The callers either set a default limit or manually handle a None-limit later on (by setting a default value). Update the callers to always instantiate PaginationConfig with a default limit and then assume the limit is non-None. --- changelog.d/14146.removal | 1 + synapse/handlers/account_data.py | 2 +- synapse/handlers/initial_sync.py | 27 ++++----------------------- synapse/handlers/pagination.py | 5 ----- synapse/handlers/presence.py | 4 +++- synapse/handlers/receipts.py | 2 +- synapse/handlers/relations.py | 3 --- synapse/handlers/room.py | 2 +- synapse/handlers/typing.py | 2 +- synapse/rest/client/events.py | 4 +++- synapse/rest/client/initial_sync.py | 4 +++- synapse/rest/client/room.py | 4 +++- synapse/storage/databases/main/stream.py | 2 -- synapse/streams/__init__.py | 2 +- synapse/streams/config.py | 12 +++++------- tests/rest/client/test_typing.py | 3 ++- 16 files changed, 29 insertions(+), 50 deletions(-) create mode 100644 changelog.d/14146.removal (limited to 'synapse/handlers') diff --git a/changelog.d/14146.removal b/changelog.d/14146.removal new file mode 100644 index 0000000000..08fa752897 --- /dev/null +++ b/changelog.d/14146.removal @@ -0,0 +1 @@ +Remove the unstable identifier for [MSC3715](https://github.com/matrix-org/matrix-doc/pull/3715). diff --git a/synapse/handlers/account_data.py b/synapse/handlers/account_data.py index 0478448b47..fc21d58001 100644 --- a/synapse/handlers/account_data.py +++ b/synapse/handlers/account_data.py @@ -225,7 +225,7 @@ class AccountDataEventSource(EventSource[int, JsonDict]): self, user: UserID, from_key: int, - limit: Optional[int], + limit: int, room_ids: Collection[str], is_guest: bool, explicit_room_id: Optional[str] = None, diff --git a/synapse/handlers/initial_sync.py b/synapse/handlers/initial_sync.py index 860c82c110..9c335e6863 100644 --- a/synapse/handlers/initial_sync.py +++ b/synapse/handlers/initial_sync.py @@ -57,13 +57,7 @@ class InitialSyncHandler: self.validator = EventValidator() self.snapshot_cache: ResponseCache[ Tuple[ - str, - Optional[StreamToken], - Optional[StreamToken], - str, - Optional[int], - bool, - bool, + str, Optional[StreamToken], Optional[StreamToken], str, int, bool, bool ] ] = ResponseCache(hs.get_clock(), "initial_sync_cache") self._event_serializer = hs.get_event_client_serializer() @@ -154,11 +148,6 @@ class InitialSyncHandler: public_room_ids = await self.store.get_public_room_ids() - if pagin_config.limit is not None: - limit = pagin_config.limit - else: - limit = 10 - serializer_options = SerializeEventConfig(as_client_event=as_client_event) async def handle_room(event: RoomsForUser) -> None: @@ -210,7 +199,7 @@ class InitialSyncHandler: run_in_background( self.store.get_recent_events_for_room, event.room_id, - limit=limit, + limit=pagin_config.limit, end_token=room_end_token, ), deferred_room_state, @@ -360,15 +349,11 @@ class InitialSyncHandler: member_event_id ) - limit = pagin_config.limit if pagin_config else None - if limit is None: - limit = 10 - leave_position = await self.store.get_position_for_event(member_event_id) stream_token = leave_position.to_room_stream_token() messages, token = await self.store.get_recent_events_for_room( - room_id, limit=limit, end_token=stream_token + room_id, limit=pagin_config.limit, end_token=stream_token ) messages = await filter_events_for_client( @@ -420,10 +405,6 @@ class InitialSyncHandler: now_token = self.hs.get_event_sources().get_current_token() - limit = pagin_config.limit if pagin_config else None - if limit is None: - limit = 10 - room_members = [ m for m in current_state.values() @@ -467,7 +448,7 @@ class InitialSyncHandler: run_in_background( self.store.get_recent_events_for_room, room_id, - limit=limit, + limit=pagin_config.limit, end_token=now_token.room_key, ), ), diff --git a/synapse/handlers/pagination.py b/synapse/handlers/pagination.py index 1f83bab836..a4ca9cb8b4 100644 --- a/synapse/handlers/pagination.py +++ b/synapse/handlers/pagination.py @@ -458,11 +458,6 @@ class PaginationHandler: # `/messages` should still works with live tokens when manually provided. assert from_token.room_key.topological is not None - if pagin_config.limit is None: - # This shouldn't happen as we've set a default limit before this - # gets called. - raise Exception("limit not set") - room_token = from_token.room_key async with self.pagination_lock.read(room_id): diff --git a/synapse/handlers/presence.py b/synapse/handlers/presence.py index 4e575ffbaa..2670e561d7 100644 --- a/synapse/handlers/presence.py +++ b/synapse/handlers/presence.py @@ -1596,7 +1596,9 @@ class PresenceEventSource(EventSource[int, UserPresenceState]): self, user: UserID, from_key: Optional[int], - limit: Optional[int] = None, + # Having a default limit doesn't match the EventSource API, but some + # callers do not provide it. It is unused in this class. + limit: int = 0, room_ids: Optional[Collection[str]] = None, is_guest: bool = False, explicit_room_id: Optional[str] = None, diff --git a/synapse/handlers/receipts.py b/synapse/handlers/receipts.py index 4a7ec9e426..ac01582442 100644 --- a/synapse/handlers/receipts.py +++ b/synapse/handlers/receipts.py @@ -257,7 +257,7 @@ class ReceiptEventSource(EventSource[int, JsonDict]): self, user: UserID, from_key: int, - limit: Optional[int], + limit: int, room_ids: Iterable[str], is_guest: bool, explicit_room_id: Optional[str] = None, diff --git a/synapse/handlers/relations.py b/synapse/handlers/relations.py index 1fdd7a10bc..0a0c6d938e 100644 --- a/synapse/handlers/relations.py +++ b/synapse/handlers/relations.py @@ -116,9 +116,6 @@ class RelationsHandler: if event is None: raise SynapseError(404, "Unknown parent event.") - # TODO Update pagination config to not allow None limits. - assert pagin_config.limit is not None - # Note that ignored users are not passed into get_relations_for_event # below. Ignored users are handled in filter_events_for_client (and by # not passing them in here we should get a better cache hit rate). diff --git a/synapse/handlers/room.py b/synapse/handlers/room.py index 57ab05ad25..4e1aacb408 100644 --- a/synapse/handlers/room.py +++ b/synapse/handlers/room.py @@ -1646,7 +1646,7 @@ class RoomEventSource(EventSource[RoomStreamToken, EventBase]): self, user: UserID, from_key: RoomStreamToken, - limit: Optional[int], + limit: int, room_ids: Collection[str], is_guest: bool, explicit_room_id: Optional[str] = None, diff --git a/synapse/handlers/typing.py b/synapse/handlers/typing.py index f953691669..a0ea719430 100644 --- a/synapse/handlers/typing.py +++ b/synapse/handlers/typing.py @@ -513,7 +513,7 @@ class TypingNotificationEventSource(EventSource[int, JsonDict]): self, user: UserID, from_key: int, - limit: Optional[int], + limit: int, room_ids: Iterable[str], is_guest: bool, explicit_room_id: Optional[str] = None, diff --git a/synapse/rest/client/events.py b/synapse/rest/client/events.py index 916f5230f1..782e7d14e8 100644 --- a/synapse/rest/client/events.py +++ b/synapse/rest/client/events.py @@ -50,7 +50,9 @@ class EventStreamRestServlet(RestServlet): raise SynapseError(400, "Guest users must specify room_id param") room_id = parse_string(request, "room_id") - pagin_config = await PaginationConfig.from_request(self.store, request) + pagin_config = await PaginationConfig.from_request( + self.store, request, default_limit=10 + ) timeout = EventStreamRestServlet.DEFAULT_LONGPOLL_TIME_MS if b"timeout" in args: try: diff --git a/synapse/rest/client/initial_sync.py b/synapse/rest/client/initial_sync.py index cfadcb8e50..9b1bb8b521 100644 --- a/synapse/rest/client/initial_sync.py +++ b/synapse/rest/client/initial_sync.py @@ -39,7 +39,9 @@ class InitialSyncRestServlet(RestServlet): requester = await self.auth.get_user_by_req(request) args: Dict[bytes, List[bytes]] = request.args # type: ignore as_client_event = b"raw" not in args - pagination_config = await PaginationConfig.from_request(self.store, request) + pagination_config = await PaginationConfig.from_request( + self.store, request, default_limit=10 + ) include_archived = parse_boolean(request, "archived", default=False) content = await self.initial_sync_handler.snapshot_all_rooms( user_id=requester.user.to_string(), diff --git a/synapse/rest/client/room.py b/synapse/rest/client/room.py index b6dedbed04..01e5079963 100644 --- a/synapse/rest/client/room.py +++ b/synapse/rest/client/room.py @@ -729,7 +729,9 @@ class RoomInitialSyncRestServlet(RestServlet): self, request: SynapseRequest, room_id: str ) -> Tuple[int, JsonDict]: requester = await self.auth.get_user_by_req(request, allow_guest=True) - pagination_config = await PaginationConfig.from_request(self.store, request) + pagination_config = await PaginationConfig.from_request( + self.store, request, default_limit=10 + ) content = await self.initial_sync_handler.room_initial_sync( room_id=room_id, requester=requester, pagin_config=pagination_config ) diff --git a/synapse/storage/databases/main/stream.py b/synapse/storage/databases/main/stream.py index ffeb2b3683..5baffbfe55 100644 --- a/synapse/storage/databases/main/stream.py +++ b/synapse/storage/databases/main/stream.py @@ -1200,8 +1200,6 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): `to_token`), or `limit` is zero. """ - assert int(limit) >= 0 - # Tokens really represent positions between elements, but we use # the convention of pointing to the event before the gap. Hence # we have a bit of asymmetry when it comes to equalities. diff --git a/synapse/streams/__init__.py b/synapse/streams/__init__.py index 806b671305..2dcd43d0a2 100644 --- a/synapse/streams/__init__.py +++ b/synapse/streams/__init__.py @@ -27,7 +27,7 @@ class EventSource(Generic[K, R]): self, user: UserID, from_key: K, - limit: Optional[int], + limit: int, room_ids: Collection[str], is_guest: bool, explicit_room_id: Optional[str] = None, diff --git a/synapse/streams/config.py b/synapse/streams/config.py index f6f7bf3d8b..6df2de919c 100644 --- a/synapse/streams/config.py +++ b/synapse/streams/config.py @@ -35,14 +35,14 @@ class PaginationConfig: from_token: Optional[StreamToken] to_token: Optional[StreamToken] direction: str - limit: Optional[int] + limit: int @classmethod async def from_request( cls, store: "DataStore", request: SynapseRequest, - default_limit: Optional[int] = None, + default_limit: int, default_dir: str = "f", ) -> "PaginationConfig": direction = parse_string( @@ -69,12 +69,10 @@ class PaginationConfig: raise SynapseError(400, "'to' parameter is invalid") limit = parse_integer(request, "limit", default=default_limit) + if limit < 0: + raise SynapseError(400, "Limit must be 0 or above") - if limit: - if limit < 0: - raise SynapseError(400, "Limit must be 0 or above") - - limit = min(int(limit), MAX_LIMIT) + limit = min(limit, MAX_LIMIT) try: return PaginationConfig(from_tok, to_tok, direction, limit) diff --git a/tests/rest/client/test_typing.py b/tests/rest/client/test_typing.py index 61b66d7685..fdc433a8b5 100644 --- a/tests/rest/client/test_typing.py +++ b/tests/rest/client/test_typing.py @@ -59,7 +59,8 @@ class RoomTypingTestCase(unittest.HomeserverTestCase): self.event_source.get_new_events( user=UserID.from_string(self.user_id), from_key=0, - limit=None, + # Limit is unused. + limit=0, room_ids=[self.room_id], is_guest=False, ) -- cgit 1.5.1 From 40bb37eb27e1841754a297ac1277748de7f6c1cb Mon Sep 17 00:00:00 2001 From: Eric Eastwood Date: Sat, 15 Oct 2022 00:36:49 -0500 Subject: Stop getting missing `prev_events` after we already know their signature is invalid (#13816) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit While https://github.com/matrix-org/synapse/pull/13635 stops us from doing the slow thing after we've already done it once, this PR stops us from doing one of the slow things in the first place. Related to - https://github.com/matrix-org/synapse/issues/13622 - https://github.com/matrix-org/synapse/pull/13635 - https://github.com/matrix-org/synapse/issues/13676 Part of https://github.com/matrix-org/synapse/issues/13356 Follow-up to https://github.com/matrix-org/synapse/pull/13815 which tracks event signature failures. With this PR, we avoid the call to the costly `_get_state_ids_after_missing_prev_event` because the signature failure will count as an attempt before and we filter events based on the backoff before calling `_get_state_ids_after_missing_prev_event` now. For example, this will save us 156s out of the 185s total that this `matrix.org` `/messages` request. If you want to see the full Jaeger trace of this, you can drag and drop this `trace.json` into your own Jaeger, https://gist.github.com/MadLittleMods/4b12d0d0afe88c2f65ffcc907306b761 To explain this exact scenario around `/messages` -> backfill, we call `/backfill` and first check the signatures of the 100 events. We see bad signature for `$luA4l7QHhf_jadH3mI-AyFqho0U2Q-IXXUbGSMq6h6M` and `$zuOn2Rd2vsC7SUia3Hp3r6JSkSFKcc5j3QTTqW_0jDw` (both member events). Then we process the 98 events remaining that have valid signatures but one of the events references `$luA4l7QHhf_jadH3mI-AyFqho0U2Q-IXXUbGSMq6h6M` as a `prev_event`. So we have to do the whole `_get_state_ids_after_missing_prev_event` rigmarole which pulls in those same events which fail again because the signatures are still invalid. - `backfill` - `outgoing-federation-request` `/backfill` - `_check_sigs_and_hash_and_fetch` - `_check_sigs_and_hash_and_fetch_one` for each event received over backfill - ❗ `$luA4l7QHhf_jadH3mI-AyFqho0U2Q-IXXUbGSMq6h6M` fails with `Signature on retrieved event was invalid.`: `unable to verify signature for sender domain xxx: 401: Failed to find any key to satisfy: _FetchKeyRequest(...)` - ❗ `$zuOn2Rd2vsC7SUia3Hp3r6JSkSFKcc5j3QTTqW_0jDw` fails with `Signature on retrieved event was invalid.`: `unable to verify signature for sender domain xxx: 401: Failed to find any key to satisfy: _FetchKeyRequest(...)` - `_process_pulled_events` - `_process_pulled_event` for each validated event - ❗ Event `$Q0iMdqtz3IJYfZQU2Xk2WjB5NDF8Gg8cFSYYyKQgKJ0` references `$luA4l7QHhf_jadH3mI-AyFqho0U2Q-IXXUbGSMq6h6M` as a `prev_event` which is missing so we try to get it - `_get_state_ids_after_missing_prev_event` - `outgoing-federation-request` `/state_ids` - ❗ `get_pdu` for `$luA4l7QHhf_jadH3mI-AyFqho0U2Q-IXXUbGSMq6h6M` which fails the signature check again - ❗ `get_pdu` for `$zuOn2Rd2vsC7SUia3Hp3r6JSkSFKcc5j3QTTqW_0jDw` which fails the signature check --- changelog.d/13816.feature | 1 + synapse/api/errors.py | 21 +++ synapse/handlers/federation.py | 16 ++ synapse/handlers/federation_event.py | 31 ++++ synapse/storage/databases/main/event_federation.py | 54 ++++++ tests/handlers/test_federation_event.py | 201 ++++++++++++++++++++- tests/storage/test_event_federation.py | 64 +++++++ 7 files changed, 386 insertions(+), 2 deletions(-) create mode 100644 changelog.d/13816.feature (limited to 'synapse/handlers') diff --git a/changelog.d/13816.feature b/changelog.d/13816.feature new file mode 100644 index 0000000000..5eaa936b08 --- /dev/null +++ b/changelog.d/13816.feature @@ -0,0 +1 @@ +Stop fetching missing `prev_events` after we already know their signature is invalid. diff --git a/synapse/api/errors.py b/synapse/api/errors.py index c606207569..e0873b1913 100644 --- a/synapse/api/errors.py +++ b/synapse/api/errors.py @@ -640,6 +640,27 @@ class FederationError(RuntimeError): } +class FederationPullAttemptBackoffError(RuntimeError): + """ + Raised to indicate that we are are deliberately not attempting to pull the given + event over federation because we've already done so recently and are backing off. + + Attributes: + event_id: The event_id which we are refusing to pull + message: A custom error message that gives more context + """ + + def __init__(self, event_ids: List[str], message: Optional[str]): + self.event_ids = event_ids + + if message: + error_message = message + else: + error_message = f"Not attempting to pull event_ids={self.event_ids} because we already tried to pull them recently (backing off)." + + super().__init__(error_message) + + class HttpResponseException(CodeMessageException): """ Represents an HTTP-level failure of an outbound request diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index 44e70c6c3c..5f7e0a1f79 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -45,6 +45,7 @@ from synapse.api.errors import ( Codes, FederationDeniedError, FederationError, + FederationPullAttemptBackoffError, HttpResponseException, LimitExceededError, NotFoundError, @@ -1720,7 +1721,22 @@ class FederationHandler: destination, event ) break + except FederationPullAttemptBackoffError as exc: + # Log a warning about why we failed to process the event (the error message + # for `FederationPullAttemptBackoffError` is pretty good) + logger.warning("_sync_partial_state_room: %s", exc) + # We do not record a failed pull attempt when we backoff fetching a missing + # `prev_event` because not being able to fetch the `prev_events` just means + # we won't be able to de-outlier the pulled event. But we can still use an + # `outlier` in the state/auth chain for another event. So we shouldn't stop + # a downstream event from trying to pull it. + # + # This avoids a cascade of backoff for all events in the DAG downstream from + # one event backoff upstream. except FederationError as e: + # TODO: We should `record_event_failed_pull_attempt` here, + # see https://github.com/matrix-org/synapse/issues/13700 + if attempt == len(destinations) - 1: # We have tried every remote server for this event. Give up. # TODO(faster_joins) giving up isn't the right thing to do diff --git a/synapse/handlers/federation_event.py b/synapse/handlers/federation_event.py index f382961099..4300e8dd40 100644 --- a/synapse/handlers/federation_event.py +++ b/synapse/handlers/federation_event.py @@ -44,6 +44,7 @@ from synapse.api.errors import ( AuthError, Codes, FederationError, + FederationPullAttemptBackoffError, HttpResponseException, RequestSendFailed, SynapseError, @@ -567,6 +568,9 @@ class FederationEventHandler: event: partial-state event to be de-partial-stated Raises: + FederationPullAttemptBackoffError if we are are deliberately not attempting + to pull the given event over federation because we've already done so + recently and are backing off. FederationError if we fail to request state from the remote server. """ logger.info("Updating state for %s", event.event_id) @@ -901,6 +905,18 @@ class FederationEventHandler: context, backfilled=backfilled, ) + except FederationPullAttemptBackoffError as exc: + # Log a warning about why we failed to process the event (the error message + # for `FederationPullAttemptBackoffError` is pretty good) + logger.warning("_process_pulled_event: %s", exc) + # We do not record a failed pull attempt when we backoff fetching a missing + # `prev_event` because not being able to fetch the `prev_events` just means + # we won't be able to de-outlier the pulled event. But we can still use an + # `outlier` in the state/auth chain for another event. So we shouldn't stop + # a downstream event from trying to pull it. + # + # This avoids a cascade of backoff for all events in the DAG downstream from + # one event backoff upstream. except FederationError as e: await self._store.record_event_failed_pull_attempt( event.room_id, event_id, str(e) @@ -947,6 +963,9 @@ class FederationEventHandler: The event context. Raises: + FederationPullAttemptBackoffError if we are are deliberately not attempting + to pull the given event over federation because we've already done so + recently and are backing off. FederationError if we fail to get the state from the remote server after any missing `prev_event`s. """ @@ -957,6 +976,18 @@ class FederationEventHandler: seen = await self._store.have_events_in_timeline(prevs) missing_prevs = prevs - seen + # If we've already recently attempted to pull this missing event, don't + # try it again so soon. Since we have to fetch all of the prev_events, we can + # bail early here if we find any to ignore. + prevs_to_ignore = await self._store.get_event_ids_to_not_pull_from_backoff( + room_id, missing_prevs + ) + if len(prevs_to_ignore) > 0: + raise FederationPullAttemptBackoffError( + event_ids=prevs_to_ignore, + message=f"While computing context for event={event_id}, not attempting to pull missing prev_event={prevs_to_ignore[0]} because we already tried to pull recently (backing off).", + ) + if not missing_prevs: return await self._state_handler.compute_event_context(event) diff --git a/synapse/storage/databases/main/event_federation.py b/synapse/storage/databases/main/event_federation.py index 6b9a629edd..309a4ba664 100644 --- a/synapse/storage/databases/main/event_federation.py +++ b/synapse/storage/databases/main/event_federation.py @@ -1501,6 +1501,12 @@ class EventFederationWorkerStore(SignatureWorkerStore, EventsWorkerStore, SQLBas event_id: The event that failed to be fetched or processed cause: The error message or reason that we failed to pull the event """ + logger.debug( + "record_event_failed_pull_attempt room_id=%s, event_id=%s, cause=%s", + room_id, + event_id, + cause, + ) await self.db_pool.runInteraction( "record_event_failed_pull_attempt", self._record_event_failed_pull_attempt_upsert_txn, @@ -1530,6 +1536,54 @@ class EventFederationWorkerStore(SignatureWorkerStore, EventsWorkerStore, SQLBas txn.execute(sql, (room_id, event_id, 1, self._clock.time_msec(), cause)) + @trace + async def get_event_ids_to_not_pull_from_backoff( + self, + room_id: str, + event_ids: Collection[str], + ) -> List[str]: + """ + Filter down the events to ones that we've failed to pull before recently. Uses + exponential backoff. + + Args: + room_id: The room that the events belong to + event_ids: A list of events to filter down + + Returns: + List of event_ids that should not be attempted to be pulled + """ + event_failed_pull_attempts = await self.db_pool.simple_select_many_batch( + table="event_failed_pull_attempts", + column="event_id", + iterable=event_ids, + keyvalues={}, + retcols=( + "event_id", + "last_attempt_ts", + "num_attempts", + ), + desc="get_event_ids_to_not_pull_from_backoff", + ) + + current_time = self._clock.time_msec() + return [ + event_failed_pull_attempt["event_id"] + for event_failed_pull_attempt in event_failed_pull_attempts + # Exponential back-off (up to the upper bound) so we don't try to + # pull the same event over and over. ex. 2hr, 4hr, 8hr, 16hr, etc. + if current_time + < event_failed_pull_attempt["last_attempt_ts"] + + ( + 2 + ** min( + event_failed_pull_attempt["num_attempts"], + BACKFILL_EVENT_EXPONENTIAL_BACKOFF_MAXIMUM_DOUBLING_STEPS, + ) + ) + * BACKFILL_EVENT_EXPONENTIAL_BACKOFF_STEP_MILLISECONDS + ] + async def get_missing_events( self, room_id: str, diff --git a/tests/handlers/test_federation_event.py b/tests/handlers/test_federation_event.py index 918010cddb..e448cb1901 100644 --- a/tests/handlers/test_federation_event.py +++ b/tests/handlers/test_federation_event.py @@ -14,7 +14,7 @@ from typing import Optional from unittest import mock -from synapse.api.errors import AuthError +from synapse.api.errors import AuthError, StoreError from synapse.api.room_versions import RoomVersion from synapse.event_auth import ( check_state_dependent_auth_rules, @@ -43,7 +43,7 @@ class FederationEventHandlerTests(unittest.FederatingHomeserverTestCase): def make_homeserver(self, reactor, clock): # mock out the federation transport client self.mock_federation_transport_client = mock.Mock( - spec=["get_room_state_ids", "get_room_state", "get_event"] + spec=["get_room_state_ids", "get_room_state", "get_event", "backfill"] ) return super().setup_test_homeserver( federation_transport_client=self.mock_federation_transport_client @@ -459,6 +459,203 @@ class FederationEventHandlerTests(unittest.FederatingHomeserverTestCase): ) self.assertIsNotNone(persisted, "pulled event was not persisted at all") + def test_backfill_signature_failure_does_not_fetch_same_prev_event_later( + self, + ) -> None: + """ + Test to make sure we backoff and don't try to fetch a missing prev_event when we + already know it has a invalid signature from checking the signatures of all of + the events in the backfill response. + """ + OTHER_USER = f"@user:{self.OTHER_SERVER_NAME}" + main_store = self.hs.get_datastores().main + + # Create the room + user_id = self.register_user("kermit", "test") + tok = self.login("kermit", "test") + room_id = self.helper.create_room_as(room_creator=user_id, tok=tok) + room_version = self.get_success(main_store.get_room_version(room_id)) + + # Allow the remote user to send state events + self.helper.send_state( + room_id, + "m.room.power_levels", + {"events_default": 0, "state_default": 0}, + tok=tok, + ) + + # Add the remote user to the room + member_event = self.get_success( + event_injection.inject_member_event(self.hs, room_id, OTHER_USER, "join") + ) + + initial_state_map = self.get_success( + main_store.get_partial_current_state_ids(room_id) + ) + + auth_event_ids = [ + initial_state_map[("m.room.create", "")], + initial_state_map[("m.room.power_levels", "")], + member_event.event_id, + ] + + # We purposely don't run `add_hashes_and_signatures_from_other_server` + # over this because we want the signature check to fail. + pulled_event_without_signatures = make_event_from_dict( + { + "type": "test_regular_type", + "room_id": room_id, + "sender": OTHER_USER, + "prev_events": [member_event.event_id], + "auth_events": auth_event_ids, + "origin_server_ts": 1, + "depth": 12, + "content": {"body": "pulled_event_without_signatures"}, + }, + room_version, + ) + + # Create a regular event that should pass except for the + # `pulled_event_without_signatures` in the `prev_event`. + pulled_event = make_event_from_dict( + self.add_hashes_and_signatures_from_other_server( + { + "type": "test_regular_type", + "room_id": room_id, + "sender": OTHER_USER, + "prev_events": [ + member_event.event_id, + pulled_event_without_signatures.event_id, + ], + "auth_events": auth_event_ids, + "origin_server_ts": 1, + "depth": 12, + "content": {"body": "pulled_event"}, + } + ), + room_version, + ) + + # We expect an outbound request to /backfill, so stub that out + self.mock_federation_transport_client.backfill.return_value = make_awaitable( + { + "origin": self.OTHER_SERVER_NAME, + "origin_server_ts": 123, + "pdus": [ + # This is one of the important aspects of this test: we include + # `pulled_event_without_signatures` so it fails the signature check + # when we filter down the backfill response down to events which + # have valid signatures in + # `_check_sigs_and_hash_for_pulled_events_and_fetch` + pulled_event_without_signatures.get_pdu_json(), + # Then later when we process this valid signature event, when we + # fetch the missing `prev_event`s, we want to make sure that we + # backoff and don't try and fetch `pulled_event_without_signatures` + # again since we know it just had an invalid signature. + pulled_event.get_pdu_json(), + ], + } + ) + + # Keep track of the count and make sure we don't make any of these requests + event_endpoint_requested_count = 0 + room_state_ids_endpoint_requested_count = 0 + room_state_endpoint_requested_count = 0 + + async def get_event( + destination: str, event_id: str, timeout: Optional[int] = None + ) -> None: + nonlocal event_endpoint_requested_count + event_endpoint_requested_count += 1 + + async def get_room_state_ids( + destination: str, room_id: str, event_id: str + ) -> None: + nonlocal room_state_ids_endpoint_requested_count + room_state_ids_endpoint_requested_count += 1 + + async def get_room_state( + room_version: RoomVersion, destination: str, room_id: str, event_id: str + ) -> None: + nonlocal room_state_endpoint_requested_count + room_state_endpoint_requested_count += 1 + + # We don't expect an outbound request to `/event`, `/state_ids`, or `/state` in + # the happy path but if the logic is sneaking around what we expect, stub that + # out so we can detect that failure + self.mock_federation_transport_client.get_event.side_effect = get_event + self.mock_federation_transport_client.get_room_state_ids.side_effect = ( + get_room_state_ids + ) + self.mock_federation_transport_client.get_room_state.side_effect = ( + get_room_state + ) + + # The function under test: try to backfill and process the pulled event + with LoggingContext("test"): + self.get_success( + self.hs.get_federation_event_handler().backfill( + self.OTHER_SERVER_NAME, + room_id, + limit=1, + extremities=["$some_extremity"], + ) + ) + + if event_endpoint_requested_count > 0: + self.fail( + "We don't expect an outbound request to /event in the happy path but if " + "the logic is sneaking around what we expect, make sure to fail the test. " + "We don't expect it because the signature failure should cause us to backoff " + "and not asking about pulled_event_without_signatures=" + f"{pulled_event_without_signatures.event_id} again" + ) + + if room_state_ids_endpoint_requested_count > 0: + self.fail( + "We don't expect an outbound request to /state_ids in the happy path but if " + "the logic is sneaking around what we expect, make sure to fail the test. " + "We don't expect it because the signature failure should cause us to backoff " + "and not asking about pulled_event_without_signatures=" + f"{pulled_event_without_signatures.event_id} again" + ) + + if room_state_endpoint_requested_count > 0: + self.fail( + "We don't expect an outbound request to /state in the happy path but if " + "the logic is sneaking around what we expect, make sure to fail the test. " + "We don't expect it because the signature failure should cause us to backoff " + "and not asking about pulled_event_without_signatures=" + f"{pulled_event_without_signatures.event_id} again" + ) + + # Make sure we only recorded a single failure which corresponds to the signature + # failure initially in `_check_sigs_and_hash_for_pulled_events_and_fetch` before + # we process all of the pulled events. + backfill_num_attempts_for_event_without_signatures = self.get_success( + main_store.db_pool.simple_select_one_onecol( + table="event_failed_pull_attempts", + keyvalues={"event_id": pulled_event_without_signatures.event_id}, + retcol="num_attempts", + ) + ) + self.assertEqual(backfill_num_attempts_for_event_without_signatures, 1) + + # And make sure we didn't record a failure for the event that has the missing + # prev_event because we don't want to cause a cascade of failures. Not being + # able to fetch the `prev_events` just means we won't be able to de-outlier the + # pulled event. But we can still use an `outlier` in the state/auth chain for + # another event. So we shouldn't stop a downstream event from trying to pull it. + self.get_failure( + main_store.db_pool.simple_select_one_onecol( + table="event_failed_pull_attempts", + keyvalues={"event_id": pulled_event.event_id}, + retcol="num_attempts", + ), + # StoreError: 404: No row found + StoreError, + ) + def test_process_pulled_event_with_rejected_missing_state(self) -> None: """Ensure that we correctly handle pulled events with missing state containing a rejected state event diff --git a/tests/storage/test_event_federation.py b/tests/storage/test_event_federation.py index 59b8910907..853db930d6 100644 --- a/tests/storage/test_event_federation.py +++ b/tests/storage/test_event_federation.py @@ -27,6 +27,8 @@ from synapse.api.room_versions import ( RoomVersion, ) from synapse.events import _EventInternalMetadata +from synapse.rest import admin +from synapse.rest.client import login, room from synapse.server import HomeServer from synapse.storage.database import LoggingTransaction from synapse.types import JsonDict @@ -43,6 +45,12 @@ class _BackfillSetupInfo: class EventFederationWorkerStoreTestCase(tests.unittest.HomeserverTestCase): + servlets = [ + admin.register_servlets, + room.register_servlets, + login.register_servlets, + ] + def prepare(self, reactor: MemoryReactor, clock: Clock, hs: HomeServer) -> None: self.store = hs.get_datastores().main @@ -1122,6 +1130,62 @@ class EventFederationWorkerStoreTestCase(tests.unittest.HomeserverTestCase): backfill_event_ids = [backfill_point[0] for backfill_point in backfill_points] self.assertEqual(backfill_event_ids, ["insertion_eventA"]) + def test_get_event_ids_to_not_pull_from_backoff( + self, + ): + """ + Test to make sure only event IDs we should backoff from are returned. + """ + # Create the room + user_id = self.register_user("alice", "test") + tok = self.login("alice", "test") + room_id = self.helper.create_room_as(room_creator=user_id, tok=tok) + + self.get_success( + self.store.record_event_failed_pull_attempt( + room_id, "$failed_event_id", "fake cause" + ) + ) + + event_ids_to_backoff = self.get_success( + self.store.get_event_ids_to_not_pull_from_backoff( + room_id=room_id, event_ids=["$failed_event_id", "$normal_event_id"] + ) + ) + + self.assertEqual(event_ids_to_backoff, ["$failed_event_id"]) + + def test_get_event_ids_to_not_pull_from_backoff_retry_after_backoff_duration( + self, + ): + """ + Test to make sure no event IDs are returned after the backoff duration has + elapsed. + """ + # Create the room + user_id = self.register_user("alice", "test") + tok = self.login("alice", "test") + room_id = self.helper.create_room_as(room_creator=user_id, tok=tok) + + self.get_success( + self.store.record_event_failed_pull_attempt( + room_id, "$failed_event_id", "fake cause" + ) + ) + + # Now advance time by 2 hours so we wait long enough for the single failed + # attempt (2^1 hours). + self.reactor.advance(datetime.timedelta(hours=2).total_seconds()) + + event_ids_to_backoff = self.get_success( + self.store.get_event_ids_to_not_pull_from_backoff( + room_id=room_id, event_ids=["$failed_event_id", "$normal_event_id"] + ) + ) + # Since this function only returns events we should backoff from, time has + # elapsed past the backoff range so there is no events to backoff from. + self.assertEqual(event_ids_to_backoff, []) + @attr.s class FakeEvent: -- cgit 1.5.1 From ccce8cdfc5e567b5b905b58e82a1d725f2647524 Mon Sep 17 00:00:00 2001 From: David Robertson Date: Mon, 17 Oct 2022 13:39:12 +0100 Subject: Use Pydantic when PUTting room aliases (#14179) --- changelog.d/14179.feature | 1 + synapse/handlers/directory.py | 19 +++++++------ synapse/rest/client/directory.py | 58 ++++++++++++++++++++++++---------------- 3 files changed, 47 insertions(+), 31 deletions(-) create mode 100644 changelog.d/14179.feature (limited to 'synapse/handlers') diff --git a/changelog.d/14179.feature b/changelog.d/14179.feature new file mode 100644 index 0000000000..48f2db91d3 --- /dev/null +++ b/changelog.d/14179.feature @@ -0,0 +1 @@ +Improve the validation of the following PUT endpoints: [`/directory/room/{roomAlias}`](https://spec.matrix.org/v1.4/client-server-api/#put_matrixclientv3directoryroomroomalias), [`/directory/list/room/{roomId}`](https://spec.matrix.org/v1.4/client-server-api/#put_matrixclientv3directorylistroomroomid) and [`/directory/list/appservice/{networkId}/{roomId}`](https://spec.matrix.org/v1.4/application-service-api/#put_matrixclientv3directorylistappservicenetworkidroomid). diff --git a/synapse/handlers/directory.py b/synapse/handlers/directory.py index 7127d5aefc..d52ebada6b 100644 --- a/synapse/handlers/directory.py +++ b/synapse/handlers/directory.py @@ -16,6 +16,8 @@ import logging import string from typing import TYPE_CHECKING, Iterable, List, Optional +from typing_extensions import Literal + from synapse.api.constants import MAX_ALIAS_LENGTH, EventTypes from synapse.api.errors import ( AuthError, @@ -429,7 +431,10 @@ class DirectoryHandler: return await self.auth.check_can_change_room_list(room_id, requester) async def edit_published_room_list( - self, requester: Requester, room_id: str, visibility: str + self, + requester: Requester, + room_id: str, + visibility: Literal["public", "private"], ) -> None: """Edit the entry of the room in the published room list. @@ -451,9 +456,6 @@ class DirectoryHandler: if requester.is_guest: raise AuthError(403, "Guests cannot edit the published room list") - if visibility not in ["public", "private"]: - raise SynapseError(400, "Invalid visibility setting") - if visibility == "public" and not self.enable_room_list_search: # The room list has been disabled. raise AuthError( @@ -505,7 +507,11 @@ class DirectoryHandler: await self.store.set_room_is_public(room_id, making_public) async def edit_published_appservice_room_list( - self, appservice_id: str, network_id: str, room_id: str, visibility: str + self, + appservice_id: str, + network_id: str, + room_id: str, + visibility: Literal["public", "private"], ) -> None: """Add or remove a room from the appservice/network specific public room list. @@ -516,9 +522,6 @@ class DirectoryHandler: room_id visibility: either "public" or "private" """ - if visibility not in ["public", "private"]: - raise SynapseError(400, "Invalid visibility setting") - await self.store.set_room_is_public_appservice( room_id, appservice_id, network_id, visibility == "public" ) diff --git a/synapse/rest/client/directory.py b/synapse/rest/client/directory.py index bc1b18c92d..f17b4c8d22 100644 --- a/synapse/rest/client/directory.py +++ b/synapse/rest/client/directory.py @@ -13,15 +13,22 @@ # limitations under the License. import logging -from typing import TYPE_CHECKING, Tuple +from typing import TYPE_CHECKING, List, Optional, Tuple + +from pydantic import StrictStr +from typing_extensions import Literal from twisted.web.server import Request from synapse.api.errors import AuthError, Codes, NotFoundError, SynapseError from synapse.http.server import HttpServer -from synapse.http.servlet import RestServlet, parse_json_object_from_request +from synapse.http.servlet import ( + RestServlet, + parse_and_validate_json_object_from_request, +) from synapse.http.site import SynapseRequest from synapse.rest.client._base import client_patterns +from synapse.rest.models import RequestBodyModel from synapse.types import JsonDict, RoomAlias if TYPE_CHECKING: @@ -54,6 +61,12 @@ class ClientDirectoryServer(RestServlet): return 200, res + class PutBody(RequestBodyModel): + # TODO: get Pydantic to validate that this is a valid room id? + room_id: StrictStr + # `servers` is unspecced + servers: Optional[List[StrictStr]] = None + async def on_PUT( self, request: SynapseRequest, room_alias: str ) -> Tuple[int, JsonDict]: @@ -61,31 +74,22 @@ class ClientDirectoryServer(RestServlet): raise SynapseError(400, "Room alias invalid", errcode=Codes.INVALID_PARAM) room_alias_obj = RoomAlias.from_string(room_alias) - content = parse_json_object_from_request(request) - if "room_id" not in content: - raise SynapseError( - 400, 'Missing params: ["room_id"]', errcode=Codes.BAD_JSON - ) + content = parse_and_validate_json_object_from_request(request, self.PutBody) logger.debug("Got content: %s", content) logger.debug("Got room name: %s", room_alias_obj.to_string()) - room_id = content["room_id"] - servers = content["servers"] if "servers" in content else None - - logger.debug("Got room_id: %s", room_id) - logger.debug("Got servers: %s", servers) + logger.debug("Got room_id: %s", content.room_id) + logger.debug("Got servers: %s", content.servers) - # TODO(erikj): Check types. - - room = await self.store.get_room(room_id) + room = await self.store.get_room(content.room_id) if room is None: raise SynapseError(400, "Room does not exist") requester = await self.auth.get_user_by_req(request) await self.directory_handler.create_association( - requester, room_alias_obj, room_id, servers + requester, room_alias_obj, content.room_id, content.servers ) return 200, {} @@ -137,16 +141,18 @@ class ClientDirectoryListServer(RestServlet): return 200, {"visibility": "public" if room["is_public"] else "private"} + class PutBody(RequestBodyModel): + visibility: Literal["public", "private"] = "public" + async def on_PUT( self, request: SynapseRequest, room_id: str ) -> Tuple[int, JsonDict]: requester = await self.auth.get_user_by_req(request) - content = parse_json_object_from_request(request) - visibility = content.get("visibility", "public") + content = parse_and_validate_json_object_from_request(request, self.PutBody) await self.directory_handler.edit_published_room_list( - requester, room_id, visibility + requester, room_id, content.visibility ) return 200, {} @@ -163,12 +169,14 @@ class ClientAppserviceDirectoryListServer(RestServlet): self.directory_handler = hs.get_directory_handler() self.auth = hs.get_auth() + class PutBody(RequestBodyModel): + visibility: Literal["public", "private"] = "public" + async def on_PUT( self, request: SynapseRequest, network_id: str, room_id: str ) -> Tuple[int, JsonDict]: - content = parse_json_object_from_request(request) - visibility = content.get("visibility", "public") - return await self._edit(request, network_id, room_id, visibility) + content = parse_and_validate_json_object_from_request(request, self.PutBody) + return await self._edit(request, network_id, room_id, content.visibility) async def on_DELETE( self, request: SynapseRequest, network_id: str, room_id: str @@ -176,7 +184,11 @@ class ClientAppserviceDirectoryListServer(RestServlet): return await self._edit(request, network_id, room_id, "private") async def _edit( - self, request: SynapseRequest, network_id: str, room_id: str, visibility: str + self, + request: SynapseRequest, + network_id: str, + room_id: str, + visibility: Literal["public", "private"], ) -> Tuple[int, JsonDict]: requester = await self.auth.get_user_by_req(request) if not requester.app_service: -- cgit 1.5.1 From dc02d9f8c54576d4b41ce51a2704fdd43b582d66 Mon Sep 17 00:00:00 2001 From: Andrew Morgan <1342360+anoadragon453@users.noreply.github.com> Date: Tue, 18 Oct 2022 10:33:35 +0100 Subject: Avoid checking the event cache when backfilling events (#14164) --- changelog.d/14164.bugfix | 1 + synapse/handlers/federation_event.py | 47 ++++++++--- synapse/storage/databases/main/events_worker.py | 2 +- tests/handlers/test_federation.py | 105 +++++++++++++++++++++++- 4 files changed, 140 insertions(+), 15 deletions(-) create mode 100644 changelog.d/14164.bugfix (limited to 'synapse/handlers') diff --git a/changelog.d/14164.bugfix b/changelog.d/14164.bugfix new file mode 100644 index 0000000000..aed4d9e386 --- /dev/null +++ b/changelog.d/14164.bugfix @@ -0,0 +1 @@ +Fix a bug introduced in 1.30.0 where purging and rejoining a room without restarting in-between would result in a broken room. \ No newline at end of file diff --git a/synapse/handlers/federation_event.py b/synapse/handlers/federation_event.py index 4300e8dd40..06e41b5cc0 100644 --- a/synapse/handlers/federation_event.py +++ b/synapse/handlers/federation_event.py @@ -798,9 +798,42 @@ class FederationEventHandler: ], ) + # Check if we already any of these have these events. + # Note: we currently make a lookup in the database directly here rather than + # checking the event cache, due to: + # https://github.com/matrix-org/synapse/issues/13476 + existing_events_map = await self._store._get_events_from_db( + [event.event_id for event in events] + ) + + new_events = [] + for event in events: + event_id = event.event_id + + # If we've already seen this event ID... + if event_id in existing_events_map: + existing_event = existing_events_map[event_id] + + # ...and the event itself was not previously stored as an outlier... + if not existing_event.event.internal_metadata.is_outlier(): + # ...then there's no need to persist it. We have it already. + logger.info( + "_process_pulled_event: Ignoring received event %s which we " + "have already seen", + event.event_id, + ) + continue + + # While we have seen this event before, it was stored as an outlier. + # We'll now persist it as a non-outlier. + logger.info("De-outliering event %s", event_id) + + # Continue on with the events that are new to us. + new_events.append(event) + # We want to sort these by depth so we process them and # tell clients about them in order. - sorted_events = sorted(events, key=lambda x: x.depth) + sorted_events = sorted(new_events, key=lambda x: x.depth) for ev in sorted_events: with nested_logging_context(ev.event_id): await self._process_pulled_event(origin, ev, backfilled=backfilled) @@ -852,18 +885,6 @@ class FederationEventHandler: event_id = event.event_id - existing = await self._store.get_event( - event_id, allow_none=True, allow_rejected=True - ) - if existing: - if not existing.internal_metadata.is_outlier(): - logger.info( - "_process_pulled_event: Ignoring received event %s which we have already seen", - event_id, - ) - return - logger.info("De-outliering event %s", event_id) - try: self._sanity_check_event(event) except SynapseError as err: diff --git a/synapse/storage/databases/main/events_worker.py b/synapse/storage/databases/main/events_worker.py index cfd4780add..7bc7f2f33e 100644 --- a/synapse/storage/databases/main/events_worker.py +++ b/synapse/storage/databases/main/events_worker.py @@ -374,7 +374,7 @@ class EventsWorkerStore(SQLBaseStore): If there is a mismatch, behave as per allow_none. Returns: - The event, or None if the event was not found. + The event, or None if the event was not found and allow_none is `True`. """ if not isinstance(event_id, str): raise TypeError("Invalid event event_id %r" % (event_id,)) diff --git a/tests/handlers/test_federation.py b/tests/handlers/test_federation.py index 745750b1d7..d00c69c229 100644 --- a/tests/handlers/test_federation.py +++ b/tests/handlers/test_federation.py @@ -19,7 +19,13 @@ from unittest.mock import Mock, patch from twisted.test.proto_helpers import MemoryReactor from synapse.api.constants import EventTypes -from synapse.api.errors import AuthError, Codes, LimitExceededError, SynapseError +from synapse.api.errors import ( + AuthError, + Codes, + LimitExceededError, + NotFoundError, + SynapseError, +) from synapse.api.room_versions import RoomVersions from synapse.events import EventBase, make_event_from_dict from synapse.federation.federation_base import event_from_pdu_json @@ -28,6 +34,7 @@ from synapse.logging.context import LoggingContext, run_in_background from synapse.rest import admin from synapse.rest.client import login, room from synapse.server import HomeServer +from synapse.storage.databases.main.events_worker import EventCacheEntry from synapse.util import Clock from synapse.util.stringutils import random_string @@ -322,6 +329,102 @@ class FederationTestCase(unittest.FederatingHomeserverTestCase): ) self.get_success(d) + def test_backfill_ignores_known_events(self) -> None: + """ + Tests that events that we already know about are ignored when backfilling. + """ + # Set up users + user_id = self.register_user("kermit", "test") + tok = self.login("kermit", "test") + + other_server = "otherserver" + other_user = "@otheruser:" + other_server + + # Create a room to backfill events into + room_id = self.helper.create_room_as(room_creator=user_id, tok=tok) + room_version = self.get_success(self.store.get_room_version(room_id)) + + # Build an event to backfill + event = event_from_pdu_json( + { + "type": EventTypes.Message, + "content": {"body": "hello world", "msgtype": "m.text"}, + "room_id": room_id, + "sender": other_user, + "depth": 32, + "prev_events": [], + "auth_events": [], + "origin_server_ts": self.clock.time_msec(), + }, + room_version, + ) + + # Ensure the event is not already in the DB + self.get_failure( + self.store.get_event(event.event_id), + NotFoundError, + ) + + # Backfill the event and check that it has entered the DB. + + # We mock out the FederationClient.backfill method, to pretend that a remote + # server has returned our fake event. + federation_client_backfill_mock = Mock(return_value=make_awaitable([event])) + self.hs.get_federation_client().backfill = federation_client_backfill_mock + + # We also mock the persist method with a side effect of itself. This allows us + # to track when it has been called while preserving its function. + persist_events_and_notify_mock = Mock( + side_effect=self.hs.get_federation_event_handler().persist_events_and_notify + ) + self.hs.get_federation_event_handler().persist_events_and_notify = ( + persist_events_and_notify_mock + ) + + # Small side-tangent. We populate the event cache with the event, even though + # it is not yet in the DB. This is an invalid scenario that can currently occur + # due to not properly invalidating the event cache. + # See https://github.com/matrix-org/synapse/issues/13476. + # + # As a result, backfill should not rely on the event cache to check whether + # we already have an event in the DB. + # TODO: Remove this bit when the event cache is properly invalidated. + cache_entry = EventCacheEntry( + event=event, + redacted_event=None, + ) + self.store._get_event_cache.set_local((event.event_id,), cache_entry) + + # We now call FederationEventHandler.backfill (a separate method) to trigger + # a backfill request. It should receive the fake event. + self.get_success( + self.hs.get_federation_event_handler().backfill( + other_user, + room_id, + limit=10, + extremities=[], + ) + ) + + # Check that our fake event was persisted. + persist_events_and_notify_mock.assert_called_once() + persist_events_and_notify_mock.reset_mock() + + # Now we repeat the backfill, having the homeserver receive the fake event + # again. + self.get_success( + self.hs.get_federation_event_handler().backfill( + other_user, + room_id, + limit=10, + extremities=[], + ), + ) + + # This time, we expect no event persistence to have occurred, as we already + # have this event. + persist_events_and_notify_mock.assert_not_called() + @unittest.override_config( {"rc_invites": {"per_user": {"per_second": 0.5, "burst_count": 3}}} ) -- cgit 1.5.1 From c3a4780080a5bcb04132283c0f32f7452655792a Mon Sep 17 00:00:00 2001 From: David Robertson Date: Tue, 18 Oct 2022 12:33:18 +0100 Subject: When restarting a partial join resync, prioritise the server which actioned a partial join (#14126) --- changelog.d/14126.misc | 1 + synapse/handlers/device.py | 5 +- synapse/handlers/federation.py | 57 +++++++++++++--------- synapse/storage/database.py | 2 +- synapse/storage/databases/main/room.py | 43 +++++++++++++--- .../delta/73/09partial_joined_via_destination.sql | 18 +++++++ 6 files changed, 95 insertions(+), 31 deletions(-) create mode 100644 changelog.d/14126.misc create mode 100644 synapse/storage/schema/main/delta/73/09partial_joined_via_destination.sql (limited to 'synapse/handlers') diff --git a/changelog.d/14126.misc b/changelog.d/14126.misc new file mode 100644 index 0000000000..30b3482fbd --- /dev/null +++ b/changelog.d/14126.misc @@ -0,0 +1 @@ +Faster joins: prioritise the server we joined by when restarting a partial join resync. diff --git a/synapse/handlers/device.py b/synapse/handlers/device.py index f9cc5bddbc..c597639a7f 100644 --- a/synapse/handlers/device.py +++ b/synapse/handlers/device.py @@ -937,7 +937,10 @@ class DeviceListUpdater: # Check if we are partially joining any rooms. If so we need to store # all device list updates so that we can handle them correctly once we # know who is in the room. - partial_rooms = await self.store.get_partial_state_rooms_and_servers() + # TODO(faster joins): this fetches and processes a bunch of data that we don't + # use. Could be replaced by a tighter query e.g. + # SELECT EXISTS(SELECT 1 FROM partial_state_rooms) + partial_rooms = await self.store.get_partial_state_room_resync_info() if partial_rooms: await self.store.add_remote_device_list_to_pending( user_id, diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index 5f7e0a1f79..ccc045d36f 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -632,6 +632,7 @@ class FederationHandler: room_id=room_id, servers=ret.servers_in_room, device_lists_stream_id=self.store.get_device_stream_token(), + joined_via=origin, ) try: @@ -1615,13 +1616,13 @@ class FederationHandler: """Resumes resyncing of all partial-state rooms after a restart.""" assert not self.config.worker.worker_app - partial_state_rooms = await self.store.get_partial_state_rooms_and_servers() - for room_id, servers_in_room in partial_state_rooms.items(): + partial_state_rooms = await self.store.get_partial_state_room_resync_info() + for room_id, resync_info in partial_state_rooms.items(): run_as_background_process( desc="sync_partial_state_room", func=self._sync_partial_state_room, - initial_destination=None, - other_destinations=servers_in_room, + initial_destination=resync_info.joined_via, + other_destinations=resync_info.servers_in_room, room_id=room_id, ) @@ -1650,28 +1651,12 @@ class FederationHandler: # really leave, that might mean we have difficulty getting the room state over # federation. # https://github.com/matrix-org/synapse/issues/12802 - # - # TODO(faster_joins): we need some way of prioritising which homeservers in - # `other_destinations` to try first, otherwise we'll spend ages trying dead - # homeservers for large rooms. - # https://github.com/matrix-org/synapse/issues/12999 - - if initial_destination is None and len(other_destinations) == 0: - raise ValueError( - f"Cannot resync state of {room_id}: no destinations provided" - ) # Make an infinite iterator of destinations to try. Once we find a working # destination, we'll stick with it until it flakes. - destinations: Collection[str] - if initial_destination is not None: - # Move `initial_destination` to the front of the list. - destinations = list(other_destinations) - if initial_destination in destinations: - destinations.remove(initial_destination) - destinations = [initial_destination] + destinations - else: - destinations = other_destinations + destinations = _prioritise_destinations_for_partial_state_resync( + initial_destination, other_destinations, room_id + ) destination_iter = itertools.cycle(destinations) # `destination` is the current remote homeserver we're pulling from. @@ -1769,3 +1754,29 @@ class FederationHandler: room_id, destination, ) + + +def _prioritise_destinations_for_partial_state_resync( + initial_destination: Optional[str], + other_destinations: Collection[str], + room_id: str, +) -> Collection[str]: + """Work out the order in which we should ask servers to resync events. + + If an `initial_destination` is given, it takes top priority. Otherwise + all servers are treated equally. + + :raises ValueError: if no destination is provided at all. + """ + if initial_destination is None and len(other_destinations) == 0: + raise ValueError(f"Cannot resync state of {room_id}: no destinations provided") + + if initial_destination is None: + return other_destinations + + # Move `initial_destination` to the front of the list. + destinations = list(other_destinations) + if initial_destination in destinations: + destinations.remove(initial_destination) + destinations = [initial_destination] + destinations + return destinations diff --git a/synapse/storage/database.py b/synapse/storage/database.py index 7bb21f8f81..4717c9728a 100644 --- a/synapse/storage/database.py +++ b/synapse/storage/database.py @@ -1658,7 +1658,7 @@ class DatabasePool: table: string giving the table name keyvalues: dict of column names and values to select the row with retcol: string giving the name of the column to return - allow_none: If true, return None instead of failing if the SELECT + allow_none: If true, return None instead of raising StoreError if the SELECT statement returns no rows desc: description of the transaction, for logging and metrics """ diff --git a/synapse/storage/databases/main/room.py b/synapse/storage/databases/main/room.py index e41c99027a..7d97f8f60e 100644 --- a/synapse/storage/databases/main/room.py +++ b/synapse/storage/databases/main/room.py @@ -97,6 +97,12 @@ class RoomSortOrder(Enum): STATE_EVENTS = "state_events" +@attr.s(slots=True, frozen=True, auto_attribs=True) +class PartialStateResyncInfo: + joined_via: Optional[str] + servers_in_room: List[str] = attr.ib(factory=list) + + class RoomWorkerStore(CacheInvalidationWorkerStore): def __init__( self, @@ -1160,17 +1166,29 @@ class RoomWorkerStore(CacheInvalidationWorkerStore): desc="get_partial_state_servers_at_join", ) - async def get_partial_state_rooms_and_servers( + async def get_partial_state_room_resync_info( self, - ) -> Mapping[str, Collection[str]]: - """Get all rooms containing events with partial state, and the servers known - to be in the room. + ) -> Mapping[str, PartialStateResyncInfo]: + """Get all rooms containing events with partial state, and the information + needed to restart a "resync" of those rooms. Returns: A dictionary of rooms with partial state, with room IDs as keys and lists of servers in rooms as values. """ - room_servers: Dict[str, List[str]] = {} + room_servers: Dict[str, PartialStateResyncInfo] = {} + + rows = await self.db_pool.simple_select_list( + table="partial_state_rooms", + keyvalues={}, + retcols=("room_id", "joined_via"), + desc="get_server_which_served_partial_join", + ) + + for row in rows: + room_id = row["room_id"] + joined_via = row["joined_via"] + room_servers[room_id] = PartialStateResyncInfo(joined_via=joined_via) rows = await self.db_pool.simple_select_list( "partial_state_rooms_servers", @@ -1182,7 +1200,15 @@ class RoomWorkerStore(CacheInvalidationWorkerStore): for row in rows: room_id = row["room_id"] server_name = row["server_name"] - room_servers.setdefault(room_id, []).append(server_name) + entry = room_servers.get(room_id) + if entry is None: + # There is a foreign key constraint which enforces that every room_id in + # partial_state_rooms_servers appears in partial_state_rooms. So we + # expect `entry` to be non-null. (This reasoning fails if we've + # partial-joined between the two SELECTs, but this is unlikely to happen + # in practice.) + continue + entry.servers_in_room.append(server_name) return room_servers @@ -1827,6 +1853,7 @@ class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore): room_id: str, servers: Collection[str], device_lists_stream_id: int, + joined_via: str, ) -> None: """Mark the given room as containing events with partial state. @@ -1842,6 +1869,7 @@ class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore): servers: other servers known to be in the room device_lists_stream_id: the device_lists stream ID at the time when we first joined the room. + joined_via: the server name we requested a partial join from. """ await self.db_pool.runInteraction( "store_partial_state_room", @@ -1849,6 +1877,7 @@ class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore): room_id, servers, device_lists_stream_id, + joined_via, ) def _store_partial_state_room_txn( @@ -1857,6 +1886,7 @@ class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore): room_id: str, servers: Collection[str], device_lists_stream_id: int, + joined_via: str, ) -> None: DatabasePool.simple_insert_txn( txn, @@ -1866,6 +1896,7 @@ class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore): "device_lists_stream_id": device_lists_stream_id, # To be updated later once the join event is persisted. "join_event_id": None, + "joined_via": joined_via, }, ) DatabasePool.simple_insert_many_txn( diff --git a/synapse/storage/schema/main/delta/73/09partial_joined_via_destination.sql b/synapse/storage/schema/main/delta/73/09partial_joined_via_destination.sql new file mode 100644 index 0000000000..066d602b18 --- /dev/null +++ b/synapse/storage/schema/main/delta/73/09partial_joined_via_destination.sql @@ -0,0 +1,18 @@ +/* Copyright 2022 The Matrix.org Foundation C.I.C + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +-- When we resync partial state, we prioritise doing so using the server we +-- partial-joined from. To do this we need to record which server that was! +ALTER TABLE partial_state_rooms ADD COLUMN joined_via TEXT; -- cgit 1.5.1 From 4eaf3eb840b8cfa78d970216c74fc128495f08a5 Mon Sep 17 00:00:00 2001 From: Hugh Nimmo-Smith Date: Tue, 18 Oct 2022 16:52:25 +0100 Subject: Implementation of HTTP 307 response for MSC3886 POST endpoint (#14018) Co-authored-by: reivilibre Co-authored-by: Andrew Morgan --- changelog.d/14018.feature | 1 + synapse/config/experimental.py | 7 +- synapse/config/server.py | 4 ++ synapse/handlers/sso.py | 2 +- synapse/http/server.py | 48 ++++++++++--- synapse/http/site.py | 3 + synapse/rest/__init__.py | 2 + synapse/rest/client/rendezvous.py | 74 +++++++++++++++++++ synapse/rest/client/versions.py | 3 + synapse/rest/key/v2/local_key_resource.py | 4 +- synapse/rest/synapse/client/new_user_consent.py | 3 +- synapse/rest/well_known.py | 3 +- tests/logging/test_terse_json.py | 1 + tests/rest/client/test_rendezvous.py | 45 ++++++++++++ tests/server.py | 8 ++- tests/test_server.py | 94 ++++++++++++++++++------- 16 files changed, 257 insertions(+), 45 deletions(-) create mode 100644 changelog.d/14018.feature create mode 100644 synapse/rest/client/rendezvous.py create mode 100644 tests/rest/client/test_rendezvous.py (limited to 'synapse/handlers') diff --git a/changelog.d/14018.feature b/changelog.d/14018.feature new file mode 100644 index 0000000000..c8454607eb --- /dev/null +++ b/changelog.d/14018.feature @@ -0,0 +1 @@ +Support for redirecting to an implementation of a [MSC3886](https://github.com/matrix-org/matrix-spec-proposals/pull/3886) HTTP rendezvous service. \ No newline at end of file diff --git a/synapse/config/experimental.py b/synapse/config/experimental.py index f9a49451d8..4009add01d 100644 --- a/synapse/config/experimental.py +++ b/synapse/config/experimental.py @@ -12,7 +12,7 @@ # See the License for the specific language governing permissions and # limitations under the License. -from typing import Any +from typing import Any, Optional import attr @@ -120,3 +120,8 @@ class ExperimentalConfig(Config): # MSC3874: Filtering /messages with rel_types / not_rel_types. self.msc3874_enabled: bool = experimental.get("msc3874_enabled", False) + + # MSC3886: Simple client rendezvous capability + self.msc3886_endpoint: Optional[str] = experimental.get( + "msc3886_endpoint", None + ) diff --git a/synapse/config/server.py b/synapse/config/server.py index f2353ce5fb..ec46ca63ad 100644 --- a/synapse/config/server.py +++ b/synapse/config/server.py @@ -207,6 +207,9 @@ class HttpListenerConfig: additional_resources: Dict[str, dict] = attr.Factory(dict) tag: Optional[str] = None request_id_header: Optional[str] = None + # If true, the listener will return CORS response headers compatible with MSC3886: + # https://github.com/matrix-org/matrix-spec-proposals/pull/3886 + experimental_cors_msc3886: bool = False @attr.s(slots=True, frozen=True, auto_attribs=True) @@ -935,6 +938,7 @@ def parse_listener_def(num: int, listener: Any) -> ListenerConfig: additional_resources=listener.get("additional_resources", {}), tag=listener.get("tag"), request_id_header=listener.get("request_id_header"), + experimental_cors_msc3886=listener.get("experimental_cors_msc3886", False), ) return ListenerConfig(port, bind_addresses, listener_type, tls, http_config) diff --git a/synapse/handlers/sso.py b/synapse/handlers/sso.py index e035677b8a..5943f08e91 100644 --- a/synapse/handlers/sso.py +++ b/synapse/handlers/sso.py @@ -874,7 +874,7 @@ class SsoHandler: ) async def handle_terms_accepted( - self, request: Request, session_id: str, terms_version: str + self, request: SynapseRequest, session_id: str, terms_version: str ) -> None: """Handle a request to the new-user 'consent' endpoint diff --git a/synapse/http/server.py b/synapse/http/server.py index bcbfac2c9f..b26e34bceb 100644 --- a/synapse/http/server.py +++ b/synapse/http/server.py @@ -19,6 +19,7 @@ import logging import types import urllib from http import HTTPStatus +from http.client import FOUND from inspect import isawaitable from typing import ( TYPE_CHECKING, @@ -339,7 +340,7 @@ class _AsyncResource(resource.Resource, metaclass=abc.ABCMeta): return callback_return - _unrecognised_request_handler(request) + return _unrecognised_request_handler(request) @abc.abstractmethod def _send_response( @@ -598,7 +599,7 @@ class RootRedirect(resource.Resource): class OptionsResource(resource.Resource): """Responds to OPTION requests for itself and all children.""" - def render_OPTIONS(self, request: Request) -> bytes: + def render_OPTIONS(self, request: SynapseRequest) -> bytes: request.setResponseCode(204) request.setHeader(b"Content-Length", b"0") @@ -763,7 +764,7 @@ def respond_with_json( def respond_with_json_bytes( - request: Request, + request: SynapseRequest, code: int, json_bytes: bytes, send_cors: bool = False, @@ -859,7 +860,7 @@ def _write_bytes_to_request(request: Request, bytes_to_write: bytes) -> None: _ByteProducer(request, bytes_generator) -def set_cors_headers(request: Request) -> None: +def set_cors_headers(request: SynapseRequest) -> None: """Set the CORS headers so that javascript running in a web browsers can use this API @@ -870,10 +871,20 @@ def set_cors_headers(request: Request) -> None: request.setHeader( b"Access-Control-Allow-Methods", b"GET, HEAD, POST, PUT, DELETE, OPTIONS" ) - request.setHeader( - b"Access-Control-Allow-Headers", - b"X-Requested-With, Content-Type, Authorization, Date", - ) + if request.experimental_cors_msc3886: + request.setHeader( + b"Access-Control-Allow-Headers", + b"X-Requested-With, Content-Type, Authorization, Date, If-Match, If-None-Match", + ) + request.setHeader( + b"Access-Control-Expose-Headers", + b"ETag, Location, X-Max-Bytes", + ) + else: + request.setHeader( + b"Access-Control-Allow-Headers", + b"X-Requested-With, Content-Type, Authorization, Date", + ) def set_corp_headers(request: Request) -> None: @@ -942,10 +953,25 @@ def set_clickjacking_protection_headers(request: Request) -> None: request.setHeader(b"Content-Security-Policy", b"frame-ancestors 'none';") -def respond_with_redirect(request: Request, url: bytes) -> None: - """Write a 302 response to the request, if it is still alive.""" +def respond_with_redirect( + request: SynapseRequest, url: bytes, statusCode: int = FOUND, cors: bool = False +) -> None: + """ + Write a 302 (or other specified status code) response to the request, if it is still alive. + + Args: + request: The http request to respond to. + url: The URL to redirect to. + statusCode: The HTTP status code to use for the redirect (defaults to 302). + cors: Whether to set CORS headers on the response. + """ logger.debug("Redirect to %s", url.decode("utf-8")) - request.redirect(url) + + if cors: + set_cors_headers(request) + + request.setResponseCode(statusCode) + request.setHeader(b"location", url) finish_request(request) diff --git a/synapse/http/site.py b/synapse/http/site.py index 55a6afce35..3dbd541fed 100644 --- a/synapse/http/site.py +++ b/synapse/http/site.py @@ -82,6 +82,7 @@ class SynapseRequest(Request): self.reactor = site.reactor self._channel = channel # this is used by the tests self.start_time = 0.0 + self.experimental_cors_msc3886 = site.experimental_cors_msc3886 # The requester, if authenticated. For federation requests this is the # server name, for client requests this is the Requester object. @@ -622,6 +623,8 @@ class SynapseSite(Site): request_id_header = config.http_options.request_id_header + self.experimental_cors_msc3886 = config.http_options.experimental_cors_msc3886 + def request_factory(channel: HTTPChannel, queued: bool) -> Request: return request_class( channel, diff --git a/synapse/rest/__init__.py b/synapse/rest/__init__.py index 9a2ab99ede..28542cd774 100644 --- a/synapse/rest/__init__.py +++ b/synapse/rest/__init__.py @@ -44,6 +44,7 @@ from synapse.rest.client import ( receipts, register, relations, + rendezvous, report_event, room, room_batch, @@ -132,3 +133,4 @@ class ClientRestResource(JsonResource): # unstable mutual_rooms.register_servlets(hs, client_resource) login_token_request.register_servlets(hs, client_resource) + rendezvous.register_servlets(hs, client_resource) diff --git a/synapse/rest/client/rendezvous.py b/synapse/rest/client/rendezvous.py new file mode 100644 index 0000000000..89176b1ffa --- /dev/null +++ b/synapse/rest/client/rendezvous.py @@ -0,0 +1,74 @@ +# Copyright 2022 The Matrix.org Foundation C.I.C. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import logging +from http.client import TEMPORARY_REDIRECT +from typing import TYPE_CHECKING, Optional + +from synapse.http.server import HttpServer, respond_with_redirect +from synapse.http.servlet import RestServlet +from synapse.http.site import SynapseRequest +from synapse.rest.client._base import client_patterns + +if TYPE_CHECKING: + from synapse.server import HomeServer + +logger = logging.getLogger(__name__) + + +class RendezvousServlet(RestServlet): + """ + This is a placeholder implementation of [MSC3886](https://github.com/matrix-org/matrix-spec-proposals/pull/3886) + simple client rendezvous capability that is used by the "Sign in with QR" functionality. + + This implementation only serves as a 307 redirect to a configured server rather than being a full implementation. + + A module that implements the full functionality is available at: https://pypi.org/project/matrix-http-rendezvous-synapse/. + + Request: + + POST /rendezvous HTTP/1.1 + Content-Type: ... + + ... + + Response: + + HTTP/1.1 307 + Location: + """ + + PATTERNS = client_patterns( + "/org.matrix.msc3886/rendezvous$", releases=[], v1=False, unstable=True + ) + + def __init__(self, hs: "HomeServer"): + super().__init__() + redirection_target: Optional[str] = hs.config.experimental.msc3886_endpoint + assert ( + redirection_target is not None + ), "Servlet is only registered if there is a redirection target" + self.endpoint = redirection_target.encode("utf-8") + + async def on_POST(self, request: SynapseRequest) -> None: + respond_with_redirect( + request, self.endpoint, statusCode=TEMPORARY_REDIRECT, cors=True + ) + + # PUT, GET and DELETE are not implemented as they should be fulfilled by the redirect target. + + +def register_servlets(hs: "HomeServer", http_server: HttpServer) -> None: + if hs.config.experimental.msc3886_endpoint is not None: + RendezvousServlet(hs).register(http_server) diff --git a/synapse/rest/client/versions.py b/synapse/rest/client/versions.py index 4b87ee978a..9b1b72c68a 100644 --- a/synapse/rest/client/versions.py +++ b/synapse/rest/client/versions.py @@ -116,6 +116,9 @@ class VersionsRestServlet(RestServlet): "org.matrix.msc3881": self.config.experimental.msc3881_enabled, # Adds support for filtering /messages by event relation. "org.matrix.msc3874": self.config.experimental.msc3874_enabled, + # Adds support for simple HTTP rendezvous as per MSC3886 + "org.matrix.msc3886": self.config.experimental.msc3886_endpoint + is not None, }, }, ) diff --git a/synapse/rest/key/v2/local_key_resource.py b/synapse/rest/key/v2/local_key_resource.py index 0c9f042c84..095993415c 100644 --- a/synapse/rest/key/v2/local_key_resource.py +++ b/synapse/rest/key/v2/local_key_resource.py @@ -20,9 +20,9 @@ from signedjson.sign import sign_json from unpaddedbase64 import encode_base64 from twisted.web.resource import Resource -from twisted.web.server import Request from synapse.http.server import respond_with_json_bytes +from synapse.http.site import SynapseRequest from synapse.types import JsonDict if TYPE_CHECKING: @@ -99,7 +99,7 @@ class LocalKey(Resource): json_object = sign_json(json_object, self.config.server.server_name, key) return json_object - def render_GET(self, request: Request) -> Optional[int]: + def render_GET(self, request: SynapseRequest) -> Optional[int]: time_now = self.clock.time_msec() # Update the expiry time if less than half the interval remains. if time_now + self.config.key.key_refresh_interval / 2 > self.valid_until_ts: diff --git a/synapse/rest/synapse/client/new_user_consent.py b/synapse/rest/synapse/client/new_user_consent.py index 1c1c7b3613..22784157e6 100644 --- a/synapse/rest/synapse/client/new_user_consent.py +++ b/synapse/rest/synapse/client/new_user_consent.py @@ -20,6 +20,7 @@ from synapse.api.errors import SynapseError from synapse.handlers.sso import get_username_mapping_session_cookie_from_request from synapse.http.server import DirectServeHtmlResource, respond_with_html from synapse.http.servlet import parse_string +from synapse.http.site import SynapseRequest from synapse.types import UserID from synapse.util.templates import build_jinja_env @@ -88,7 +89,7 @@ class NewUserConsentResource(DirectServeHtmlResource): html = template.render(template_params) respond_with_html(request, 200, html) - async def _async_render_POST(self, request: Request) -> None: + async def _async_render_POST(self, request: SynapseRequest) -> None: try: session_id = get_username_mapping_session_cookie_from_request(request) except SynapseError as e: diff --git a/synapse/rest/well_known.py b/synapse/rest/well_known.py index 6f7ac54c65..e2174fdfea 100644 --- a/synapse/rest/well_known.py +++ b/synapse/rest/well_known.py @@ -18,6 +18,7 @@ from twisted.web.resource import Resource from twisted.web.server import Request from synapse.http.server import set_cors_headers +from synapse.http.site import SynapseRequest from synapse.types import JsonDict from synapse.util import json_encoder from synapse.util.stringutils import parse_server_name @@ -63,7 +64,7 @@ class ClientWellKnownResource(Resource): Resource.__init__(self) self._well_known_builder = WellKnownBuilder(hs) - def render_GET(self, request: Request) -> bytes: + def render_GET(self, request: SynapseRequest) -> bytes: set_cors_headers(request) r = self._well_known_builder.get_well_known() if not r: diff --git a/tests/logging/test_terse_json.py b/tests/logging/test_terse_json.py index 96f399b7ab..0b0d8737c1 100644 --- a/tests/logging/test_terse_json.py +++ b/tests/logging/test_terse_json.py @@ -153,6 +153,7 @@ class TerseJsonTestCase(LoggerCleanupMixin, TestCase): site.site_tag = "test-site" site.server_version_string = "Server v1" site.reactor = Mock() + site.experimental_cors_msc3886 = False request = SynapseRequest(FakeChannel(site, None), site) # Call requestReceived to finish instantiating the object. request.content = BytesIO() diff --git a/tests/rest/client/test_rendezvous.py b/tests/rest/client/test_rendezvous.py new file mode 100644 index 0000000000..ad00a476e1 --- /dev/null +++ b/tests/rest/client/test_rendezvous.py @@ -0,0 +1,45 @@ +# Copyright 2022 The Matrix.org Foundation C.I.C. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from twisted.test.proto_helpers import MemoryReactor + +from synapse.rest.client import rendezvous +from synapse.server import HomeServer +from synapse.util import Clock + +from tests import unittest +from tests.unittest import override_config + +endpoint = "/_matrix/client/unstable/org.matrix.msc3886/rendezvous" + + +class RendezvousServletTestCase(unittest.HomeserverTestCase): + + servlets = [ + rendezvous.register_servlets, + ] + + def make_homeserver(self, reactor: MemoryReactor, clock: Clock) -> HomeServer: + self.hs = self.setup_test_homeserver() + return self.hs + + def test_disabled(self) -> None: + channel = self.make_request("POST", endpoint, {}, access_token=None) + self.assertEqual(channel.code, 400) + + @override_config({"experimental_features": {"msc3886_endpoint": "/asd"}}) + def test_redirect(self) -> None: + channel = self.make_request("POST", endpoint, {}, access_token=None) + self.assertEqual(channel.code, 307) + self.assertEqual(channel.headers.getRawHeaders("Location"), ["/asd"]) diff --git a/tests/server.py b/tests/server.py index c447d5e4c4..8b1d186219 100644 --- a/tests/server.py +++ b/tests/server.py @@ -266,7 +266,12 @@ class FakeSite: site_tag = "test" access_logger = logging.getLogger("synapse.access.http.fake") - def __init__(self, resource: IResource, reactor: IReactorTime): + def __init__( + self, + resource: IResource, + reactor: IReactorTime, + experimental_cors_msc3886: bool = False, + ): """ Args: @@ -274,6 +279,7 @@ class FakeSite: """ self._resource = resource self.reactor = reactor + self.experimental_cors_msc3886 = experimental_cors_msc3886 def getResourceFor(self, request): return self._resource diff --git a/tests/test_server.py b/tests/test_server.py index 7c66448245..2d9a0257d4 100644 --- a/tests/test_server.py +++ b/tests/test_server.py @@ -222,13 +222,22 @@ class OptionsResourceTests(unittest.TestCase): self.resource = OptionsResource() self.resource.putChild(b"res", DummyResource()) - def _make_request(self, method: bytes, path: bytes) -> FakeChannel: + def _make_request( + self, method: bytes, path: bytes, experimental_cors_msc3886: bool = False + ) -> FakeChannel: """Create a request from the method/path and return a channel with the response.""" # Create a site and query for the resource. site = SynapseSite( "test", "site_tag", - parse_listener_def(0, {"type": "http", "port": 0}), + parse_listener_def( + 0, + { + "type": "http", + "port": 0, + "experimental_cors_msc3886": experimental_cors_msc3886, + }, + ), self.resource, "1.0", max_request_body_size=4096, @@ -239,25 +248,58 @@ class OptionsResourceTests(unittest.TestCase): channel = make_request(self.reactor, site, method, path, shorthand=False) return channel + def _check_cors_standard_headers(self, channel: FakeChannel) -> None: + # Ensure the correct CORS headers have been added + # as per https://spec.matrix.org/v1.4/client-server-api/#web-browser-clients + self.assertEqual( + channel.headers.getRawHeaders(b"Access-Control-Allow-Origin"), + [b"*"], + "has correct CORS Origin header", + ) + self.assertEqual( + channel.headers.getRawHeaders(b"Access-Control-Allow-Methods"), + [b"GET, HEAD, POST, PUT, DELETE, OPTIONS"], # HEAD isn't in the spec + "has correct CORS Methods header", + ) + self.assertEqual( + channel.headers.getRawHeaders(b"Access-Control-Allow-Headers"), + [b"X-Requested-With, Content-Type, Authorization, Date"], + "has correct CORS Headers header", + ) + + def _check_cors_msc3886_headers(self, channel: FakeChannel) -> None: + # Ensure the correct CORS headers have been added + # as per https://github.com/matrix-org/matrix-spec-proposals/blob/hughns/simple-rendezvous-capability/proposals/3886-simple-rendezvous-capability.md#cors + self.assertEqual( + channel.headers.getRawHeaders(b"Access-Control-Allow-Origin"), + [b"*"], + "has correct CORS Origin header", + ) + self.assertEqual( + channel.headers.getRawHeaders(b"Access-Control-Allow-Methods"), + [b"GET, HEAD, POST, PUT, DELETE, OPTIONS"], # HEAD isn't in the spec + "has correct CORS Methods header", + ) + self.assertEqual( + channel.headers.getRawHeaders(b"Access-Control-Allow-Headers"), + [ + b"X-Requested-With, Content-Type, Authorization, Date, If-Match, If-None-Match" + ], + "has correct CORS Headers header", + ) + self.assertEqual( + channel.headers.getRawHeaders(b"Access-Control-Expose-Headers"), + [b"ETag, Location, X-Max-Bytes"], + "has correct CORS Expose Headers header", + ) + def test_unknown_options_request(self) -> None: """An OPTIONS requests to an unknown URL still returns 204 No Content.""" channel = self._make_request(b"OPTIONS", b"/foo/") self.assertEqual(channel.code, 204) self.assertNotIn("body", channel.result) - # Ensure the correct CORS headers have been added - self.assertTrue( - channel.headers.hasHeader(b"Access-Control-Allow-Origin"), - "has CORS Origin header", - ) - self.assertTrue( - channel.headers.hasHeader(b"Access-Control-Allow-Methods"), - "has CORS Methods header", - ) - self.assertTrue( - channel.headers.hasHeader(b"Access-Control-Allow-Headers"), - "has CORS Headers header", - ) + self._check_cors_standard_headers(channel) def test_known_options_request(self) -> None: """An OPTIONS requests to an known URL still returns 204 No Content.""" @@ -265,19 +307,17 @@ class OptionsResourceTests(unittest.TestCase): self.assertEqual(channel.code, 204) self.assertNotIn("body", channel.result) - # Ensure the correct CORS headers have been added - self.assertTrue( - channel.headers.hasHeader(b"Access-Control-Allow-Origin"), - "has CORS Origin header", - ) - self.assertTrue( - channel.headers.hasHeader(b"Access-Control-Allow-Methods"), - "has CORS Methods header", - ) - self.assertTrue( - channel.headers.hasHeader(b"Access-Control-Allow-Headers"), - "has CORS Headers header", + self._check_cors_standard_headers(channel) + + def test_known_options_request_msc3886(self) -> None: + """An OPTIONS requests to an known URL still returns 204 No Content.""" + channel = self._make_request( + b"OPTIONS", b"/res/", experimental_cors_msc3886=True ) + self.assertEqual(channel.code, 204) + self.assertNotIn("body", channel.result) + + self._check_cors_msc3886_headers(channel) def test_unknown_request(self) -> None: """A non-OPTIONS request to an unknown URL should 404.""" -- cgit 1.5.1 From 847e2393f3198b88809c9b99de5c681efbf1c92e Mon Sep 17 00:00:00 2001 From: Shay Date: Tue, 18 Oct 2022 09:58:47 -0700 Subject: Prepatory work for adding power level event to batched events (#14214) --- changelog.d/14214.misc | 1 + synapse/event_auth.py | 19 ++++++++++++++++++- synapse/handlers/event_auth.py | 18 +++++++++++++----- synapse/handlers/federation.py | 12 +++++------- synapse/handlers/message.py | 10 +++++++++- synapse/handlers/room.py | 4 +--- 6 files changed, 47 insertions(+), 17 deletions(-) create mode 100644 changelog.d/14214.misc (limited to 'synapse/handlers') diff --git a/changelog.d/14214.misc b/changelog.d/14214.misc new file mode 100644 index 0000000000..102928b575 --- /dev/null +++ b/changelog.d/14214.misc @@ -0,0 +1 @@ +When authenticating batched events, check for auth events in batch as well as DB. diff --git a/synapse/event_auth.py b/synapse/event_auth.py index c7d5ef92fc..bab31e33c5 100644 --- a/synapse/event_auth.py +++ b/synapse/event_auth.py @@ -15,7 +15,18 @@ import logging import typing -from typing import Any, Collection, Dict, Iterable, List, Optional, Set, Tuple, Union +from typing import ( + Any, + Collection, + Dict, + Iterable, + List, + Mapping, + Optional, + Set, + Tuple, + Union, +) from canonicaljson import encode_canonical_json from signedjson.key import decode_verify_key_bytes @@ -134,6 +145,7 @@ def validate_event_for_room_version(event: "EventBase") -> None: async def check_state_independent_auth_rules( store: _EventSourceStore, event: "EventBase", + batched_auth_events: Optional[Mapping[str, "EventBase"]] = None, ) -> None: """Check that an event complies with auth rules that are independent of room state @@ -143,6 +155,8 @@ async def check_state_independent_auth_rules( Args: store: the datastore; used to fetch the auth events for validation event: the event being checked. + batched_auth_events: if the event being authed is part of a batch, any events + from the same batch that may be necessary to auth the current event Raises: AuthError if the checks fail @@ -162,6 +176,9 @@ async def check_state_independent_auth_rules( redact_behaviour=EventRedactBehaviour.as_is, allow_rejected=True, ) + if batched_auth_events: + auth_events.update(batched_auth_events) + room_id = event.room_id auth_dict: MutableStateMap[str] = {} expected_auth_types = auth_types_for_event(event.room_version, event) diff --git a/synapse/handlers/event_auth.py b/synapse/handlers/event_auth.py index 8249ca1ed2..3bbad0271b 100644 --- a/synapse/handlers/event_auth.py +++ b/synapse/handlers/event_auth.py @@ -12,7 +12,7 @@ # See the License for the specific language governing permissions and # limitations under the License. import logging -from typing import TYPE_CHECKING, Collection, List, Optional, Union +from typing import TYPE_CHECKING, Collection, List, Mapping, Optional, Union from synapse import event_auth from synapse.api.constants import ( @@ -29,7 +29,6 @@ from synapse.event_auth import ( ) from synapse.events import EventBase from synapse.events.builder import EventBuilder -from synapse.events.snapshot import EventContext from synapse.types import StateMap, get_domain_from_id if TYPE_CHECKING: @@ -51,12 +50,21 @@ class EventAuthHandler: async def check_auth_rules_from_context( self, event: EventBase, - context: EventContext, + batched_auth_events: Optional[Mapping[str, EventBase]] = None, ) -> None: - """Check an event passes the auth rules at its own auth events""" - await check_state_independent_auth_rules(self._store, event) + """Check an event passes the auth rules at its own auth events + Args: + event: event to be authed + batched_auth_events: if the event being authed is part of a batch, any events + from the same batch that may be necessary to auth the current event + """ + await check_state_independent_auth_rules( + self._store, event, batched_auth_events + ) auth_event_ids = event.auth_event_ids() auth_events_by_id = await self._store.get_events(auth_event_ids) + if batched_auth_events: + auth_events_by_id.update(batched_auth_events) check_state_dependent_auth_rules(event, auth_events_by_id.values()) def compute_auth_events( diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index ccc045d36f..275a37a575 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -942,7 +942,7 @@ class FederationHandler: # The remote hasn't signed it yet, obviously. We'll do the full checks # when we get the event back in `on_send_join_request` - await self._event_auth_handler.check_auth_rules_from_context(event, context) + await self._event_auth_handler.check_auth_rules_from_context(event) return event async def on_invite_request( @@ -1123,7 +1123,7 @@ class FederationHandler: try: # The remote hasn't signed it yet, obviously. We'll do the full checks # when we get the event back in `on_send_leave_request` - await self._event_auth_handler.check_auth_rules_from_context(event, context) + await self._event_auth_handler.check_auth_rules_from_context(event) except AuthError as e: logger.warning("Failed to create new leave %r because %s", event, e) raise e @@ -1182,7 +1182,7 @@ class FederationHandler: try: # The remote hasn't signed it yet, obviously. We'll do the full checks # when we get the event back in `on_send_knock_request` - await self._event_auth_handler.check_auth_rules_from_context(event, context) + await self._event_auth_handler.check_auth_rules_from_context(event) except AuthError as e: logger.warning("Failed to create new knock %r because %s", event, e) raise e @@ -1348,9 +1348,7 @@ class FederationHandler: try: validate_event_for_room_version(event) - await self._event_auth_handler.check_auth_rules_from_context( - event, context - ) + await self._event_auth_handler.check_auth_rules_from_context(event) except AuthError as e: logger.warning("Denying new third party invite %r because %s", event, e) raise e @@ -1400,7 +1398,7 @@ class FederationHandler: try: validate_event_for_room_version(event) - await self._event_auth_handler.check_auth_rules_from_context(event, context) + await self._event_auth_handler.check_auth_rules_from_context(event) except AuthError as e: logger.warning("Denying third party invite %r because %s", event, e) raise e diff --git a/synapse/handlers/message.py b/synapse/handlers/message.py index 4e55ebba0b..15b828dd74 100644 --- a/synapse/handlers/message.py +++ b/synapse/handlers/message.py @@ -1360,8 +1360,16 @@ class EventCreationHandler: else: try: validate_event_for_room_version(event) + # If we are persisting a batch of events the event(s) needed to auth the + # current event may be part of the batch and will not be in the DB yet + event_id_to_event = {e.event_id: e for e, _ in events_and_context} + batched_auth_events = {} + for event_id in event.auth_event_ids(): + auth_event = event_id_to_event.get(event_id) + if auth_event: + batched_auth_events[event_id] = auth_event await self._event_auth_handler.check_auth_rules_from_context( - event, context + event, batched_auth_events ) except AuthError as err: logger.warning("Denying new event %r because %s", event, err) diff --git a/synapse/handlers/room.py b/synapse/handlers/room.py index 4e1aacb408..638f54051a 100644 --- a/synapse/handlers/room.py +++ b/synapse/handlers/room.py @@ -229,9 +229,7 @@ class RoomCreationHandler: }, ) validate_event_for_room_version(tombstone_event) - await self._event_auth_handler.check_auth_rules_from_context( - tombstone_event, tombstone_context - ) + await self._event_auth_handler.check_auth_rules_from_context(tombstone_event) # Upgrade the room # -- cgit 1.5.1 From 1433b5d5b64c3a6624e6e4ff4fef22127c49df86 Mon Sep 17 00:00:00 2001 From: Tadeusz Sośnierz Date: Fri, 21 Oct 2022 14:52:44 +0200 Subject: Show erasure status when listing users in the Admin API (#14205) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit * Show erasure status when listing users in the Admin API * Use USING when joining erased_users * Add changelog entry * Revert "Use USING when joining erased_users" This reverts commit 30bd2bf106415caadcfdbdd1b234ef2b106cc394. * Make the erased check work on postgres * Add a testcase for showing erased user status * Appease the style linter * Explicitly convert `erased` to bool to make SQLite consistent with Postgres This also adds us an easy way in to fix the other accidentally integered columns. * Move erasure status test to UsersListTestCase * Include user erased status when fetching user info via the admin API * Document the erase status in user_admin_api * Appease the linter and mypy * Signpost comments in tests Co-authored-by: Tadeusz Sośnierz Co-authored-by: David Robertson --- changelog.d/14205.feature | 1 + docs/admin_api/user_admin_api.md | 4 ++++ synapse/handlers/admin.py | 1 + synapse/storage/databases/main/__init__.py | 13 +++++++++-- tests/rest/admin/test_user.py | 35 +++++++++++++++++++++++++++++- 5 files changed, 51 insertions(+), 3 deletions(-) create mode 100644 changelog.d/14205.feature (limited to 'synapse/handlers') diff --git a/changelog.d/14205.feature b/changelog.d/14205.feature new file mode 100644 index 0000000000..6692063352 --- /dev/null +++ b/changelog.d/14205.feature @@ -0,0 +1 @@ +Show erasure status when listing users in the Admin API. diff --git a/docs/admin_api/user_admin_api.md b/docs/admin_api/user_admin_api.md index 3625c7b6c5..c95d6c9b05 100644 --- a/docs/admin_api/user_admin_api.md +++ b/docs/admin_api/user_admin_api.md @@ -37,6 +37,7 @@ It returns a JSON body like the following: "is_guest": 0, "admin": 0, "deactivated": 0, + "erased": false, "shadow_banned": 0, "creation_ts": 1560432506, "appservice_id": null, @@ -167,6 +168,7 @@ A response body like the following is returned: "admin": 0, "user_type": null, "deactivated": 0, + "erased": false, "shadow_banned": 0, "displayname": "", "avatar_url": null, @@ -177,6 +179,7 @@ A response body like the following is returned: "admin": 1, "user_type": null, "deactivated": 0, + "erased": false, "shadow_banned": 0, "displayname": "", "avatar_url": "", @@ -247,6 +250,7 @@ The following fields are returned in the JSON response body: - `user_type` - string - Type of the user. Normal users are type `None`. This allows user type specific behaviour. There are also types `support` and `bot`. - `deactivated` - bool - Status if that user has been marked as deactivated. + - `erased` - bool - Status if that user has been marked as erased. - `shadow_banned` - bool - Status if that user has been marked as shadow banned. - `displayname` - string - The user's display name if they have set one. - `avatar_url` - string - The user's avatar URL if they have set one. diff --git a/synapse/handlers/admin.py b/synapse/handlers/admin.py index f2989cc4a2..5bf8e86387 100644 --- a/synapse/handlers/admin.py +++ b/synapse/handlers/admin.py @@ -100,6 +100,7 @@ class AdminHandler: user_info_dict["avatar_url"] = profile.avatar_url user_info_dict["threepids"] = threepids user_info_dict["external_ids"] = external_ids + user_info_dict["erased"] = await self.store.is_user_erased(user.to_string()) return user_info_dict diff --git a/synapse/storage/databases/main/__init__.py b/synapse/storage/databases/main/__init__.py index a62b4abd4e..cfaedf5e0c 100644 --- a/synapse/storage/databases/main/__init__.py +++ b/synapse/storage/databases/main/__init__.py @@ -201,7 +201,7 @@ class DataStore( name: Optional[str] = None, guests: bool = True, deactivated: bool = False, - order_by: str = UserSortOrder.USER_ID.value, + order_by: str = UserSortOrder.NAME.value, direction: str = "f", approved: bool = True, ) -> Tuple[List[JsonDict], int]: @@ -261,6 +261,7 @@ class DataStore( sql_base = f""" FROM users as u LEFT JOIN profiles AS p ON u.name = '@' || p.user_id || ':' || ? + LEFT JOIN erased_users AS eu ON u.name = eu.user_id {where_clause} """ sql = "SELECT COUNT(*) as total_users " + sql_base @@ -269,7 +270,8 @@ class DataStore( sql = f""" SELECT name, user_type, is_guest, admin, deactivated, shadow_banned, - displayname, avatar_url, creation_ts * 1000 as creation_ts, approved + displayname, avatar_url, creation_ts * 1000 as creation_ts, approved, + eu.user_id is not null as erased {sql_base} ORDER BY {order_by_column} {order}, u.name ASC LIMIT ? OFFSET ? @@ -277,6 +279,13 @@ class DataStore( args += [limit, start] txn.execute(sql, args) users = self.db_pool.cursor_to_dict(txn) + + # some of those boolean values are returned as integers when we're on SQLite + columns_to_boolify = ["erased"] + for user in users: + for column in columns_to_boolify: + user[column] = bool(user[column]) + return users, count return await self.db_pool.runInteraction( diff --git a/tests/rest/admin/test_user.py b/tests/rest/admin/test_user.py index 4c1ce33463..63410ffdf1 100644 --- a/tests/rest/admin/test_user.py +++ b/tests/rest/admin/test_user.py @@ -31,7 +31,7 @@ from synapse.api.room_versions import RoomVersions from synapse.rest.client import devices, login, logout, profile, register, room, sync from synapse.rest.media.v1.filepath import MediaFilePaths from synapse.server import HomeServer -from synapse.types import JsonDict, UserID +from synapse.types import JsonDict, UserID, create_requester from synapse.util import Clock from tests import unittest @@ -924,6 +924,36 @@ class UsersListTestCase(unittest.HomeserverTestCase): self.assertEqual(1, len(non_admin_user_ids), non_admin_user_ids) self.assertEqual(not_approved_user, non_admin_user_ids[0]) + def test_erasure_status(self) -> None: + # Create a new user. + user_id = self.register_user("eraseme", "eraseme") + + # They should appear in the list users API, marked as not erased. + channel = self.make_request( + "GET", + self.url + "?deactivated=true", + access_token=self.admin_user_tok, + ) + users = {user["name"]: user for user in channel.json_body["users"]} + self.assertIs(users[user_id]["erased"], False) + + # Deactivate that user, requesting erasure. + deactivate_account_handler = self.hs.get_deactivate_account_handler() + self.get_success( + deactivate_account_handler.deactivate_account( + user_id, erase_data=True, requester=create_requester(user_id) + ) + ) + + # Repeat the list users query. They should now be marked as erased. + channel = self.make_request( + "GET", + self.url + "?deactivated=true", + access_token=self.admin_user_tok, + ) + users = {user["name"]: user for user in channel.json_body["users"]} + self.assertIs(users[user_id]["erased"], True) + def _order_test( self, expected_user_list: List[str], @@ -1195,6 +1225,7 @@ class DeactivateAccountTestCase(unittest.HomeserverTestCase): self.assertEqual("foo@bar.com", channel.json_body["threepids"][0]["address"]) self.assertEqual("mxc://servername/mediaid", channel.json_body["avatar_url"]) self.assertEqual("User1", channel.json_body["displayname"]) + self.assertFalse(channel.json_body["erased"]) # Deactivate and erase user channel = self.make_request( @@ -1219,6 +1250,7 @@ class DeactivateAccountTestCase(unittest.HomeserverTestCase): self.assertEqual(0, len(channel.json_body["threepids"])) self.assertIsNone(channel.json_body["avatar_url"]) self.assertIsNone(channel.json_body["displayname"]) + self.assertTrue(channel.json_body["erased"]) self._is_erased("@user:test", True) @@ -2757,6 +2789,7 @@ class UserRestTestCase(unittest.HomeserverTestCase): self.assertIn("avatar_url", content) self.assertIn("admin", content) self.assertIn("deactivated", content) + self.assertIn("erased", content) self.assertIn("shadow_banned", content) self.assertIn("creation_ts", content) self.assertIn("appservice_id", content) -- cgit 1.5.1 From b7a7ff6ee39da4981dcfdce61bf8ac4735e3d047 Mon Sep 17 00:00:00 2001 From: Shay Date: Fri, 21 Oct 2022 10:46:22 -0700 Subject: Add initial power level event to batch of bulk persisted events when creating a new room. (#14228) --- changelog.d/14228.misc | 1 + synapse/handlers/federation.py | 4 +- synapse/handlers/federation_event.py | 4 +- synapse/handlers/message.py | 14 ++---- synapse/handlers/room.py | 39 ++++----------- synapse/push/bulk_push_rule_evaluator.py | 74 ++++++++++++++++++++++++----- tests/push/test_bulk_push_rule_evaluator.py | 2 +- tests/replication/_base.py | 2 +- 8 files changed, 82 insertions(+), 58 deletions(-) create mode 100644 changelog.d/14228.misc (limited to 'synapse/handlers') diff --git a/changelog.d/14228.misc b/changelog.d/14228.misc new file mode 100644 index 0000000000..14fe31a8bc --- /dev/null +++ b/changelog.d/14228.misc @@ -0,0 +1 @@ +Add initial power level event to batch of bulk persisted events when creating a new room. diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index 275a37a575..4fbc79a6cb 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -1017,7 +1017,9 @@ class FederationHandler: context = EventContext.for_outlier(self._storage_controllers) - await self._bulk_push_rule_evaluator.action_for_event_by_user(event, context) + await self._bulk_push_rule_evaluator.action_for_events_by_user( + [(event, context)] + ) try: await self._federation_event_handler.persist_events_and_notify( event.room_id, [(event, context)] diff --git a/synapse/handlers/federation_event.py b/synapse/handlers/federation_event.py index 06e41b5cc0..7da6316a82 100644 --- a/synapse/handlers/federation_event.py +++ b/synapse/handlers/federation_event.py @@ -2171,8 +2171,8 @@ class FederationEventHandler: min_depth, ) else: - await self._bulk_push_rule_evaluator.action_for_event_by_user( - event, context + await self._bulk_push_rule_evaluator.action_for_events_by_user( + [(event, context)] ) try: diff --git a/synapse/handlers/message.py b/synapse/handlers/message.py index 15b828dd74..468900a07f 100644 --- a/synapse/handlers/message.py +++ b/synapse/handlers/message.py @@ -1433,17 +1433,9 @@ class EventCreationHandler: a room that has been un-partial stated. """ - for event, context in events_and_context: - # Skip push notification actions for historical messages - # because we don't want to notify people about old history back in time. - # The historical messages also do not have the proper `context.current_state_ids` - # and `state_groups` because they have `prev_events` that aren't persisted yet - # (historical messages persisted in reverse-chronological order). - if not event.internal_metadata.is_historical(): - with opentracing.start_active_span("calculate_push_actions"): - await self._bulk_push_rule_evaluator.action_for_event_by_user( - event, context - ) + await self._bulk_push_rule_evaluator.action_for_events_by_user( + events_and_context + ) try: # If we're a worker we need to hit out to the master. diff --git a/synapse/handlers/room.py b/synapse/handlers/room.py index 638f54051a..cc1e5c8f97 100644 --- a/synapse/handlers/room.py +++ b/synapse/handlers/room.py @@ -1055,9 +1055,6 @@ class RoomCreationHandler: event_keys = {"room_id": room_id, "sender": creator_id, "state_key": ""} depth = 1 - # the last event sent/persisted to the db - last_sent_event_id: Optional[str] = None - # the most recently created event prev_event: List[str] = [] # a map of event types, state keys -> event_ids. We collect these mappings this as events are @@ -1102,26 +1099,6 @@ class RoomCreationHandler: return new_event, new_context - async def send( - event: EventBase, - context: synapse.events.snapshot.EventContext, - creator: Requester, - ) -> int: - nonlocal last_sent_event_id - - ev = await self.event_creation_handler.handle_new_client_event( - requester=creator, - events_and_context=[(event, context)], - ratelimit=False, - ignore_shadow_ban=True, - ) - - last_sent_event_id = ev.event_id - - # we know it was persisted, so must have a stream ordering - assert ev.internal_metadata.stream_ordering - return ev.internal_metadata.stream_ordering - try: config = self._presets_dict[preset_config] except KeyError: @@ -1135,10 +1112,14 @@ class RoomCreationHandler: ) logger.debug("Sending %s in new room", EventTypes.Member) - await send(creation_event, creation_context, creator) + ev = await self.event_creation_handler.handle_new_client_event( + requester=creator, + events_and_context=[(creation_event, creation_context)], + ratelimit=False, + ignore_shadow_ban=True, + ) + last_sent_event_id = ev.event_id - # Room create event must exist at this point - assert last_sent_event_id is not None member_event_id, _ = await self.room_member_handler.update_membership( creator, creator.user, @@ -1157,6 +1138,7 @@ class RoomCreationHandler: depth += 1 state_map[(EventTypes.Member, creator.user.to_string())] = member_event_id + events_to_send = [] # We treat the power levels override specially as this needs to be one # of the first events that get sent into a room. pl_content = initial_state.pop((EventTypes.PowerLevels, ""), None) @@ -1165,7 +1147,7 @@ class RoomCreationHandler: EventTypes.PowerLevels, pl_content, False ) current_state_group = power_context._state_group - await send(power_event, power_context, creator) + events_to_send.append((power_event, power_context)) else: power_level_content: JsonDict = { "users": {creator_id: 100}, @@ -1214,9 +1196,8 @@ class RoomCreationHandler: False, ) current_state_group = pl_context._state_group - await send(pl_event, pl_context, creator) + events_to_send.append((pl_event, pl_context)) - events_to_send = [] if room_alias and (EventTypes.CanonicalAlias, "") not in initial_state: room_alias_event, room_alias_context = await create_event( EventTypes.CanonicalAlias, {"alias": room_alias.to_string()}, True diff --git a/synapse/push/bulk_push_rule_evaluator.py b/synapse/push/bulk_push_rule_evaluator.py index a75386f6a0..d7795a9080 100644 --- a/synapse/push/bulk_push_rule_evaluator.py +++ b/synapse/push/bulk_push_rule_evaluator.py @@ -165,8 +165,21 @@ class BulkPushRuleEvaluator: return rules_by_user async def _get_power_levels_and_sender_level( - self, event: EventBase, context: EventContext + self, + event: EventBase, + context: EventContext, + event_id_to_event: Mapping[str, EventBase], ) -> Tuple[dict, Optional[int]]: + """ + Given an event and an event context, get the power level event relevant to the event + and the power level of the sender of the event. + Args: + event: event to check + context: context of event to check + event_id_to_event: a mapping of event_id to event for a set of events being + batch persisted. This is needed as the sought-after power level event may + be in this batch rather than the DB + """ # There are no power levels and sender levels possible to get from outlier if event.internal_metadata.is_outlier(): return {}, None @@ -177,15 +190,26 @@ class BulkPushRuleEvaluator: ) pl_event_id = prev_state_ids.get(POWER_KEY) + # fastpath: if there's a power level event, that's all we need, and + # not having a power level event is an extreme edge case if pl_event_id: - # fastpath: if there's a power level event, that's all we need, and - # not having a power level event is an extreme edge case - auth_events = {POWER_KEY: await self.store.get_event(pl_event_id)} + # Get the power level event from the batch, or fall back to the database. + pl_event = event_id_to_event.get(pl_event_id) + if pl_event: + auth_events = {POWER_KEY: pl_event} + else: + auth_events = {POWER_KEY: await self.store.get_event(pl_event_id)} else: auth_events_ids = self._event_auth_handler.compute_auth_events( event, prev_state_ids, for_verification=False ) auth_events_dict = await self.store.get_events(auth_events_ids) + # Some needed auth events might be in the batch, combine them with those + # fetched from the database. + for auth_event_id in auth_events_ids: + auth_event = event_id_to_event.get(auth_event_id) + if auth_event: + auth_events_dict[auth_event_id] = auth_event auth_events = {(e.type, e.state_key): e for e in auth_events_dict.values()} sender_level = get_user_power_level(event.sender, auth_events) @@ -194,16 +218,38 @@ class BulkPushRuleEvaluator: return pl_event.content if pl_event else {}, sender_level - @measure_func("action_for_event_by_user") - async def action_for_event_by_user( - self, event: EventBase, context: EventContext + async def action_for_events_by_user( + self, events_and_context: List[Tuple[EventBase, EventContext]] ) -> None: - """Given an event and context, evaluate the push rules, check if the message - should increment the unread count, and insert the results into the - event_push_actions_staging table. + """Given a list of events and their associated contexts, evaluate the push rules + for each event, check if the message should increment the unread count, and + insert the results into the event_push_actions_staging table. """ - if not event.internal_metadata.is_notifiable(): - # Push rules for events that aren't notifiable can't be processed by this + # For batched events the power level events may not have been persisted yet, + # so we pass in the batched events. Thus if the event cannot be found in the + # database we can check in the batch. + event_id_to_event = {e.event_id: e for e, _ in events_and_context} + for event, context in events_and_context: + await self._action_for_event_by_user(event, context, event_id_to_event) + + @measure_func("action_for_event_by_user") + async def _action_for_event_by_user( + self, + event: EventBase, + context: EventContext, + event_id_to_event: Mapping[str, EventBase], + ) -> None: + + if ( + not event.internal_metadata.is_notifiable() + or event.internal_metadata.is_historical() + ): + # Push rules for events that aren't notifiable can't be processed by this and + # we want to skip push notification actions for historical messages + # because we don't want to notify people about old history back in time. + # The historical messages also do not have the proper `context.current_state_ids` + # and `state_groups` because they have `prev_events` that aren't persisted yet + # (historical messages persisted in reverse-chronological order). return # Disable counting as unread unless the experimental configuration is @@ -223,7 +269,9 @@ class BulkPushRuleEvaluator: ( power_levels, sender_power_level, - ) = await self._get_power_levels_and_sender_level(event, context) + ) = await self._get_power_levels_and_sender_level( + event, context, event_id_to_event + ) # Find the event's thread ID. relation = relation_from_event(event) diff --git a/tests/push/test_bulk_push_rule_evaluator.py b/tests/push/test_bulk_push_rule_evaluator.py index 675d7df2ac..594e7937a8 100644 --- a/tests/push/test_bulk_push_rule_evaluator.py +++ b/tests/push/test_bulk_push_rule_evaluator.py @@ -71,4 +71,4 @@ class TestBulkPushRuleEvaluator(unittest.HomeserverTestCase): bulk_evaluator = BulkPushRuleEvaluator(self.hs) # should not raise - self.get_success(bulk_evaluator.action_for_event_by_user(event, context)) + self.get_success(bulk_evaluator.action_for_events_by_user([(event, context)])) diff --git a/tests/replication/_base.py b/tests/replication/_base.py index ce53f808db..121f3d8d65 100644 --- a/tests/replication/_base.py +++ b/tests/replication/_base.py @@ -371,7 +371,7 @@ class BaseMultiWorkerStreamTestCase(unittest.HomeserverTestCase): config=worker_hs.config.server.listeners[0], resource=resource, server_version_string="1", - max_request_body_size=4096, + max_request_body_size=8192, reactor=self.reactor, ) -- cgit 1.5.1 From 9192d74b0bf2f87b00d3e106a18baa9ce27acda1 Mon Sep 17 00:00:00 2001 From: Quentin Gliech Date: Tue, 25 Oct 2022 16:25:02 +0200 Subject: Refactor OIDC tests to better mimic an actual OIDC provider. (#13910) This implements a fake OIDC server, which intercepts calls to the HTTP client. Improves accuracy of tests by covering more internal methods. One particular example was the ID token validation, which previously mocked. This uncovered an incorrect dependency: Synapse actually requires at least authlib 0.15.1, not 0.14.0. --- changelog.d/13910.misc | 1 + pyproject.toml | 2 +- synapse/handlers/oidc.py | 15 +- tests/federation/test_federation_client.py | 36 +- tests/handlers/test_oidc.py | 580 +++++++++++++---------------- tests/rest/client/test_auth.py | 32 +- tests/rest/client/test_login.py | 40 +- tests/rest/client/utils.py | 136 +++---- tests/test_utils/__init__.py | 40 +- tests/test_utils/oidc.py | 325 ++++++++++++++++ 10 files changed, 747 insertions(+), 460 deletions(-) create mode 100644 changelog.d/13910.misc create mode 100644 tests/test_utils/oidc.py (limited to 'synapse/handlers') diff --git a/changelog.d/13910.misc b/changelog.d/13910.misc new file mode 100644 index 0000000000..e906952aab --- /dev/null +++ b/changelog.d/13910.misc @@ -0,0 +1 @@ +Refactor OIDC tests to better mimic an actual OIDC provider. diff --git a/pyproject.toml b/pyproject.toml index 6ebac41ed1..7e0feb75aa 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -192,7 +192,7 @@ psycopg2 = { version = ">=2.8", markers = "platform_python_implementation != 'Py psycopg2cffi = { version = ">=2.8", markers = "platform_python_implementation == 'PyPy'", optional = true } psycopg2cffi-compat = { version = "==1.1", markers = "platform_python_implementation == 'PyPy'", optional = true } pysaml2 = { version = ">=4.5.0", optional = true } -authlib = { version = ">=0.14.0", optional = true } +authlib = { version = ">=0.15.1", optional = true } # systemd-python is necessary for logging to the systemd journal via # `systemd.journal.JournalHandler`, as is documented in # `contrib/systemd/log_config.yaml`. diff --git a/synapse/handlers/oidc.py b/synapse/handlers/oidc.py index d7a8226900..9759daf043 100644 --- a/synapse/handlers/oidc.py +++ b/synapse/handlers/oidc.py @@ -275,6 +275,7 @@ class OidcProvider: provider: OidcProviderConfig, ): self._store = hs.get_datastores().main + self._clock = hs.get_clock() self._macaroon_generaton = macaroon_generator @@ -673,6 +674,13 @@ class OidcProvider: Returns: The decoded claims in the ID token. """ + id_token = token.get("id_token") + logger.debug("Attempting to decode JWT id_token %r", id_token) + + # That has been theoritically been checked by the caller, so even though + # assertion are not enabled in production, it is mainly here to appease mypy + assert id_token is not None + metadata = await self.load_metadata() claims_params = { "nonce": nonce, @@ -688,9 +696,6 @@ class OidcProvider: claim_options = {"iss": {"values": [metadata["issuer"]]}} - id_token = token["id_token"] - logger.debug("Attempting to decode JWT id_token %r", id_token) - # Try to decode the keys in cache first, then retry by forcing the keys # to be reloaded jwk_set = await self.load_jwks() @@ -715,7 +720,9 @@ class OidcProvider: logger.debug("Decoded id_token JWT %r; validating", claims) - claims.validate(leeway=120) # allows 2 min of clock skew + claims.validate( + now=self._clock.time(), leeway=120 + ) # allows 2 min of clock skew return claims diff --git a/tests/federation/test_federation_client.py b/tests/federation/test_federation_client.py index a538215931..51d3bb8fff 100644 --- a/tests/federation/test_federation_client.py +++ b/tests/federation/test_federation_client.py @@ -12,13 +12,10 @@ # See the License for the specific language governing permissions and # limitations under the License. -import json from unittest import mock import twisted.web.client from twisted.internet import defer -from twisted.internet.protocol import Protocol -from twisted.python.failure import Failure from twisted.test.proto_helpers import MemoryReactor from synapse.api.room_versions import RoomVersions @@ -26,10 +23,9 @@ from synapse.events import EventBase from synapse.rest import admin from synapse.rest.client import login, room from synapse.server import HomeServer -from synapse.types import JsonDict from synapse.util import Clock -from tests.test_utils import event_injection +from tests.test_utils import FakeResponse, event_injection from tests.unittest import FederatingHomeserverTestCase @@ -98,8 +94,8 @@ class FederationClientTest(FederatingHomeserverTestCase): # mock up the response, and have the agent return it self._mock_agent.request.side_effect = lambda *args, **kwargs: defer.succeed( - _mock_response( - { + FakeResponse.json( + payload={ "pdus": [ create_event_dict, member_event_dict, @@ -208,8 +204,8 @@ class FederationClientTest(FederatingHomeserverTestCase): # mock up the response, and have the agent return it self._mock_agent.request.side_effect = lambda *args, **kwargs: defer.succeed( - _mock_response( - { + FakeResponse.json( + payload={ "origin": "yet.another.server", "origin_server_ts": 900, "pdus": [ @@ -269,8 +265,8 @@ class FederationClientTest(FederatingHomeserverTestCase): # We expect an outbound request to /backfill, so stub that out self._mock_agent.request.side_effect = lambda *args, **kwargs: defer.succeed( - _mock_response( - { + FakeResponse.json( + payload={ "origin": "yet.another.server", "origin_server_ts": 900, # Mimic the other server returning our new `pulled_event` @@ -305,21 +301,3 @@ class FederationClientTest(FederatingHomeserverTestCase): # This is 2 because it failed once from `self.OTHER_SERVER_NAME` and the # other from "yet.another.server" self.assertEqual(backfill_num_attempts, 2) - - -def _mock_response(resp: JsonDict): - body = json.dumps(resp).encode("utf-8") - - def deliver_body(p: Protocol): - p.dataReceived(body) - p.connectionLost(Failure(twisted.web.client.ResponseDone())) - - response = mock.Mock( - code=200, - phrase=b"OK", - headers=twisted.web.client.Headers({"content-Type": ["application/json"]}), - length=len(body), - deliverBody=deliver_body, - ) - mock.seal(response) - return response diff --git a/tests/handlers/test_oidc.py b/tests/handlers/test_oidc.py index e6cd3af7b7..5955410524 100644 --- a/tests/handlers/test_oidc.py +++ b/tests/handlers/test_oidc.py @@ -11,9 +11,8 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. -import json import os -from typing import Any, Dict +from typing import Any, Dict, Tuple from unittest.mock import ANY, Mock, patch from urllib.parse import parse_qs, urlparse @@ -22,12 +21,15 @@ import pymacaroons from twisted.test.proto_helpers import MemoryReactor from synapse.handlers.sso import MappingException +from synapse.http.site import SynapseRequest from synapse.server import HomeServer -from synapse.types import JsonDict, UserID +from synapse.types import UserID from synapse.util import Clock -from synapse.util.macaroons import OidcSessionData, get_value_from_macaroon +from synapse.util.macaroons import get_value_from_macaroon +from synapse.util.stringutils import random_string from tests.test_utils import FakeResponse, get_awaitable_result, simple_async_mock +from tests.test_utils.oidc import FakeAuthorizationGrant, FakeOidcServer from tests.unittest import HomeserverTestCase, override_config try: @@ -46,12 +48,6 @@ BASE_URL = "https://synapse/" CALLBACK_URL = BASE_URL + "_synapse/client/oidc/callback" SCOPES = ["openid"] -AUTHORIZATION_ENDPOINT = ISSUER + "authorize" -TOKEN_ENDPOINT = ISSUER + "token" -USERINFO_ENDPOINT = ISSUER + "userinfo" -WELL_KNOWN = ISSUER + ".well-known/openid-configuration" -JWKS_URI = ISSUER + ".well-known/jwks.json" - # config for common cases DEFAULT_CONFIG = { "enabled": True, @@ -66,9 +62,9 @@ DEFAULT_CONFIG = { EXPLICIT_ENDPOINT_CONFIG = { **DEFAULT_CONFIG, "discover": False, - "authorization_endpoint": AUTHORIZATION_ENDPOINT, - "token_endpoint": TOKEN_ENDPOINT, - "jwks_uri": JWKS_URI, + "authorization_endpoint": ISSUER + "authorize", + "token_endpoint": ISSUER + "token", + "jwks_uri": ISSUER + "jwks", } @@ -102,27 +98,6 @@ class TestMappingProviderFailures(TestMappingProvider): } -async def get_json(url: str) -> JsonDict: - # Mock get_json calls to handle jwks & oidc discovery endpoints - if url == WELL_KNOWN: - # Minimal discovery document, as defined in OpenID.Discovery - # https://openid.net/specs/openid-connect-discovery-1_0.html#ProviderMetadata - return { - "issuer": ISSUER, - "authorization_endpoint": AUTHORIZATION_ENDPOINT, - "token_endpoint": TOKEN_ENDPOINT, - "jwks_uri": JWKS_URI, - "userinfo_endpoint": USERINFO_ENDPOINT, - "response_types_supported": ["code"], - "subject_types_supported": ["public"], - "id_token_signing_alg_values_supported": ["RS256"], - } - elif url == JWKS_URI: - return {"keys": []} - - return {} - - def _key_file_path() -> str: """path to a file containing the private half of a test key""" @@ -159,11 +134,11 @@ class OidcHandlerTestCase(HomeserverTestCase): return config def make_homeserver(self, reactor: MemoryReactor, clock: Clock) -> HomeServer: - self.http_client = Mock(spec=["get_json"]) - self.http_client.get_json.side_effect = get_json - self.http_client.user_agent = b"Synapse Test" + self.fake_server = FakeOidcServer(clock=clock, issuer=ISSUER) - hs = self.setup_test_homeserver(proxied_http_client=self.http_client) + hs = self.setup_test_homeserver() + self.hs_patcher = self.fake_server.patch_homeserver(hs=hs) + self.hs_patcher.start() self.handler = hs.get_oidc_handler() self.provider = self.handler._providers["oidc"] @@ -175,18 +150,51 @@ class OidcHandlerTestCase(HomeserverTestCase): # Reduce the number of attempts when generating MXIDs. sso_handler._MAP_USERNAME_RETRIES = 3 + auth_handler = hs.get_auth_handler() + # Mock the complete SSO login method. + self.complete_sso_login = simple_async_mock() + auth_handler.complete_sso_login = self.complete_sso_login # type: ignore[assignment] + return hs + def tearDown(self) -> None: + self.hs_patcher.stop() + return super().tearDown() + + def reset_mocks(self): + """Reset all the Mocks.""" + self.fake_server.reset_mocks() + self.render_error.reset_mock() + self.complete_sso_login.reset_mock() + def metadata_edit(self, values): """Modify the result that will be returned by the well-known query""" - async def patched_get_json(uri): - res = await get_json(uri) - if uri == WELL_KNOWN: - res.update(values) - return res + metadata = self.fake_server.get_metadata() + metadata.update(values) + return patch.object(self.fake_server, "get_metadata", return_value=metadata) - return patch.object(self.http_client, "get_json", patched_get_json) + def start_authorization( + self, + userinfo: dict, + client_redirect_url: str = "http://client/redirect", + scope: str = "openid", + with_sid: bool = False, + ) -> Tuple[SynapseRequest, FakeAuthorizationGrant]: + """Start an authorization request, and get the callback request back.""" + nonce = random_string(10) + state = random_string(10) + + code, grant = self.fake_server.start_authorization( + userinfo=userinfo, + scope=scope, + client_id=self.provider._client_auth.client_id, + redirect_uri=self.provider._callback_url, + nonce=nonce, + with_sid=with_sid, + ) + session = self._generate_oidc_session_token(state, nonce, client_redirect_url) + return _build_callback_request(code, state, session), grant def assertRenderedError(self, error, error_description=None): self.render_error.assert_called_once() @@ -210,52 +218,54 @@ class OidcHandlerTestCase(HomeserverTestCase): """The handler should discover the endpoints from OIDC discovery document.""" # This would throw if some metadata were invalid metadata = self.get_success(self.provider.load_metadata()) - self.http_client.get_json.assert_called_once_with(WELL_KNOWN) + self.fake_server.get_metadata_handler.assert_called_once() - self.assertEqual(metadata.issuer, ISSUER) - self.assertEqual(metadata.authorization_endpoint, AUTHORIZATION_ENDPOINT) - self.assertEqual(metadata.token_endpoint, TOKEN_ENDPOINT) - self.assertEqual(metadata.jwks_uri, JWKS_URI) - # FIXME: it seems like authlib does not have that defined in its metadata models - # self.assertEqual(metadata.userinfo_endpoint, USERINFO_ENDPOINT) + self.assertEqual(metadata.issuer, self.fake_server.issuer) + self.assertEqual( + metadata.authorization_endpoint, + self.fake_server.authorization_endpoint, + ) + self.assertEqual(metadata.token_endpoint, self.fake_server.token_endpoint) + self.assertEqual(metadata.jwks_uri, self.fake_server.jwks_uri) + # It seems like authlib does not have that defined in its metadata models + self.assertEqual( + metadata.get("userinfo_endpoint"), + self.fake_server.userinfo_endpoint, + ) # subsequent calls should be cached - self.http_client.reset_mock() + self.reset_mocks() self.get_success(self.provider.load_metadata()) - self.http_client.get_json.assert_not_called() + self.fake_server.get_metadata_handler.assert_not_called() @override_config({"oidc_config": EXPLICIT_ENDPOINT_CONFIG}) def test_no_discovery(self) -> None: """When discovery is disabled, it should not try to load from discovery document.""" self.get_success(self.provider.load_metadata()) - self.http_client.get_json.assert_not_called() + self.fake_server.get_metadata_handler.assert_not_called() - @override_config({"oidc_config": EXPLICIT_ENDPOINT_CONFIG}) + @override_config({"oidc_config": DEFAULT_CONFIG}) def test_load_jwks(self) -> None: """JWKS loading is done once (then cached) if used.""" jwks = self.get_success(self.provider.load_jwks()) - self.http_client.get_json.assert_called_once_with(JWKS_URI) - self.assertEqual(jwks, {"keys": []}) + self.fake_server.get_jwks_handler.assert_called_once() + self.assertEqual(jwks, self.fake_server.get_jwks()) # subsequent calls should be cached… - self.http_client.reset_mock() + self.reset_mocks() self.get_success(self.provider.load_jwks()) - self.http_client.get_json.assert_not_called() + self.fake_server.get_jwks_handler.assert_not_called() # …unless forced - self.http_client.reset_mock() + self.reset_mocks() self.get_success(self.provider.load_jwks(force=True)) - self.http_client.get_json.assert_called_once_with(JWKS_URI) + self.fake_server.get_jwks_handler.assert_called_once() - # Throw if the JWKS uri is missing - original = self.provider.load_metadata - - async def patched_load_metadata(): - m = (await original()).copy() - m.update({"jwks_uri": None}) - return m - - with patch.object(self.provider, "load_metadata", patched_load_metadata): + with self.metadata_edit({"jwks_uri": None}): + # If we don't do this, the load_metadata call will throw because of the + # missing jwks_uri + self.provider._user_profile_method = "userinfo_endpoint" + self.get_success(self.provider.load_metadata(force=True)) self.get_failure(self.provider.load_jwks(force=True), RuntimeError) @override_config({"oidc_config": DEFAULT_CONFIG}) @@ -359,7 +369,7 @@ class OidcHandlerTestCase(HomeserverTestCase): self.provider.handle_redirect_request(req, b"http://client/redirect") ) ) - auth_endpoint = urlparse(AUTHORIZATION_ENDPOINT) + auth_endpoint = urlparse(self.fake_server.authorization_endpoint) self.assertEqual(url.scheme, auth_endpoint.scheme) self.assertEqual(url.netloc, auth_endpoint.netloc) @@ -424,48 +434,34 @@ class OidcHandlerTestCase(HomeserverTestCase): with self.assertRaises(AttributeError): _ = mapping_provider.get_extra_attributes - token = { - "type": "bearer", - "id_token": "id_token", - "access_token": "access_token", - } username = "bar" userinfo = { "sub": "foo", "username": username, } expected_user_id = "@%s:%s" % (username, self.hs.hostname) - self.provider._exchange_code = simple_async_mock(return_value=token) # type: ignore[assignment] - self.provider._parse_id_token = simple_async_mock(return_value=userinfo) # type: ignore[assignment] - self.provider._fetch_userinfo = simple_async_mock(return_value=userinfo) # type: ignore[assignment] - auth_handler = self.hs.get_auth_handler() - auth_handler.complete_sso_login = simple_async_mock() - code = "code" - state = "state" - nonce = "nonce" client_redirect_url = "http://client/redirect" - ip_address = "10.0.0.1" - session = self._generate_oidc_session_token(state, nonce, client_redirect_url) - request = _build_callback_request(code, state, session, ip_address=ip_address) - + request, _ = self.start_authorization( + userinfo, client_redirect_url=client_redirect_url + ) self.get_success(self.handler.handle_oidc_callback(request)) - auth_handler.complete_sso_login.assert_called_once_with( + self.complete_sso_login.assert_called_once_with( expected_user_id, - "oidc", + self.provider.idp_id, request, client_redirect_url, None, new_user=True, auth_provider_session_id=None, ) - self.provider._exchange_code.assert_called_once_with(code) - self.provider._parse_id_token.assert_called_once_with(token, nonce=nonce) - self.provider._fetch_userinfo.assert_not_called() + self.fake_server.post_token_handler.assert_called_once() + self.fake_server.get_userinfo_handler.assert_not_called() self.render_error.assert_not_called() # Handle mapping errors + request, _ = self.start_authorization(userinfo) with patch.object( self.provider, "_remote_id_from_userinfo", @@ -475,81 +471,63 @@ class OidcHandlerTestCase(HomeserverTestCase): self.assertRenderedError("mapping_error") # Handle ID token errors - self.provider._parse_id_token = simple_async_mock(raises=Exception()) # type: ignore[assignment] - self.get_success(self.handler.handle_oidc_callback(request)) + request, _ = self.start_authorization(userinfo) + with self.fake_server.id_token_override({"iss": "https://bad.issuer/"}): + self.get_success(self.handler.handle_oidc_callback(request)) self.assertRenderedError("invalid_token") - auth_handler.complete_sso_login.reset_mock() - self.provider._exchange_code.reset_mock() - self.provider._parse_id_token.reset_mock() - self.provider._fetch_userinfo.reset_mock() + self.reset_mocks() # With userinfo fetching self.provider._user_profile_method = "userinfo_endpoint" - token = { - "type": "bearer", - "access_token": "access_token", - } - self.provider._exchange_code = simple_async_mock(return_value=token) # type: ignore[assignment] + # Without the "openid" scope, the FakeProvider does not generate an id_token + request, _ = self.start_authorization(userinfo, scope="") self.get_success(self.handler.handle_oidc_callback(request)) - auth_handler.complete_sso_login.assert_called_once_with( + self.complete_sso_login.assert_called_once_with( expected_user_id, - "oidc", + self.provider.idp_id, request, - client_redirect_url, + ANY, None, new_user=False, auth_provider_session_id=None, ) - self.provider._exchange_code.assert_called_once_with(code) - self.provider._parse_id_token.assert_not_called() - self.provider._fetch_userinfo.assert_called_once_with(token) + self.fake_server.post_token_handler.assert_called_once() + self.fake_server.get_userinfo_handler.assert_called_once() self.render_error.assert_not_called() + self.reset_mocks() + # With an ID token, userinfo fetching and sid in the ID token self.provider._user_profile_method = "userinfo_endpoint" - token = { - "type": "bearer", - "access_token": "access_token", - "id_token": "id_token", - } - id_token = { - "sid": "abcdefgh", - } - self.provider._parse_id_token = simple_async_mock(return_value=id_token) # type: ignore[assignment] - self.provider._exchange_code = simple_async_mock(return_value=token) # type: ignore[assignment] - auth_handler.complete_sso_login.reset_mock() - self.provider._fetch_userinfo.reset_mock() + request, grant = self.start_authorization(userinfo, with_sid=True) + self.assertIsNotNone(grant.sid) self.get_success(self.handler.handle_oidc_callback(request)) - auth_handler.complete_sso_login.assert_called_once_with( + self.complete_sso_login.assert_called_once_with( expected_user_id, - "oidc", + self.provider.idp_id, request, - client_redirect_url, + ANY, None, new_user=False, - auth_provider_session_id=id_token["sid"], + auth_provider_session_id=grant.sid, ) - self.provider._exchange_code.assert_called_once_with(code) - self.provider._parse_id_token.assert_called_once_with(token, nonce=nonce) - self.provider._fetch_userinfo.assert_called_once_with(token) + self.fake_server.post_token_handler.assert_called_once() + self.fake_server.get_userinfo_handler.assert_called_once() self.render_error.assert_not_called() # Handle userinfo fetching error - self.provider._fetch_userinfo = simple_async_mock(raises=Exception()) # type: ignore[assignment] - self.get_success(self.handler.handle_oidc_callback(request)) + request, _ = self.start_authorization(userinfo) + with self.fake_server.buggy_endpoint(userinfo=True): + self.get_success(self.handler.handle_oidc_callback(request)) self.assertRenderedError("fetch_error") - # Handle code exchange failure - from synapse.handlers.oidc import OidcError - - self.provider._exchange_code = simple_async_mock( # type: ignore[assignment] - raises=OidcError("invalid_request") - ) - self.get_success(self.handler.handle_oidc_callback(request)) - self.assertRenderedError("invalid_request") + request, _ = self.start_authorization(userinfo) + with self.fake_server.buggy_endpoint(token=True): + self.get_success(self.handler.handle_oidc_callback(request)) + self.assertRenderedError("server_error") @override_config({"oidc_config": DEFAULT_CONFIG}) def test_callback_session(self) -> None: @@ -599,18 +577,22 @@ class OidcHandlerTestCase(HomeserverTestCase): ) def test_exchange_code(self) -> None: """Code exchange behaves correctly and handles various error scenarios.""" - token = {"type": "bearer"} - token_json = json.dumps(token).encode("utf-8") - self.http_client.request = simple_async_mock( - return_value=FakeResponse(code=200, phrase=b"OK", body=token_json) + token = { + "type": "Bearer", + "access_token": "aabbcc", + } + + self.fake_server.post_token_handler.side_effect = None + self.fake_server.post_token_handler.return_value = FakeResponse.json( + payload=token ) code = "code" ret = self.get_success(self.provider._exchange_code(code)) - kwargs = self.http_client.request.call_args[1] + kwargs = self.fake_server.request.call_args[1] self.assertEqual(ret, token) self.assertEqual(kwargs["method"], "POST") - self.assertEqual(kwargs["uri"], TOKEN_ENDPOINT) + self.assertEqual(kwargs["uri"], self.fake_server.token_endpoint) args = parse_qs(kwargs["data"].decode("utf-8")) self.assertEqual(args["grant_type"], ["authorization_code"]) @@ -620,12 +602,8 @@ class OidcHandlerTestCase(HomeserverTestCase): self.assertEqual(args["redirect_uri"], [CALLBACK_URL]) # Test error handling - self.http_client.request = simple_async_mock( - return_value=FakeResponse( - code=400, - phrase=b"Bad Request", - body=b'{"error": "foo", "error_description": "bar"}', - ) + self.fake_server.post_token_handler.return_value = FakeResponse.json( + code=400, payload={"error": "foo", "error_description": "bar"} ) from synapse.handlers.oidc import OidcError @@ -634,46 +612,30 @@ class OidcHandlerTestCase(HomeserverTestCase): self.assertEqual(exc.value.error_description, "bar") # Internal server error with no JSON body - self.http_client.request = simple_async_mock( - return_value=FakeResponse( - code=500, - phrase=b"Internal Server Error", - body=b"Not JSON", - ) + self.fake_server.post_token_handler.return_value = FakeResponse( + code=500, body=b"Not JSON" ) exc = self.get_failure(self.provider._exchange_code(code), OidcError) self.assertEqual(exc.value.error, "server_error") # Internal server error with JSON body - self.http_client.request = simple_async_mock( - return_value=FakeResponse( - code=500, - phrase=b"Internal Server Error", - body=b'{"error": "internal_server_error"}', - ) + self.fake_server.post_token_handler.return_value = FakeResponse.json( + code=500, payload={"error": "internal_server_error"} ) exc = self.get_failure(self.provider._exchange_code(code), OidcError) self.assertEqual(exc.value.error, "internal_server_error") # 4xx error without "error" field - self.http_client.request = simple_async_mock( - return_value=FakeResponse( - code=400, - phrase=b"Bad request", - body=b"{}", - ) + self.fake_server.post_token_handler.return_value = FakeResponse.json( + code=400, payload={} ) exc = self.get_failure(self.provider._exchange_code(code), OidcError) self.assertEqual(exc.value.error, "server_error") # 2xx error with "error" field - self.http_client.request = simple_async_mock( - return_value=FakeResponse( - code=200, - phrase=b"OK", - body=b'{"error": "some_error"}', - ) + self.fake_server.post_token_handler.return_value = FakeResponse.json( + code=200, payload={"error": "some_error"} ) exc = self.get_failure(self.provider._exchange_code(code), OidcError) self.assertEqual(exc.value.error, "some_error") @@ -697,11 +659,14 @@ class OidcHandlerTestCase(HomeserverTestCase): """Test that code exchange works with a JWK client secret.""" from authlib.jose import jwt - token = {"type": "bearer"} - self.http_client.request = simple_async_mock( - return_value=FakeResponse( - code=200, phrase=b"OK", body=json.dumps(token).encode("utf-8") - ) + token = { + "type": "Bearer", + "access_token": "aabbcc", + } + + self.fake_server.post_token_handler.side_effect = None + self.fake_server.post_token_handler.return_value = FakeResponse.json( + payload=token ) code = "code" @@ -714,9 +679,9 @@ class OidcHandlerTestCase(HomeserverTestCase): self.assertEqual(ret, token) # the request should have hit the token endpoint - kwargs = self.http_client.request.call_args[1] + kwargs = self.fake_server.request.call_args[1] self.assertEqual(kwargs["method"], "POST") - self.assertEqual(kwargs["uri"], TOKEN_ENDPOINT) + self.assertEqual(kwargs["uri"], self.fake_server.token_endpoint) # the client secret provided to the should be a jwt which can be checked with # the public key @@ -750,11 +715,14 @@ class OidcHandlerTestCase(HomeserverTestCase): ) def test_exchange_code_no_auth(self) -> None: """Test that code exchange works with no client secret.""" - token = {"type": "bearer"} - self.http_client.request = simple_async_mock( - return_value=FakeResponse( - code=200, phrase=b"OK", body=json.dumps(token).encode("utf-8") - ) + token = { + "type": "Bearer", + "access_token": "aabbcc", + } + + self.fake_server.post_token_handler.side_effect = None + self.fake_server.post_token_handler.return_value = FakeResponse.json( + payload=token ) code = "code" ret = self.get_success(self.provider._exchange_code(code)) @@ -762,9 +730,9 @@ class OidcHandlerTestCase(HomeserverTestCase): self.assertEqual(ret, token) # the request should have hit the token endpoint - kwargs = self.http_client.request.call_args[1] + kwargs = self.fake_server.request.call_args[1] self.assertEqual(kwargs["method"], "POST") - self.assertEqual(kwargs["uri"], TOKEN_ENDPOINT) + self.assertEqual(kwargs["uri"], self.fake_server.token_endpoint) # check the POSTed data args = parse_qs(kwargs["data"].decode("utf-8")) @@ -787,37 +755,19 @@ class OidcHandlerTestCase(HomeserverTestCase): """ Login while using a mapping provider that implements get_extra_attributes. """ - token = { - "type": "bearer", - "id_token": "id_token", - "access_token": "access_token", - } userinfo = { "sub": "foo", "username": "foo", "phone": "1234567", } - self.provider._exchange_code = simple_async_mock(return_value=token) # type: ignore[assignment] - self.provider._parse_id_token = simple_async_mock(return_value=userinfo) # type: ignore[assignment] - auth_handler = self.hs.get_auth_handler() - auth_handler.complete_sso_login = simple_async_mock() - - state = "state" - client_redirect_url = "http://client/redirect" - session = self._generate_oidc_session_token( - state=state, - nonce="nonce", - client_redirect_url=client_redirect_url, - ) - request = _build_callback_request("code", state, session) - + request, _ = self.start_authorization(userinfo) self.get_success(self.handler.handle_oidc_callback(request)) - auth_handler.complete_sso_login.assert_called_once_with( + self.complete_sso_login.assert_called_once_with( "@foo:test", - "oidc", + self.provider.idp_id, request, - client_redirect_url, + ANY, {"phone": "1234567"}, new_user=True, auth_provider_session_id=None, @@ -826,41 +776,40 @@ class OidcHandlerTestCase(HomeserverTestCase): @override_config({"oidc_config": DEFAULT_CONFIG}) def test_map_userinfo_to_user(self) -> None: """Ensure that mapping the userinfo returned from a provider to an MXID works properly.""" - auth_handler = self.hs.get_auth_handler() - auth_handler.complete_sso_login = simple_async_mock() - userinfo: dict = { "sub": "test_user", "username": "test_user", } - self.get_success(_make_callback_with_userinfo(self.hs, userinfo)) - auth_handler.complete_sso_login.assert_called_once_with( + request, _ = self.start_authorization(userinfo) + self.get_success(self.handler.handle_oidc_callback(request)) + self.complete_sso_login.assert_called_once_with( "@test_user:test", - "oidc", - ANY, + self.provider.idp_id, + request, ANY, None, new_user=True, auth_provider_session_id=None, ) - auth_handler.complete_sso_login.reset_mock() + self.reset_mocks() # Some providers return an integer ID. userinfo = { "sub": 1234, "username": "test_user_2", } - self.get_success(_make_callback_with_userinfo(self.hs, userinfo)) - auth_handler.complete_sso_login.assert_called_once_with( + request, _ = self.start_authorization(userinfo) + self.get_success(self.handler.handle_oidc_callback(request)) + self.complete_sso_login.assert_called_once_with( "@test_user_2:test", - "oidc", - ANY, + self.provider.idp_id, + request, ANY, None, new_user=True, auth_provider_session_id=None, ) - auth_handler.complete_sso_login.reset_mock() + self.reset_mocks() # Test if the mxid is already taken store = self.hs.get_datastores().main @@ -869,8 +818,9 @@ class OidcHandlerTestCase(HomeserverTestCase): store.register_user(user_id=user3.to_string(), password_hash=None) ) userinfo = {"sub": "test3", "username": "test_user_3"} - self.get_success(_make_callback_with_userinfo(self.hs, userinfo)) - auth_handler.complete_sso_login.assert_not_called() + request, _ = self.start_authorization(userinfo) + self.get_success(self.handler.handle_oidc_callback(request)) + self.complete_sso_login.assert_not_called() self.assertRenderedError( "mapping_error", "Mapping provider does not support de-duplicating Matrix IDs", @@ -885,38 +835,37 @@ class OidcHandlerTestCase(HomeserverTestCase): store.register_user(user_id=user.to_string(), password_hash=None) ) - auth_handler = self.hs.get_auth_handler() - auth_handler.complete_sso_login = simple_async_mock() - # Map a user via SSO. userinfo = { "sub": "test", "username": "test_user", } - self.get_success(_make_callback_with_userinfo(self.hs, userinfo)) - auth_handler.complete_sso_login.assert_called_once_with( + request, _ = self.start_authorization(userinfo) + self.get_success(self.handler.handle_oidc_callback(request)) + self.complete_sso_login.assert_called_once_with( user.to_string(), - "oidc", - ANY, + self.provider.idp_id, + request, ANY, None, new_user=False, auth_provider_session_id=None, ) - auth_handler.complete_sso_login.reset_mock() + self.reset_mocks() # Subsequent calls should map to the same mxid. - self.get_success(_make_callback_with_userinfo(self.hs, userinfo)) - auth_handler.complete_sso_login.assert_called_once_with( + request, _ = self.start_authorization(userinfo) + self.get_success(self.handler.handle_oidc_callback(request)) + self.complete_sso_login.assert_called_once_with( user.to_string(), - "oidc", - ANY, + self.provider.idp_id, + request, ANY, None, new_user=False, auth_provider_session_id=None, ) - auth_handler.complete_sso_login.reset_mock() + self.reset_mocks() # Note that a second SSO user can be mapped to the same Matrix ID. (This # requires a unique sub, but something that maps to the same matrix ID, @@ -927,17 +876,18 @@ class OidcHandlerTestCase(HomeserverTestCase): "sub": "test1", "username": "test_user", } - self.get_success(_make_callback_with_userinfo(self.hs, userinfo)) - auth_handler.complete_sso_login.assert_called_once_with( + request, _ = self.start_authorization(userinfo) + self.get_success(self.handler.handle_oidc_callback(request)) + self.complete_sso_login.assert_called_once_with( user.to_string(), - "oidc", - ANY, + self.provider.idp_id, + request, ANY, None, new_user=False, auth_provider_session_id=None, ) - auth_handler.complete_sso_login.reset_mock() + self.reset_mocks() # Register some non-exact matching cases. user2 = UserID.from_string("@TEST_user_2:test") @@ -954,8 +904,9 @@ class OidcHandlerTestCase(HomeserverTestCase): "sub": "test2", "username": "TEST_USER_2", } - self.get_success(_make_callback_with_userinfo(self.hs, userinfo)) - auth_handler.complete_sso_login.assert_not_called() + request, _ = self.start_authorization(userinfo) + self.get_success(self.handler.handle_oidc_callback(request)) + self.complete_sso_login.assert_not_called() args = self.assertRenderedError("mapping_error") self.assertTrue( args[2].startswith( @@ -969,11 +920,12 @@ class OidcHandlerTestCase(HomeserverTestCase): store.register_user(user_id=user2.to_string(), password_hash=None) ) - self.get_success(_make_callback_with_userinfo(self.hs, userinfo)) - auth_handler.complete_sso_login.assert_called_once_with( + request, _ = self.start_authorization(userinfo) + self.get_success(self.handler.handle_oidc_callback(request)) + self.complete_sso_login.assert_called_once_with( "@TEST_USER_2:test", - "oidc", - ANY, + self.provider.idp_id, + request, ANY, None, new_user=False, @@ -983,9 +935,9 @@ class OidcHandlerTestCase(HomeserverTestCase): @override_config({"oidc_config": DEFAULT_CONFIG}) def test_map_userinfo_to_invalid_localpart(self) -> None: """If the mapping provider generates an invalid localpart it should be rejected.""" - self.get_success( - _make_callback_with_userinfo(self.hs, {"sub": "test2", "username": "föö"}) - ) + userinfo = {"sub": "test2", "username": "föö"} + request, _ = self.start_authorization(userinfo) + self.get_success(self.handler.handle_oidc_callback(request)) self.assertRenderedError("mapping_error", "localpart is invalid: föö") @override_config( @@ -1000,9 +952,6 @@ class OidcHandlerTestCase(HomeserverTestCase): ) def test_map_userinfo_to_user_retries(self) -> None: """The mapping provider can retry generating an MXID if the MXID is already in use.""" - auth_handler = self.hs.get_auth_handler() - auth_handler.complete_sso_login = simple_async_mock() - store = self.hs.get_datastores().main self.get_success( store.register_user(user_id="@test_user:test", password_hash=None) @@ -1011,19 +960,20 @@ class OidcHandlerTestCase(HomeserverTestCase): "sub": "test", "username": "test_user", } - self.get_success(_make_callback_with_userinfo(self.hs, userinfo)) + request, _ = self.start_authorization(userinfo) + self.get_success(self.handler.handle_oidc_callback(request)) # test_user is already taken, so test_user1 gets registered instead. - auth_handler.complete_sso_login.assert_called_once_with( + self.complete_sso_login.assert_called_once_with( "@test_user1:test", - "oidc", - ANY, + self.provider.idp_id, + request, ANY, None, new_user=True, auth_provider_session_id=None, ) - auth_handler.complete_sso_login.reset_mock() + self.reset_mocks() # Register all of the potential mxids for a particular OIDC username. self.get_success( @@ -1039,8 +989,9 @@ class OidcHandlerTestCase(HomeserverTestCase): "sub": "tester", "username": "tester", } - self.get_success(_make_callback_with_userinfo(self.hs, userinfo)) - auth_handler.complete_sso_login.assert_not_called() + request, _ = self.start_authorization(userinfo) + self.get_success(self.handler.handle_oidc_callback(request)) + self.complete_sso_login.assert_not_called() self.assertRenderedError( "mapping_error", "Unable to generate a Matrix ID from the SSO response" ) @@ -1052,7 +1003,8 @@ class OidcHandlerTestCase(HomeserverTestCase): "sub": "tester", "username": "", } - self.get_success(_make_callback_with_userinfo(self.hs, userinfo)) + request, _ = self.start_authorization(userinfo) + self.get_success(self.handler.handle_oidc_callback(request)) self.assertRenderedError("mapping_error", "localpart is invalid: ") @override_config( @@ -1071,7 +1023,8 @@ class OidcHandlerTestCase(HomeserverTestCase): "sub": "tester", "username": None, } - self.get_success(_make_callback_with_userinfo(self.hs, userinfo)) + request, _ = self.start_authorization(userinfo) + self.get_success(self.handler.handle_oidc_callback(request)) self.assertRenderedError("mapping_error", "localpart is invalid: ") @override_config( @@ -1084,16 +1037,14 @@ class OidcHandlerTestCase(HomeserverTestCase): ) def test_attribute_requirements(self) -> None: """The required attributes must be met from the OIDC userinfo response.""" - auth_handler = self.hs.get_auth_handler() - auth_handler.complete_sso_login = simple_async_mock() - # userinfo lacking "test": "foobar" attribute should fail. userinfo = { "sub": "tester", "username": "tester", } - self.get_success(_make_callback_with_userinfo(self.hs, userinfo)) - auth_handler.complete_sso_login.assert_not_called() + request, _ = self.start_authorization(userinfo) + self.get_success(self.handler.handle_oidc_callback(request)) + self.complete_sso_login.assert_not_called() # userinfo with "test": "foobar" attribute should succeed. userinfo = { @@ -1101,13 +1052,14 @@ class OidcHandlerTestCase(HomeserverTestCase): "username": "tester", "test": "foobar", } - self.get_success(_make_callback_with_userinfo(self.hs, userinfo)) + request, _ = self.start_authorization(userinfo) + self.get_success(self.handler.handle_oidc_callback(request)) # check that the auth handler got called as expected - auth_handler.complete_sso_login.assert_called_once_with( + self.complete_sso_login.assert_called_once_with( "@tester:test", - "oidc", - ANY, + self.provider.idp_id, + request, ANY, None, new_user=True, @@ -1124,21 +1076,20 @@ class OidcHandlerTestCase(HomeserverTestCase): ) def test_attribute_requirements_contains(self) -> None: """Test that auth succeeds if userinfo attribute CONTAINS required value""" - auth_handler = self.hs.get_auth_handler() - auth_handler.complete_sso_login = simple_async_mock() # userinfo with "test": ["foobar", "foo", "bar"] attribute should succeed. userinfo = { "sub": "tester", "username": "tester", "test": ["foobar", "foo", "bar"], } - self.get_success(_make_callback_with_userinfo(self.hs, userinfo)) + request, _ = self.start_authorization(userinfo) + self.get_success(self.handler.handle_oidc_callback(request)) # check that the auth handler got called as expected - auth_handler.complete_sso_login.assert_called_once_with( + self.complete_sso_login.assert_called_once_with( "@tester:test", - "oidc", - ANY, + self.provider.idp_id, + request, ANY, None, new_user=True, @@ -1158,16 +1109,15 @@ class OidcHandlerTestCase(HomeserverTestCase): Test that auth fails if attributes exist but don't match, or are non-string values. """ - auth_handler = self.hs.get_auth_handler() - auth_handler.complete_sso_login = simple_async_mock() # userinfo with "test": "not_foobar" attribute should fail userinfo: dict = { "sub": "tester", "username": "tester", "test": "not_foobar", } - self.get_success(_make_callback_with_userinfo(self.hs, userinfo)) - auth_handler.complete_sso_login.assert_not_called() + request, _ = self.start_authorization(userinfo) + self.get_success(self.handler.handle_oidc_callback(request)) + self.complete_sso_login.assert_not_called() # userinfo with "test": ["foo", "bar"] attribute should fail userinfo = { @@ -1175,8 +1125,9 @@ class OidcHandlerTestCase(HomeserverTestCase): "username": "tester", "test": ["foo", "bar"], } - self.get_success(_make_callback_with_userinfo(self.hs, userinfo)) - auth_handler.complete_sso_login.assert_not_called() + request, _ = self.start_authorization(userinfo) + self.get_success(self.handler.handle_oidc_callback(request)) + self.complete_sso_login.assert_not_called() # userinfo with "test": False attribute should fail # this is largely just to ensure we don't crash here @@ -1185,8 +1136,9 @@ class OidcHandlerTestCase(HomeserverTestCase): "username": "tester", "test": False, } - self.get_success(_make_callback_with_userinfo(self.hs, userinfo)) - auth_handler.complete_sso_login.assert_not_called() + request, _ = self.start_authorization(userinfo) + self.get_success(self.handler.handle_oidc_callback(request)) + self.complete_sso_login.assert_not_called() # userinfo with "test": None attribute should fail # a value of None breaks the OIDC spec, but it's important to not crash here @@ -1195,8 +1147,9 @@ class OidcHandlerTestCase(HomeserverTestCase): "username": "tester", "test": None, } - self.get_success(_make_callback_with_userinfo(self.hs, userinfo)) - auth_handler.complete_sso_login.assert_not_called() + request, _ = self.start_authorization(userinfo) + self.get_success(self.handler.handle_oidc_callback(request)) + self.complete_sso_login.assert_not_called() # userinfo with "test": 1 attribute should fail # this is largely just to ensure we don't crash here @@ -1205,8 +1158,9 @@ class OidcHandlerTestCase(HomeserverTestCase): "username": "tester", "test": 1, } - self.get_success(_make_callback_with_userinfo(self.hs, userinfo)) - auth_handler.complete_sso_login.assert_not_called() + request, _ = self.start_authorization(userinfo) + self.get_success(self.handler.handle_oidc_callback(request)) + self.complete_sso_login.assert_not_called() # userinfo with "test": 3.14 attribute should fail # this is largely just to ensure we don't crash here @@ -1215,8 +1169,9 @@ class OidcHandlerTestCase(HomeserverTestCase): "username": "tester", "test": 3.14, } - self.get_success(_make_callback_with_userinfo(self.hs, userinfo)) - auth_handler.complete_sso_login.assert_not_called() + request, _ = self.start_authorization(userinfo) + self.get_success(self.handler.handle_oidc_callback(request)) + self.complete_sso_login.assert_not_called() def _generate_oidc_session_token( self, @@ -1230,7 +1185,7 @@ class OidcHandlerTestCase(HomeserverTestCase): return self.handler._macaroon_generator.generate_oidc_session_token( state=state, session_data=OidcSessionData( - idp_id="oidc", + idp_id=self.provider.idp_id, nonce=nonce, client_redirect_url=client_redirect_url, ui_auth_session_id=ui_auth_session_id, @@ -1238,41 +1193,6 @@ class OidcHandlerTestCase(HomeserverTestCase): ) -async def _make_callback_with_userinfo( - hs: HomeServer, userinfo: dict, client_redirect_url: str = "http://client/redirect" -) -> None: - """Mock up an OIDC callback with the given userinfo dict - - We'll pull out the OIDC handler from the homeserver, stub out a couple of methods, - and poke in the userinfo dict as if it were the response to an OIDC userinfo call. - - Args: - hs: the HomeServer impl to send the callback to. - userinfo: the OIDC userinfo dict - client_redirect_url: the URL to redirect to on success. - """ - - handler = hs.get_oidc_handler() - provider = handler._providers["oidc"] - provider._exchange_code = simple_async_mock(return_value={"id_token": ""}) # type: ignore[assignment] - provider._parse_id_token = simple_async_mock(return_value=userinfo) # type: ignore[assignment] - provider._fetch_userinfo = simple_async_mock(return_value=userinfo) # type: ignore[assignment] - - state = "state" - session = handler._macaroon_generator.generate_oidc_session_token( - state=state, - session_data=OidcSessionData( - idp_id="oidc", - nonce="nonce", - client_redirect_url=client_redirect_url, - ui_auth_session_id="", - ), - ) - request = _build_callback_request("code", state, session) - - await handler.handle_oidc_callback(request) - - def _build_callback_request( code: str, state: str, diff --git a/tests/rest/client/test_auth.py b/tests/rest/client/test_auth.py index 090cef5216..ebf653d018 100644 --- a/tests/rest/client/test_auth.py +++ b/tests/rest/client/test_auth.py @@ -465,9 +465,11 @@ class UIAuthTests(unittest.HomeserverTestCase): * checking that the original operation succeeds """ + fake_oidc_server = self.helper.fake_oidc_server() + # log the user in remote_user_id = UserID.from_string(self.user).localpart - login_resp = self.helper.login_via_oidc(remote_user_id) + login_resp, _ = self.helper.login_via_oidc(fake_oidc_server, remote_user_id) self.assertEqual(login_resp["user_id"], self.user) # initiate a UI Auth process by attempting to delete the device @@ -481,8 +483,8 @@ class UIAuthTests(unittest.HomeserverTestCase): # run the UIA-via-SSO flow session_id = channel.json_body["session"] - channel = self.helper.auth_via_oidc( - {"sub": remote_user_id}, ui_auth_session_id=session_id + channel, _ = self.helper.auth_via_oidc( + fake_oidc_server, {"sub": remote_user_id}, ui_auth_session_id=session_id ) # that should serve a confirmation page @@ -499,7 +501,8 @@ class UIAuthTests(unittest.HomeserverTestCase): @skip_unless(HAS_OIDC, "requires OIDC") @override_config({"oidc_config": TEST_OIDC_CONFIG}) def test_does_not_offer_password_for_sso_user(self) -> None: - login_resp = self.helper.login_via_oidc("username") + fake_oidc_server = self.helper.fake_oidc_server() + login_resp, _ = self.helper.login_via_oidc(fake_oidc_server, "username") user_tok = login_resp["access_token"] device_id = login_resp["device_id"] @@ -522,7 +525,10 @@ class UIAuthTests(unittest.HomeserverTestCase): @override_config({"oidc_config": TEST_OIDC_CONFIG}) def test_offers_both_flows_for_upgraded_user(self) -> None: """A user that had a password and then logged in with SSO should get both flows""" - login_resp = self.helper.login_via_oidc(UserID.from_string(self.user).localpart) + fake_oidc_server = self.helper.fake_oidc_server() + login_resp, _ = self.helper.login_via_oidc( + fake_oidc_server, UserID.from_string(self.user).localpart + ) self.assertEqual(login_resp["user_id"], self.user) channel = self.delete_device( @@ -539,8 +545,13 @@ class UIAuthTests(unittest.HomeserverTestCase): @override_config({"oidc_config": TEST_OIDC_CONFIG}) def test_ui_auth_fails_for_incorrect_sso_user(self) -> None: """If the user tries to authenticate with the wrong SSO user, they get an error""" + + fake_oidc_server = self.helper.fake_oidc_server() + # log the user in - login_resp = self.helper.login_via_oidc(UserID.from_string(self.user).localpart) + login_resp, _ = self.helper.login_via_oidc( + fake_oidc_server, UserID.from_string(self.user).localpart + ) self.assertEqual(login_resp["user_id"], self.user) # start a UI Auth flow by attempting to delete a device @@ -553,8 +564,8 @@ class UIAuthTests(unittest.HomeserverTestCase): session_id = channel.json_body["session"] # do the OIDC auth, but auth as the wrong user - channel = self.helper.auth_via_oidc( - {"sub": "wrong_user"}, ui_auth_session_id=session_id + channel, _ = self.helper.auth_via_oidc( + fake_oidc_server, {"sub": "wrong_user"}, ui_auth_session_id=session_id ) # that should return a failure message @@ -584,7 +595,10 @@ class UIAuthTests(unittest.HomeserverTestCase): """Tests that if we register a user via SSO while requiring approval for new accounts, we still raise the correct error before logging the user in. """ - login_resp = self.helper.login_via_oidc("username", expected_status=403) + fake_oidc_server = self.helper.fake_oidc_server() + login_resp, _ = self.helper.login_via_oidc( + fake_oidc_server, "username", expected_status=403 + ) self.assertEqual(login_resp["errcode"], Codes.USER_AWAITING_APPROVAL) self.assertEqual( diff --git a/tests/rest/client/test_login.py b/tests/rest/client/test_login.py index e801ba8c8b..ff5baa9f0a 100644 --- a/tests/rest/client/test_login.py +++ b/tests/rest/client/test_login.py @@ -36,7 +36,7 @@ from synapse.util import Clock from tests import unittest from tests.handlers.test_oidc import HAS_OIDC from tests.handlers.test_saml import has_saml2 -from tests.rest.client.utils import TEST_OIDC_AUTH_ENDPOINT, TEST_OIDC_CONFIG +from tests.rest.client.utils import TEST_OIDC_CONFIG from tests.server import FakeChannel from tests.test_utils.html_parsers import TestHtmlParser from tests.unittest import HomeserverTestCase, override_config, skip_unless @@ -612,13 +612,16 @@ class MultiSSOTestCase(unittest.HomeserverTestCase): def test_login_via_oidc(self) -> None: """If OIDC is chosen, should redirect to the OIDC auth endpoint""" - # pick the default OIDC provider - channel = self.make_request( - "GET", - "/_synapse/client/pick_idp?redirectUrl=" - + urllib.parse.quote_plus(TEST_CLIENT_REDIRECT_URL) - + "&idp=oidc", - ) + fake_oidc_server = self.helper.fake_oidc_server() + + with fake_oidc_server.patch_homeserver(hs=self.hs): + # pick the default OIDC provider + channel = self.make_request( + "GET", + "/_synapse/client/pick_idp?redirectUrl=" + + urllib.parse.quote_plus(TEST_CLIENT_REDIRECT_URL) + + "&idp=oidc", + ) self.assertEqual(channel.code, 302, channel.result) location_headers = channel.headers.getRawHeaders("Location") assert location_headers @@ -626,7 +629,7 @@ class MultiSSOTestCase(unittest.HomeserverTestCase): oidc_uri_path, oidc_uri_query = oidc_uri.split("?", 1) # it should redirect us to the auth page of the OIDC server - self.assertEqual(oidc_uri_path, TEST_OIDC_AUTH_ENDPOINT) + self.assertEqual(oidc_uri_path, fake_oidc_server.authorization_endpoint) # ... and should have set a cookie including the redirect url cookie_headers = channel.headers.getRawHeaders("Set-Cookie") @@ -643,7 +646,9 @@ class MultiSSOTestCase(unittest.HomeserverTestCase): TEST_CLIENT_REDIRECT_URL, ) - channel = self.helper.complete_oidc_auth(oidc_uri, cookies, {"sub": "user1"}) + channel, _ = self.helper.complete_oidc_auth( + fake_oidc_server, oidc_uri, cookies, {"sub": "user1"} + ) # that should serve a confirmation page self.assertEqual(channel.code, 200, channel.result) @@ -693,7 +698,10 @@ class MultiSSOTestCase(unittest.HomeserverTestCase): def test_client_idp_redirect_to_oidc(self) -> None: """If the client pick a known IdP, redirect to it""" - channel = self._make_sso_redirect_request("oidc") + fake_oidc_server = self.helper.fake_oidc_server() + + with fake_oidc_server.patch_homeserver(hs=self.hs): + channel = self._make_sso_redirect_request("oidc") self.assertEqual(channel.code, 302, channel.result) location_headers = channel.headers.getRawHeaders("Location") assert location_headers @@ -701,7 +709,7 @@ class MultiSSOTestCase(unittest.HomeserverTestCase): oidc_uri_path, oidc_uri_query = oidc_uri.split("?", 1) # it should redirect us to the auth page of the OIDC server - self.assertEqual(oidc_uri_path, TEST_OIDC_AUTH_ENDPOINT) + self.assertEqual(oidc_uri_path, fake_oidc_server.authorization_endpoint) def _make_sso_redirect_request(self, idp_prov: Optional[str] = None) -> FakeChannel: """Send a request to /_matrix/client/r0/login/sso/redirect @@ -1280,9 +1288,13 @@ class UsernamePickerTestCase(HomeserverTestCase): def test_username_picker(self) -> None: """Test the happy path of a username picker flow.""" + fake_oidc_server = self.helper.fake_oidc_server() + # do the start of the login flow - channel = self.helper.auth_via_oidc( - {"sub": "tester", "displayname": "Jonny"}, TEST_CLIENT_REDIRECT_URL + channel, _ = self.helper.auth_via_oidc( + fake_oidc_server, + {"sub": "tester", "displayname": "Jonny"}, + TEST_CLIENT_REDIRECT_URL, ) # that should redirect to the username picker diff --git a/tests/rest/client/utils.py b/tests/rest/client/utils.py index c249a42bb6..967d229223 100644 --- a/tests/rest/client/utils.py +++ b/tests/rest/client/utils.py @@ -31,7 +31,6 @@ from typing import ( Tuple, overload, ) -from unittest.mock import patch from urllib.parse import urlencode import attr @@ -46,8 +45,19 @@ from synapse.server import HomeServer from synapse.types import JsonDict from tests.server import FakeChannel, FakeSite, make_request -from tests.test_utils import FakeResponse from tests.test_utils.html_parsers import TestHtmlParser +from tests.test_utils.oidc import FakeAuthorizationGrant, FakeOidcServer + +# an 'oidc_config' suitable for login_via_oidc. +TEST_OIDC_ISSUER = "https://issuer.test/" +TEST_OIDC_CONFIG = { + "enabled": True, + "issuer": TEST_OIDC_ISSUER, + "client_id": "test-client-id", + "client_secret": "test-client-secret", + "scopes": ["openid"], + "user_mapping_provider": {"config": {"localpart_template": "{{ user.sub }}"}}, +} @attr.s(auto_attribs=True) @@ -543,12 +553,28 @@ class RestHelper: return channel.json_body + def fake_oidc_server(self, issuer: str = TEST_OIDC_ISSUER) -> FakeOidcServer: + """Create a ``FakeOidcServer``. + + This can be used in conjuction with ``login_via_oidc``:: + + fake_oidc_server = self.helper.fake_oidc_server() + login_data, _ = self.helper.login_via_oidc(fake_oidc_server, "user") + """ + + return FakeOidcServer( + clock=self.hs.get_clock(), + issuer=issuer, + ) + def login_via_oidc( self, + fake_server: FakeOidcServer, remote_user_id: str, + with_sid: bool = False, expected_status: int = 200, - ) -> JsonDict: - """Log in via OIDC + ) -> Tuple[JsonDict, FakeAuthorizationGrant]: + """Log in (as a new user) via OIDC Returns the result of the final token login. @@ -560,7 +586,10 @@ class RestHelper: the normal places. """ client_redirect_url = "https://x" - channel = self.auth_via_oidc({"sub": remote_user_id}, client_redirect_url) + userinfo = {"sub": remote_user_id} + channel, grant = self.auth_via_oidc( + fake_server, userinfo, client_redirect_url, with_sid=with_sid + ) # expect a confirmation page assert channel.code == HTTPStatus.OK, channel.result @@ -585,14 +614,16 @@ class RestHelper: assert ( channel.code == expected_status ), f"unexpected status in response: {channel.code}" - return channel.json_body + return channel.json_body, grant def auth_via_oidc( self, + fake_server: FakeOidcServer, user_info_dict: JsonDict, client_redirect_url: Optional[str] = None, ui_auth_session_id: Optional[str] = None, - ) -> FakeChannel: + with_sid: bool = False, + ) -> Tuple[FakeChannel, FakeAuthorizationGrant]: """Perform an OIDC authentication flow via a mock OIDC provider. This can be used for either login or user-interactive auth. @@ -616,6 +647,7 @@ class RestHelper: the login redirect endpoint ui_auth_session_id: if set, we will perform a UI Auth flow. The session id of the UI auth. + with_sid: if True, generates a random `sid` (OIDC session ID) Returns: A FakeChannel containing the result of calling the OIDC callback endpoint. @@ -625,14 +657,15 @@ class RestHelper: cookies: Dict[str, str] = {} - # if we're doing a ui auth, hit the ui auth redirect endpoint - if ui_auth_session_id: - # can't set the client redirect url for UI Auth - assert client_redirect_url is None - oauth_uri = self.initiate_sso_ui_auth(ui_auth_session_id, cookies) - else: - # otherwise, hit the login redirect endpoint - oauth_uri = self.initiate_sso_login(client_redirect_url, cookies) + with fake_server.patch_homeserver(hs=self.hs): + # if we're doing a ui auth, hit the ui auth redirect endpoint + if ui_auth_session_id: + # can't set the client redirect url for UI Auth + assert client_redirect_url is None + oauth_uri = self.initiate_sso_ui_auth(ui_auth_session_id, cookies) + else: + # otherwise, hit the login redirect endpoint + oauth_uri = self.initiate_sso_login(client_redirect_url, cookies) # we now have a URI for the OIDC IdP, but we skip that and go straight # back to synapse's OIDC callback resource. However, we do need the "state" @@ -640,17 +673,21 @@ class RestHelper: # that synapse passes to the client. oauth_uri_path, _ = oauth_uri.split("?", 1) - assert oauth_uri_path == TEST_OIDC_AUTH_ENDPOINT, ( + assert oauth_uri_path == fake_server.authorization_endpoint, ( "unexpected SSO URI " + oauth_uri_path ) - return self.complete_oidc_auth(oauth_uri, cookies, user_info_dict) + return self.complete_oidc_auth( + fake_server, oauth_uri, cookies, user_info_dict, with_sid=with_sid + ) def complete_oidc_auth( self, + fake_serer: FakeOidcServer, oauth_uri: str, cookies: Mapping[str, str], user_info_dict: JsonDict, - ) -> FakeChannel: + with_sid: bool = False, + ) -> Tuple[FakeChannel, FakeAuthorizationGrant]: """Mock out an OIDC authentication flow Assumes that an OIDC auth has been initiated by one of initiate_sso_login or @@ -661,50 +698,37 @@ class RestHelper: Requires the OIDC callback resource to be mounted at the normal place. Args: + fake_server: the fake OIDC server with which the auth should be done oauth_uri: the OIDC URI returned by synapse's redirect endpoint (ie, from initiate_sso_login or initiate_sso_ui_auth). cookies: the cookies set by synapse's redirect endpoint, which will be sent back to the callback endpoint. user_info_dict: the remote userinfo that the OIDC provider should present. Typically this should be '{"sub": ""}'. + with_sid: if True, generates a random `sid` (OIDC session ID) Returns: A FakeChannel containing the result of calling the OIDC callback endpoint. """ _, oauth_uri_qs = oauth_uri.split("?", 1) params = urllib.parse.parse_qs(oauth_uri_qs) + + code, grant = fake_serer.start_authorization( + scope=params["scope"][0], + userinfo=user_info_dict, + client_id=params["client_id"][0], + redirect_uri=params["redirect_uri"][0], + nonce=params["nonce"][0], + with_sid=with_sid, + ) + state = params["state"][0] + callback_uri = "%s?%s" % ( urllib.parse.urlparse(params["redirect_uri"][0]).path, - urllib.parse.urlencode({"state": params["state"][0], "code": "TEST_CODE"}), + urllib.parse.urlencode({"state": state, "code": code}), ) - # before we hit the callback uri, stub out some methods in the http client so - # that we don't have to handle full HTTPS requests. - # (expected url, json response) pairs, in the order we expect them. - expected_requests = [ - # first we get a hit to the token endpoint, which we tell to return - # a dummy OIDC access token - (TEST_OIDC_TOKEN_ENDPOINT, {"access_token": "TEST"}), - # and then one to the user_info endpoint, which returns our remote user id. - (TEST_OIDC_USERINFO_ENDPOINT, user_info_dict), - ] - - async def mock_req( - method: str, - uri: str, - data: Optional[dict] = None, - headers: Optional[Iterable[Tuple[AnyStr, AnyStr]]] = None, - ): - (expected_uri, resp_obj) = expected_requests.pop(0) - assert uri == expected_uri - resp = FakeResponse( - code=HTTPStatus.OK, - phrase=b"OK", - body=json.dumps(resp_obj).encode("utf-8"), - ) - return resp - - with patch.object(self.hs.get_proxied_http_client(), "request", mock_req): + with fake_serer.patch_homeserver(hs=self.hs): # now hit the callback URI with the right params and a made-up code channel = make_request( self.hs.get_reactor(), @@ -715,7 +739,7 @@ class RestHelper: ("Cookie", "%s=%s" % (k, v)) for (k, v) in cookies.items() ], ) - return channel + return channel, grant def initiate_sso_login( self, client_redirect_url: Optional[str], cookies: MutableMapping[str, str] @@ -806,21 +830,3 @@ class RestHelper: assert len(p.links) == 1, "not exactly one link in confirmation page" oauth_uri = p.links[0] return oauth_uri - - -# an 'oidc_config' suitable for login_via_oidc. -TEST_OIDC_AUTH_ENDPOINT = "https://issuer.test/auth" -TEST_OIDC_TOKEN_ENDPOINT = "https://issuer.test/token" -TEST_OIDC_USERINFO_ENDPOINT = "https://issuer.test/userinfo" -TEST_OIDC_CONFIG = { - "enabled": True, - "discover": False, - "issuer": "https://issuer.test", - "client_id": "test-client-id", - "client_secret": "test-client-secret", - "scopes": ["profile"], - "authorization_endpoint": TEST_OIDC_AUTH_ENDPOINT, - "token_endpoint": TEST_OIDC_TOKEN_ENDPOINT, - "userinfo_endpoint": TEST_OIDC_USERINFO_ENDPOINT, - "user_mapping_provider": {"config": {"localpart_template": "{{ user.sub }}"}}, -} diff --git a/tests/test_utils/__init__.py b/tests/test_utils/__init__.py index 0d0d6faf0d..e62ebcc6a5 100644 --- a/tests/test_utils/__init__.py +++ b/tests/test_utils/__init__.py @@ -15,17 +15,24 @@ """ Utilities for running the unit tests """ +import json import sys import warnings from asyncio import Future from binascii import unhexlify -from typing import Awaitable, Callable, TypeVar +from typing import Awaitable, Callable, Tuple, TypeVar from unittest.mock import Mock import attr +import zope.interface from twisted.python.failure import Failure from twisted.web.client import ResponseDone +from twisted.web.http import RESPONSES +from twisted.web.http_headers import Headers +from twisted.web.iweb import IResponse + +from synapse.types import JsonDict TV = TypeVar("TV") @@ -97,27 +104,44 @@ def simple_async_mock(return_value=None, raises=None) -> Mock: return Mock(side_effect=cb) -@attr.s -class FakeResponse: +# Type ignore: it does not fully implement IResponse, but is good enough for tests +@zope.interface.implementer(IResponse) +@attr.s(slots=True, frozen=True, auto_attribs=True) +class FakeResponse: # type: ignore[misc] """A fake twisted.web.IResponse object there is a similar class at treq.test.test_response, but it lacks a `phrase` attribute, and didn't support deliverBody until recently. """ - # HTTP response code - code = attr.ib(type=int) + version: Tuple[bytes, int, int] = (b"HTTP", 1, 1) - # HTTP response phrase (eg b'OK' for a 200) - phrase = attr.ib(type=bytes) + # HTTP response code + code: int = 200 # body of the response - body = attr.ib(type=bytes) + body: bytes = b"" + + headers: Headers = attr.Factory(Headers) + + @property + def phrase(self): + return RESPONSES.get(self.code, b"Unknown Status") + + @property + def length(self): + return len(self.body) def deliverBody(self, protocol): protocol.dataReceived(self.body) protocol.connectionLost(Failure(ResponseDone())) + @classmethod + def json(cls, *, code: int = 200, payload: JsonDict) -> "FakeResponse": + headers = Headers({"Content-Type": ["application/json"]}) + body = json.dumps(payload).encode("utf-8") + return cls(code=code, body=body, headers=headers) + # A small image used in some tests. # diff --git a/tests/test_utils/oidc.py b/tests/test_utils/oidc.py new file mode 100644 index 0000000000..de134bbc89 --- /dev/null +++ b/tests/test_utils/oidc.py @@ -0,0 +1,325 @@ +# Copyright 2022 The Matrix.org Foundation C.I.C. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + + +import json +from typing import Any, Dict, List, Optional, Tuple +from unittest.mock import Mock, patch +from urllib.parse import parse_qs + +import attr + +from twisted.web.http_headers import Headers +from twisted.web.iweb import IResponse + +from synapse.server import HomeServer +from synapse.util import Clock +from synapse.util.stringutils import random_string + +from tests.test_utils import FakeResponse + + +@attr.s(slots=True, frozen=True, auto_attribs=True) +class FakeAuthorizationGrant: + userinfo: dict + client_id: str + redirect_uri: str + scope: str + nonce: Optional[str] + sid: Optional[str] + + +class FakeOidcServer: + """A fake OpenID Connect Provider.""" + + # All methods here are mocks, so we can track when they are called, and override + # their values + request: Mock + get_jwks_handler: Mock + get_metadata_handler: Mock + get_userinfo_handler: Mock + post_token_handler: Mock + + def __init__(self, clock: Clock, issuer: str): + from authlib.jose import ECKey, KeySet + + self._clock = clock + self.issuer = issuer + + self.request = Mock(side_effect=self._request) + self.get_jwks_handler = Mock(side_effect=self._get_jwks_handler) + self.get_metadata_handler = Mock(side_effect=self._get_metadata_handler) + self.get_userinfo_handler = Mock(side_effect=self._get_userinfo_handler) + self.post_token_handler = Mock(side_effect=self._post_token_handler) + + # A code -> grant mapping + self._authorization_grants: Dict[str, FakeAuthorizationGrant] = {} + # An access token -> grant mapping + self._sessions: Dict[str, FakeAuthorizationGrant] = {} + + # We generate here an ECDSA key with the P-256 curve (ES256 algorithm) used for + # signing JWTs. ECDSA keys are really quick to generate compared to RSA. + self._key = ECKey.generate_key(crv="P-256", is_private=True) + self._jwks = KeySet([ECKey.import_key(self._key.as_pem(is_private=False))]) + + self._id_token_overrides: Dict[str, Any] = {} + + def reset_mocks(self): + self.request.reset_mock() + self.get_jwks_handler.reset_mock() + self.get_metadata_handler.reset_mock() + self.get_userinfo_handler.reset_mock() + self.post_token_handler.reset_mock() + + def patch_homeserver(self, hs: HomeServer): + """Patch the ``HomeServer`` HTTP client to handle requests through the ``FakeOidcServer``. + + This patch should be used whenever the HS is expected to perform request to the + OIDC provider, e.g.:: + + fake_oidc_server = self.helper.fake_oidc_server() + with fake_oidc_server.patch_homeserver(hs): + self.make_request("GET", "/_matrix/client/r0/login/sso/redirect") + """ + return patch.object(hs.get_proxied_http_client(), "request", self.request) + + @property + def authorization_endpoint(self) -> str: + return self.issuer + "authorize" + + @property + def token_endpoint(self) -> str: + return self.issuer + "token" + + @property + def userinfo_endpoint(self) -> str: + return self.issuer + "userinfo" + + @property + def metadata_endpoint(self) -> str: + return self.issuer + ".well-known/openid-configuration" + + @property + def jwks_uri(self) -> str: + return self.issuer + "jwks" + + def get_metadata(self) -> dict: + return { + "issuer": self.issuer, + "authorization_endpoint": self.authorization_endpoint, + "token_endpoint": self.token_endpoint, + "jwks_uri": self.jwks_uri, + "userinfo_endpoint": self.userinfo_endpoint, + "response_types_supported": ["code"], + "subject_types_supported": ["public"], + "id_token_signing_alg_values_supported": ["ES256"], + } + + def get_jwks(self) -> dict: + return self._jwks.as_dict() + + def get_userinfo(self, access_token: str) -> Optional[dict]: + """Given an access token, get the userinfo of the associated session.""" + session = self._sessions.get(access_token, None) + if session is None: + return None + return session.userinfo + + def _sign(self, payload: dict) -> str: + from authlib.jose import JsonWebSignature + + jws = JsonWebSignature() + kid = self.get_jwks()["keys"][0]["kid"] + protected = {"alg": "ES256", "kid": kid} + json_payload = json.dumps(payload) + return jws.serialize_compact(protected, json_payload, self._key).decode("utf-8") + + def generate_id_token(self, grant: FakeAuthorizationGrant) -> str: + now = self._clock.time() + id_token = { + **grant.userinfo, + "iss": self.issuer, + "aud": grant.client_id, + "iat": now, + "nbf": now, + "exp": now + 600, + } + + if grant.nonce is not None: + id_token["nonce"] = grant.nonce + + if grant.sid is not None: + id_token["sid"] = grant.sid + + id_token.update(self._id_token_overrides) + + return self._sign(id_token) + + def id_token_override(self, overrides: dict): + """Temporarily patch the ID token generated by the token endpoint.""" + return patch.object(self, "_id_token_overrides", overrides) + + def start_authorization( + self, + client_id: str, + scope: str, + redirect_uri: str, + userinfo: dict, + nonce: Optional[str] = None, + with_sid: bool = False, + ) -> Tuple[str, FakeAuthorizationGrant]: + """Start an authorization request, and get back the code to use on the authorization endpoint.""" + code = random_string(10) + sid = None + if with_sid: + sid = random_string(10) + + grant = FakeAuthorizationGrant( + userinfo=userinfo, + scope=scope, + redirect_uri=redirect_uri, + nonce=nonce, + client_id=client_id, + sid=sid, + ) + self._authorization_grants[code] = grant + + return code, grant + + def exchange_code(self, code: str) -> Optional[Dict[str, Any]]: + grant = self._authorization_grants.pop(code, None) + if grant is None: + return None + + access_token = random_string(10) + self._sessions[access_token] = grant + + token = { + "token_type": "Bearer", + "access_token": access_token, + "expires_in": 3600, + "scope": grant.scope, + } + + if "openid" in grant.scope: + token["id_token"] = self.generate_id_token(grant) + + return dict(token) + + def buggy_endpoint( + self, + *, + jwks: bool = False, + metadata: bool = False, + token: bool = False, + userinfo: bool = False, + ): + """A context which makes a set of endpoints return a 500 error. + + Args: + jwks: If True, makes the JWKS endpoint return a 500 error. + metadata: If True, makes the OIDC Discovery endpoint return a 500 error. + token: If True, makes the token endpoint return a 500 error. + userinfo: If True, makes the userinfo endpoint return a 500 error. + """ + buggy = FakeResponse(code=500, body=b"Internal server error") + + patches = {} + if jwks: + patches["get_jwks_handler"] = Mock(return_value=buggy) + if metadata: + patches["get_metadata_handler"] = Mock(return_value=buggy) + if token: + patches["post_token_handler"] = Mock(return_value=buggy) + if userinfo: + patches["get_userinfo_handler"] = Mock(return_value=buggy) + + return patch.multiple(self, **patches) + + async def _request( + self, + method: str, + uri: str, + data: Optional[bytes] = None, + headers: Optional[Headers] = None, + ) -> IResponse: + """The override of the SimpleHttpClient#request() method""" + access_token: Optional[str] = None + + if headers is None: + headers = Headers() + + # Try to find the access token in the headers if any + auth_headers = headers.getRawHeaders(b"Authorization") + if auth_headers: + parts = auth_headers[0].split(b" ") + if parts[0] == b"Bearer" and len(parts) == 2: + access_token = parts[1].decode("ascii") + + if method == "POST": + # If the method is POST, assume it has an url-encoded body + if data is None or headers.getRawHeaders(b"Content-Type") != [ + b"application/x-www-form-urlencoded" + ]: + return FakeResponse.json(code=400, payload={"error": "invalid_request"}) + + params = parse_qs(data.decode("utf-8")) + + if uri == self.token_endpoint: + # Even though this endpoint should be protected, this does not check + # for client authentication. We're not checking it for simplicity, + # and because client authentication is tested in other standalone tests. + return self.post_token_handler(params) + + elif method == "GET": + if uri == self.jwks_uri: + return self.get_jwks_handler() + elif uri == self.metadata_endpoint: + return self.get_metadata_handler() + elif uri == self.userinfo_endpoint: + return self.get_userinfo_handler(access_token=access_token) + + return FakeResponse(code=404, body=b"404 not found") + + # Request handlers + def _get_jwks_handler(self) -> IResponse: + """Handles requests to the JWKS URI.""" + return FakeResponse.json(payload=self.get_jwks()) + + def _get_metadata_handler(self) -> IResponse: + """Handles requests to the OIDC well-known document.""" + return FakeResponse.json(payload=self.get_metadata()) + + def _get_userinfo_handler(self, access_token: Optional[str]) -> IResponse: + """Handles requests to the userinfo endpoint.""" + if access_token is None: + return FakeResponse(code=401) + user_info = self.get_userinfo(access_token) + if user_info is None: + return FakeResponse(code=401) + + return FakeResponse.json(payload=user_info) + + def _post_token_handler(self, params: Dict[str, List[str]]) -> IResponse: + """Handles requests to the token endpoint.""" + code = params.get("code", []) + + if len(code) != 1: + return FakeResponse.json(code=400, payload={"error": "invalid_request"}) + + grant = self.exchange_code(code=code[0]) + if grant is None: + return FakeResponse.json(code=400, payload={"error": "invalid_grant"}) + + return FakeResponse.json(payload=grant) -- cgit 1.5.1 From 8756d5c87efc5637da55c9e21d2a4eb2369ba693 Mon Sep 17 00:00:00 2001 From: Quentin Gliech Date: Wed, 26 Oct 2022 12:45:41 +0200 Subject: Save login tokens in database (#13844) * Save login tokens in database Signed-off-by: Quentin Gliech * Add upgrade notes * Track login token reuse in a Prometheus metric Signed-off-by: Quentin Gliech --- changelog.d/13844.misc | 1 + docs/upgrade.md | 9 ++ synapse/handlers/auth.py | 64 +++++++-- synapse/module_api/__init__.py | 41 +----- synapse/rest/client/login.py | 3 +- synapse/rest/client/login_token_request.py | 5 +- synapse/storage/databases/main/registration.py | 156 ++++++++++++++++++++- .../schema/main/delta/73/10login_tokens.sql | 35 +++++ synapse/util/macaroons.py | 87 +----------- tests/handlers/test_auth.py | 135 ++++++++++-------- tests/util/test_macaroons.py | 28 ---- 11 files changed, 337 insertions(+), 227 deletions(-) create mode 100644 changelog.d/13844.misc create mode 100644 synapse/storage/schema/main/delta/73/10login_tokens.sql (limited to 'synapse/handlers') diff --git a/changelog.d/13844.misc b/changelog.d/13844.misc new file mode 100644 index 0000000000..66f4414df7 --- /dev/null +++ b/changelog.d/13844.misc @@ -0,0 +1 @@ +Save login tokens in database and prevent login token reuse. diff --git a/docs/upgrade.md b/docs/upgrade.md index b81385b191..78c34d0c15 100644 --- a/docs/upgrade.md +++ b/docs/upgrade.md @@ -88,6 +88,15 @@ process, for example: dpkg -i matrix-synapse-py3_1.3.0+stretch1_amd64.deb ``` +# Upgrading to v1.71.0 + +## Removal of the `generate_short_term_login_token` module API method + +As announced with the release of [Synapse 1.69.0](#deprecation-of-the-generate_short_term_login_token-module-api-method), the deprecated `generate_short_term_login_token` module method has been removed. + +Modules relying on it can instead use the `create_login_token` method. + + # Upgrading to v1.69.0 ## Changes to the receipts replication streams diff --git a/synapse/handlers/auth.py b/synapse/handlers/auth.py index f5f0e0e7a7..8b9ef25d29 100644 --- a/synapse/handlers/auth.py +++ b/synapse/handlers/auth.py @@ -38,6 +38,7 @@ from typing import ( import attr import bcrypt import unpaddedbase64 +from prometheus_client import Counter from twisted.internet.defer import CancelledError from twisted.web.server import Request @@ -48,6 +49,7 @@ from synapse.api.errors import ( Codes, InteractiveAuthIncompleteError, LoginError, + NotFoundError, StoreError, SynapseError, UserDeactivatedError, @@ -63,10 +65,14 @@ from synapse.http.server import finish_request, respond_with_html from synapse.http.site import SynapseRequest from synapse.logging.context import defer_to_thread from synapse.metrics.background_process_metrics import run_as_background_process +from synapse.storage.databases.main.registration import ( + LoginTokenExpired, + LoginTokenLookupResult, + LoginTokenReused, +) from synapse.types import JsonDict, Requester, UserID from synapse.util import stringutils as stringutils from synapse.util.async_helpers import delay_cancellation, maybe_awaitable -from synapse.util.macaroons import LoginTokenAttributes from synapse.util.msisdn import phone_number_to_msisdn from synapse.util.stringutils import base62_encode from synapse.util.threepids import canonicalise_email @@ -80,6 +86,12 @@ logger = logging.getLogger(__name__) INVALID_USERNAME_OR_PASSWORD = "Invalid username or password" +invalid_login_token_counter = Counter( + "synapse_user_login_invalid_login_tokens", + "Counts the number of rejected m.login.token on /login", + ["reason"], +) + def convert_client_dict_legacy_fields_to_identifier( submission: JsonDict, @@ -883,6 +895,25 @@ class AuthHandler: return True + async def create_login_token_for_user_id( + self, + user_id: str, + duration_ms: int = (2 * 60 * 1000), + auth_provider_id: Optional[str] = None, + auth_provider_session_id: Optional[str] = None, + ) -> str: + login_token = self.generate_login_token() + now = self._clock.time_msec() + expiry_ts = now + duration_ms + await self.store.add_login_token_to_user( + user_id=user_id, + token=login_token, + expiry_ts=expiry_ts, + auth_provider_id=auth_provider_id, + auth_provider_session_id=auth_provider_session_id, + ) + return login_token + async def create_refresh_token_for_user_id( self, user_id: str, @@ -1401,6 +1432,18 @@ class AuthHandler: return None return user_id + def generate_login_token(self) -> str: + """Generates an opaque string, for use as an short-term login token""" + + # we use the following format for access tokens: + # syl__ + + random_string = stringutils.random_string(20) + base = f"syl_{random_string}" + + crc = base62_encode(crc32(base.encode("ascii")), minwidth=6) + return f"{base}_{crc}" + def generate_access_token(self, for_user: UserID) -> str: """Generates an opaque string, for use as an access token""" @@ -1427,16 +1470,17 @@ class AuthHandler: crc = base62_encode(crc32(base.encode("ascii")), minwidth=6) return f"{base}_{crc}" - async def validate_short_term_login_token( - self, login_token: str - ) -> LoginTokenAttributes: + async def consume_login_token(self, login_token: str) -> LoginTokenLookupResult: try: - res = self.macaroon_gen.verify_short_term_login_token(login_token) - except Exception: - raise AuthError(403, "Invalid login token", errcode=Codes.FORBIDDEN) + return await self.store.consume_login_token(login_token) + except LoginTokenExpired: + invalid_login_token_counter.labels("expired").inc() + except LoginTokenReused: + invalid_login_token_counter.labels("reused").inc() + except NotFoundError: + invalid_login_token_counter.labels("not found").inc() - await self.auth_blocking.check_auth_blocking(res.user_id) - return res + raise AuthError(403, "Invalid login token", errcode=Codes.FORBIDDEN) async def delete_access_token(self, access_token: str) -> None: """Invalidate a single access token @@ -1711,7 +1755,7 @@ class AuthHandler: ) # Create a login token - login_token = self.macaroon_gen.generate_short_term_login_token( + login_token = await self.create_login_token_for_user_id( registered_user_id, auth_provider_id=auth_provider_id, auth_provider_session_id=auth_provider_session_id, diff --git a/synapse/module_api/__init__.py b/synapse/module_api/__init__.py index 6a6ae208d1..30e689d00d 100644 --- a/synapse/module_api/__init__.py +++ b/synapse/module_api/__init__.py @@ -771,50 +771,11 @@ class ModuleApi: auth_provider_session_id: The session ID got during login from the SSO IdP, if any. """ - # The deprecated `generate_short_term_login_token` method defaulted to an empty - # string for the `auth_provider_id` because of how the underlying macaroon was - # generated. This will change to a proper NULL-able field when the tokens get - # moved to the database. - return self._hs.get_macaroon_generator().generate_short_term_login_token( + return await self._hs.get_auth_handler().create_login_token_for_user_id( user_id, - auth_provider_id or "", - auth_provider_session_id, duration_in_ms, - ) - - def generate_short_term_login_token( - self, - user_id: str, - duration_in_ms: int = (2 * 60 * 1000), - auth_provider_id: str = "", - auth_provider_session_id: Optional[str] = None, - ) -> str: - """Generate a login token suitable for m.login.token authentication - - Added in Synapse v1.9.0. - - This was deprecated in Synapse v1.69.0 in favor of create_login_token, and will - be removed in Synapse 1.71.0. - - Args: - user_id: gives the ID of the user that the token is for - - duration_in_ms: the time that the token will be valid for - - auth_provider_id: the ID of the SSO IdP that the user used to authenticate - to get this token, if any. This is encoded in the token so that - /login can report stats on number of successful logins by IdP. - """ - logger.warn( - "A module configured on this server uses ModuleApi.generate_short_term_login_token(), " - "which is deprecated in favor of ModuleApi.create_login_token(), and will be removed in " - "Synapse 1.71.0", - ) - return self._hs.get_macaroon_generator().generate_short_term_login_token( - user_id, auth_provider_id, auth_provider_session_id, - duration_in_ms, ) @defer.inlineCallbacks diff --git a/synapse/rest/client/login.py b/synapse/rest/client/login.py index f554586ac3..7774f1967d 100644 --- a/synapse/rest/client/login.py +++ b/synapse/rest/client/login.py @@ -436,8 +436,7 @@ class LoginRestServlet(RestServlet): The body of the JSON response. """ token = login_submission["token"] - auth_handler = self.auth_handler - res = await auth_handler.validate_short_term_login_token(token) + res = await self.auth_handler.consume_login_token(token) return await self._complete_login( res.user_id, diff --git a/synapse/rest/client/login_token_request.py b/synapse/rest/client/login_token_request.py index 277b20fb63..43ea21d5e6 100644 --- a/synapse/rest/client/login_token_request.py +++ b/synapse/rest/client/login_token_request.py @@ -57,7 +57,6 @@ class LoginTokenRequestServlet(RestServlet): self.store = hs.get_datastores().main self.clock = hs.get_clock() self.server_name = hs.config.server.server_name - self.macaroon_gen = hs.get_macaroon_generator() self.auth_handler = hs.get_auth_handler() self.token_timeout = hs.config.experimental.msc3882_token_timeout self.ui_auth = hs.config.experimental.msc3882_ui_auth @@ -76,10 +75,10 @@ class LoginTokenRequestServlet(RestServlet): can_skip_ui_auth=False, # Don't allow skipping of UI auth ) - login_token = self.macaroon_gen.generate_short_term_login_token( + login_token = await self.auth_handler.create_login_token_for_user_id( user_id=requester.user.to_string(), auth_provider_id="org.matrix.msc3882.login_token_request", - duration_in_ms=self.token_timeout, + duration_ms=self.token_timeout, ) return ( diff --git a/synapse/storage/databases/main/registration.py b/synapse/storage/databases/main/registration.py index 2996d6bb4d..0255295317 100644 --- a/synapse/storage/databases/main/registration.py +++ b/synapse/storage/databases/main/registration.py @@ -21,7 +21,13 @@ from typing import TYPE_CHECKING, Any, Dict, List, Optional, Tuple, Union, cast import attr from synapse.api.constants import UserTypes -from synapse.api.errors import Codes, StoreError, SynapseError, ThreepidValidationError +from synapse.api.errors import ( + Codes, + NotFoundError, + StoreError, + SynapseError, + ThreepidValidationError, +) from synapse.config.homeserver import HomeServerConfig from synapse.metrics.background_process_metrics import wrap_as_background_process from synapse.storage.database import ( @@ -50,6 +56,14 @@ class ExternalIDReuseException(Exception): because this external id is given to an other user.""" +class LoginTokenExpired(Exception): + """Exception if the login token sent expired""" + + +class LoginTokenReused(Exception): + """Exception if the login token sent was already used""" + + @attr.s(frozen=True, slots=True, auto_attribs=True) class TokenLookupResult: """Result of looking up an access token. @@ -115,6 +129,20 @@ class RefreshTokenLookupResult: If None, the session can be refreshed indefinitely.""" +@attr.s(auto_attribs=True, frozen=True, slots=True) +class LoginTokenLookupResult: + """Result of looking up a login token.""" + + user_id: str + """The user this token belongs to.""" + + auth_provider_id: Optional[str] + """The SSO Identity Provider that the user authenticated with, to get this token.""" + + auth_provider_session_id: Optional[str] + """The session ID advertised by the SSO Identity Provider.""" + + class RegistrationWorkerStore(CacheInvalidationWorkerStore): def __init__( self, @@ -1789,6 +1817,109 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore): "replace_refresh_token", _replace_refresh_token_txn ) + async def add_login_token_to_user( + self, + user_id: str, + token: str, + expiry_ts: int, + auth_provider_id: Optional[str], + auth_provider_session_id: Optional[str], + ) -> None: + """Adds a short-term login token for the given user. + + Args: + user_id: The user ID. + token: The new login token to add. + expiry_ts (milliseconds since the epoch): Time after which the login token + cannot be used. + auth_provider_id: The SSO Identity Provider that the user authenticated with + to get this token, if any + auth_provider_session_id: The session ID advertised by the SSO Identity + Provider, if any. + """ + await self.db_pool.simple_insert( + "login_tokens", + { + "token": token, + "user_id": user_id, + "expiry_ts": expiry_ts, + "auth_provider_id": auth_provider_id, + "auth_provider_session_id": auth_provider_session_id, + }, + desc="add_login_token_to_user", + ) + + def _consume_login_token( + self, + txn: LoggingTransaction, + token: str, + ts: int, + ) -> LoginTokenLookupResult: + values = self.db_pool.simple_select_one_txn( + txn, + "login_tokens", + keyvalues={"token": token}, + retcols=( + "user_id", + "expiry_ts", + "used_ts", + "auth_provider_id", + "auth_provider_session_id", + ), + allow_none=True, + ) + + if values is None: + raise NotFoundError() + + self.db_pool.simple_update_one_txn( + txn, + "login_tokens", + keyvalues={"token": token}, + updatevalues={"used_ts": ts}, + ) + user_id = values["user_id"] + expiry_ts = values["expiry_ts"] + used_ts = values["used_ts"] + auth_provider_id = values["auth_provider_id"] + auth_provider_session_id = values["auth_provider_session_id"] + + # Token was already used + if used_ts is not None: + raise LoginTokenReused() + + # Token expired + if ts > int(expiry_ts): + raise LoginTokenExpired() + + return LoginTokenLookupResult( + user_id=user_id, + auth_provider_id=auth_provider_id, + auth_provider_session_id=auth_provider_session_id, + ) + + async def consume_login_token(self, token: str) -> LoginTokenLookupResult: + """Lookup a login token and consume it. + + Args: + token: The login token. + + Returns: + The data stored with that token, including the `user_id`. Returns `None` if + the token does not exist or if it expired. + + Raises: + NotFound if the login token was not found in database + LoginTokenExpired if the login token expired + LoginTokenReused if the login token was already used + """ + return await self.db_pool.runInteraction( + "consume_login_token", + self._consume_login_token, + token, + self._clock.time_msec(), + ) + @cached() async def is_guest(self, user_id: str) -> bool: res = await self.db_pool.simple_select_one_onecol( @@ -2019,6 +2150,12 @@ class RegistrationStore(StatsStore, RegistrationBackgroundUpdateStore): and hs.config.experimental.msc3866.require_approval_for_new_accounts ) + # Create a background job for removing expired login tokens + if hs.config.worker.run_background_tasks: + self._clock.looping_call( + self._delete_expired_login_tokens, THIRTY_MINUTES_IN_MS + ) + async def add_access_token_to_user( self, user_id: str, @@ -2617,6 +2754,23 @@ class RegistrationStore(StatsStore, RegistrationBackgroundUpdateStore): approved, ) + @wrap_as_background_process("delete_expired_login_tokens") + async def _delete_expired_login_tokens(self) -> None: + """Remove login tokens with expiry dates that have passed.""" + + def _delete_expired_login_tokens_txn(txn: LoggingTransaction, ts: int) -> None: + sql = "DELETE FROM login_tokens WHERE expiry_ts <= ?" + txn.execute(sql, (ts,)) + + # We keep the expired tokens for an extra 5 minutes so we can measure how many + # times a token is being used after its expiry + now = self._clock.time_msec() + await self.db_pool.runInteraction( + "delete_expired_login_tokens", + _delete_expired_login_tokens_txn, + now - (5 * 60 * 1000), + ) + def find_max_generated_user_id_localpart(cur: Cursor) -> int: """ diff --git a/synapse/storage/schema/main/delta/73/10login_tokens.sql b/synapse/storage/schema/main/delta/73/10login_tokens.sql new file mode 100644 index 0000000000..a39b7bcece --- /dev/null +++ b/synapse/storage/schema/main/delta/73/10login_tokens.sql @@ -0,0 +1,35 @@ +/* + * Copyright 2022 The Matrix.org Foundation C.I.C. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +-- Login tokens are short-lived tokens that are used for the m.login.token +-- login method, mainly during SSO logins +CREATE TABLE login_tokens ( + token TEXT PRIMARY KEY, + user_id TEXT NOT NULL, + expiry_ts BIGINT NOT NULL, + used_ts BIGINT, + auth_provider_id TEXT, + auth_provider_session_id TEXT +); + +-- We're sometimes querying them by their session ID we got from their IDP +CREATE INDEX login_tokens_auth_provider_idx + ON login_tokens (auth_provider_id, auth_provider_session_id); + +-- We're deleting them by their expiration time +CREATE INDEX login_tokens_expiry_time_idx + ON login_tokens (expiry_ts); + diff --git a/synapse/util/macaroons.py b/synapse/util/macaroons.py index df77edcce2..5df03d3ddc 100644 --- a/synapse/util/macaroons.py +++ b/synapse/util/macaroons.py @@ -24,7 +24,7 @@ from typing_extensions import Literal from synapse.util import Clock, stringutils -MacaroonType = Literal["access", "delete_pusher", "session", "login"] +MacaroonType = Literal["access", "delete_pusher", "session"] def get_value_from_macaroon(macaroon: pymacaroons.Macaroon, key: str) -> str: @@ -111,19 +111,6 @@ class OidcSessionData: """The session ID of the ongoing UI Auth ("" if this is a login)""" -@attr.s(slots=True, frozen=True, auto_attribs=True) -class LoginTokenAttributes: - """Data we store in a short-term login token""" - - user_id: str - - auth_provider_id: str - """The SSO Identity Provider that the user authenticated with, to get this token.""" - - auth_provider_session_id: Optional[str] - """The session ID advertised by the SSO Identity Provider.""" - - class MacaroonGenerator: def __init__(self, clock: Clock, location: str, secret_key: bytes): self._clock = clock @@ -165,35 +152,6 @@ class MacaroonGenerator: macaroon.add_first_party_caveat(f"pushkey = {pushkey}") return macaroon.serialize() - def generate_short_term_login_token( - self, - user_id: str, - auth_provider_id: str, - auth_provider_session_id: Optional[str] = None, - duration_in_ms: int = (2 * 60 * 1000), - ) -> str: - """Generate a short-term login token used during SSO logins - - Args: - user_id: The user for which the token is valid. - auth_provider_id: The SSO IdP the user used. - auth_provider_session_id: The session ID got during login from the SSO IdP. - - Returns: - A signed token valid for using as a ``m.login.token`` token. - """ - now = self._clock.time_msec() - expiry = now + duration_in_ms - macaroon = self._generate_base_macaroon("login") - macaroon.add_first_party_caveat(f"user_id = {user_id}") - macaroon.add_first_party_caveat(f"time < {expiry}") - macaroon.add_first_party_caveat(f"auth_provider_id = {auth_provider_id}") - if auth_provider_session_id is not None: - macaroon.add_first_party_caveat( - f"auth_provider_session_id = {auth_provider_session_id}" - ) - return macaroon.serialize() - def generate_oidc_session_token( self, state: str, @@ -233,49 +191,6 @@ class MacaroonGenerator: return macaroon.serialize() - def verify_short_term_login_token(self, token: str) -> LoginTokenAttributes: - """Verify a short-term-login macaroon - - Checks that the given token is a valid, unexpired short-term-login token - minted by this server. - - Args: - token: The login token to verify. - - Returns: - A set of attributes carried by this token, including the - ``user_id`` and informations about the SSO IDP used during that - login. - - Raises: - MacaroonVerificationFailedException if the verification failed - """ - macaroon = pymacaroons.Macaroon.deserialize(token) - - v = self._base_verifier("login") - v.satisfy_general(lambda c: c.startswith("user_id = ")) - v.satisfy_general(lambda c: c.startswith("auth_provider_id = ")) - v.satisfy_general(lambda c: c.startswith("auth_provider_session_id = ")) - satisfy_expiry(v, self._clock.time_msec) - v.verify(macaroon, self._secret_key) - - user_id = get_value_from_macaroon(macaroon, "user_id") - auth_provider_id = get_value_from_macaroon(macaroon, "auth_provider_id") - - auth_provider_session_id: Optional[str] = None - try: - auth_provider_session_id = get_value_from_macaroon( - macaroon, "auth_provider_session_id" - ) - except MacaroonVerificationFailedException: - pass - - return LoginTokenAttributes( - user_id=user_id, - auth_provider_id=auth_provider_id, - auth_provider_session_id=auth_provider_session_id, - ) - def verify_guest_token(self, token: str) -> str: """Verify a guest access token macaroon diff --git a/tests/handlers/test_auth.py b/tests/handlers/test_auth.py index 7106799d44..036dbbc45b 100644 --- a/tests/handlers/test_auth.py +++ b/tests/handlers/test_auth.py @@ -11,6 +11,7 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. +from typing import Optional from unittest.mock import Mock import pymacaroons @@ -19,6 +20,7 @@ from twisted.test.proto_helpers import MemoryReactor from synapse.api.errors import AuthError, ResourceLimitError from synapse.rest import admin +from synapse.rest.client import login from synapse.server import HomeServer from synapse.util import Clock @@ -29,6 +31,7 @@ from tests.test_utils import make_awaitable class AuthTestCase(unittest.HomeserverTestCase): servlets = [ admin.register_servlets, + login.register_servlets, ] def prepare(self, reactor: MemoryReactor, clock: Clock, hs: HomeServer) -> None: @@ -46,6 +49,23 @@ class AuthTestCase(unittest.HomeserverTestCase): self.user1 = self.register_user("a_user", "pass") + def token_login(self, token: str) -> Optional[str]: + body = { + "type": "m.login.token", + "token": token, + } + + channel = self.make_request( + "POST", + "/_matrix/client/v3/login", + body, + ) + + if channel.code == 200: + return channel.json_body["user_id"] + + return None + def test_macaroon_caveats(self) -> None: token = self.macaroon_generator.generate_guest_access_token("a_user") macaroon = pymacaroons.Macaroon.deserialize(token) @@ -73,49 +93,62 @@ class AuthTestCase(unittest.HomeserverTestCase): v.satisfy_general(verify_guest) v.verify(macaroon, self.hs.config.key.macaroon_secret_key) - def test_short_term_login_token_gives_user_id(self) -> None: - token = self.macaroon_generator.generate_short_term_login_token( - self.user1, "", duration_in_ms=5000 + def test_login_token_gives_user_id(self) -> None: + token = self.get_success( + self.auth_handler.create_login_token_for_user_id( + self.user1, + duration_ms=(5 * 1000), + ) ) - res = self.get_success(self.auth_handler.validate_short_term_login_token(token)) + + res = self.get_success(self.auth_handler.consume_login_token(token)) self.assertEqual(self.user1, res.user_id) - self.assertEqual("", res.auth_provider_id) + self.assertEqual(None, res.auth_provider_id) - # when we advance the clock, the token should be rejected - self.reactor.advance(6) - self.get_failure( - self.auth_handler.validate_short_term_login_token(token), - AuthError, + def test_login_token_reuse_fails(self) -> None: + token = self.get_success( + self.auth_handler.create_login_token_for_user_id( + self.user1, + duration_ms=(5 * 1000), + ) ) - def test_short_term_login_token_gives_auth_provider(self) -> None: - token = self.macaroon_generator.generate_short_term_login_token( - self.user1, auth_provider_id="my_idp" - ) - res = self.get_success(self.auth_handler.validate_short_term_login_token(token)) - self.assertEqual(self.user1, res.user_id) - self.assertEqual("my_idp", res.auth_provider_id) + self.get_success(self.auth_handler.consume_login_token(token)) - def test_short_term_login_token_cannot_replace_user_id(self) -> None: - token = self.macaroon_generator.generate_short_term_login_token( - self.user1, "", duration_in_ms=5000 + self.get_failure( + self.auth_handler.consume_login_token(token), + AuthError, ) - macaroon = pymacaroons.Macaroon.deserialize(token) - res = self.get_success( - self.auth_handler.validate_short_term_login_token(macaroon.serialize()) + def test_login_token_expires(self) -> None: + token = self.get_success( + self.auth_handler.create_login_token_for_user_id( + self.user1, + duration_ms=(5 * 1000), + ) ) - self.assertEqual(self.user1, res.user_id) - - # add another "user_id" caveat, which might allow us to override the - # user_id. - macaroon.add_first_party_caveat("user_id = b_user") + # when we advance the clock, the token should be rejected + self.reactor.advance(6) self.get_failure( - self.auth_handler.validate_short_term_login_token(macaroon.serialize()), + self.auth_handler.consume_login_token(token), AuthError, ) + def test_login_token_gives_auth_provider(self) -> None: + token = self.get_success( + self.auth_handler.create_login_token_for_user_id( + self.user1, + auth_provider_id="my_idp", + auth_provider_session_id="11-22-33-44", + duration_ms=(5 * 1000), + ) + ) + res = self.get_success(self.auth_handler.consume_login_token(token)) + self.assertEqual(self.user1, res.user_id) + self.assertEqual("my_idp", res.auth_provider_id) + self.assertEqual("11-22-33-44", res.auth_provider_session_id) + def test_mau_limits_disabled(self) -> None: self.auth_blocking._limit_usage_by_mau = False # Ensure does not throw exception @@ -125,12 +158,12 @@ class AuthTestCase(unittest.HomeserverTestCase): ) ) - self.get_success( - self.auth_handler.validate_short_term_login_token( - self._get_macaroon().serialize() - ) + token = self.get_success( + self.auth_handler.create_login_token_for_user_id(self.user1) ) + self.assertIsNotNone(self.token_login(token)) + def test_mau_limits_exceeded_large(self) -> None: self.auth_blocking._limit_usage_by_mau = True self.hs.get_datastores().main.get_monthly_active_count = Mock( @@ -147,12 +180,10 @@ class AuthTestCase(unittest.HomeserverTestCase): self.hs.get_datastores().main.get_monthly_active_count = Mock( return_value=make_awaitable(self.large_number_of_users) ) - self.get_failure( - self.auth_handler.validate_short_term_login_token( - self._get_macaroon().serialize() - ), - ResourceLimitError, + token = self.get_success( + self.auth_handler.create_login_token_for_user_id(self.user1) ) + self.assertIsNone(self.token_login(token)) def test_mau_limits_parity(self) -> None: # Ensure we're not at the unix epoch. @@ -171,12 +202,10 @@ class AuthTestCase(unittest.HomeserverTestCase): ), ResourceLimitError, ) - self.get_failure( - self.auth_handler.validate_short_term_login_token( - self._get_macaroon().serialize() - ), - ResourceLimitError, + token = self.get_success( + self.auth_handler.create_login_token_for_user_id(self.user1) ) + self.assertIsNone(self.token_login(token)) # If in monthly active cohort self.hs.get_datastores().main.user_last_seen_monthly_active = Mock( @@ -187,11 +216,10 @@ class AuthTestCase(unittest.HomeserverTestCase): self.user1, device_id=None, valid_until_ms=None ) ) - self.get_success( - self.auth_handler.validate_short_term_login_token( - self._get_macaroon().serialize() - ) + token = self.get_success( + self.auth_handler.create_login_token_for_user_id(self.user1) ) + self.assertIsNotNone(self.token_login(token)) def test_mau_limits_not_exceeded(self) -> None: self.auth_blocking._limit_usage_by_mau = True @@ -209,14 +237,7 @@ class AuthTestCase(unittest.HomeserverTestCase): self.hs.get_datastores().main.get_monthly_active_count = Mock( return_value=make_awaitable(self.small_number_of_users) ) - self.get_success( - self.auth_handler.validate_short_term_login_token( - self._get_macaroon().serialize() - ) - ) - - def _get_macaroon(self) -> pymacaroons.Macaroon: - token = self.macaroon_generator.generate_short_term_login_token( - self.user1, "", duration_in_ms=5000 + token = self.get_success( + self.auth_handler.create_login_token_for_user_id(self.user1) ) - return pymacaroons.Macaroon.deserialize(token) + self.assertIsNotNone(self.token_login(token)) diff --git a/tests/util/test_macaroons.py b/tests/util/test_macaroons.py index 32125f7bb7..40754a4711 100644 --- a/tests/util/test_macaroons.py +++ b/tests/util/test_macaroons.py @@ -84,34 +84,6 @@ class MacaroonGeneratorTestCase(TestCase): ) self.assertEqual(user_id, "@user:tesths") - def test_short_term_login_token(self): - """Test the generation and verification of short-term login tokens""" - token = self.macaroon_generator.generate_short_term_login_token( - user_id="@user:tesths", - auth_provider_id="oidc", - auth_provider_session_id="sid", - duration_in_ms=2 * 60 * 1000, - ) - - info = self.macaroon_generator.verify_short_term_login_token(token) - self.assertEqual(info.user_id, "@user:tesths") - self.assertEqual(info.auth_provider_id, "oidc") - self.assertEqual(info.auth_provider_session_id, "sid") - - # Raises with another secret key - with self.assertRaises(MacaroonVerificationFailedException): - self.other_macaroon_generator.verify_short_term_login_token(token) - - # Wait a minute - self.reactor.pump([60]) - # Shouldn't raise - self.macaroon_generator.verify_short_term_login_token(token) - # Wait another minute - self.reactor.pump([60]) - # Should raise since it expired - with self.assertRaises(MacaroonVerificationFailedException): - self.macaroon_generator.verify_short_term_login_token(token) - def test_oidc_session_token(self): """Test the generation and verification of OIDC session cookies""" state = "arandomstate" -- cgit 1.5.1 From 0cfbb3513152b8360155c2d75df50e06ea861fa4 Mon Sep 17 00:00:00 2001 From: Ashish Kumar Date: Wed, 26 Oct 2022 18:51:23 +0400 Subject: fix broken avatar checks when server_name contains a port (#13927) Fixes check_avatar_size_and_mime_type() to successfully update avatars on homeservers running on non-default ports which it would mistakenly treat as remote homeserver while validating the avatar's size and mime type. Signed-off-by: Ashish Kumar ashfame@users.noreply.github.com --- changelog.d/13927.bugfix | 1 + synapse/handlers/profile.py | 6 +++++- tests/handlers/test_profile.py | 49 ++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 55 insertions(+), 1 deletion(-) create mode 100644 changelog.d/13927.bugfix (limited to 'synapse/handlers') diff --git a/changelog.d/13927.bugfix b/changelog.d/13927.bugfix new file mode 100644 index 0000000000..119cd128e7 --- /dev/null +++ b/changelog.d/13927.bugfix @@ -0,0 +1 @@ +Fix a bug which prevented setting an avatar on homeservers which have an explicit port in their `server_name` and have `max_avatar_size` and/or `allowed_avatar_mimetypes` configuration. Contributed by @ashfame. diff --git a/synapse/handlers/profile.py b/synapse/handlers/profile.py index d8ff5289b5..4bf9a047a3 100644 --- a/synapse/handlers/profile.py +++ b/synapse/handlers/profile.py @@ -307,7 +307,11 @@ class ProfileHandler: if not self.max_avatar_size and not self.allowed_avatar_mimetypes: return True - server_name, _, media_id = parse_and_validate_mxc_uri(mxc) + host, port, media_id = parse_and_validate_mxc_uri(mxc) + if port is not None: + server_name = host + ":" + str(port) + else: + server_name = host if server_name == self.server_name: media_info = await self.store.get_local_media(media_id) diff --git a/tests/handlers/test_profile.py b/tests/handlers/test_profile.py index f88c725a42..675aa023ac 100644 --- a/tests/handlers/test_profile.py +++ b/tests/handlers/test_profile.py @@ -14,6 +14,8 @@ from typing import Any, Awaitable, Callable, Dict from unittest.mock import Mock +from parameterized import parameterized + from twisted.test.proto_helpers import MemoryReactor import synapse.types @@ -327,6 +329,53 @@ class ProfileTestCase(unittest.HomeserverTestCase): ) self.assertFalse(res) + @unittest.override_config( + {"server_name": "test:8888", "allowed_avatar_mimetypes": ["image/png"]} + ) + def test_avatar_constraint_on_local_server_with_port(self): + """Test that avatar metadata is correctly fetched when the media is on a local + server and the server has an explicit port. + + (This was previously a bug) + """ + local_server_name = self.hs.config.server.server_name + media_id = "local" + local_mxc = f"mxc://{local_server_name}/{media_id}" + + # mock up the existence of the avatar file + self._setup_local_files({media_id: {"mimetype": "image/png"}}) + + # and now check that check_avatar_size_and_mime_type is happy + self.assertTrue( + self.get_success(self.handler.check_avatar_size_and_mime_type(local_mxc)) + ) + + @parameterized.expand([("remote",), ("remote:1234",)]) + @unittest.override_config({"allowed_avatar_mimetypes": ["image/png"]}) + def test_check_avatar_on_remote_server(self, remote_server_name: str) -> None: + """Test that avatar metadata is correctly fetched from a remote server""" + media_id = "remote" + remote_mxc = f"mxc://{remote_server_name}/{media_id}" + + # if the media is remote, check_avatar_size_and_mime_type just checks the + # media cache, so we don't need to instantiate a real remote server. It is + # sufficient to poke an entry into the db. + self.get_success( + self.hs.get_datastores().main.store_cached_remote_media( + media_id=media_id, + media_type="image/png", + media_length=50, + origin=remote_server_name, + time_now_ms=self.clock.time_msec(), + upload_name=None, + filesystem_id="xyz", + ) + ) + + self.assertTrue( + self.get_success(self.handler.check_avatar_size_and_mime_type(remote_mxc)) + ) + def _setup_local_files(self, names_and_props: Dict[str, Dict[str, Any]]): """Stores metadata about files in the database. -- cgit 1.5.1 From 40fa8294e3096132819287dd0c6d6bd71a408902 Mon Sep 17 00:00:00 2001 From: Eric Eastwood Date: Wed, 26 Oct 2022 16:10:55 -0500 Subject: Refactor MSC3030 `/timestamp_to_event` to move away from our snowflake pull from `destination` pattern (#14096) 1. `federation_client.timestamp_to_event(...)` now handles all `destination` looping and uses our generic `_try_destination_list(...)` helper. 2. Consistently handling `NotRetryingDestination` and `FederationDeniedError` across `get_pdu` , backfill, and the generic `_try_destination_list` which is used for many places we use this pattern. 3. `get_pdu(...)` now returns `PulledPduInfo` so we know which `destination` we ended up pulling the PDU from --- changelog.d/14096.misc | 1 + synapse/federation/federation_client.py | 130 ++++++++++++++++++++++++----- synapse/handlers/federation.py | 15 ++-- synapse/handlers/federation_event.py | 31 ++++--- synapse/handlers/room.py | 126 +++++++++++----------------- synapse/util/retryutils.py | 2 +- tests/federation/test_federation_client.py | 12 ++- 7 files changed, 191 insertions(+), 126 deletions(-) create mode 100644 changelog.d/14096.misc (limited to 'synapse/handlers') diff --git a/changelog.d/14096.misc b/changelog.d/14096.misc new file mode 100644 index 0000000000..2c07dc673b --- /dev/null +++ b/changelog.d/14096.misc @@ -0,0 +1 @@ +Refactor [MSC3030](https://github.com/matrix-org/matrix-spec-proposals/pull/3030) `/timestamp_to_event` endpoint to loop over federation destinations with standard pattern and error handling. diff --git a/synapse/federation/federation_client.py b/synapse/federation/federation_client.py index b220ab43fc..fa225182be 100644 --- a/synapse/federation/federation_client.py +++ b/synapse/federation/federation_client.py @@ -80,6 +80,18 @@ PDU_RETRY_TIME_MS = 1 * 60 * 1000 T = TypeVar("T") +@attr.s(frozen=True, slots=True, auto_attribs=True) +class PulledPduInfo: + """ + A result object that stores the PDU and info about it like which homeserver we + pulled it from (`pull_origin`) + """ + + pdu: EventBase + # Which homeserver we pulled the PDU from + pull_origin: str + + class InvalidResponseError(RuntimeError): """Helper for _try_destination_list: indicates that the server returned a response we couldn't parse @@ -114,7 +126,9 @@ class FederationClient(FederationBase): self.hostname = hs.hostname self.signing_key = hs.signing_key - self._get_pdu_cache: ExpiringCache[str, EventBase] = ExpiringCache( + # Cache mapping `event_id` to a tuple of the event itself and the `pull_origin` + # (which server we pulled the event from) + self._get_pdu_cache: ExpiringCache[str, Tuple[EventBase, str]] = ExpiringCache( cache_name="get_pdu_cache", clock=self._clock, max_len=1000, @@ -352,11 +366,11 @@ class FederationClient(FederationBase): @tag_args async def get_pdu( self, - destinations: Iterable[str], + destinations: Collection[str], event_id: str, room_version: RoomVersion, timeout: Optional[int] = None, - ) -> Optional[EventBase]: + ) -> Optional[PulledPduInfo]: """Requests the PDU with given origin and ID from the remote home servers. @@ -371,11 +385,11 @@ class FederationClient(FederationBase): moving to the next destination. None indicates no timeout. Returns: - The requested PDU, or None if we were unable to find it. + The requested PDU wrapped in `PulledPduInfo`, or None if we were unable to find it. """ logger.debug( - "get_pdu: event_id=%s from destinations=%s", event_id, destinations + "get_pdu(event_id=%s): from destinations=%s", event_id, destinations ) # TODO: Rate limit the number of times we try and get the same event. @@ -384,19 +398,25 @@ class FederationClient(FederationBase): # it gets persisted to the database), so we cache the results of the lookup. # Note that this is separate to the regular get_event cache which caches # events once they have been persisted. - event = self._get_pdu_cache.get(event_id) + get_pdu_cache_entry = self._get_pdu_cache.get(event_id) + event = None + pull_origin = None + if get_pdu_cache_entry: + event, pull_origin = get_pdu_cache_entry # If we don't see the event in the cache, go try to fetch it from the # provided remote federated destinations - if not event: + else: pdu_attempts = self.pdu_destination_tried.setdefault(event_id, {}) + # TODO: We can probably refactor this to use `_try_destination_list` for destination in destinations: now = self._clock.time_msec() last_attempt = pdu_attempts.get(destination, 0) if last_attempt + PDU_RETRY_TIME_MS > now: logger.debug( - "get_pdu: skipping destination=%s because we tried it recently last_attempt=%s and we only check every %s (now=%s)", + "get_pdu(event_id=%s): skipping destination=%s because we tried it recently last_attempt=%s and we only check every %s (now=%s)", + event_id, destination, last_attempt, PDU_RETRY_TIME_MS, @@ -411,43 +431,48 @@ class FederationClient(FederationBase): room_version=room_version, timeout=timeout, ) + pull_origin = destination pdu_attempts[destination] = now if event: # Prime the cache - self._get_pdu_cache[event.event_id] = event + self._get_pdu_cache[event.event_id] = (event, pull_origin) # Now that we have an event, we can break out of this # loop and stop asking other destinations. break + except NotRetryingDestination as e: + logger.info("get_pdu(event_id=%s): %s", event_id, e) + continue + except FederationDeniedError: + logger.info( + "get_pdu(event_id=%s): Not attempting to fetch PDU from %s because the homeserver is not on our federation whitelist", + event_id, + destination, + ) + continue except SynapseError as e: logger.info( - "Failed to get PDU %s from %s because %s", + "get_pdu(event_id=%s): Failed to get PDU from %s because %s", event_id, destination, e, ) continue - except NotRetryingDestination as e: - logger.info(str(e)) - continue - except FederationDeniedError as e: - logger.info(str(e)) - continue except Exception as e: pdu_attempts[destination] = now logger.info( - "Failed to get PDU %s from %s because %s", + "get_pdu(event_id=): Failed to get PDU from %s because %s", event_id, destination, e, ) continue - if not event: + if not event or not pull_origin: return None # `event` now refers to an object stored in `get_pdu_cache`. Our @@ -459,7 +484,7 @@ class FederationClient(FederationBase): event.room_version, ) - return event_copy + return PulledPduInfo(event_copy, pull_origin) @trace @tag_args @@ -699,12 +724,14 @@ class FederationClient(FederationBase): pdu_origin = get_domain_from_id(pdu.sender) if not res and pdu_origin != origin: try: - res = await self.get_pdu( + pulled_pdu_info = await self.get_pdu( destinations=[pdu_origin], event_id=pdu.event_id, room_version=room_version, timeout=10000, ) + if pulled_pdu_info is not None: + res = pulled_pdu_info.pdu except SynapseError: pass @@ -806,6 +833,7 @@ class FederationClient(FederationBase): ) for destination in destinations: + # We don't want to ask our own server for information we don't have if destination == self.server_name: continue @@ -814,9 +842,21 @@ class FederationClient(FederationBase): except ( RequestSendFailed, InvalidResponseError, - NotRetryingDestination, ) as e: logger.warning("Failed to %s via %s: %s", description, destination, e) + # Skip to the next homeserver in the list to try. + continue + except NotRetryingDestination as e: + logger.info("%s: %s", description, e) + continue + except FederationDeniedError: + logger.info( + "%s: Not attempting to %s from %s because the homeserver is not on our federation whitelist", + description, + description, + destination, + ) + continue except UnsupportedRoomVersionError: raise except HttpResponseException as e: @@ -1609,6 +1649,54 @@ class FederationClient(FederationBase): return result async def timestamp_to_event( + self, *, destinations: List[str], room_id: str, timestamp: int, direction: str + ) -> Optional["TimestampToEventResponse"]: + """ + Calls each remote federating server from `destinations` asking for their closest + event to the given timestamp in the given direction until we get a response. + Also validates the response to always return the expected keys or raises an + error. + + Args: + destinations: The domains of homeservers to try fetching from + room_id: Room to fetch the event from + timestamp: The point in time (inclusive) we should navigate from in + the given direction to find the closest event. + direction: ["f"|"b"] to indicate whether we should navigate forward + or backward from the given timestamp to find the closest event. + + Returns: + A parsed TimestampToEventResponse including the closest event_id + and origin_server_ts or None if no destination has a response. + """ + + async def _timestamp_to_event_from_destination( + destination: str, + ) -> TimestampToEventResponse: + return await self._timestamp_to_event_from_destination( + destination, room_id, timestamp, direction + ) + + try: + # Loop through each homeserver candidate until we get a succesful response + timestamp_to_event_response = await self._try_destination_list( + "timestamp_to_event", + destinations, + # TODO: The requested timestamp may lie in a part of the + # event graph that the remote server *also* didn't have, + # in which case they will have returned another event + # which may be nowhere near the requested timestamp. In + # the future, we may need to reconcile that gap and ask + # other homeservers, and/or extend `/timestamp_to_event` + # to return events on *both* sides of the timestamp to + # help reconcile the gap faster. + _timestamp_to_event_from_destination, + ) + return timestamp_to_event_response + except SynapseError: + return None + + async def _timestamp_to_event_from_destination( self, destination: str, room_id: str, timestamp: int, direction: str ) -> "TimestampToEventResponse": """ diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index 4fbc79a6cb..5fc3b8bc8c 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -442,6 +442,15 @@ class FederationHandler: # appropriate stuff. # TODO: We can probably do something more intelligent here. return True + except NotRetryingDestination as e: + logger.info("_maybe_backfill_inner: %s", e) + continue + except FederationDeniedError: + logger.info( + "_maybe_backfill_inner: Not attempting to backfill from %s because the homeserver is not on our federation whitelist", + dom, + ) + continue except (SynapseError, InvalidResponseError) as e: logger.info("Failed to backfill from %s because %s", dom, e) continue @@ -477,15 +486,9 @@ class FederationHandler: logger.info("Failed to backfill from %s because %s", dom, e) continue - except NotRetryingDestination as e: - logger.info(str(e)) - continue except RequestSendFailed as e: logger.info("Failed to get backfill from %s because %s", dom, e) continue - except FederationDeniedError as e: - logger.info(e) - continue except Exception as e: logger.exception("Failed to backfill from %s because %s", dom, e) continue diff --git a/synapse/handlers/federation_event.py b/synapse/handlers/federation_event.py index 7da6316a82..9ca5df7c78 100644 --- a/synapse/handlers/federation_event.py +++ b/synapse/handlers/federation_event.py @@ -58,7 +58,7 @@ from synapse.event_auth import ( ) from synapse.events import EventBase from synapse.events.snapshot import EventContext -from synapse.federation.federation_client import InvalidResponseError +from synapse.federation.federation_client import InvalidResponseError, PulledPduInfo from synapse.logging.context import nested_logging_context from synapse.logging.opentracing import ( SynapseTags, @@ -1517,8 +1517,8 @@ class FederationEventHandler: ) async def backfill_event_id( - self, destination: str, room_id: str, event_id: str - ) -> EventBase: + self, destinations: List[str], room_id: str, event_id: str + ) -> PulledPduInfo: """Backfill a single event and persist it as a non-outlier which means we also pull in all of the state and auth events necessary for it. @@ -1530,24 +1530,21 @@ class FederationEventHandler: Raises: FederationError if we are unable to find the event from the destination """ - logger.info( - "backfill_event_id: event_id=%s from destination=%s", event_id, destination - ) + logger.info("backfill_event_id: event_id=%s", event_id) room_version = await self._store.get_room_version(room_id) - event_from_response = await self._federation_client.get_pdu( - [destination], + pulled_pdu_info = await self._federation_client.get_pdu( + destinations, event_id, room_version, ) - if not event_from_response: + if not pulled_pdu_info: raise FederationError( "ERROR", 404, - "Unable to find event_id=%s from destination=%s to backfill." - % (event_id, destination), + f"Unable to find event_id={event_id} from remote servers to backfill.", affected=event_id, ) @@ -1555,13 +1552,13 @@ class FederationEventHandler: # and auth events to de-outlier it. This also sets up the necessary # `state_groups` for the event. await self._process_pulled_events( - destination, - [event_from_response], + pulled_pdu_info.pull_origin, + [pulled_pdu_info.pdu], # Prevent notifications going to clients backfilled=True, ) - return event_from_response + return pulled_pdu_info @trace @tag_args @@ -1584,19 +1581,19 @@ class FederationEventHandler: async def get_event(event_id: str) -> None: with nested_logging_context(event_id): try: - event = await self._federation_client.get_pdu( + pulled_pdu_info = await self._federation_client.get_pdu( [destination], event_id, room_version, ) - if event is None: + if pulled_pdu_info is None: logger.warning( "Server %s didn't return event %s", destination, event_id, ) return - events.append(event) + events.append(pulled_pdu_info.pdu) except Exception as e: logger.warning( diff --git a/synapse/handlers/room.py b/synapse/handlers/room.py index cc1e5c8f97..de97886ea9 100644 --- a/synapse/handlers/room.py +++ b/synapse/handlers/room.py @@ -49,7 +49,6 @@ from synapse.api.constants import ( from synapse.api.errors import ( AuthError, Codes, - HttpResponseException, LimitExceededError, NotFoundError, StoreError, @@ -60,7 +59,6 @@ from synapse.api.room_versions import KNOWN_ROOM_VERSIONS, RoomVersion from synapse.event_auth import validate_event_for_room_version from synapse.events import EventBase from synapse.events.utils import copy_and_fixup_power_levels_contents -from synapse.federation.federation_client import InvalidResponseError from synapse.handlers.relations import BundledAggregations from synapse.module_api import NOT_SPAM from synapse.rest.admin._base import assert_user_is_admin @@ -1472,7 +1470,12 @@ class TimestampLookupHandler: Raises: SynapseError if unable to find any event locally in the given direction """ - + logger.debug( + "get_event_for_timestamp(room_id=%s, timestamp=%s, direction=%s) Finding closest event...", + room_id, + timestamp, + direction, + ) local_event_id = await self.store.get_event_id_for_timestamp( room_id, timestamp, direction ) @@ -1524,85 +1527,54 @@ class TimestampLookupHandler: ) ) - # Loop through each homeserver candidate until we get a succesful response - for domain in likely_domains: - # We don't want to ask our own server for information we don't have - if domain == self.server_name: - continue + remote_response = await self.federation_client.timestamp_to_event( + destinations=likely_domains, + room_id=room_id, + timestamp=timestamp, + direction=direction, + ) + if remote_response is not None: + logger.debug( + "get_event_for_timestamp: remote_response=%s", + remote_response, + ) - try: - remote_response = await self.federation_client.timestamp_to_event( - domain, room_id, timestamp, direction - ) - logger.debug( - "get_event_for_timestamp: response from domain(%s)=%s", - domain, - remote_response, - ) + remote_event_id = remote_response.event_id + remote_origin_server_ts = remote_response.origin_server_ts - remote_event_id = remote_response.event_id - remote_origin_server_ts = remote_response.origin_server_ts - - # Backfill this event so we can get a pagination token for - # it with `/context` and paginate `/messages` from this - # point. - # - # TODO: The requested timestamp may lie in a part of the - # event graph that the remote server *also* didn't have, - # in which case they will have returned another event - # which may be nowhere near the requested timestamp. In - # the future, we may need to reconcile that gap and ask - # other homeservers, and/or extend `/timestamp_to_event` - # to return events on *both* sides of the timestamp to - # help reconcile the gap faster. - remote_event = ( - await self.federation_event_handler.backfill_event_id( - domain, room_id, remote_event_id - ) - ) + # Backfill this event so we can get a pagination token for + # it with `/context` and paginate `/messages` from this + # point. + pulled_pdu_info = await self.federation_event_handler.backfill_event_id( + likely_domains, room_id, remote_event_id + ) + remote_event = pulled_pdu_info.pdu - # XXX: When we see that the remote server is not trustworthy, - # maybe we should not ask them first in the future. - if remote_origin_server_ts != remote_event.origin_server_ts: - logger.info( - "get_event_for_timestamp: Remote server (%s) claimed that remote_event_id=%s occured at remote_origin_server_ts=%s but that isn't true (actually occured at %s). Their claims are dubious and we should consider not trusting them.", - domain, - remote_event_id, - remote_origin_server_ts, - remote_event.origin_server_ts, - ) - - # Only return the remote event if it's closer than the local event - if not local_event or ( - abs(remote_event.origin_server_ts - timestamp) - < abs(local_event.origin_server_ts - timestamp) - ): - logger.info( - "get_event_for_timestamp: returning remote_event_id=%s (%s) since it's closer to timestamp=%s than local_event=%s (%s)", - remote_event_id, - remote_event.origin_server_ts, - timestamp, - local_event.event_id if local_event else None, - local_event.origin_server_ts if local_event else None, - ) - return remote_event_id, remote_origin_server_ts - except (HttpResponseException, InvalidResponseError) as ex: - # Let's not put a high priority on some other homeserver - # failing to respond or giving a random response - logger.debug( - "get_event_for_timestamp: Failed to fetch /timestamp_to_event from %s because of exception(%s) %s args=%s", - domain, - type(ex).__name__, - ex, - ex.args, + # XXX: When we see that the remote server is not trustworthy, + # maybe we should not ask them first in the future. + if remote_origin_server_ts != remote_event.origin_server_ts: + logger.info( + "get_event_for_timestamp: Remote server (%s) claimed that remote_event_id=%s occured at remote_origin_server_ts=%s but that isn't true (actually occured at %s). Their claims are dubious and we should consider not trusting them.", + pulled_pdu_info.pull_origin, + remote_event_id, + remote_origin_server_ts, + remote_event.origin_server_ts, ) - except Exception: - # But we do want to see some exceptions in our code - logger.warning( - "get_event_for_timestamp: Failed to fetch /timestamp_to_event from %s because of exception", - domain, - exc_info=True, + + # Only return the remote event if it's closer than the local event + if not local_event or ( + abs(remote_event.origin_server_ts - timestamp) + < abs(local_event.origin_server_ts - timestamp) + ): + logger.info( + "get_event_for_timestamp: returning remote_event_id=%s (%s) since it's closer to timestamp=%s than local_event=%s (%s)", + remote_event_id, + remote_event.origin_server_ts, + timestamp, + local_event.event_id if local_event else None, + local_event.origin_server_ts if local_event else None, ) + return remote_event_id, remote_origin_server_ts # To appease mypy, we have to add both of these conditions to check for # `None`. We only expect `local_event` to be `None` when diff --git a/synapse/util/retryutils.py b/synapse/util/retryutils.py index d0a69ff843..dcc037b982 100644 --- a/synapse/util/retryutils.py +++ b/synapse/util/retryutils.py @@ -51,7 +51,7 @@ class NotRetryingDestination(Exception): destination: the domain in question """ - msg = "Not retrying server %s." % (destination,) + msg = f"Not retrying server {destination} because we tried it recently retry_last_ts={retry_last_ts} and we won't check for another retry_interval={retry_interval}ms." super().__init__(msg) self.retry_last_ts = retry_last_ts diff --git a/tests/federation/test_federation_client.py b/tests/federation/test_federation_client.py index 51d3bb8fff..e67f405826 100644 --- a/tests/federation/test_federation_client.py +++ b/tests/federation/test_federation_client.py @@ -142,14 +142,14 @@ class FederationClientTest(FederatingHomeserverTestCase): def test_get_pdu_returns_nothing_when_event_does_not_exist(self): """No event should be returned when the event does not exist""" - remote_pdu = self.get_success( + pulled_pdu_info = self.get_success( self.hs.get_federation_client().get_pdu( ["yet.another.server"], "event_should_not_exist", RoomVersions.V9, ) ) - self.assertEqual(remote_pdu, None) + self.assertEqual(pulled_pdu_info, None) def test_get_pdu(self): """Test to make sure an event is returned by `get_pdu()`""" @@ -169,13 +169,15 @@ class FederationClientTest(FederatingHomeserverTestCase): remote_pdu.internal_metadata.outlier = True # Get the event again. This time it should read it from cache. - remote_pdu2 = self.get_success( + pulled_pdu_info2 = self.get_success( self.hs.get_federation_client().get_pdu( ["yet.another.server"], remote_pdu.event_id, RoomVersions.V9, ) ) + self.assertIsNotNone(pulled_pdu_info2) + remote_pdu2 = pulled_pdu_info2.pdu # Sanity check that we are working against the same event self.assertEqual(remote_pdu.event_id, remote_pdu2.event_id) @@ -215,13 +217,15 @@ class FederationClientTest(FederatingHomeserverTestCase): ) ) - remote_pdu = self.get_success( + pulled_pdu_info = self.get_success( self.hs.get_federation_client().get_pdu( ["yet.another.server"], "event_id", RoomVersions.V9, ) ) + self.assertIsNotNone(pulled_pdu_info) + remote_pdu = pulled_pdu_info.pdu # check the right call got made to the agent self._mock_agent.request.assert_called_once_with( -- cgit 1.5.1 From 6a6e1e8c0711939338f25d8d41d1e4d33d984949 Mon Sep 17 00:00:00 2001 From: reivilibre Date: Fri, 28 Oct 2022 10:53:34 +0000 Subject: Fix room creation being rate limited too aggressively since Synapse v1.69.0. (#14314) * Introduce a test for the old behaviour which we want to restore * Reintroduce the old behaviour in a simpler way * Newsfile Signed-off-by: Olivier Wilkinson (reivilibre) * Use 1 credit instead of 2 for creating a room: be more lenient than before Notably, the UI in Element Web was still broken after restoring to prior behaviour. After discussion, we agreed that it would be sensible to increase the limit. Signed-off-by: Olivier Wilkinson (reivilibre) --- changelog.d/14314.bugfix | 1 + synapse/api/ratelimiting.py | 8 +++++- synapse/handlers/room.py | 16 ++++++++---- tests/rest/client/test_rooms.py | 54 ++++++++++++++++++++++++++++++++++++++--- 4 files changed, 70 insertions(+), 9 deletions(-) create mode 100644 changelog.d/14314.bugfix (limited to 'synapse/handlers') diff --git a/changelog.d/14314.bugfix b/changelog.d/14314.bugfix new file mode 100644 index 0000000000..8be47ee083 --- /dev/null +++ b/changelog.d/14314.bugfix @@ -0,0 +1 @@ +Fix room creation being rate limited too aggressively since Synapse v1.69.0. \ No newline at end of file diff --git a/synapse/api/ratelimiting.py b/synapse/api/ratelimiting.py index 044c7d4926..511790c7c5 100644 --- a/synapse/api/ratelimiting.py +++ b/synapse/api/ratelimiting.py @@ -343,6 +343,7 @@ class RequestRatelimiter: requester: Requester, update: bool = True, is_admin_redaction: bool = False, + n_actions: int = 1, ) -> None: """Ratelimits requests. @@ -355,6 +356,8 @@ class RequestRatelimiter: is_admin_redaction: Whether this is a room admin/moderator redacting an event. If so then we may apply different ratelimits depending on config. + n_actions: Multiplier for the number of actions to apply to the + rate limiter at once. Raises: LimitExceededError if the request should be ratelimited @@ -383,7 +386,9 @@ class RequestRatelimiter: if is_admin_redaction and self.admin_redaction_ratelimiter: # If we have separate config for admin redactions, use a separate # ratelimiter as to not have user_ids clash - await self.admin_redaction_ratelimiter.ratelimit(requester, update=update) + await self.admin_redaction_ratelimiter.ratelimit( + requester, update=update, n_actions=n_actions + ) else: # Override rate and burst count per-user await self.request_ratelimiter.ratelimit( @@ -391,4 +396,5 @@ class RequestRatelimiter: rate_hz=messages_per_second, burst_count=burst_count, update=update, + n_actions=n_actions, ) diff --git a/synapse/handlers/room.py b/synapse/handlers/room.py index 638f54051a..d74b675adc 100644 --- a/synapse/handlers/room.py +++ b/synapse/handlers/room.py @@ -559,7 +559,6 @@ class RoomCreationHandler: invite_list=[], initial_state=initial_state, creation_content=creation_content, - ratelimit=False, ) # Transfer membership events @@ -753,6 +752,10 @@ class RoomCreationHandler: ) if ratelimit: + # Rate limit once in advance, but don't rate limit the individual + # events in the room — room creation isn't atomic and it's very + # janky if half the events in the initial state don't make it because + # of rate limiting. await self.request_ratelimiter.ratelimit(requester) room_version_id = config.get( @@ -913,7 +916,6 @@ class RoomCreationHandler: room_alias=room_alias, power_level_content_override=power_level_content_override, creator_join_profile=creator_join_profile, - ratelimit=ratelimit, ) if "name" in config: @@ -1037,7 +1039,6 @@ class RoomCreationHandler: room_alias: Optional[RoomAlias] = None, power_level_content_override: Optional[JsonDict] = None, creator_join_profile: Optional[JsonDict] = None, - ratelimit: bool = True, ) -> Tuple[int, str, int]: """Sends the initial events into a new room. Sends the room creation, membership, and power level events into the room sequentially, then creates and batches up the @@ -1046,6 +1047,8 @@ class RoomCreationHandler: `power_level_content_override` doesn't apply when initial state has power level state event content. + Rate limiting should already have been applied by this point. + Returns: A tuple containing the stream ID, event ID and depth of the last event sent to the room. @@ -1144,7 +1147,7 @@ class RoomCreationHandler: creator.user, room_id, "join", - ratelimit=ratelimit, + ratelimit=False, content=creator_join_profile, new_room=True, prev_event_ids=[last_sent_event_id], @@ -1269,7 +1272,10 @@ class RoomCreationHandler: events_to_send.append((encryption_event, encryption_context)) last_event = await self.event_creation_handler.handle_new_client_event( - creator, events_to_send, ignore_shadow_ban=True + creator, + events_to_send, + ignore_shadow_ban=True, + ratelimit=False, ) assert last_event.internal_metadata.stream_ordering is not None return last_event.internal_metadata.stream_ordering, last_event.event_id, depth diff --git a/tests/rest/client/test_rooms.py b/tests/rest/client/test_rooms.py index 716366eb90..1084d4ad9d 100644 --- a/tests/rest/client/test_rooms.py +++ b/tests/rest/client/test_rooms.py @@ -54,6 +54,7 @@ from tests.http.server._base import make_request_with_cancellation_test from tests.storage.test_stream import PaginationTestCase from tests.test_utils import make_awaitable from tests.test_utils.event_injection import create_event +from tests.unittest import override_config PATH_PREFIX = b"/_matrix/client/api/v1" @@ -871,6 +872,41 @@ class RoomsCreateTestCase(RoomBase): self.assertEqual(channel.code, HTTPStatus.OK, channel.json_body) self.assertEqual(join_mock.call_count, 0) + def _create_basic_room(self) -> Tuple[int, object]: + """ + Tries to create a basic room and returns the response code. + """ + channel = self.make_request( + "POST", + "/createRoom", + {}, + ) + return channel.code, channel.json_body + + @override_config( + { + "rc_message": {"per_second": 0.2, "burst_count": 10}, + } + ) + def test_room_creation_ratelimiting(self) -> None: + """ + Regression test for #14312, where ratelimiting was made too strict. + Clients should be able to create 10 rooms in a row + without hitting rate limits, using default rate limit config. + (We override rate limiting config back to its default value.) + + To ensure we don't make ratelimiting too generous accidentally, + also check that we can't create an 11th room. + """ + + for _ in range(10): + code, json_body = self._create_basic_room() + self.assertEqual(code, HTTPStatus.OK, json_body) + + # The 6th room hits the rate limit. + code, json_body = self._create_basic_room() + self.assertEqual(code, HTTPStatus.TOO_MANY_REQUESTS, json_body) + class RoomTopicTestCase(RoomBase): """Tests /rooms/$room_id/topic REST events.""" @@ -1390,10 +1426,22 @@ class RoomJoinRatelimitTestCase(RoomBase): ) def test_join_local_ratelimit(self) -> None: """Tests that local joins are actually rate-limited.""" - for _ in range(3): - self.helper.create_room_as(self.user_id) + # Create 4 rooms + room_ids = [ + self.helper.create_room_as(self.user_id, is_public=True) for _ in range(4) + ] + + joiner_user_id = self.register_user("joiner", "secret") + # Now make a new user try to join some of them. - self.helper.create_room_as(self.user_id, expect_code=429) + # The user can join 3 rooms + for room_id in room_ids[0:3]: + self.helper.join(room_id, joiner_user_id) + + # But the user cannot join a 4th room + self.helper.join( + room_ids[3], joiner_user_id, expect_code=HTTPStatus.TOO_MANY_REQUESTS + ) @unittest.override_config( {"rc_joins": {"local": {"per_second": 0.5, "burst_count": 3}}} -- cgit 1.5.1 From 7911e2835df7b4bf1dec98b09da89beda65e2ab2 Mon Sep 17 00:00:00 2001 From: Andrew Morgan <1342360+anoadragon453@users.noreply.github.com> Date: Fri, 28 Oct 2022 18:06:02 +0100 Subject: Prevent federation user keys query from returning device names if disallowed (#14304) --- changelog.d/14304.bugfix | 1 + synapse/handlers/e2e_keys.py | 37 ++++++++++++++++++++--- synapse/storage/databases/main/end_to_end_keys.py | 17 ++++++++--- 3 files changed, 46 insertions(+), 9 deletions(-) create mode 100644 changelog.d/14304.bugfix (limited to 'synapse/handlers') diff --git a/changelog.d/14304.bugfix b/changelog.d/14304.bugfix new file mode 100644 index 0000000000..b8d4d91034 --- /dev/null +++ b/changelog.d/14304.bugfix @@ -0,0 +1 @@ +Fix a bug introduced in 1.34.0 where device names would be returned via a federation user key query request when `allow_device_name_lookup_over_federation` was set to `false`. \ No newline at end of file diff --git a/synapse/handlers/e2e_keys.py b/synapse/handlers/e2e_keys.py index 09a2492afc..a9912c467d 100644 --- a/synapse/handlers/e2e_keys.py +++ b/synapse/handlers/e2e_keys.py @@ -49,6 +49,7 @@ logger = logging.getLogger(__name__) class E2eKeysHandler: def __init__(self, hs: "HomeServer"): + self.config = hs.config self.store = hs.get_datastores().main self.federation = hs.get_federation_client() self.device_handler = hs.get_device_handler() @@ -431,13 +432,17 @@ class E2eKeysHandler: @trace @cancellable async def query_local_devices( - self, query: Mapping[str, Optional[List[str]]] + self, + query: Mapping[str, Optional[List[str]]], + include_displaynames: bool = True, ) -> Dict[str, Dict[str, dict]]: """Get E2E device keys for local users Args: query: map from user_id to a list of devices to query (None for all devices) + include_displaynames: Whether to include device displaynames in the returned + device details. Returns: A map from user_id -> device_id -> device details @@ -469,7 +474,9 @@ class E2eKeysHandler: # make sure that each queried user appears in the result dict result_dict[user_id] = {} - results = await self.store.get_e2e_device_keys_for_cs_api(local_query) + results = await self.store.get_e2e_device_keys_for_cs_api( + local_query, include_displaynames + ) # Build the result structure for user_id, device_keys in results.items(): @@ -482,11 +489,33 @@ class E2eKeysHandler: async def on_federation_query_client_keys( self, query_body: Dict[str, Dict[str, Optional[List[str]]]] ) -> JsonDict: - """Handle a device key query from a federated server""" + """Handle a device key query from a federated server: + + Handles the path: GET /_matrix/federation/v1/users/keys/query + + Args: + query_body: The body of the query request. Should contain a key + "device_keys" that map to a dictionary of user ID's -> list of + device IDs. If the list of device IDs is empty, all devices of + that user will be queried. + + Returns: + A json dictionary containing the following: + - device_keys: A dictionary containing the requested device information. + - master_keys: An optional dictionary of user ID -> master cross-signing + key info. + - self_signing_key: An optional dictionary of user ID -> self-signing + key info. + """ device_keys_query: Dict[str, Optional[List[str]]] = query_body.get( "device_keys", {} ) - res = await self.query_local_devices(device_keys_query) + res = await self.query_local_devices( + device_keys_query, + include_displaynames=( + self.config.federation.allow_device_name_lookup_over_federation + ), + ) ret = {"device_keys": res} # add in the cross-signing keys diff --git a/synapse/storage/databases/main/end_to_end_keys.py b/synapse/storage/databases/main/end_to_end_keys.py index 8a10ae800c..2a4f58ed92 100644 --- a/synapse/storage/databases/main/end_to_end_keys.py +++ b/synapse/storage/databases/main/end_to_end_keys.py @@ -139,11 +139,15 @@ class EndToEndKeyWorkerStore(EndToEndKeyBackgroundStore, CacheInvalidationWorker @trace @cancellable async def get_e2e_device_keys_for_cs_api( - self, query_list: List[Tuple[str, Optional[str]]] + self, + query_list: List[Tuple[str, Optional[str]]], + include_displaynames: bool = True, ) -> Dict[str, Dict[str, JsonDict]]: """Fetch a list of device keys, formatted suitably for the C/S API. Args: - query_list(list): List of pairs of user_ids and device_ids. + query_list: List of pairs of user_ids and device_ids. + include_displaynames: Whether to include the displayname of returned devices + (if one exists). Returns: Dict mapping from user-id to dict mapping from device_id to key data. The key data will be a dict in the same format as the @@ -166,9 +170,12 @@ class EndToEndKeyWorkerStore(EndToEndKeyBackgroundStore, CacheInvalidationWorker continue r["unsigned"] = {} - display_name = device_info.display_name - if display_name is not None: - r["unsigned"]["device_display_name"] = display_name + if include_displaynames: + # Include the device's display name in the "unsigned" dictionary + display_name = device_info.display_name + if display_name is not None: + r["unsigned"]["device_display_name"] = display_name + rv[user_id][device_id] = r return rv -- cgit 1.5.1 From cc3a52b33df72bb4230367536b924a6d1f510d36 Mon Sep 17 00:00:00 2001 From: Quentin Gliech Date: Mon, 31 Oct 2022 18:07:30 +0100 Subject: Support OIDC backchannel logouts (#11414) If configured an OIDC IdP can log a user's session out of Synapse when they log out of the identity provider. The IdP sends a request directly to Synapse (and must be configured with an endpoint) when a user logs out. --- changelog.d/11414.feature | 1 + docs/openid.md | 14 + docs/usage/configuration/config_documentation.md | 9 + synapse/config/oidc.py | 12 + synapse/handlers/oidc.py | 381 ++++++++++++++++++-- synapse/handlers/sso.py | 71 ++++ synapse/rest/synapse/client/oidc/__init__.py | 4 + .../client/oidc/backchannel_logout_resource.py | 35 ++ synapse/storage/databases/main/registration.py | 21 ++ tests/rest/client/test_auth.py | 390 +++++++++++++++++++-- tests/rest/client/utils.py | 55 ++- tests/server.py | 6 + tests/test_utils/oidc.py | 27 +- 13 files changed, 960 insertions(+), 66 deletions(-) create mode 100644 changelog.d/11414.feature create mode 100644 synapse/rest/synapse/client/oidc/backchannel_logout_resource.py (limited to 'synapse/handlers') diff --git a/changelog.d/11414.feature b/changelog.d/11414.feature new file mode 100644 index 0000000000..fc035e50a7 --- /dev/null +++ b/changelog.d/11414.feature @@ -0,0 +1 @@ +Support back-channel logouts from OpenID Connect providers. diff --git a/docs/openid.md b/docs/openid.md index 87ebea4c29..37c5eb244d 100644 --- a/docs/openid.md +++ b/docs/openid.md @@ -49,6 +49,13 @@ setting in your configuration file. See the [configuration manual](usage/configuration/config_documentation.md#oidc_providers) for some sample settings, as well as the text below for example configurations for specific providers. +## OIDC Back-Channel Logout + +Synapse supports receiving [OpenID Connect Back-Channel Logout](https://openid.net/specs/openid-connect-backchannel-1_0.html) notifications. + +This lets the OpenID Connect Provider notify Synapse when a user logs out, so that Synapse can end that user session. +This feature can be enabled by setting the `backchannel_logout_enabled` property to `true` in the provider configuration, and setting the following URL as destination for Back-Channel Logout notifications in your OpenID Connect Provider: `[synapse public baseurl]/_synapse/client/oidc/backchannel_logout` + ## Sample configs Here are a few configs for providers that should work with Synapse. @@ -123,6 +130,9 @@ oidc_providers: [Keycloak][keycloak-idp] is an opensource IdP maintained by Red Hat. +Keycloak supports OIDC Back-Channel Logout, which sends logout notification to Synapse, so that Synapse users get logged out when they log out from Keycloak. +This can be optionally enabled by setting `backchannel_logout_enabled` to `true` in the Synapse configuration, and by setting the "Backchannel Logout URL" in Keycloak. + Follow the [Getting Started Guide](https://www.keycloak.org/getting-started) to install Keycloak and set up a realm. 1. Click `Clients` in the sidebar and click `Create` @@ -144,6 +154,8 @@ Follow the [Getting Started Guide](https://www.keycloak.org/getting-started) to | Client Protocol | `openid-connect` | | Access Type | `confidential` | | Valid Redirect URIs | `[synapse public baseurl]/_synapse/client/oidc/callback` | +| Backchannel Logout URL (optional) | `[synapse public baseurl]/_synapse/client/oidc/backchannel_logout` | +| Backchannel Logout Session Required (optional) | `On` | 5. Click `Save` 6. On the Credentials tab, update the fields: @@ -167,7 +179,9 @@ oidc_providers: config: localpart_template: "{{ user.preferred_username }}" display_name_template: "{{ user.name }}" + backchannel_logout_enabled: true # Optional ``` + ### Auth0 [Auth0][auth0] is a hosted SaaS IdP solution. diff --git a/docs/usage/configuration/config_documentation.md b/docs/usage/configuration/config_documentation.md index 97fb505a5f..44358faf59 100644 --- a/docs/usage/configuration/config_documentation.md +++ b/docs/usage/configuration/config_documentation.md @@ -3021,6 +3021,15 @@ Options for each entry include: which is set to the claims returned by the UserInfo Endpoint and/or in the ID Token. +* `backchannel_logout_enabled`: set to `true` to process OIDC Back-Channel Logout notifications. + Those notifications are expected to be received on `/_synapse/client/oidc/backchannel_logout`. + Defaults to `false`. + +* `backchannel_logout_ignore_sub`: by default, the OIDC Back-Channel Logout feature checks that the + `sub` claim matches the subject claim received during login. This check can be disabled by setting + this to `true`. Defaults to `false`. + + You might want to disable this if the `subject_claim` returned by the mapping provider is not `sub`. It is possible to configure Synapse to only allow logins if certain attributes match particular values in the OIDC userinfo. The requirements can be listed under diff --git a/synapse/config/oidc.py b/synapse/config/oidc.py index 5418a332da..0bd83f4010 100644 --- a/synapse/config/oidc.py +++ b/synapse/config/oidc.py @@ -123,6 +123,8 @@ OIDC_PROVIDER_CONFIG_SCHEMA = { "userinfo_endpoint": {"type": "string"}, "jwks_uri": {"type": "string"}, "skip_verification": {"type": "boolean"}, + "backchannel_logout_enabled": {"type": "boolean"}, + "backchannel_logout_ignore_sub": {"type": "boolean"}, "user_profile_method": { "type": "string", "enum": ["auto", "userinfo_endpoint"], @@ -292,6 +294,10 @@ def _parse_oidc_config_dict( token_endpoint=oidc_config.get("token_endpoint"), userinfo_endpoint=oidc_config.get("userinfo_endpoint"), jwks_uri=oidc_config.get("jwks_uri"), + backchannel_logout_enabled=oidc_config.get("backchannel_logout_enabled", False), + backchannel_logout_ignore_sub=oidc_config.get( + "backchannel_logout_ignore_sub", False + ), skip_verification=oidc_config.get("skip_verification", False), user_profile_method=oidc_config.get("user_profile_method", "auto"), allow_existing_users=oidc_config.get("allow_existing_users", False), @@ -368,6 +374,12 @@ class OidcProviderConfig: # "openid" scope is used. jwks_uri: Optional[str] + # Whether Synapse should react to backchannel logouts + backchannel_logout_enabled: bool + + # Whether Synapse should ignore the `sub` claim in backchannel logouts or not. + backchannel_logout_ignore_sub: bool + # Whether to skip metadata verification skip_verification: bool diff --git a/synapse/handlers/oidc.py b/synapse/handlers/oidc.py index 9759daf043..867973dcca 100644 --- a/synapse/handlers/oidc.py +++ b/synapse/handlers/oidc.py @@ -12,14 +12,28 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. +import binascii import inspect +import json import logging -from typing import TYPE_CHECKING, Any, Dict, Generic, List, Optional, TypeVar, Union +from typing import ( + TYPE_CHECKING, + Any, + Dict, + Generic, + List, + Optional, + Type, + TypeVar, + Union, +) from urllib.parse import urlencode, urlparse import attr +import unpaddedbase64 from authlib.common.security import generate_token -from authlib.jose import JsonWebToken, jwt +from authlib.jose import JsonWebToken, JWTClaims +from authlib.jose.errors import InvalidClaimError, JoseError, MissingClaimError from authlib.oauth2.auth import ClientAuth from authlib.oauth2.rfc6749.parameters import prepare_grant_uri from authlib.oidc.core import CodeIDToken, UserInfo @@ -35,9 +49,12 @@ from typing_extensions import TypedDict from twisted.web.client import readBody from twisted.web.http_headers import Headers +from synapse.api.errors import SynapseError from synapse.config import ConfigError from synapse.config.oidc import OidcProviderClientSecretJwtKey, OidcProviderConfig from synapse.handlers.sso import MappingException, UserAttributes +from synapse.http.server import finish_request +from synapse.http.servlet import parse_string from synapse.http.site import SynapseRequest from synapse.logging.context import make_deferred_yieldable from synapse.types import JsonDict, UserID, map_username_to_mxid_localpart @@ -88,6 +105,8 @@ class Token(TypedDict): #: there is no real point of doing this in our case. JWK = Dict[str, str] +C = TypeVar("C") + #: A JWK Set, as per RFC7517 sec 5. class JWKS(TypedDict): @@ -247,6 +266,80 @@ class OidcHandler: await oidc_provider.handle_oidc_callback(request, session_data, code) + async def handle_backchannel_logout(self, request: SynapseRequest) -> None: + """Handle an incoming request to /_synapse/client/oidc/backchannel_logout + + This extracts the logout_token from the request and tries to figure out + which OpenID Provider it is comming from. This works by matching the iss claim + with the issuer and the aud claim with the client_id. + + Since at this point we don't know who signed the JWT, we can't just + decode it using authlib since it will always verifies the signature. We + have to decode it manually without validating the signature. The actual JWT + verification is done in the `OidcProvider.handler_backchannel_logout` method, + once we figured out which provider sent the request. + + Args: + request: the incoming request from the browser. + """ + logout_token = parse_string(request, "logout_token") + if logout_token is None: + raise SynapseError(400, "Missing logout_token in request") + + # A JWT looks like this: + # header.payload.signature + # where all parts are encoded with urlsafe base64. + # The aud and iss claims we care about are in the payload part, which + # is a JSON object. + try: + # By destructuring the list after splitting, we ensure that we have + # exactly 3 segments + _, payload, _ = logout_token.split(".") + except ValueError: + raise SynapseError(400, "Invalid logout_token in request") + + try: + payload_bytes = unpaddedbase64.decode_base64(payload) + claims = json_decoder.decode(payload_bytes.decode("utf-8")) + except (json.JSONDecodeError, binascii.Error, UnicodeError): + raise SynapseError(400, "Invalid logout_token payload in request") + + try: + # Let's extract the iss and aud claims + iss = claims["iss"] + aud = claims["aud"] + # The aud claim can be either a string or a list of string. Here we + # normalize it as a list of strings. + if isinstance(aud, str): + aud = [aud] + + # Check that we have the right types for the aud and the iss claims + if not isinstance(iss, str) or not isinstance(aud, list): + raise TypeError() + for a in aud: + if not isinstance(a, str): + raise TypeError() + + # At this point we properly checked both claims types + issuer: str = iss + audience: List[str] = aud + except (TypeError, KeyError): + raise SynapseError(400, "Invalid issuer/audience in logout_token") + + # Now that we know the audience and the issuer, we can figure out from + # what provider it is coming from + oidc_provider: Optional[OidcProvider] = None + for provider in self._providers.values(): + if provider.issuer == issuer and provider.client_id in audience: + oidc_provider = provider + break + + if oidc_provider is None: + raise SynapseError(400, "Could not find the OP that issued this event") + + # Ask the provider to handle the logout request. + await oidc_provider.handle_backchannel_logout(request, logout_token) + class OidcError(Exception): """Used to catch errors when calling the token_endpoint""" @@ -342,6 +435,7 @@ class OidcProvider: self.idp_brand = provider.idp_brand self._sso_handler = hs.get_sso_handler() + self._device_handler = hs.get_device_handler() self._sso_handler.register_identity_provider(self) @@ -400,6 +494,41 @@ class OidcProvider: # If we're not using userinfo, we need a valid jwks to validate the ID token m.validate_jwks_uri() + if self._config.backchannel_logout_enabled: + if not m.get("backchannel_logout_supported", False): + logger.warning( + "OIDC Back-Channel Logout is enabled for issuer %r" + "but it does not advertise support for it", + self.issuer, + ) + + elif not m.get("backchannel_logout_session_supported", False): + logger.warning( + "OIDC Back-Channel Logout is enabled and supported " + "by issuer %r but it might not send a session ID with " + "logout tokens, which is required for the logouts to work", + self.issuer, + ) + + if not self._config.backchannel_logout_ignore_sub: + # If OIDC backchannel logouts are enabled, the provider mapping provider + # should use the `sub` claim. We verify that by mapping a dumb user and + # see if we get back the sub claim + user = UserInfo({"sub": "thisisasubject"}) + try: + subject = self._user_mapping_provider.get_remote_user_id(user) + if subject != user["sub"]: + raise ValueError("Unexpected subject") + except Exception: + logger.warning( + f"OIDC Back-Channel Logout is enabled for issuer {self.issuer!r} " + "but it looks like the configured `user_mapping_provider` " + "does not use the `sub` claim as subject. If it is the case, " + "and you want Synapse to ignore the `sub` claim in OIDC " + "Back-Channel Logouts, set `backchannel_logout_ignore_sub` " + "to `true` in the issuer config." + ) + @property def _uses_userinfo(self) -> bool: """Returns True if the ``userinfo_endpoint`` should be used. @@ -415,6 +544,16 @@ class OidcProvider: or self._user_profile_method == "userinfo_endpoint" ) + @property + def issuer(self) -> str: + """The issuer identifying this provider.""" + return self._config.issuer + + @property + def client_id(self) -> str: + """The client_id used when interacting with this provider.""" + return self._config.client_id + async def load_metadata(self, force: bool = False) -> OpenIDProviderMetadata: """Return the provider metadata. @@ -662,6 +801,59 @@ class OidcProvider: return UserInfo(resp) + async def _verify_jwt( + self, + alg_values: List[str], + token: str, + claims_cls: Type[C], + claims_options: Optional[dict] = None, + claims_params: Optional[dict] = None, + ) -> C: + """Decode and validate a JWT, re-fetching the JWKS as needed. + + Args: + alg_values: list of `alg` values allowed when verifying the JWT. + token: the JWT. + claims_cls: the JWTClaims class to use to validate the claims. + claims_options: dict of options passed to the `claims_cls` constructor. + claims_params: dict of params passed to the `claims_cls` constructor. + + Returns: + The decoded claims in the JWT. + """ + jwt = JsonWebToken(alg_values) + + logger.debug("Attempting to decode JWT (%s) %r", claims_cls.__name__, token) + + # Try to decode the keys in cache first, then retry by forcing the keys + # to be reloaded + jwk_set = await self.load_jwks() + try: + claims = jwt.decode( + token, + key=jwk_set, + claims_cls=claims_cls, + claims_options=claims_options, + claims_params=claims_params, + ) + except ValueError: + logger.info("Reloading JWKS after decode error") + jwk_set = await self.load_jwks(force=True) # try reloading the jwks + claims = jwt.decode( + token, + key=jwk_set, + claims_cls=claims_cls, + claims_options=claims_options, + claims_params=claims_params, + ) + + logger.debug("Decoded JWT (%s) %r; validating", claims_cls.__name__, claims) + + claims.validate( + now=self._clock.time(), leeway=120 + ) # allows 2 min of clock skew + return claims + async def _parse_id_token(self, token: Token, nonce: str) -> CodeIDToken: """Return an instance of UserInfo from token's ``id_token``. @@ -675,13 +867,13 @@ class OidcProvider: The decoded claims in the ID token. """ id_token = token.get("id_token") - logger.debug("Attempting to decode JWT id_token %r", id_token) # That has been theoritically been checked by the caller, so even though # assertion are not enabled in production, it is mainly here to appease mypy assert id_token is not None metadata = await self.load_metadata() + claims_params = { "nonce": nonce, "client_id": self._client_auth.client_id, @@ -691,38 +883,17 @@ class OidcProvider: # in the `id_token` that we can check against. claims_params["access_token"] = token["access_token"] - alg_values = metadata.get("id_token_signing_alg_values_supported", ["RS256"]) - jwt = JsonWebToken(alg_values) - - claim_options = {"iss": {"values": [metadata["issuer"]]}} + claims_options = {"iss": {"values": [metadata["issuer"]]}} - # Try to decode the keys in cache first, then retry by forcing the keys - # to be reloaded - jwk_set = await self.load_jwks() - try: - claims = jwt.decode( - id_token, - key=jwk_set, - claims_cls=CodeIDToken, - claims_options=claim_options, - claims_params=claims_params, - ) - except ValueError: - logger.info("Reloading JWKS after decode error") - jwk_set = await self.load_jwks(force=True) # try reloading the jwks - claims = jwt.decode( - id_token, - key=jwk_set, - claims_cls=CodeIDToken, - claims_options=claim_options, - claims_params=claims_params, - ) - - logger.debug("Decoded id_token JWT %r; validating", claims) + alg_values = metadata.get("id_token_signing_alg_values_supported", ["RS256"]) - claims.validate( - now=self._clock.time(), leeway=120 - ) # allows 2 min of clock skew + claims = await self._verify_jwt( + alg_values=alg_values, + token=id_token, + claims_cls=CodeIDToken, + claims_options=claims_options, + claims_params=claims_params, + ) return claims @@ -1043,6 +1214,146 @@ class OidcProvider: # to be strings. return str(remote_user_id) + async def handle_backchannel_logout( + self, request: SynapseRequest, logout_token: str + ) -> None: + """Handle an incoming request to /_synapse/client/oidc/backchannel_logout + + The OIDC Provider posts a logout token to this endpoint when a user + session ends. That token is a JWT signed with the same keys as + ID tokens. The OpenID Connect Back-Channel Logout draft explains how to + validate the JWT and figure out what session to end. + + Args: + request: The request to respond to + logout_token: The logout token (a JWT) extracted from the request body + """ + # Back-Channel Logout can be disabled in the config, hence this check. + # This is not that important for now since Synapse is registered + # manually to the OP, so not specifying the backchannel-logout URI is + # as effective than disabling it here. It might make more sense if we + # support dynamic registration in Synapse at some point. + if not self._config.backchannel_logout_enabled: + logger.warning( + f"Received an OIDC Back-Channel Logout request from issuer {self.issuer!r} but it is disabled in config" + ) + + # TODO: this responds with a 400 status code, which is what the OIDC + # Back-Channel Logout spec expects, but spec also suggests answering with + # a JSON object, with the `error` and `error_description` fields set, which + # we are not doing here. + # See https://openid.net/specs/openid-connect-backchannel-1_0.html#BCResponse + raise SynapseError( + 400, "OpenID Connect Back-Channel Logout is disabled for this provider" + ) + + metadata = await self.load_metadata() + + # As per OIDC Back-Channel Logout 1.0 sec. 2.4: + # A Logout Token MUST be signed and MAY also be encrypted. The same + # keys are used to sign and encrypt Logout Tokens as are used for ID + # Tokens. If the Logout Token is encrypted, it SHOULD replicate the + # iss (issuer) claim in the JWT Header Parameters, as specified in + # Section 5.3 of [JWT]. + alg_values = metadata.get("id_token_signing_alg_values_supported", ["RS256"]) + + # As per sec. 2.6: + # 3. Validate the iss, aud, and iat Claims in the same way they are + # validated in ID Tokens. + # Which means the audience should contain Synapse's client_id and the + # issuer should be the IdP issuer + claims_options = { + "iss": {"values": [metadata["issuer"]]}, + "aud": {"values": [self.client_id]}, + } + + try: + claims = await self._verify_jwt( + alg_values=alg_values, + token=logout_token, + claims_cls=LogoutToken, + claims_options=claims_options, + ) + except JoseError: + logger.exception("Invalid logout_token") + raise SynapseError(400, "Invalid logout_token") + + # As per sec. 2.6: + # 4. Verify that the Logout Token contains a sub Claim, a sid Claim, + # or both. + # 5. Verify that the Logout Token contains an events Claim whose + # value is JSON object containing the member name + # http://schemas.openid.net/event/backchannel-logout. + # 6. Verify that the Logout Token does not contain a nonce Claim. + # This is all verified by the LogoutToken claims class, so at this + # point the `sid` claim exists and is a string. + sid: str = claims.get("sid") + + # If the `sub` claim was included in the logout token, we check that it matches + # that it matches the right user. We can have cases where the `sub` claim is not + # the ID saved in database, so we let admins disable this check in config. + sub: Optional[str] = claims.get("sub") + expected_user_id: Optional[str] = None + if sub is not None and not self._config.backchannel_logout_ignore_sub: + expected_user_id = await self._store.get_user_by_external_id( + self.idp_id, sub + ) + + # Invalidate any running user-mapping sessions, in-flight login tokens and + # active devices + await self._sso_handler.revoke_sessions_for_provider_session_id( + auth_provider_id=self.idp_id, + auth_provider_session_id=sid, + expected_user_id=expected_user_id, + ) + + request.setResponseCode(200) + request.setHeader(b"Cache-Control", b"no-cache, no-store") + request.setHeader(b"Pragma", b"no-cache") + finish_request(request) + + +class LogoutToken(JWTClaims): + """ + Holds and verify claims of a logout token, as per + https://openid.net/specs/openid-connect-backchannel-1_0.html#LogoutToken + """ + + REGISTERED_CLAIMS = ["iss", "sub", "aud", "iat", "jti", "events", "sid"] + + def validate(self, now: Optional[int] = None, leeway: int = 0) -> None: + """Validate everything in claims payload.""" + super().validate(now, leeway) + self.validate_sid() + self.validate_events() + self.validate_nonce() + + def validate_sid(self) -> None: + """Ensure the sid claim is present""" + sid = self.get("sid") + if not sid: + raise MissingClaimError("sid") + + if not isinstance(sid, str): + raise InvalidClaimError("sid") + + def validate_nonce(self) -> None: + """Ensure the nonce claim is absent""" + if "nonce" in self: + raise InvalidClaimError("nonce") + + def validate_events(self) -> None: + """Ensure the events claim is present and with the right value""" + events = self.get("events") + if not events: + raise MissingClaimError("events") + + if not isinstance(events, dict): + raise InvalidClaimError("events") + + if "http://schemas.openid.net/event/backchannel-logout" not in events: + raise InvalidClaimError("events") + # number of seconds a newly-generated client secret should be valid for CLIENT_SECRET_VALIDITY_SECONDS = 3600 @@ -1112,6 +1423,7 @@ class JwtClientSecret: logger.info( "Generating new JWT for %s: %s %s", self._oauth_issuer, header, payload ) + jwt = JsonWebToken(header["alg"]) self._cached_secret = jwt.encode(header, payload, self._key.key) self._cached_secret_replacement_time = ( expires_at - CLIENT_SECRET_MIN_VALIDITY_SECONDS @@ -1126,9 +1438,6 @@ class UserAttributeDict(TypedDict): emails: List[str] -C = TypeVar("C") - - class OidcMappingProvider(Generic[C]): """A mapping provider maps a UserInfo object to user attributes. diff --git a/synapse/handlers/sso.py b/synapse/handlers/sso.py index 5943f08e91..749d7e93b0 100644 --- a/synapse/handlers/sso.py +++ b/synapse/handlers/sso.py @@ -191,6 +191,7 @@ class SsoHandler: self._server_name = hs.hostname self._registration_handler = hs.get_registration_handler() self._auth_handler = hs.get_auth_handler() + self._device_handler = hs.get_device_handler() self._error_template = hs.config.sso.sso_error_template self._bad_user_template = hs.config.sso.sso_auth_bad_user_template self._profile_handler = hs.get_profile_handler() @@ -1026,6 +1027,76 @@ class SsoHandler: return True + async def revoke_sessions_for_provider_session_id( + self, + auth_provider_id: str, + auth_provider_session_id: str, + expected_user_id: Optional[str] = None, + ) -> None: + """Revoke any devices and in-flight logins tied to a provider session. + + Args: + auth_provider_id: A unique identifier for this SSO provider, e.g. + "oidc" or "saml". + auth_provider_session_id: The session ID from the provider to logout + expected_user_id: The user we're expecting to logout. If set, it will ignore + sessions belonging to other users and log an error. + """ + # Invalidate any running user-mapping sessions + to_delete = [] + for session_id, session in self._username_mapping_sessions.items(): + if ( + session.auth_provider_id == auth_provider_id + and session.auth_provider_session_id == auth_provider_session_id + ): + to_delete.append(session_id) + + for session_id in to_delete: + logger.info("Revoking mapping session %s", session_id) + del self._username_mapping_sessions[session_id] + + # Invalidate any in-flight login tokens + await self._store.invalidate_login_tokens_by_session_id( + auth_provider_id=auth_provider_id, + auth_provider_session_id=auth_provider_session_id, + ) + + # Fetch any device(s) in the store associated with the session ID. + devices = await self._store.get_devices_by_auth_provider_session_id( + auth_provider_id=auth_provider_id, + auth_provider_session_id=auth_provider_session_id, + ) + + # We have no guarantee that all the devices of that session are for the same + # `user_id`. Hence, we have to iterate over the list of devices and log them out + # one by one. + for device in devices: + user_id = device["user_id"] + device_id = device["device_id"] + + # If the user_id associated with that device/session is not the one we got + # out of the `sub` claim, skip that device and show log an error. + if expected_user_id is not None and user_id != expected_user_id: + logger.error( + "Received a logout notification from SSO provider " + f"{auth_provider_id!r} for the user {expected_user_id!r}, but with " + f"a session ID ({auth_provider_session_id!r}) which belongs to " + f"{user_id!r}. This may happen when the SSO provider user mapper " + "uses something else than the standard attribute as mapping ID. " + "For OIDC providers, set `backchannel_logout_ignore_sub` to `true` " + "in the provider config if that is the case." + ) + continue + + logger.info( + "Logging out %r (device %r) via SSO (%r) logout notification (session %r).", + user_id, + device_id, + auth_provider_id, + auth_provider_session_id, + ) + await self._device_handler.delete_devices(user_id, [device_id]) + def get_username_mapping_session_cookie_from_request(request: IRequest) -> str: """Extract the session ID from the cookie diff --git a/synapse/rest/synapse/client/oidc/__init__.py b/synapse/rest/synapse/client/oidc/__init__.py index 81fec39659..e4b28ce3df 100644 --- a/synapse/rest/synapse/client/oidc/__init__.py +++ b/synapse/rest/synapse/client/oidc/__init__.py @@ -17,6 +17,9 @@ from typing import TYPE_CHECKING from twisted.web.resource import Resource +from synapse.rest.synapse.client.oidc.backchannel_logout_resource import ( + OIDCBackchannelLogoutResource, +) from synapse.rest.synapse.client.oidc.callback_resource import OIDCCallbackResource if TYPE_CHECKING: @@ -29,6 +32,7 @@ class OIDCResource(Resource): def __init__(self, hs: "HomeServer"): Resource.__init__(self) self.putChild(b"callback", OIDCCallbackResource(hs)) + self.putChild(b"backchannel_logout", OIDCBackchannelLogoutResource(hs)) __all__ = ["OIDCResource"] diff --git a/synapse/rest/synapse/client/oidc/backchannel_logout_resource.py b/synapse/rest/synapse/client/oidc/backchannel_logout_resource.py new file mode 100644 index 0000000000..e07e76855a --- /dev/null +++ b/synapse/rest/synapse/client/oidc/backchannel_logout_resource.py @@ -0,0 +1,35 @@ +# Copyright 2022 The Matrix.org Foundation C.I.C. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import logging +from typing import TYPE_CHECKING + +from synapse.http.server import DirectServeJsonResource +from synapse.http.site import SynapseRequest + +if TYPE_CHECKING: + from synapse.server import HomeServer + +logger = logging.getLogger(__name__) + + +class OIDCBackchannelLogoutResource(DirectServeJsonResource): + isLeaf = 1 + + def __init__(self, hs: "HomeServer"): + super().__init__() + self._oidc_handler = hs.get_oidc_handler() + + async def _async_render_POST(self, request: SynapseRequest) -> None: + await self._oidc_handler.handle_backchannel_logout(request) diff --git a/synapse/storage/databases/main/registration.py b/synapse/storage/databases/main/registration.py index 0255295317..5167089e03 100644 --- a/synapse/storage/databases/main/registration.py +++ b/synapse/storage/databases/main/registration.py @@ -1920,6 +1920,27 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore): self._clock.time_msec(), ) + async def invalidate_login_tokens_by_session_id( + self, auth_provider_id: str, auth_provider_session_id: str + ) -> None: + """Invalidate login tokens with the given IdP session ID. + + Args: + auth_provider_id: The SSO Identity Provider that the user authenticated with + to get this token + auth_provider_session_id: The session ID advertised by the SSO Identity + Provider + """ + await self.db_pool.simple_update( + table="login_tokens", + keyvalues={ + "auth_provider_id": auth_provider_id, + "auth_provider_session_id": auth_provider_session_id, + }, + updatevalues={"used_ts": self._clock.time_msec()}, + desc="invalidate_login_tokens_by_session_id", + ) + @cached() async def is_guest(self, user_id: str) -> bool: res = await self.db_pool.simple_select_one_onecol( diff --git a/tests/rest/client/test_auth.py b/tests/rest/client/test_auth.py index ebf653d018..847294dc8e 100644 --- a/tests/rest/client/test_auth.py +++ b/tests/rest/client/test_auth.py @@ -12,6 +12,7 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. +import re from http import HTTPStatus from typing import Any, Dict, List, Optional, Tuple, Union @@ -21,7 +22,7 @@ from twisted.web.resource import Resource import synapse.rest.admin from synapse.api.constants import ApprovalNoticeMedium, LoginType -from synapse.api.errors import Codes +from synapse.api.errors import Codes, SynapseError from synapse.handlers.ui_auth.checkers import UserInteractiveAuthChecker from synapse.rest.client import account, auth, devices, login, logout, register from synapse.rest.synapse.client import build_synapse_client_resource_tree @@ -32,8 +33,8 @@ from synapse.util import Clock from tests import unittest from tests.handlers.test_oidc import HAS_OIDC -from tests.rest.client.utils import TEST_OIDC_CONFIG -from tests.server import FakeChannel +from tests.rest.client.utils import TEST_OIDC_CONFIG, TEST_OIDC_ISSUER +from tests.server import FakeChannel, make_request from tests.unittest import override_config, skip_unless @@ -638,19 +639,6 @@ class RefreshAuthTests(unittest.HomeserverTestCase): {"refresh_token": refresh_token}, ) - def is_access_token_valid(self, access_token: str) -> bool: - """ - Checks whether an access token is valid, returning whether it is or not. - """ - code = self.make_request( - "GET", "/_matrix/client/v3/account/whoami", access_token=access_token - ).code - - # Either 200 or 401 is what we get back; anything else is a bug. - assert code in {HTTPStatus.OK, HTTPStatus.UNAUTHORIZED} - - return code == HTTPStatus.OK - def test_login_issue_refresh_token(self) -> None: """ A login response should include a refresh_token only if asked. @@ -847,29 +835,37 @@ class RefreshAuthTests(unittest.HomeserverTestCase): self.reactor.advance(59.0) # Both tokens should still be valid. - self.assertTrue(self.is_access_token_valid(refreshable_access_token)) - self.assertTrue(self.is_access_token_valid(nonrefreshable_access_token)) + self.helper.whoami(refreshable_access_token, expect_code=HTTPStatus.OK) + self.helper.whoami(nonrefreshable_access_token, expect_code=HTTPStatus.OK) # Advance to 61 s (just past 1 minute, the time of expiry) self.reactor.advance(2.0) # Only the non-refreshable token is still valid. - self.assertFalse(self.is_access_token_valid(refreshable_access_token)) - self.assertTrue(self.is_access_token_valid(nonrefreshable_access_token)) + self.helper.whoami( + refreshable_access_token, expect_code=HTTPStatus.UNAUTHORIZED + ) + self.helper.whoami(nonrefreshable_access_token, expect_code=HTTPStatus.OK) # Advance to 599 s (just shy of 10 minutes, the time of expiry) self.reactor.advance(599.0 - 61.0) # It's still the case that only the non-refreshable token is still valid. - self.assertFalse(self.is_access_token_valid(refreshable_access_token)) - self.assertTrue(self.is_access_token_valid(nonrefreshable_access_token)) + self.helper.whoami( + refreshable_access_token, expect_code=HTTPStatus.UNAUTHORIZED + ) + self.helper.whoami(nonrefreshable_access_token, expect_code=HTTPStatus.OK) # Advance to 601 s (just past 10 minutes, the time of expiry) self.reactor.advance(2.0) # Now neither token is valid. - self.assertFalse(self.is_access_token_valid(refreshable_access_token)) - self.assertFalse(self.is_access_token_valid(nonrefreshable_access_token)) + self.helper.whoami( + refreshable_access_token, expect_code=HTTPStatus.UNAUTHORIZED + ) + self.helper.whoami( + nonrefreshable_access_token, expect_code=HTTPStatus.UNAUTHORIZED + ) @override_config( {"refreshable_access_token_lifetime": "1m", "refresh_token_lifetime": "2m"} @@ -1165,3 +1161,349 @@ class RefreshAuthTests(unittest.HomeserverTestCase): # and no refresh token self.assertEqual(_table_length("access_tokens"), 0) self.assertEqual(_table_length("refresh_tokens"), 0) + + +def oidc_config( + id: str, with_localpart_template: bool, **kwargs: Any +) -> Dict[str, Any]: + """Sample OIDC provider config used in backchannel logout tests. + + Args: + id: IDP ID for this provider + with_localpart_template: Set to `true` to have a default localpart_template in + the `user_mapping_provider` config and skip the user mapping session + **kwargs: rest of the config + + Returns: + A dict suitable for the `oidc_config` or the `oidc_providers[]` parts of + the HS config + """ + config: Dict[str, Any] = { + "idp_id": id, + "idp_name": id, + "issuer": TEST_OIDC_ISSUER, + "client_id": "test-client-id", + "client_secret": "test-client-secret", + "scopes": ["openid"], + } + + if with_localpart_template: + config["user_mapping_provider"] = { + "config": {"localpart_template": "{{ user.sub }}"} + } + else: + config["user_mapping_provider"] = {"config": {}} + + config.update(kwargs) + + return config + + +@skip_unless(HAS_OIDC, "Requires OIDC") +class OidcBackchannelLogoutTests(unittest.HomeserverTestCase): + servlets = [ + account.register_servlets, + login.register_servlets, + ] + + def default_config(self) -> Dict[str, Any]: + config = super().default_config() + + # public_baseurl uses an http:// scheme because FakeChannel.isSecure() returns + # False, so synapse will see the requested uri as http://..., so using http in + # the public_baseurl stops Synapse trying to redirect to https. + config["public_baseurl"] = "http://synapse.test" + + return config + + def create_resource_dict(self) -> Dict[str, Resource]: + resource_dict = super().create_resource_dict() + resource_dict.update(build_synapse_client_resource_tree(self.hs)) + return resource_dict + + def submit_logout_token(self, logout_token: str) -> FakeChannel: + return self.make_request( + "POST", + "/_synapse/client/oidc/backchannel_logout", + content=f"logout_token={logout_token}", + content_is_form=True, + ) + + @override_config( + { + "oidc_providers": [ + oidc_config( + id="oidc", + with_localpart_template=True, + backchannel_logout_enabled=True, + ) + ] + } + ) + def test_simple_logout(self) -> None: + """ + Receiving a logout token should logout the user + """ + fake_oidc_server = self.helper.fake_oidc_server() + user = "john" + + login_resp, first_grant = self.helper.login_via_oidc( + fake_oidc_server, user, with_sid=True + ) + first_access_token: str = login_resp["access_token"] + self.helper.whoami(first_access_token, expect_code=HTTPStatus.OK) + + login_resp, second_grant = self.helper.login_via_oidc( + fake_oidc_server, user, with_sid=True + ) + second_access_token: str = login_resp["access_token"] + self.helper.whoami(second_access_token, expect_code=HTTPStatus.OK) + + self.assertNotEqual(first_grant.sid, second_grant.sid) + self.assertEqual(first_grant.userinfo["sub"], second_grant.userinfo["sub"]) + + # Logging out of the first session + logout_token = fake_oidc_server.generate_logout_token(first_grant) + channel = self.submit_logout_token(logout_token) + self.assertEqual(channel.code, 200) + + self.helper.whoami(first_access_token, expect_code=HTTPStatus.UNAUTHORIZED) + self.helper.whoami(second_access_token, expect_code=HTTPStatus.OK) + + # Logging out of the second session + logout_token = fake_oidc_server.generate_logout_token(second_grant) + channel = self.submit_logout_token(logout_token) + self.assertEqual(channel.code, 200) + + @override_config( + { + "oidc_providers": [ + oidc_config( + id="oidc", + with_localpart_template=True, + backchannel_logout_enabled=True, + ) + ] + } + ) + def test_logout_during_login(self) -> None: + """ + It should revoke login tokens when receiving a logout token + """ + fake_oidc_server = self.helper.fake_oidc_server() + user = "john" + + # Get an authentication, and logout before submitting the logout token + client_redirect_url = "https://x" + userinfo = {"sub": user} + channel, grant = self.helper.auth_via_oidc( + fake_oidc_server, + userinfo, + client_redirect_url, + with_sid=True, + ) + + # expect a confirmation page + self.assertEqual(channel.code, HTTPStatus.OK, channel.result) + + # fish the matrix login token out of the body of the confirmation page + m = re.search( + 'a href="%s.*loginToken=([^"]*)"' % (client_redirect_url,), + channel.text_body, + ) + assert m, channel.text_body + login_token = m.group(1) + + # Submit a logout + logout_token = fake_oidc_server.generate_logout_token(grant) + channel = self.submit_logout_token(logout_token) + self.assertEqual(channel.code, 200) + + # Now try to exchange the login token + channel = make_request( + self.hs.get_reactor(), + self.site, + "POST", + "/login", + content={"type": "m.login.token", "token": login_token}, + ) + # It should have failed + self.assertEqual(channel.code, 403) + + @override_config( + { + "oidc_providers": [ + oidc_config( + id="oidc", + with_localpart_template=False, + backchannel_logout_enabled=True, + ) + ] + } + ) + def test_logout_during_mapping(self) -> None: + """ + It should stop ongoing user mapping session when receiving a logout token + """ + fake_oidc_server = self.helper.fake_oidc_server() + user = "john" + + # Get an authentication, and logout before submitting the logout token + client_redirect_url = "https://x" + userinfo = {"sub": user} + channel, grant = self.helper.auth_via_oidc( + fake_oidc_server, + userinfo, + client_redirect_url, + with_sid=True, + ) + + # Expect a user mapping page + self.assertEqual(channel.code, HTTPStatus.FOUND, channel.result) + + # We should have a user_mapping_session cookie + cookie_headers = channel.headers.getRawHeaders("Set-Cookie") + assert cookie_headers + cookies: Dict[str, str] = {} + for h in cookie_headers: + key, value = h.split(";")[0].split("=", maxsplit=1) + cookies[key] = value + + user_mapping_session_id = cookies["username_mapping_session"] + + # Getting that session should not raise + session = self.hs.get_sso_handler().get_mapping_session(user_mapping_session_id) + self.assertIsNotNone(session) + + # Submit a logout + logout_token = fake_oidc_server.generate_logout_token(grant) + channel = self.submit_logout_token(logout_token) + self.assertEqual(channel.code, 200) + + # Now it should raise + with self.assertRaises(SynapseError): + self.hs.get_sso_handler().get_mapping_session(user_mapping_session_id) + + @override_config( + { + "oidc_providers": [ + oidc_config( + id="oidc", + with_localpart_template=True, + backchannel_logout_enabled=False, + ) + ] + } + ) + def test_disabled(self) -> None: + """ + Receiving a logout token should do nothing if it is disabled in the config + """ + fake_oidc_server = self.helper.fake_oidc_server() + user = "john" + + login_resp, grant = self.helper.login_via_oidc( + fake_oidc_server, user, with_sid=True + ) + access_token: str = login_resp["access_token"] + self.helper.whoami(access_token, expect_code=HTTPStatus.OK) + + # Logging out shouldn't work + logout_token = fake_oidc_server.generate_logout_token(grant) + channel = self.submit_logout_token(logout_token) + self.assertEqual(channel.code, 400) + + # And the token should still be valid + self.helper.whoami(access_token, expect_code=HTTPStatus.OK) + + @override_config( + { + "oidc_providers": [ + oidc_config( + id="oidc", + with_localpart_template=True, + backchannel_logout_enabled=True, + ) + ] + } + ) + def test_no_sid(self) -> None: + """ + Receiving a logout token without `sid` during the login should do nothing + """ + fake_oidc_server = self.helper.fake_oidc_server() + user = "john" + + login_resp, grant = self.helper.login_via_oidc( + fake_oidc_server, user, with_sid=False + ) + access_token: str = login_resp["access_token"] + self.helper.whoami(access_token, expect_code=HTTPStatus.OK) + + # Logging out shouldn't work + logout_token = fake_oidc_server.generate_logout_token(grant) + channel = self.submit_logout_token(logout_token) + self.assertEqual(channel.code, 400) + + # And the token should still be valid + self.helper.whoami(access_token, expect_code=HTTPStatus.OK) + + @override_config( + { + "oidc_providers": [ + oidc_config( + "first", + issuer="https://first-issuer.com/", + with_localpart_template=True, + backchannel_logout_enabled=True, + ), + oidc_config( + "second", + issuer="https://second-issuer.com/", + with_localpart_template=True, + backchannel_logout_enabled=True, + ), + ] + } + ) + def test_multiple_providers(self) -> None: + """ + It should be able to distinguish login tokens from two different IdPs + """ + first_server = self.helper.fake_oidc_server(issuer="https://first-issuer.com/") + second_server = self.helper.fake_oidc_server( + issuer="https://second-issuer.com/" + ) + user = "john" + + login_resp, first_grant = self.helper.login_via_oidc( + first_server, user, with_sid=True, idp_id="oidc-first" + ) + first_access_token: str = login_resp["access_token"] + self.helper.whoami(first_access_token, expect_code=HTTPStatus.OK) + + login_resp, second_grant = self.helper.login_via_oidc( + second_server, user, with_sid=True, idp_id="oidc-second" + ) + second_access_token: str = login_resp["access_token"] + self.helper.whoami(second_access_token, expect_code=HTTPStatus.OK) + + # `sid` in the fake providers are generated by a counter, so the first grant of + # each provider should give the same SID + self.assertEqual(first_grant.sid, second_grant.sid) + self.assertEqual(first_grant.userinfo["sub"], second_grant.userinfo["sub"]) + + # Logging out of the first session + logout_token = first_server.generate_logout_token(first_grant) + channel = self.submit_logout_token(logout_token) + self.assertEqual(channel.code, 200) + + self.helper.whoami(first_access_token, expect_code=HTTPStatus.UNAUTHORIZED) + self.helper.whoami(second_access_token, expect_code=HTTPStatus.OK) + + # Logging out of the second session + logout_token = second_server.generate_logout_token(second_grant) + channel = self.submit_logout_token(logout_token) + self.assertEqual(channel.code, 200) + + self.helper.whoami(second_access_token, expect_code=HTTPStatus.UNAUTHORIZED) diff --git a/tests/rest/client/utils.py b/tests/rest/client/utils.py index 967d229223..706399fae5 100644 --- a/tests/rest/client/utils.py +++ b/tests/rest/client/utils.py @@ -553,6 +553,34 @@ class RestHelper: return channel.json_body + def whoami( + self, + access_token: str, + expect_code: Literal[HTTPStatus.OK, HTTPStatus.UNAUTHORIZED] = HTTPStatus.OK, + ) -> JsonDict: + """Perform a 'whoami' request, which can be a quick way to check for access + token validity + + Args: + access_token: The user token to use during the request + expect_code: The return code to expect from attempting the whoami request + """ + channel = make_request( + self.hs.get_reactor(), + self.site, + "GET", + "account/whoami", + access_token=access_token, + ) + + assert channel.code == expect_code, "Exepcted: %d, got %d, resp: %r" % ( + expect_code, + channel.code, + channel.result["body"], + ) + + return channel.json_body + def fake_oidc_server(self, issuer: str = TEST_OIDC_ISSUER) -> FakeOidcServer: """Create a ``FakeOidcServer``. @@ -572,6 +600,7 @@ class RestHelper: fake_server: FakeOidcServer, remote_user_id: str, with_sid: bool = False, + idp_id: Optional[str] = None, expected_status: int = 200, ) -> Tuple[JsonDict, FakeAuthorizationGrant]: """Log in (as a new user) via OIDC @@ -588,7 +617,11 @@ class RestHelper: client_redirect_url = "https://x" userinfo = {"sub": remote_user_id} channel, grant = self.auth_via_oidc( - fake_server, userinfo, client_redirect_url, with_sid=with_sid + fake_server, + userinfo, + client_redirect_url, + with_sid=with_sid, + idp_id=idp_id, ) # expect a confirmation page @@ -623,6 +656,7 @@ class RestHelper: client_redirect_url: Optional[str] = None, ui_auth_session_id: Optional[str] = None, with_sid: bool = False, + idp_id: Optional[str] = None, ) -> Tuple[FakeChannel, FakeAuthorizationGrant]: """Perform an OIDC authentication flow via a mock OIDC provider. @@ -648,6 +682,7 @@ class RestHelper: ui_auth_session_id: if set, we will perform a UI Auth flow. The session id of the UI auth. with_sid: if True, generates a random `sid` (OIDC session ID) + idp_id: if set, explicitely chooses one specific IDP Returns: A FakeChannel containing the result of calling the OIDC callback endpoint. @@ -665,7 +700,9 @@ class RestHelper: oauth_uri = self.initiate_sso_ui_auth(ui_auth_session_id, cookies) else: # otherwise, hit the login redirect endpoint - oauth_uri = self.initiate_sso_login(client_redirect_url, cookies) + oauth_uri = self.initiate_sso_login( + client_redirect_url, cookies, idp_id=idp_id + ) # we now have a URI for the OIDC IdP, but we skip that and go straight # back to synapse's OIDC callback resource. However, we do need the "state" @@ -742,7 +779,10 @@ class RestHelper: return channel, grant def initiate_sso_login( - self, client_redirect_url: Optional[str], cookies: MutableMapping[str, str] + self, + client_redirect_url: Optional[str], + cookies: MutableMapping[str, str], + idp_id: Optional[str] = None, ) -> str: """Make a request to the login-via-sso redirect endpoint, and return the target @@ -753,6 +793,7 @@ class RestHelper: client_redirect_url: the client redirect URL to pass to the login redirect endpoint cookies: any cookies returned will be added to this dict + idp_id: if set, explicitely chooses one specific IDP Returns: the URI that the client gets redirected to (ie, the SSO server) @@ -761,6 +802,12 @@ class RestHelper: if client_redirect_url: params["redirectUrl"] = client_redirect_url + uri = "/_matrix/client/r0/login/sso/redirect" + if idp_id is not None: + uri = f"{uri}/{idp_id}" + + uri = f"{uri}?{urllib.parse.urlencode(params)}" + # hit the redirect url (which should redirect back to the redirect url. This # is the easiest way of figuring out what the Host header ought to be set to # to keep Synapse happy. @@ -768,7 +815,7 @@ class RestHelper: self.hs.get_reactor(), self.site, "GET", - "/_matrix/client/r0/login/sso/redirect?" + urllib.parse.urlencode(params), + uri, ) assert channel.code == 302 diff --git a/tests/server.py b/tests/server.py index 8b1d186219..b1730fcc8d 100644 --- a/tests/server.py +++ b/tests/server.py @@ -362,6 +362,12 @@ def make_request( # Twisted expects to be at the end of the content when parsing the request. req.content.seek(0, SEEK_END) + # Old version of Twisted (<20.3.0) have issues with parsing x-www-form-urlencoded + # bodies if the Content-Length header is missing + req.requestHeaders.addRawHeader( + b"Content-Length", str(len(content)).encode("ascii") + ) + if access_token: req.requestHeaders.addRawHeader( b"Authorization", b"Bearer " + access_token.encode("ascii") diff --git a/tests/test_utils/oidc.py b/tests/test_utils/oidc.py index de134bbc89..1461d23ee8 100644 --- a/tests/test_utils/oidc.py +++ b/tests/test_utils/oidc.py @@ -51,6 +51,8 @@ class FakeOidcServer: get_userinfo_handler: Mock post_token_handler: Mock + sid_counter: int = 0 + def __init__(self, clock: Clock, issuer: str): from authlib.jose import ECKey, KeySet @@ -146,7 +148,7 @@ class FakeOidcServer: return jws.serialize_compact(protected, json_payload, self._key).decode("utf-8") def generate_id_token(self, grant: FakeAuthorizationGrant) -> str: - now = self._clock.time() + now = int(self._clock.time()) id_token = { **grant.userinfo, "iss": self.issuer, @@ -166,6 +168,26 @@ class FakeOidcServer: return self._sign(id_token) + def generate_logout_token(self, grant: FakeAuthorizationGrant) -> str: + now = int(self._clock.time()) + logout_token = { + "iss": self.issuer, + "aud": grant.client_id, + "iat": now, + "jti": random_string(10), + "events": { + "http://schemas.openid.net/event/backchannel-logout": {}, + }, + } + + if grant.sid is not None: + logout_token["sid"] = grant.sid + + if "sub" in grant.userinfo: + logout_token["sub"] = grant.userinfo["sub"] + + return self._sign(logout_token) + def id_token_override(self, overrides: dict): """Temporarily patch the ID token generated by the token endpoint.""" return patch.object(self, "_id_token_overrides", overrides) @@ -183,7 +205,8 @@ class FakeOidcServer: code = random_string(10) sid = None if with_sid: - sid = random_string(10) + sid = str(self.sid_counter) + self.sid_counter += 1 grant = FakeAuthorizationGrant( userinfo=userinfo, -- cgit 1.5.1 From 2bd7f3eeab1a4818359c9f585b660ff3f3d8bc6c Mon Sep 17 00:00:00 2001 From: David Robertson Date: Tue, 1 Nov 2022 15:02:39 +0000 Subject: Allow PUT/GET of aliases during faster join (#14292) without blocking on full state. --- changelog.d/14292.bugfix | 1 + synapse/handlers/directory.py | 4 ++-- 2 files changed, 3 insertions(+), 2 deletions(-) create mode 100644 changelog.d/14292.bugfix (limited to 'synapse/handlers') diff --git a/changelog.d/14292.bugfix b/changelog.d/14292.bugfix new file mode 100644 index 0000000000..4ed92f5cf2 --- /dev/null +++ b/changelog.d/14292.bugfix @@ -0,0 +1 @@ +Faster joins: do not block creation of or queries for room aliases during the resync. diff --git a/synapse/handlers/directory.py b/synapse/handlers/directory.py index d52ebada6b..2ea52257cb 100644 --- a/synapse/handlers/directory.py +++ b/synapse/handlers/directory.py @@ -85,7 +85,7 @@ class DirectoryHandler: # TODO(erikj): Add transactions. # TODO(erikj): Check if there is a current association. if not servers: - servers = await self._storage_controllers.state.get_current_hosts_in_room( + servers = await self._storage_controllers.state.get_current_hosts_in_room_or_partial_state_approximation( room_id ) @@ -290,7 +290,7 @@ class DirectoryHandler: Codes.NOT_FOUND, ) - extra_servers = await self._storage_controllers.state.get_current_hosts_in_room( + extra_servers = await self._storage_controllers.state.get_current_hosts_in_room_or_partial_state_approximation( room_id ) servers_set = set(extra_servers) | set(servers) -- cgit 1.5.1 From 86c5a710d8b4212f8a8a668d7d4a79c0bb371508 Mon Sep 17 00:00:00 2001 From: Brendan Abolivier Date: Thu, 3 Nov 2022 16:21:31 +0000 Subject: Implement MSC3912: Relation-based redactions (#14260) Co-authored-by: Sean Quah <8349537+squahtx@users.noreply.github.com> --- changelog.d/14260.feature | 1 + synapse/api/constants.py | 2 + synapse/config/experimental.py | 3 + synapse/handlers/message.py | 47 ++++- synapse/handlers/relations.py | 56 +++++- synapse/rest/client/room.py | 57 ++++-- synapse/rest/client/versions.py | 2 + synapse/storage/databases/main/relations.py | 36 ++++ tests/rest/client/test_redactions.py | 273 +++++++++++++++++++++++++++- tests/rest/client/utils.py | 37 ++++ 10 files changed, 486 insertions(+), 28 deletions(-) create mode 100644 changelog.d/14260.feature (limited to 'synapse/handlers') diff --git a/changelog.d/14260.feature b/changelog.d/14260.feature new file mode 100644 index 0000000000..102dc7b3e0 --- /dev/null +++ b/changelog.d/14260.feature @@ -0,0 +1 @@ +Add experimental support for [MSC3912](https://github.com/matrix-org/matrix-spec-proposals/pull/3912): Relation-based redactions. diff --git a/synapse/api/constants.py b/synapse/api/constants.py index 44c5ffc6a5..bc04a0755b 100644 --- a/synapse/api/constants.py +++ b/synapse/api/constants.py @@ -125,6 +125,8 @@ class EventTypes: MSC2716_BATCH: Final = "org.matrix.msc2716.batch" MSC2716_MARKER: Final = "org.matrix.msc2716.marker" + Reaction: Final = "m.reaction" + class ToDeviceEventTypes: RoomKeyRequest: Final = "m.room_key_request" diff --git a/synapse/config/experimental.py b/synapse/config/experimental.py index d9bdd66d55..d4b71d1673 100644 --- a/synapse/config/experimental.py +++ b/synapse/config/experimental.py @@ -128,3 +128,6 @@ class ExperimentalConfig(Config): self.msc3886_endpoint: Optional[str] = experimental.get( "msc3886_endpoint", None ) + + # MSC3912: Relation-based redactions. + self.msc3912_enabled: bool = experimental.get("msc3912_enabled", False) diff --git a/synapse/handlers/message.py b/synapse/handlers/message.py index 468900a07f..4cf593cfdc 100644 --- a/synapse/handlers/message.py +++ b/synapse/handlers/message.py @@ -877,6 +877,36 @@ class EventCreationHandler: return prev_event return None + async def get_event_from_transaction( + self, + requester: Requester, + txn_id: str, + room_id: str, + ) -> Optional[EventBase]: + """For the given transaction ID and room ID, check if there is a matching event. + If so, fetch it and return it. + + Args: + requester: The requester making the request in the context of which we want + to fetch the event. + txn_id: The transaction ID. + room_id: The room ID. + + Returns: + An event if one could be found, None otherwise. + """ + if requester.access_token_id: + existing_event_id = await self.store.get_event_id_from_transaction_id( + room_id, + requester.user.to_string(), + requester.access_token_id, + txn_id, + ) + if existing_event_id: + return await self.store.get_event(existing_event_id) + + return None + async def create_and_send_nonmember_event( self, requester: Requester, @@ -956,18 +986,17 @@ class EventCreationHandler: # extremities to pile up, which in turn leads to state resolution # taking longer. async with self.limiter.queue(event_dict["room_id"]): - if txn_id and requester.access_token_id: - existing_event_id = await self.store.get_event_id_from_transaction_id( - event_dict["room_id"], - requester.user.to_string(), - requester.access_token_id, - txn_id, + if txn_id: + event = await self.get_event_from_transaction( + requester, txn_id, event_dict["room_id"] ) - if existing_event_id: - event = await self.store.get_event(existing_event_id) + if event: # we know it was persisted, so must have a stream ordering assert event.internal_metadata.stream_ordering - return event, event.internal_metadata.stream_ordering + return ( + event, + event.internal_metadata.stream_ordering, + ) event, context = await self.create_event( requester, diff --git a/synapse/handlers/relations.py b/synapse/handlers/relations.py index 0a0c6d938e..8e71dda970 100644 --- a/synapse/handlers/relations.py +++ b/synapse/handlers/relations.py @@ -17,7 +17,7 @@ from typing import TYPE_CHECKING, Dict, FrozenSet, Iterable, List, Optional, Tup import attr -from synapse.api.constants import RelationTypes +from synapse.api.constants import EventTypes, RelationTypes from synapse.api.errors import SynapseError from synapse.events import EventBase, relation_from_event from synapse.logging.opentracing import trace @@ -75,6 +75,7 @@ class RelationsHandler: self._clock = hs.get_clock() self._event_handler = hs.get_event_handler() self._event_serializer = hs.get_event_client_serializer() + self._event_creation_handler = hs.get_event_creation_handler() async def get_relations( self, @@ -205,6 +206,59 @@ class RelationsHandler: return related_events, next_token + async def redact_events_related_to( + self, + requester: Requester, + event_id: str, + initial_redaction_event: EventBase, + relation_types: List[str], + ) -> None: + """Redacts all events related to the given event ID with one of the given + relation types. + + This method is expected to be called when redacting the event referred to by + the given event ID. + + If an event cannot be redacted (e.g. because of insufficient permissions), log + the error and try to redact the next one. + + Args: + requester: The requester to redact events on behalf of. + event_id: The event IDs to look and redact relations of. + initial_redaction_event: The redaction for the event referred to by + event_id. + relation_types: The types of relations to look for. + + Raises: + ShadowBanError if the requester is shadow-banned + """ + related_event_ids = ( + await self._main_store.get_all_relations_for_event_with_types( + event_id, relation_types + ) + ) + + for related_event_id in related_event_ids: + try: + await self._event_creation_handler.create_and_send_nonmember_event( + requester, + { + "type": EventTypes.Redaction, + "content": initial_redaction_event.content, + "room_id": initial_redaction_event.room_id, + "sender": requester.user.to_string(), + "redacts": related_event_id, + }, + ratelimit=False, + ) + except SynapseError as e: + logger.warning( + "Failed to redact event %s (related to event %s): %s", + related_event_id, + event_id, + e.msg, + ) + async def get_annotations_for_event( self, event_id: str, diff --git a/synapse/rest/client/room.py b/synapse/rest/client/room.py index 01e5079963..91cb791139 100644 --- a/synapse/rest/client/room.py +++ b/synapse/rest/client/room.py @@ -52,6 +52,7 @@ from synapse.http.servlet import ( from synapse.http.site import SynapseRequest from synapse.logging.context import make_deferred_yieldable, run_in_background from synapse.logging.opentracing import set_tag +from synapse.metrics.background_process_metrics import run_as_background_process from synapse.rest.client._base import client_patterns from synapse.rest.client.transactions import HttpTransactionCache from synapse.storage.state import StateFilter @@ -1029,6 +1030,8 @@ class RoomRedactEventRestServlet(TransactionRestServlet): super().__init__(hs) self.event_creation_handler = hs.get_event_creation_handler() self.auth = hs.get_auth() + self._relation_handler = hs.get_relations_handler() + self._msc3912_enabled = hs.config.experimental.msc3912_enabled def register(self, http_server: HttpServer) -> None: PATTERNS = "/rooms/(?P[^/]*)/redact/(?P[^/]*)" @@ -1045,20 +1048,46 @@ class RoomRedactEventRestServlet(TransactionRestServlet): content = parse_json_object_from_request(request) try: - ( - event, - _, - ) = await self.event_creation_handler.create_and_send_nonmember_event( - requester, - { - "type": EventTypes.Redaction, - "content": content, - "room_id": room_id, - "sender": requester.user.to_string(), - "redacts": event_id, - }, - txn_id=txn_id, - ) + with_relations = None + if self._msc3912_enabled and "org.matrix.msc3912.with_relations" in content: + with_relations = content["org.matrix.msc3912.with_relations"] + del content["org.matrix.msc3912.with_relations"] + + # Check if there's an existing event for this transaction now (even though + # create_and_send_nonmember_event also does it) because, if there's one, + # then we want to skip the call to redact_events_related_to. + event = None + if txn_id: + event = await self.event_creation_handler.get_event_from_transaction( + requester, txn_id, room_id + ) + + if event is None: + ( + event, + _, + ) = await self.event_creation_handler.create_and_send_nonmember_event( + requester, + { + "type": EventTypes.Redaction, + "content": content, + "room_id": room_id, + "sender": requester.user.to_string(), + "redacts": event_id, + }, + txn_id=txn_id, + ) + + if with_relations: + run_as_background_process( + "redact_related_events", + self._relation_handler.redact_events_related_to, + requester=requester, + event_id=event_id, + initial_redaction_event=event, + relation_types=with_relations, + ) + event_id = event.event_id except ShadowBanError: event_id = "$" + random_string(43) diff --git a/synapse/rest/client/versions.py b/synapse/rest/client/versions.py index 9b1b72c68a..180a11ef88 100644 --- a/synapse/rest/client/versions.py +++ b/synapse/rest/client/versions.py @@ -119,6 +119,8 @@ class VersionsRestServlet(RestServlet): # Adds support for simple HTTP rendezvous as per MSC3886 "org.matrix.msc3886": self.config.experimental.msc3886_endpoint is not None, + # Adds support for relation-based redactions as per MSC3912. + "org.matrix.msc3912": self.config.experimental.msc3912_enabled, }, }, ) diff --git a/synapse/storage/databases/main/relations.py b/synapse/storage/databases/main/relations.py index c022510e76..ca431002c8 100644 --- a/synapse/storage/databases/main/relations.py +++ b/synapse/storage/databases/main/relations.py @@ -295,6 +295,42 @@ class RelationsWorkerStore(SQLBaseStore): "get_recent_references_for_event", _get_recent_references_for_event_txn ) + async def get_all_relations_for_event_with_types( + self, + event_id: str, + relation_types: List[str], + ) -> List[str]: + """Get the event IDs of all events that have a relation to the given event with + one of the given relation types. + + Args: + event_id: The event for which to look for related events. + relation_types: The types of relations to look for. + + Returns: + A list of the IDs of the events that relate to the given event with one of + the given relation types. + """ + + def get_all_relation_ids_for_event_with_types_txn( + txn: LoggingTransaction, + ) -> List[str]: + rows = self.db_pool.simple_select_many_txn( + txn=txn, + table="event_relations", + column="relation_type", + iterable=relation_types, + keyvalues={"relates_to_id": event_id}, + retcols=["event_id"], + ) + + return [row["event_id"] for row in rows] + + return await self.db_pool.runInteraction( + desc="get_all_relation_ids_for_event_with_types", + func=get_all_relation_ids_for_event_with_types_txn, + ) + async def event_includes_relation(self, event_id: str) -> bool: """Check if the given event relates to another event. diff --git a/tests/rest/client/test_redactions.py b/tests/rest/client/test_redactions.py index be4c67d68e..5dfe44defb 100644 --- a/tests/rest/client/test_redactions.py +++ b/tests/rest/client/test_redactions.py @@ -11,17 +11,18 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. -from typing import List +from typing import List, Optional from twisted.test.proto_helpers import MemoryReactor +from synapse.api.constants import EventTypes, RelationTypes from synapse.rest import admin from synapse.rest.client import login, room, sync from synapse.server import HomeServer from synapse.types import JsonDict from synapse.util import Clock -from tests.unittest import HomeserverTestCase +from tests.unittest import HomeserverTestCase, override_config class RedactionsTestCase(HomeserverTestCase): @@ -67,7 +68,12 @@ class RedactionsTestCase(HomeserverTestCase): ) def _redact_event( - self, access_token: str, room_id: str, event_id: str, expect_code: int = 200 + self, + access_token: str, + room_id: str, + event_id: str, + expect_code: int = 200, + with_relations: Optional[List[str]] = None, ) -> JsonDict: """Helper function to send a redaction event. @@ -75,7 +81,13 @@ class RedactionsTestCase(HomeserverTestCase): """ path = "/_matrix/client/r0/rooms/%s/redact/%s" % (room_id, event_id) - channel = self.make_request("POST", path, content={}, access_token=access_token) + request_content = {} + if with_relations: + request_content["org.matrix.msc3912.with_relations"] = with_relations + + channel = self.make_request( + "POST", path, request_content, access_token=access_token + ) self.assertEqual(channel.code, expect_code) return channel.json_body @@ -201,3 +213,256 @@ class RedactionsTestCase(HomeserverTestCase): # These should all succeed, even though this would be denied by # the standard message ratelimiter self._redact_event(self.mod_access_token, self.room_id, msg_id) + + @override_config({"experimental_features": {"msc3912_enabled": True}}) + def test_redact_relations(self) -> None: + """Tests that we can redact the relations of an event at the same time as the + event itself. + """ + # Send a root event. + res = self.helper.send_event( + room_id=self.room_id, + type=EventTypes.Message, + content={"msgtype": "m.text", "body": "hello"}, + tok=self.mod_access_token, + ) + root_event_id = res["event_id"] + + # Send an edit to this root event. + res = self.helper.send_event( + room_id=self.room_id, + type=EventTypes.Message, + content={ + "body": " * hello world", + "m.new_content": { + "body": "hello world", + "msgtype": "m.text", + }, + "m.relates_to": { + "event_id": root_event_id, + "rel_type": RelationTypes.REPLACE, + }, + "msgtype": "m.text", + }, + tok=self.mod_access_token, + ) + edit_event_id = res["event_id"] + + # Also send a threaded message whose root is the same as the edit's. + res = self.helper.send_event( + room_id=self.room_id, + type=EventTypes.Message, + content={ + "msgtype": "m.text", + "body": "message 1", + "m.relates_to": { + "event_id": root_event_id, + "rel_type": RelationTypes.THREAD, + }, + }, + tok=self.mod_access_token, + ) + threaded_event_id = res["event_id"] + + # Also send a reaction, again with the same root. + res = self.helper.send_event( + room_id=self.room_id, + type=EventTypes.Reaction, + content={ + "m.relates_to": { + "rel_type": RelationTypes.ANNOTATION, + "event_id": root_event_id, + "key": "👍", + } + }, + tok=self.mod_access_token, + ) + reaction_event_id = res["event_id"] + + # Redact the root event, specifying that we also want to delete events that + # relate to it with m.replace. + self._redact_event( + self.mod_access_token, + self.room_id, + root_event_id, + with_relations=[ + RelationTypes.REPLACE, + RelationTypes.THREAD, + ], + ) + + # Check that the root event got redacted. + event_dict = self.helper.get_event( + self.room_id, root_event_id, self.mod_access_token + ) + self.assertIn("redacted_because", event_dict, event_dict) + + # Check that the edit got redacted. + event_dict = self.helper.get_event( + self.room_id, edit_event_id, self.mod_access_token + ) + self.assertIn("redacted_because", event_dict, event_dict) + + # Check that the threaded message got redacted. + event_dict = self.helper.get_event( + self.room_id, threaded_event_id, self.mod_access_token + ) + self.assertIn("redacted_because", event_dict, event_dict) + + # Check that the reaction did not get redacted. + event_dict = self.helper.get_event( + self.room_id, reaction_event_id, self.mod_access_token + ) + self.assertNotIn("redacted_because", event_dict, event_dict) + + @override_config({"experimental_features": {"msc3912_enabled": True}}) + def test_redact_relations_no_perms(self) -> None: + """Tests that, when redacting a message along with its relations, if not all + the related messages can be redacted because of insufficient permissions, the + server still redacts all the ones that can be. + """ + # Send a root event. + res = self.helper.send_event( + room_id=self.room_id, + type=EventTypes.Message, + content={ + "msgtype": "m.text", + "body": "root", + }, + tok=self.other_access_token, + ) + root_event_id = res["event_id"] + + # Send a first threaded message, this one from the moderator. We do this for the + # first message with the m.thread relation (and not the last one) to ensure + # that, when the server fails to redact it, it doesn't stop there, and it + # instead goes on to redact the other one. + res = self.helper.send_event( + room_id=self.room_id, + type=EventTypes.Message, + content={ + "msgtype": "m.text", + "body": "message 1", + "m.relates_to": { + "event_id": root_event_id, + "rel_type": RelationTypes.THREAD, + }, + }, + tok=self.mod_access_token, + ) + first_threaded_event_id = res["event_id"] + + # Send a second threaded message, this time from the user who'll perform the + # redaction. + res = self.helper.send_event( + room_id=self.room_id, + type=EventTypes.Message, + content={ + "msgtype": "m.text", + "body": "message 2", + "m.relates_to": { + "event_id": root_event_id, + "rel_type": RelationTypes.THREAD, + }, + }, + tok=self.other_access_token, + ) + second_threaded_event_id = res["event_id"] + + # Redact the thread's root, and request that all threaded messages are also + # redacted. Send that request from the non-mod user, so that the first threaded + # event cannot be redacted. + self._redact_event( + self.other_access_token, + self.room_id, + root_event_id, + with_relations=[RelationTypes.THREAD], + ) + + # Check that the thread root got redacted. + event_dict = self.helper.get_event( + self.room_id, root_event_id, self.other_access_token + ) + self.assertIn("redacted_because", event_dict, event_dict) + + # Check that the last message in the thread got redacted, despite failing to + # redact the one before it. + event_dict = self.helper.get_event( + self.room_id, second_threaded_event_id, self.other_access_token + ) + self.assertIn("redacted_because", event_dict, event_dict) + + # Check that the message that was sent into the tread by the mod user is not + # redacted. + event_dict = self.helper.get_event( + self.room_id, first_threaded_event_id, self.other_access_token + ) + self.assertIn("body", event_dict["content"], event_dict) + self.assertEqual("message 1", event_dict["content"]["body"]) + + @override_config({"experimental_features": {"msc3912_enabled": True}}) + def test_redact_relations_txn_id_reuse(self) -> None: + """Tests that redacting a message using a transaction ID, then reusing the same + transaction ID but providing an additional list of relations to redact, is + effectively a no-op. + """ + # Send a root event. + res = self.helper.send_event( + room_id=self.room_id, + type=EventTypes.Message, + content={ + "msgtype": "m.text", + "body": "root", + }, + tok=self.mod_access_token, + ) + root_event_id = res["event_id"] + + # Send a first threaded message. + res = self.helper.send_event( + room_id=self.room_id, + type=EventTypes.Message, + content={ + "msgtype": "m.text", + "body": "I'm in a thread!", + "m.relates_to": { + "event_id": root_event_id, + "rel_type": RelationTypes.THREAD, + }, + }, + tok=self.mod_access_token, + ) + threaded_event_id = res["event_id"] + + # Send a first redaction request which redacts only the root event. + channel = self.make_request( + method="PUT", + path=f"/rooms/{self.room_id}/redact/{root_event_id}/foo", + content={}, + access_token=self.mod_access_token, + ) + self.assertEqual(channel.code, 200) + + # Send a second redaction request which redacts the root event as well as + # threaded messages. + channel = self.make_request( + method="PUT", + path=f"/rooms/{self.room_id}/redact/{root_event_id}/foo", + content={"org.matrix.msc3912.with_relations": [RelationTypes.THREAD]}, + access_token=self.mod_access_token, + ) + self.assertEqual(channel.code, 200) + + # Check that the root event got redacted. + event_dict = self.helper.get_event( + self.room_id, root_event_id, self.mod_access_token + ) + self.assertIn("redacted_because", event_dict) + + # Check that the threaded message didn't get redacted (since that wasn't part of + # the original redaction). + event_dict = self.helper.get_event( + self.room_id, threaded_event_id, self.mod_access_token + ) + self.assertIn("body", event_dict["content"], event_dict) + self.assertEqual("I'm in a thread!", event_dict["content"]["body"]) diff --git a/tests/rest/client/utils.py b/tests/rest/client/utils.py index 706399fae5..8d6f2b6ff9 100644 --- a/tests/rest/client/utils.py +++ b/tests/rest/client/utils.py @@ -410,6 +410,43 @@ class RestHelper: return channel.json_body + def get_event( + self, + room_id: str, + event_id: str, + tok: Optional[str] = None, + expect_code: int = HTTPStatus.OK, + ) -> JsonDict: + """Request a specific event from the server. + + Args: + room_id: the room in which the event was sent. + event_id: the event's ID. + tok: the token to request the event with. + expect_code: the expected HTTP status for the response. + + Returns: + The event as a dict. + """ + path = f"/_matrix/client/v3/rooms/{room_id}/event/{event_id}" + if tok: + path = path + f"?access_token={tok}" + + channel = make_request( + self.hs.get_reactor(), + self.site, + "GET", + path, + ) + + assert channel.code == expect_code, "Expected: %d, got: %d, resp: %r" % ( + expect_code, + channel.code, + channel.result["body"], + ) + + return channel.json_body + def _read_write_state( self, room_id: str, -- cgit 1.5.1 From 8bcdd712b8ba471b3489d41e569276677cf6c2bd Mon Sep 17 00:00:00 2001 From: "dependabot[bot]" <49699333+dependabot[bot]@users.noreply.github.com> Date: Fri, 4 Nov 2022 18:43:14 +0000 Subject: Bump flake8-bugbear from 22.9.23 to 22.10.27 (#14329) Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: GitHub Actions Co-authored-by: Olivier Wilkinson (reivilibre) --- changelog.d/14329.misc | 1 + poetry.lock | 10 +++++----- synapse/handlers/presence.py | 6 ++++-- synapse/server.py | 2 +- synapse/storage/_base.py | 2 +- 5 files changed, 12 insertions(+), 9 deletions(-) create mode 100644 changelog.d/14329.misc (limited to 'synapse/handlers') diff --git a/changelog.d/14329.misc b/changelog.d/14329.misc new file mode 100644 index 0000000000..2f6bbd3af7 --- /dev/null +++ b/changelog.d/14329.misc @@ -0,0 +1 @@ +Bump flake8-bugbear from 22.9.23 to 22.10.27. diff --git a/poetry.lock b/poetry.lock index b945463299..f6e462e6ae 100644 --- a/poetry.lock +++ b/poetry.lock @@ -260,18 +260,18 @@ pyflakes = ">=2.4.0,<2.5.0" [[package]] name = "flake8-bugbear" -version = "22.9.23" +version = "22.10.27" description = "A plugin for flake8 finding likely bugs and design problems in your program. Contains warnings that don't belong in pyflakes and pycodestyle." category = "dev" optional = false -python-versions = ">=3.6" +python-versions = ">=3.7" [package.dependencies] attrs = ">=19.2.0" flake8 = ">=3.0.0" [package.extras] -dev = ["coverage", "hypothesis", "hypothesmith (>=0.2)", "pre-commit"] +dev = ["coverage", "hypothesis", "hypothesmith (>=0.2)", "pre-commit", "tox"] [[package]] name = "flake8-comprehensions" @@ -1829,8 +1829,8 @@ flake8 = [ {file = "flake8-4.0.1.tar.gz", hash = "sha256:806e034dda44114815e23c16ef92f95c91e4c71100ff52813adf7132a6ad870d"}, ] flake8-bugbear = [ - {file = "flake8-bugbear-22.9.23.tar.gz", hash = "sha256:17b9623325e6e0dcdcc80ed9e4aa811287fcc81d7e03313b8736ea5733759937"}, - {file = "flake8_bugbear-22.9.23-py3-none-any.whl", hash = "sha256:cd2779b2b7ada212d7a322814a1e5651f1868ab0d3f24cc9da66169ab8fda474"}, + {file = "flake8-bugbear-22.10.27.tar.gz", hash = "sha256:a6708608965c9e0de5fff13904fed82e0ba21ac929fe4896459226a797e11cd5"}, + {file = "flake8_bugbear-22.10.27-py3-none-any.whl", hash = "sha256:6ad0ab754507319060695e2f2be80e6d8977cfcea082293089a9226276bd825d"}, ] flake8-comprehensions = [ {file = "flake8-comprehensions-3.8.0.tar.gz", hash = "sha256:8e108707637b1d13734f38e03435984f6b7854fa6b5a4e34f93e69534be8e521"}, diff --git a/synapse/handlers/presence.py b/synapse/handlers/presence.py index 2670e561d7..0066d63987 100644 --- a/synapse/handlers/presence.py +++ b/synapse/handlers/presence.py @@ -256,7 +256,7 @@ class BasePresenceHandler(abc.ABC): with the app. """ - async def update_external_syncs_row( + async def update_external_syncs_row( # noqa: B027 (no-op by design) self, process_id: str, user_id: str, is_syncing: bool, sync_time_msec: int ) -> None: """Update the syncing users for an external process as a delta. @@ -272,7 +272,9 @@ class BasePresenceHandler(abc.ABC): sync_time_msec: Time in ms when the user was last syncing """ - async def update_external_syncs_clear(self, process_id: str) -> None: + async def update_external_syncs_clear( # noqa: B027 (no-op by design) + self, process_id: str + ) -> None: """Marks all users that had been marked as syncing by a given process as offline. diff --git a/synapse/server.py b/synapse/server.py index df3a1cb405..c4e025af22 100644 --- a/synapse/server.py +++ b/synapse/server.py @@ -315,7 +315,7 @@ class HomeServer(metaclass=abc.ABCMeta): if self.config.worker.run_background_tasks: self.setup_background_tasks() - def start_listening(self) -> None: + def start_listening(self) -> None: # noqa: B027 (no-op by design) """Start the HTTP, manhole, metrics, etc listeners Does nothing in this base class; overridden in derived classes to start the diff --git a/synapse/storage/_base.py b/synapse/storage/_base.py index bf42aeb8d1..69abf6fa87 100644 --- a/synapse/storage/_base.py +++ b/synapse/storage/_base.py @@ -50,7 +50,7 @@ class SQLBaseStore(metaclass=ABCMeta): self.external_cached_functions: Dict[str, CachedFunction] = {} - def process_replication_rows( + def process_replication_rows( # noqa: B027 (no-op by design) self, stream_name: str, instance_name: str, -- cgit 1.5.1 From 7894251bcea7714b47e3849e509ea717bb18e9f5 Mon Sep 17 00:00:00 2001 From: Shay Date: Mon, 7 Nov 2022 13:38:50 -0800 Subject: Correctly create power level event during initial room creation (#14361) --- changelog.d/14361.bugfix | 1 + synapse/handlers/room.py | 25 +++++++++++++++++++++++-- tests/rest/client/test_rooms.py | 4 ++-- 3 files changed, 26 insertions(+), 4 deletions(-) create mode 100644 changelog.d/14361.bugfix (limited to 'synapse/handlers') diff --git a/changelog.d/14361.bugfix b/changelog.d/14361.bugfix new file mode 100644 index 0000000000..33ba1d92af --- /dev/null +++ b/changelog.d/14361.bugfix @@ -0,0 +1 @@ +Fix a bug introduced in v1.71.0rc1 where the power level event was incorrectly created during initial room creation. \ No newline at end of file diff --git a/synapse/handlers/room.py b/synapse/handlers/room.py index f10cfca073..66a50bca6e 100644 --- a/synapse/handlers/room.py +++ b/synapse/handlers/room.py @@ -1080,6 +1080,19 @@ class RoomCreationHandler: for_batch: bool, **kwargs: Any, ) -> Tuple[EventBase, synapse.events.snapshot.EventContext]: + """ + Creates an event and associated event context. + Args: + etype: the type of event to be created + content: content of the event + for_batch: whether the event is being created for batch persisting. If + bool for_batch is true, this will create an event using the prev_event_ids, + and will create an event context for the event using the parameters state_map + and current_state_group, thus these parameters must be provided in this + case if for_batch is True. The subsequently created event and context + are suitable for being batched up and bulk persisted to the database + with other similarly created events. + """ nonlocal depth nonlocal prev_event @@ -1139,13 +1152,21 @@ class RoomCreationHandler: depth += 1 state_map[(EventTypes.Member, creator.user.to_string())] = member_event_id + # we need the state group of the membership event as it is the current state group + event_to_state = ( + await self._storage_controllers.state.get_state_group_for_events( + [member_event_id] + ) + ) + current_state_group = event_to_state[member_event_id] + events_to_send = [] # We treat the power levels override specially as this needs to be one # of the first events that get sent into a room. pl_content = initial_state.pop((EventTypes.PowerLevels, ""), None) if pl_content is not None: power_event, power_context = await create_event( - EventTypes.PowerLevels, pl_content, False + EventTypes.PowerLevels, pl_content, True ) current_state_group = power_context._state_group events_to_send.append((power_event, power_context)) @@ -1194,7 +1215,7 @@ class RoomCreationHandler: pl_event, pl_context = await create_event( EventTypes.PowerLevels, power_level_content, - False, + True, ) current_state_group = pl_context._state_group events_to_send.append((pl_event, pl_context)) diff --git a/tests/rest/client/test_rooms.py b/tests/rest/client/test_rooms.py index 1084d4ad9d..e919e089cb 100644 --- a/tests/rest/client/test_rooms.py +++ b/tests/rest/client/test_rooms.py @@ -715,7 +715,7 @@ class RoomsCreateTestCase(RoomBase): self.assertEqual(HTTPStatus.OK, channel.code, channel.result) self.assertTrue("room_id" in channel.json_body) assert channel.resource_usage is not None - self.assertEqual(34, channel.resource_usage.db_txn_count) + self.assertEqual(33, channel.resource_usage.db_txn_count) def test_post_room_initial_state(self) -> None: # POST with initial_state config key, expect new room id @@ -728,7 +728,7 @@ class RoomsCreateTestCase(RoomBase): self.assertEqual(HTTPStatus.OK, channel.code, channel.result) self.assertTrue("room_id" in channel.json_body) assert channel.resource_usage is not None - self.assertEqual(37, channel.resource_usage.db_txn_count) + self.assertEqual(36, channel.resource_usage.db_txn_count) def test_post_room_visibility_key(self) -> None: # POST with visibility config key, expect new room id -- cgit 1.5.1 From d10a85ec9eac6f31aa82a5f07d74e5914b18b320 Mon Sep 17 00:00:00 2001 From: David Robertson Date: Thu, 10 Nov 2022 12:17:46 +0000 Subject: Quieter logging for stateres failure at missing prev events (#14346) --- changelog.d/14346.misc | 1 + synapse/handlers/federation_event.py | 5 ++--- 2 files changed, 3 insertions(+), 3 deletions(-) create mode 100644 changelog.d/14346.misc (limited to 'synapse/handlers') diff --git a/changelog.d/14346.misc b/changelog.d/14346.misc new file mode 100644 index 0000000000..9833b0733a --- /dev/null +++ b/changelog.d/14346.misc @@ -0,0 +1 @@ +Concisely log a failure to resolve state due to missing `prev_events`. diff --git a/synapse/handlers/federation_event.py b/synapse/handlers/federation_event.py index 9ca5df7c78..f7223b03c3 100644 --- a/synapse/handlers/federation_event.py +++ b/synapse/handlers/federation_event.py @@ -1065,10 +1065,9 @@ class FederationEventHandler: state_res_store=StateResolutionStore(self._store), ) - except Exception: + except Exception as e: logger.warning( - "Error attempting to resolve state at missing prev_events", - exc_info=True, + "Error attempting to resolve state at missing prev_events: %s", e ) raise FederationError( "ERROR", -- cgit 1.5.1 From 1eed795fc56d95df3968e37f3a4db92f24513e15 Mon Sep 17 00:00:00 2001 From: David Robertson Date: Tue, 15 Nov 2022 17:35:19 +0000 Subject: Include heroes in partial join responses' state (#14442) * Pull out hero selection logic * Include heroes in partial join response's state * Changelog * Fixup trial test * Remove TODO --- changelog.d/14442.feature | 1 + synapse/federation/federation_server.py | 23 +++++++++++++++++---- synapse/handlers/sync.py | 20 +++---------------- synapse/storage/databases/main/roommember.py | 30 ++++++++++++++++++++++++++++ tests/federation/test_federation_server.py | 11 ++++++---- 5 files changed, 60 insertions(+), 25 deletions(-) create mode 100644 changelog.d/14442.feature (limited to 'synapse/handlers') diff --git a/changelog.d/14442.feature b/changelog.d/14442.feature new file mode 100644 index 0000000000..917e7edfb3 --- /dev/null +++ b/changelog.d/14442.feature @@ -0,0 +1 @@ +Faster joins: include heroes' membership events in the partial join response, for rooms without a name or canonical alias. diff --git a/synapse/federation/federation_server.py b/synapse/federation/federation_server.py index 59e351595b..bb20af6e91 100644 --- a/synapse/federation/federation_server.py +++ b/synapse/federation/federation_server.py @@ -74,6 +74,8 @@ from synapse.replication.http.federation import ( ) from synapse.storage.databases.main.events import PartialStateConflictError from synapse.storage.databases.main.lock import Lock +from synapse.storage.databases.main.roommember import extract_heroes_from_room_summary +from synapse.storage.roommember import MemberSummary from synapse.types import JsonDict, StateMap, get_domain_from_id from synapse.util import json_decoder, unwrapFirstError from synapse.util.async_helpers import Linearizer, concurrently_execute, gather_results @@ -691,8 +693,9 @@ class FederationServer(FederationBase): state_event_ids: Collection[str] servers_in_room: Optional[Collection[str]] if caller_supports_partial_state: + summary = await self.store.get_room_summary(room_id) state_event_ids = _get_event_ids_for_partial_state_join( - event, prev_state_ids + event, prev_state_ids, summary ) servers_in_room = await self.state.get_hosts_in_room_at_events( room_id, event_ids=event.prev_event_ids() @@ -1495,6 +1498,7 @@ class FederationHandlerRegistry: def _get_event_ids_for_partial_state_join( join_event: EventBase, prev_state_ids: StateMap[str], + summary: Dict[str, MemberSummary], ) -> Collection[str]: """Calculate state to be retuned in a partial_state send_join @@ -1521,8 +1525,19 @@ def _get_event_ids_for_partial_state_join( if current_membership_event_id is not None: state_event_ids.add(current_membership_event_id) - # TODO: return a few more members: - # - those with invites - # - those that are kicked? / banned + name_id = prev_state_ids.get((EventTypes.Name, "")) + canonical_alias_id = prev_state_ids.get((EventTypes.CanonicalAlias, "")) + if not name_id and not canonical_alias_id: + # Also include the hero members of the room (for DM rooms without a title). + # To do this properly, we should select the correct subset of membership events + # from `prev_state_ids`. Instead, we are lazier and use the (cached) + # `get_room_summary` function, which is based on the current state of the room. + # This introduces races; we choose to ignore them because a) they should be rare + # and b) even if it's wrong, joining servers will get the full state eventually. + heroes = extract_heroes_from_room_summary(summary, join_event.state_key) + for hero in heroes: + membership_event_id = prev_state_ids.get((EventTypes.Member, hero)) + if membership_event_id: + state_event_ids.add(membership_event_id) return state_event_ids diff --git a/synapse/handlers/sync.py b/synapse/handlers/sync.py index 1db5d68021..259456b55d 100644 --- a/synapse/handlers/sync.py +++ b/synapse/handlers/sync.py @@ -41,6 +41,7 @@ from synapse.logging.context import current_context from synapse.logging.opentracing import SynapseTags, log_kv, set_tag, start_active_span from synapse.push.clientformat import format_push_rules_for_user from synapse.storage.databases.main.event_push_actions import RoomNotifCounts +from synapse.storage.databases.main.roommember import extract_heroes_from_room_summary from synapse.storage.roommember import MemberSummary from synapse.storage.state import StateFilter from synapse.types import ( @@ -805,18 +806,6 @@ class SyncHandler: if canonical_alias and canonical_alias.content.get("alias"): return summary - me = sync_config.user.to_string() - - joined_user_ids = [ - r[0] for r in details.get(Membership.JOIN, empty_ms).members if r[0] != me - ] - invited_user_ids = [ - r[0] for r in details.get(Membership.INVITE, empty_ms).members if r[0] != me - ] - gone_user_ids = [ - r[0] for r in details.get(Membership.LEAVE, empty_ms).members if r[0] != me - ] + [r[0] for r in details.get(Membership.BAN, empty_ms).members if r[0] != me] - # FIXME: only build up a member_ids list for our heroes member_ids = {} for membership in ( @@ -828,11 +817,8 @@ class SyncHandler: for user_id, event_id in details.get(membership, empty_ms).members: member_ids[user_id] = event_id - # FIXME: order by stream ordering rather than as returned by SQL - if joined_user_ids or invited_user_ids: - summary["m.heroes"] = sorted(joined_user_ids + invited_user_ids)[0:5] - else: - summary["m.heroes"] = sorted(gone_user_ids)[0:5] + me = sync_config.user.to_string() + summary["m.heroes"] = extract_heroes_from_room_summary(details, me) if not sync_config.filter_collection.lazy_load_members(): return summary diff --git a/synapse/storage/databases/main/roommember.py b/synapse/storage/databases/main/roommember.py index e56a13f21e..f02c1d7ea7 100644 --- a/synapse/storage/databases/main/roommember.py +++ b/synapse/storage/databases/main/roommember.py @@ -1517,6 +1517,36 @@ class RoomMemberStore( await self.db_pool.runInteraction("forget_membership", f) +def extract_heroes_from_room_summary( + details: Mapping[str, MemberSummary], me: str +) -> List[str]: + """Determine the users that represent a room, from the perspective of the `me` user. + + The rules which say which users we select are specified in the "Room Summary" + section of + https://spec.matrix.org/v1.4/client-server-api/#get_matrixclientv3sync + + Returns a list (possibly empty) of heroes' mxids. + """ + empty_ms = MemberSummary([], 0) + + joined_user_ids = [ + r[0] for r in details.get(Membership.JOIN, empty_ms).members if r[0] != me + ] + invited_user_ids = [ + r[0] for r in details.get(Membership.INVITE, empty_ms).members if r[0] != me + ] + gone_user_ids = [ + r[0] for r in details.get(Membership.LEAVE, empty_ms).members if r[0] != me + ] + [r[0] for r in details.get(Membership.BAN, empty_ms).members if r[0] != me] + + # FIXME: order by stream ordering rather than as returned by SQL + if joined_user_ids or invited_user_ids: + return sorted(joined_user_ids + invited_user_ids)[0:5] + else: + return sorted(gone_user_ids)[0:5] + + @attr.s(slots=True, auto_attribs=True) class _JoinedHostsCache: """The cached data used by the `_get_joined_hosts_cache`.""" diff --git a/tests/federation/test_federation_server.py b/tests/federation/test_federation_server.py index 3a6ef221ae..177e5b5afc 100644 --- a/tests/federation/test_federation_server.py +++ b/tests/federation/test_federation_server.py @@ -212,7 +212,7 @@ class SendJoinFederationTests(unittest.FederatingHomeserverTestCase): self.assertEqual(r[("m.room.member", joining_user)].membership, "join") @override_config({"experimental_features": {"msc3706_enabled": True}}) - def test_send_join_partial_state(self): + def test_send_join_partial_state(self) -> None: """When MSC3706 support is enabled, /send_join should return partial state""" joining_user = "@misspiggy:" + self.OTHER_SERVER_NAME join_result = self._make_join(joining_user) @@ -240,6 +240,9 @@ class SendJoinFederationTests(unittest.FederatingHomeserverTestCase): ("m.room.power_levels", ""), ("m.room.join_rules", ""), ("m.room.history_visibility", ""), + # Users included here because they're heroes. + ("m.room.member", "@kermit:test"), + ("m.room.member", "@fozzie:test"), ], ) @@ -249,9 +252,9 @@ class SendJoinFederationTests(unittest.FederatingHomeserverTestCase): ] self.assertCountEqual( returned_auth_chain_events, - [ - ("m.room.member", "@kermit:test"), - ], + # TODO: change the test so that we get at least one event in the auth chain + # here. + [], ) # the room should show that the new user is a member -- cgit 1.5.1 From 945a0928c793c0bd8573e179583d983187e5f392 Mon Sep 17 00:00:00 2001 From: Erik Johnston Date: Wed, 16 Nov 2022 12:09:33 +0000 Subject: Don't filter state in /context response (#14461) We don't filter state usually, so doing so here is a waste of time. This is not much of an issue for clients that enable lazy loading of members, since there will be fewer state events. --- changelog.d/14461.misc | 1 + synapse/handlers/room.py | 2 +- 2 files changed, 2 insertions(+), 1 deletion(-) create mode 100644 changelog.d/14461.misc (limited to 'synapse/handlers') diff --git a/changelog.d/14461.misc b/changelog.d/14461.misc new file mode 100644 index 0000000000..cdfa577a4c --- /dev/null +++ b/changelog.d/14461.misc @@ -0,0 +1 @@ +Improve performance of `/context` in large rooms. diff --git a/synapse/handlers/room.py b/synapse/handlers/room.py index 66a50bca6e..6dcfd86fdf 100644 --- a/synapse/handlers/room.py +++ b/synapse/handlers/room.py @@ -1451,7 +1451,7 @@ class RoomContextHandler: events_before=events_before, event=event, events_after=events_after, - state=await filter_evts(state_events), + state=state_events, aggregations=aggregations, start=await token.copy_and_replace( StreamKeyType.ROOM, results.start -- cgit 1.5.1 From d8cc86eff484b6f570f55a5badb337080c6e4dcd Mon Sep 17 00:00:00 2001 From: Patrick Cloke Date: Wed, 16 Nov 2022 10:25:24 -0500 Subject: Remove redundant types from comments. (#14412) Remove type hints from comments which have been added as Python type hints. This helps avoid drift between comments and reality, as well as removing redundant information. Also adds some missing type hints which were simple to fill in. --- changelog.d/14412.misc | 1 + synapse/api/errors.py | 2 +- synapse/config/logger.py | 5 ++- synapse/crypto/keyring.py | 9 +++-- synapse/events/__init__.py | 3 +- synapse/federation/transport/client.py | 11 +++--- synapse/federation/transport/server/_base.py | 4 +-- synapse/handlers/e2e_keys.py | 2 +- synapse/handlers/e2e_room_keys.py | 5 +-- synapse/handlers/federation.py | 4 +-- synapse/handlers/identity.py | 2 +- synapse/handlers/oidc.py | 2 +- synapse/handlers/presence.py | 2 +- synapse/handlers/saml.py | 4 +-- synapse/http/additional_resource.py | 3 +- synapse/http/federation/matrix_federation_agent.py | 9 +++-- synapse/http/matrixfederationclient.py | 3 +- synapse/http/proxyagent.py | 20 +++++------ synapse/http/server.py | 2 +- synapse/http/site.py | 2 +- synapse/logging/context.py | 39 +++++++++++----------- synapse/logging/opentracing.py | 4 +-- synapse/module_api/__init__.py | 7 ++-- synapse/replication/http/_base.py | 2 +- synapse/rest/admin/users.py | 5 +-- synapse/rest/client/login.py | 2 +- synapse/rest/media/v1/media_repository.py | 4 +-- synapse/rest/media/v1/thumbnailer.py | 4 +-- synapse/server_notices/consent_server_notices.py | 5 ++- .../resource_limits_server_notices.py | 12 ++++--- synapse/storage/controllers/persist_events.py | 5 ++- synapse/storage/databases/main/devices.py | 2 +- synapse/storage/databases/main/e2e_room_keys.py | 8 ++--- synapse/storage/databases/main/end_to_end_keys.py | 7 ++-- synapse/storage/databases/main/events.py | 22 ++++++------ synapse/storage/databases/main/events_worker.py | 2 +- .../storage/databases/main/monthly_active_users.py | 8 ++--- synapse/storage/databases/main/registration.py | 6 ++-- synapse/storage/databases/main/room.py | 8 +++-- synapse/storage/databases/main/user_directory.py | 9 +++-- synapse/types.py | 4 +-- synapse/util/async_helpers.py | 3 +- synapse/util/caches/__init__.py | 2 +- synapse/util/caches/deferred_cache.py | 2 +- synapse/util/caches/dictionary_cache.py | 9 ++--- synapse/util/caches/expiringcache.py | 2 +- synapse/util/caches/lrucache.py | 8 ++--- synapse/util/ratelimitutils.py | 2 +- synapse/util/threepids.py | 2 +- synapse/util/wheel_timer.py | 4 +-- tests/http/__init__.py | 7 ++-- tests/replication/slave/storage/test_events.py | 7 ++-- tests/replication/test_multi_media_repo.py | 14 ++++---- .../test_resource_limits_server_notices.py | 10 +++--- tests/unittest.py | 18 +++++----- 55 files changed, 174 insertions(+), 176 deletions(-) create mode 100644 changelog.d/14412.misc (limited to 'synapse/handlers') diff --git a/changelog.d/14412.misc b/changelog.d/14412.misc new file mode 100644 index 0000000000..4da061d461 --- /dev/null +++ b/changelog.d/14412.misc @@ -0,0 +1 @@ +Remove duplicated type information from type hints. diff --git a/synapse/api/errors.py b/synapse/api/errors.py index 400dd12aba..e2cfcea0f2 100644 --- a/synapse/api/errors.py +++ b/synapse/api/errors.py @@ -713,7 +713,7 @@ class HttpResponseException(CodeMessageException): set to the reason code from the HTTP response. Returns: - SynapseError: + The error converted to a SynapseError. """ # try to parse the body as json, to get better errcode/msg, but # default to M_UNKNOWN with the HTTP status as the error text diff --git a/synapse/config/logger.py b/synapse/config/logger.py index 94d1150415..5468b963a2 100644 --- a/synapse/config/logger.py +++ b/synapse/config/logger.py @@ -317,10 +317,9 @@ def setup_logging( Set up the logging subsystem. Args: - config (LoggingConfig | synapse.config.worker.WorkerConfig): - configuration data + config: configuration data - use_worker_options (bool): True to use the 'worker_log_config' option + use_worker_options: True to use the 'worker_log_config' option instead of 'log_config'. logBeginner: The Twisted logBeginner to use. diff --git a/synapse/crypto/keyring.py b/synapse/crypto/keyring.py index c88afb2986..dd9b8089ec 100644 --- a/synapse/crypto/keyring.py +++ b/synapse/crypto/keyring.py @@ -213,7 +213,7 @@ class Keyring: def verify_json_objects_for_server( self, server_and_json: Iterable[Tuple[str, dict, int]] - ) -> List[defer.Deferred]: + ) -> List["defer.Deferred[None]"]: """Bulk verifies signatures of json objects, bulk fetching keys as necessary. @@ -226,10 +226,9 @@ class Keyring: valid. Returns: - List: for each input triplet, a deferred indicating success - or failure to verify each json object's signature for the given - server_name. The deferreds run their callbacks in the sentinel - logcontext. + For each input triplet, a deferred indicating success or failure to + verify each json object's signature for the given server_name. The + deferreds run their callbacks in the sentinel logcontext. """ return [ run_in_background( diff --git a/synapse/events/__init__.py b/synapse/events/__init__.py index 030c3ca408..8aca9a3ab9 100644 --- a/synapse/events/__init__.py +++ b/synapse/events/__init__.py @@ -597,8 +597,7 @@ def _event_type_from_format_version( format_version: The event format version Returns: - type: A type that can be initialized as per the initializer of - `FrozenEvent` + A type that can be initialized as per the initializer of `FrozenEvent` """ if format_version == EventFormatVersions.ROOM_V1_V2: diff --git a/synapse/federation/transport/client.py b/synapse/federation/transport/client.py index cd39d4d111..a3cfc701cd 100644 --- a/synapse/federation/transport/client.py +++ b/synapse/federation/transport/client.py @@ -280,12 +280,11 @@ class TransportLayerClient: Note that this does not append any events to any graphs. Args: - destination (str): address of remote homeserver - room_id (str): room to join/leave - user_id (str): user to be joined/left - membership (str): one of join/leave - params (dict[str, str|Iterable[str]]): Query parameters to include in the - request. + destination: address of remote homeserver + room_id: room to join/leave + user_id: user to be joined/left + membership: one of join/leave + params: Query parameters to include in the request. Returns: Succeeds when we get a 2xx HTTP response. The result diff --git a/synapse/federation/transport/server/_base.py b/synapse/federation/transport/server/_base.py index 1db8009d6c..cdaf0d5de7 100644 --- a/synapse/federation/transport/server/_base.py +++ b/synapse/federation/transport/server/_base.py @@ -224,10 +224,10 @@ class BaseFederationServlet: With arguments: - origin (unicode|None): The authenticated server_name of the calling server, + origin (str|None): The authenticated server_name of the calling server, unless REQUIRE_AUTH is set to False and authentication failed. - content (unicode|None): decoded json body of the request. None if the + content (str|None): decoded json body of the request. None if the request was a GET. query (dict[bytes, list[bytes]]): Query params from the request. url-decoded diff --git a/synapse/handlers/e2e_keys.py b/synapse/handlers/e2e_keys.py index a9912c467d..bf1221f523 100644 --- a/synapse/handlers/e2e_keys.py +++ b/synapse/handlers/e2e_keys.py @@ -870,7 +870,7 @@ class E2eKeysHandler: - signatures of the user's master key by the user's devices. Args: - user_id (string): the user uploading the keys + user_id: the user uploading the keys signatures (dict[string, dict]): map of devices to signed keys Returns: diff --git a/synapse/handlers/e2e_room_keys.py b/synapse/handlers/e2e_room_keys.py index 28dc08c22a..83f53ceb88 100644 --- a/synapse/handlers/e2e_room_keys.py +++ b/synapse/handlers/e2e_room_keys.py @@ -377,8 +377,9 @@ class E2eRoomKeysHandler: """Deletes a given version of the user's e2e_room_keys backup Args: - user_id(str): the user whose current backup version we're deleting - version(str): the version id of the backup being deleted + user_id: the user whose current backup version we're deleting + version: Optional. the version ID of the backup version we're deleting + If missing, we delete the current backup version info. Raises: NotFoundError: if this backup version doesn't exist """ diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index 5fc3b8bc8c..188f0956ef 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -1596,8 +1596,8 @@ class FederationHandler: Fetch the complexity of a remote room over federation. Args: - remote_room_hosts (list[str]): The remote servers to ask. - room_id (str): The room ID to ask about. + remote_room_hosts: The remote servers to ask. + room_id: The room ID to ask about. Returns: Dict contains the complexity diff --git a/synapse/handlers/identity.py b/synapse/handlers/identity.py index 93d09e9939..848e46eb9b 100644 --- a/synapse/handlers/identity.py +++ b/synapse/handlers/identity.py @@ -711,7 +711,7 @@ class IdentityHandler: inviter_display_name: The current display name of the inviter. inviter_avatar_url: The URL of the inviter's avatar. - id_access_token (str): The access token to authenticate to the identity + id_access_token: The access token to authenticate to the identity server with Returns: diff --git a/synapse/handlers/oidc.py b/synapse/handlers/oidc.py index 867973dcca..41c675f408 100644 --- a/synapse/handlers/oidc.py +++ b/synapse/handlers/oidc.py @@ -787,7 +787,7 @@ class OidcProvider: Must include an ``access_token`` field. Returns: - UserInfo: an object representing the user. + an object representing the user. """ logger.debug("Using the OAuth2 access_token to request userinfo") metadata = await self.load_metadata() diff --git a/synapse/handlers/presence.py b/synapse/handlers/presence.py index 0066d63987..b7bc787636 100644 --- a/synapse/handlers/presence.py +++ b/synapse/handlers/presence.py @@ -201,7 +201,7 @@ class BasePresenceHandler(abc.ABC): """Get the current presence state for multiple users. Returns: - dict: `user_id` -> `UserPresenceState` + A mapping of `user_id` -> `UserPresenceState` """ states = {} missing = [] diff --git a/synapse/handlers/saml.py b/synapse/handlers/saml.py index 9602f0d0bb..874860d461 100644 --- a/synapse/handlers/saml.py +++ b/synapse/handlers/saml.py @@ -441,7 +441,7 @@ class DefaultSamlMappingProvider: client_redirect_url: where the client wants to redirect to Returns: - dict: A dict containing new user attributes. Possible keys: + A dict containing new user attributes. Possible keys: * mxid_localpart (str): Required. The localpart of the user's mxid * displayname (str): The displayname of the user * emails (list[str]): Any emails for the user @@ -483,7 +483,7 @@ class DefaultSamlMappingProvider: Args: config: A dictionary containing configuration options for this provider Returns: - SamlConfig: A custom config object for this module + A custom config object for this module """ # Parse config options and use defaults where necessary mxid_source_attribute = config.get("mxid_source_attribute", "uid") diff --git a/synapse/http/additional_resource.py b/synapse/http/additional_resource.py index 6a9f6635d2..8729630581 100644 --- a/synapse/http/additional_resource.py +++ b/synapse/http/additional_resource.py @@ -45,8 +45,7 @@ class AdditionalResource(DirectServeJsonResource): Args: hs: homeserver - handler ((twisted.web.server.Request) -> twisted.internet.defer.Deferred): - function to be called to handle the request. + handler: function to be called to handle the request. """ super().__init__() self._handler = handler diff --git a/synapse/http/federation/matrix_federation_agent.py b/synapse/http/federation/matrix_federation_agent.py index 2f0177f1e2..0359231e7d 100644 --- a/synapse/http/federation/matrix_federation_agent.py +++ b/synapse/http/federation/matrix_federation_agent.py @@ -155,11 +155,10 @@ class MatrixFederationAgent: a file for a file upload). Or None if the request is to have no body. Returns: - Deferred[twisted.web.iweb.IResponse]: - fires when the header of the response has been received (regardless of the - response status code). Fails if there is any problem which prevents that - response from being received (including problems that prevent the request - from being sent). + A deferred which fires when the header of the response has been received + (regardless of the response status code). Fails if there is any problem + which prevents that response from being received (including problems that + prevent the request from being sent). """ # We use urlparse as that will set `port` to None if there is no # explicit port. diff --git a/synapse/http/matrixfederationclient.py b/synapse/http/matrixfederationclient.py index 3c35b1d2c7..b92f1d3d1a 100644 --- a/synapse/http/matrixfederationclient.py +++ b/synapse/http/matrixfederationclient.py @@ -951,8 +951,7 @@ class MatrixFederationHttpClient: args: query params Returns: - dict|list: Succeeds when we get a 2xx HTTP response. The - result will be the decoded JSON body. + Succeeds when we get a 2xx HTTP response. The result will be the decoded JSON body. Raises: HttpResponseException: If we get an HTTP response code >= 300 diff --git a/synapse/http/proxyagent.py b/synapse/http/proxyagent.py index 1f8227896f..18899bc6d1 100644 --- a/synapse/http/proxyagent.py +++ b/synapse/http/proxyagent.py @@ -34,7 +34,7 @@ from twisted.web.client import ( ) from twisted.web.error import SchemeNotSupported from twisted.web.http_headers import Headers -from twisted.web.iweb import IAgent, IBodyProducer, IPolicyForHTTPS +from twisted.web.iweb import IAgent, IBodyProducer, IPolicyForHTTPS, IResponse from synapse.http import redact_uri from synapse.http.connectproxyclient import HTTPConnectProxyEndpoint, ProxyCredentials @@ -134,7 +134,7 @@ class ProxyAgent(_AgentBase): uri: bytes, headers: Optional[Headers] = None, bodyProducer: Optional[IBodyProducer] = None, - ) -> defer.Deferred: + ) -> "defer.Deferred[IResponse]": """ Issue a request to the server indicated by the given uri. @@ -157,17 +157,17 @@ class ProxyAgent(_AgentBase): a file upload). Or, None if the request is to have no body. Returns: - Deferred[IResponse]: completes when the header of the response has - been received (regardless of the response status code). + A deferred which completes when the header of the response has + been received (regardless of the response status code). - Can fail with: - SchemeNotSupported: if the uri is not http or https + Can fail with: + SchemeNotSupported: if the uri is not http or https - twisted.internet.error.TimeoutError if the server we are connecting - to (proxy or destination) does not accept a connection before - connectTimeout. + twisted.internet.error.TimeoutError if the server we are connecting + to (proxy or destination) does not accept a connection before + connectTimeout. - ... other things too. + ... other things too. """ uri = uri.strip() if not _VALID_URI.match(uri): diff --git a/synapse/http/server.py b/synapse/http/server.py index b26e34bceb..051a1899a0 100644 --- a/synapse/http/server.py +++ b/synapse/http/server.py @@ -267,7 +267,7 @@ class HttpServer(Protocol): request. The first argument will be the request object and subsequent arguments will be any matched groups from the regex. This should return either tuple of (code, response), or None. - servlet_classname (str): The name of the handler to be used in prometheus + servlet_classname: The name of the handler to be used in prometheus and opentracing logs. """ diff --git a/synapse/http/site.py b/synapse/http/site.py index 3dbd541fed..6a1dbf7f33 100644 --- a/synapse/http/site.py +++ b/synapse/http/site.py @@ -400,7 +400,7 @@ class SynapseRequest(Request): be sure to call finished_processing. Args: - servlet_name (str): the name of the servlet which will be + servlet_name: the name of the servlet which will be processing this request. This is used in the metrics. It is possible to update this afterwards by updating diff --git a/synapse/logging/context.py b/synapse/logging/context.py index 6a08ffed64..f62bea968f 100644 --- a/synapse/logging/context.py +++ b/synapse/logging/context.py @@ -117,8 +117,7 @@ class ContextResourceUsage: """Create a new ContextResourceUsage Args: - copy_from (ContextResourceUsage|None): if not None, an object to - copy stats from + copy_from: if not None, an object to copy stats from """ if copy_from is None: self.reset() @@ -162,7 +161,7 @@ class ContextResourceUsage: """Add another ContextResourceUsage's stats to this one's. Args: - other (ContextResourceUsage): the other resource usage object + other: the other resource usage object """ self.ru_utime += other.ru_utime self.ru_stime += other.ru_stime @@ -342,7 +341,7 @@ class LoggingContext: called directly. Returns: - LoggingContext: the current logging context + The current logging context """ warnings.warn( "synapse.logging.context.LoggingContext.current_context() is deprecated " @@ -362,7 +361,8 @@ class LoggingContext: called directly. Args: - context(LoggingContext): The context to activate. + context: The context to activate. + Returns: The context that was previously active """ @@ -474,8 +474,7 @@ class LoggingContext: """Get resources used by this logcontext so far. Returns: - ContextResourceUsage: a *copy* of the object tracking resource - usage so far + A *copy* of the object tracking resource usage so far """ # we always return a copy, for consistency res = self._resource_usage.copy() @@ -663,7 +662,8 @@ def current_context() -> LoggingContextOrSentinel: def set_current_context(context: LoggingContextOrSentinel) -> LoggingContextOrSentinel: """Set the current logging context in thread local storage Args: - context(LoggingContext): The context to activate. + context: The context to activate. + Returns: The context that was previously active """ @@ -700,7 +700,7 @@ def nested_logging_context(suffix: str) -> LoggingContext: suffix: suffix to add to the parent context's 'name'. Returns: - LoggingContext: new logging context. + A new logging context. """ curr_context = current_context() if not curr_context: @@ -898,20 +898,19 @@ def defer_to_thread( on it. Args: - reactor (twisted.internet.base.ReactorBase): The reactor in whose main thread - the Deferred will be invoked, and whose threadpool we should use for the - function. + reactor: The reactor in whose main thread the Deferred will be invoked, + and whose threadpool we should use for the function. Normally this will be hs.get_reactor(). - f (callable): The function to call. + f: The function to call. args: positional arguments to pass to f. kwargs: keyword arguments to pass to f. Returns: - Deferred: A Deferred which fires a callback with the result of `f`, or an + A Deferred which fires a callback with the result of `f`, or an errback if `f` throws an exception. """ return defer_to_threadpool(reactor, reactor.getThreadPool(), f, *args, **kwargs) @@ -939,20 +938,20 @@ def defer_to_threadpool( on it. Args: - reactor (twisted.internet.base.ReactorBase): The reactor in whose main thread - the Deferred will be invoked. Normally this will be hs.get_reactor(). + reactor: The reactor in whose main thread the Deferred will be invoked. + Normally this will be hs.get_reactor(). - threadpool (twisted.python.threadpool.ThreadPool): The threadpool to use for - running `f`. Normally this will be hs.get_reactor().getThreadPool(). + threadpool: The threadpool to use for running `f`. Normally this will be + hs.get_reactor().getThreadPool(). - f (callable): The function to call. + f: The function to call. args: positional arguments to pass to f. kwargs: keyword arguments to pass to f. Returns: - Deferred: A Deferred which fires a callback with the result of `f`, or an + A Deferred which fires a callback with the result of `f`, or an errback if `f` throws an exception. """ curr_context = current_context() diff --git a/synapse/logging/opentracing.py b/synapse/logging/opentracing.py index 8ce5a2a338..b69060854f 100644 --- a/synapse/logging/opentracing.py +++ b/synapse/logging/opentracing.py @@ -721,7 +721,7 @@ def inject_header_dict( destination: address of entity receiving the span context. Must be given unless check_destination is False. The context will only be injected if the destination matches the opentracing whitelist - check_destination (bool): If false, destination will be ignored and the context + check_destination: If false, destination will be ignored and the context will always be injected. Note: @@ -780,7 +780,7 @@ def get_active_span_text_map(destination: Optional[str] = None) -> Dict[str, str destination: the name of the remote server. Returns: - dict: the active span's context if opentracing is enabled, otherwise empty. + the active span's context if opentracing is enabled, otherwise empty. """ if destination and not whitelisted_homeserver(destination): diff --git a/synapse/module_api/__init__.py b/synapse/module_api/__init__.py index 30e689d00d..1adc1fd64f 100644 --- a/synapse/module_api/__init__.py +++ b/synapse/module_api/__init__.py @@ -787,7 +787,7 @@ class ModuleApi: Added in Synapse v0.25.0. Args: - access_token(str): access token + access_token: access token Returns: twisted.internet.defer.Deferred - resolves once the access token @@ -832,7 +832,7 @@ class ModuleApi: **kwargs: named args to be passed to func Returns: - Deferred[object]: result of func + Result of func """ # type-ignore: See https://github.com/python/mypy/issues/8862 return defer.ensureDeferred( @@ -924,8 +924,7 @@ class ModuleApi: to represent 'any') of the room state to acquire. Returns: - twisted.internet.defer.Deferred[list(synapse.events.FrozenEvent)]: - The filtered state events in the room. + The filtered state events in the room. """ state_ids = yield defer.ensureDeferred( self._storage_controllers.state.get_current_state_ids( diff --git a/synapse/replication/http/_base.py b/synapse/replication/http/_base.py index 5e661f8c73..3f4d3fc51a 100644 --- a/synapse/replication/http/_base.py +++ b/synapse/replication/http/_base.py @@ -153,7 +153,7 @@ class ReplicationEndpoint(metaclass=abc.ABCMeta): argument list. Returns: - dict: If POST/PUT request then dictionary must be JSON serialisable, + If POST/PUT request then dictionary must be JSON serialisable, otherwise must be appropriate for adding as query args. """ return {} diff --git a/synapse/rest/admin/users.py b/synapse/rest/admin/users.py index 1951b8a9f2..6e0c44be2a 100644 --- a/synapse/rest/admin/users.py +++ b/synapse/rest/admin/users.py @@ -903,8 +903,9 @@ class PushersRestServlet(RestServlet): @user:server/pushers Returns: - pushers: Dictionary containing pushers information. - total: Number of pushers in dictionary `pushers`. + A dictionary with keys: + pushers: Dictionary containing pushers information. + total: Number of pushers in dictionary `pushers`. """ PATTERNS = admin_patterns("/users/(?P[^/]*)/pushers$") diff --git a/synapse/rest/client/login.py b/synapse/rest/client/login.py index 05706b598c..8adced41e5 100644 --- a/synapse/rest/client/login.py +++ b/synapse/rest/client/login.py @@ -350,7 +350,7 @@ class LoginRestServlet(RestServlet): auth_provider_session_id: The session ID got during login from the SSO IdP. Returns: - result: Dictionary of account information after successful login. + Dictionary of account information after successful login. """ # Before we actually log them in we check if they've already logged in diff --git a/synapse/rest/media/v1/media_repository.py b/synapse/rest/media/v1/media_repository.py index 328c0c5477..40b0d39eb2 100644 --- a/synapse/rest/media/v1/media_repository.py +++ b/synapse/rest/media/v1/media_repository.py @@ -344,8 +344,8 @@ class MediaRepository: download from remote server. Args: - server_name (str): Remote server_name where the media originated. - media_id (str): The media ID of the content (as defined by the + server_name: Remote server_name where the media originated. + media_id: The media ID of the content (as defined by the remote server). Returns: diff --git a/synapse/rest/media/v1/thumbnailer.py b/synapse/rest/media/v1/thumbnailer.py index 9b93b9b4f6..a48a4de92a 100644 --- a/synapse/rest/media/v1/thumbnailer.py +++ b/synapse/rest/media/v1/thumbnailer.py @@ -138,7 +138,7 @@ class Thumbnailer: """Rescales the image to the given dimensions. Returns: - BytesIO: the bytes of the encoded image ready to be written to disk + The bytes of the encoded image ready to be written to disk """ with self._resize(width, height) as scaled: return self._encode_image(scaled, output_type) @@ -155,7 +155,7 @@ class Thumbnailer: max_height: The largest possible height. Returns: - BytesIO: the bytes of the encoded image ready to be written to disk + The bytes of the encoded image ready to be written to disk """ if width * self.height > height * self.width: scaled_width = width diff --git a/synapse/server_notices/consent_server_notices.py b/synapse/server_notices/consent_server_notices.py index 698ca742ed..94025ba41f 100644 --- a/synapse/server_notices/consent_server_notices.py +++ b/synapse/server_notices/consent_server_notices.py @@ -113,9 +113,8 @@ def copy_with_str_subst(x: Any, substitutions: Any) -> Any: """Deep-copy a structure, carrying out string substitutions on any strings Args: - x (object): structure to be copied - substitutions (object): substitutions to be made - passed into the - string '%' operator + x: structure to be copied + substitutions: substitutions to be made - passed into the string '%' operator Returns: copy of x diff --git a/synapse/server_notices/resource_limits_server_notices.py b/synapse/server_notices/resource_limits_server_notices.py index 3134cd2d3d..a31a2c99a7 100644 --- a/synapse/server_notices/resource_limits_server_notices.py +++ b/synapse/server_notices/resource_limits_server_notices.py @@ -170,11 +170,13 @@ class ResourceLimitsServerNotices: room_id: The room id of the server notices room Returns: - bool: Is the room currently blocked - list: The list of pinned event IDs that are unrelated to limit blocking - This list can be used as a convenience in the case where the block - is to be lifted and the remaining pinned event references need to be - preserved + Tuple of: + Is the room currently blocked + + The list of pinned event IDs that are unrelated to limit blocking + This list can be used as a convenience in the case where the block + is to be lifted and the remaining pinned event references need to be + preserved """ currently_blocked = False pinned_state_event = None diff --git a/synapse/storage/controllers/persist_events.py b/synapse/storage/controllers/persist_events.py index 48976dc570..33ffef521b 100644 --- a/synapse/storage/controllers/persist_events.py +++ b/synapse/storage/controllers/persist_events.py @@ -204,9 +204,8 @@ class _EventPeristenceQueue(Generic[_PersistResult]): process to to so, calling the per_item_callback for each item. Args: - room_id (str): - task (_EventPersistQueueTask): A _PersistEventsTask or - _UpdateCurrentStateTask to process. + room_id: + task: A _PersistEventsTask or _UpdateCurrentStateTask to process. Returns: the result returned by the `_per_item_callback` passed to diff --git a/synapse/storage/databases/main/devices.py b/synapse/storage/databases/main/devices.py index aa58c2adc3..e114c733d1 100644 --- a/synapse/storage/databases/main/devices.py +++ b/synapse/storage/databases/main/devices.py @@ -535,7 +535,7 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): limit: Maximum number of device updates to return Returns: - List: List of device update tuples: + List of device update tuples: - user_id - device_id - stream_id diff --git a/synapse/storage/databases/main/e2e_room_keys.py b/synapse/storage/databases/main/e2e_room_keys.py index af59be6b48..6240f9a75e 100644 --- a/synapse/storage/databases/main/e2e_room_keys.py +++ b/synapse/storage/databases/main/e2e_room_keys.py @@ -391,10 +391,10 @@ class EndToEndRoomKeyStore(SQLBaseStore): Returns: A dict giving the info metadata for this backup version, with fields including: - version(str) - algorithm(str) - auth_data(object): opaque dict supplied by the client - etag(int): tag of the keys in the backup + version (str) + algorithm (str) + auth_data (object): opaque dict supplied by the client + etag (int): tag of the keys in the backup """ def _get_e2e_room_keys_version_info_txn(txn: LoggingTransaction) -> JsonDict: diff --git a/synapse/storage/databases/main/end_to_end_keys.py b/synapse/storage/databases/main/end_to_end_keys.py index 2a4f58ed92..cf33e73e2b 100644 --- a/synapse/storage/databases/main/end_to_end_keys.py +++ b/synapse/storage/databases/main/end_to_end_keys.py @@ -412,10 +412,9 @@ class EndToEndKeyWorkerStore(EndToEndKeyBackgroundStore, CacheInvalidationWorker """Retrieve a number of one-time keys for a user Args: - user_id(str): id of user to get keys for - device_id(str): id of device to get keys for - key_ids(list[str]): list of key ids (excluding algorithm) to - retrieve + user_id: id of user to get keys for + device_id: id of device to get keys for + key_ids: list of key ids (excluding algorithm) to retrieve Returns: A map from (algorithm, key_id) to json string for key diff --git a/synapse/storage/databases/main/events.py b/synapse/storage/databases/main/events.py index c4acff5be6..d68f127f9b 100644 --- a/synapse/storage/databases/main/events.py +++ b/synapse/storage/databases/main/events.py @@ -1279,9 +1279,10 @@ class PersistEventsStore: Pick the earliest non-outlier if there is one, else the earliest one. Args: - events_and_contexts (list[(EventBase, EventContext)]): + events_and_contexts: + Returns: - list[(EventBase, EventContext)]: filtered list + filtered list """ new_events_and_contexts: OrderedDict[ str, Tuple[EventBase, EventContext] @@ -1307,9 +1308,8 @@ class PersistEventsStore: """Update min_depth for each room Args: - txn (twisted.enterprise.adbapi.Connection): db connection - events_and_contexts (list[(EventBase, EventContext)]): events - we are persisting + txn: db connection + events_and_contexts: events we are persisting """ depth_updates: Dict[str, int] = {} for event, context in events_and_contexts: @@ -1580,13 +1580,11 @@ class PersistEventsStore: """Update all the miscellaneous tables for new events Args: - txn (twisted.enterprise.adbapi.Connection): db connection - events_and_contexts (list[(EventBase, EventContext)]): events - we are persisting - all_events_and_contexts (list[(EventBase, EventContext)]): all - events that we were going to persist. This includes events - we've already persisted, etc, that wouldn't appear in - events_and_context. + txn: db connection + events_and_contexts: events we are persisting + all_events_and_contexts: all events that we were going to persist. + This includes events we've already persisted, etc, that wouldn't + appear in events_and_context. inhibit_local_membership_updates: Stop the local_current_membership from being updated by these events. This should be set to True for backfilled events because backfilled events in the past do diff --git a/synapse/storage/databases/main/events_worker.py b/synapse/storage/databases/main/events_worker.py index 467d20253d..8a104f7e93 100644 --- a/synapse/storage/databases/main/events_worker.py +++ b/synapse/storage/databases/main/events_worker.py @@ -1589,7 +1589,7 @@ class EventsWorkerStore(SQLBaseStore): room_id: The room ID to query. Returns: - dict[str:float] of complexity version to complexity. + Map of complexity version to complexity. """ state_events = await self.get_current_state_event_counts(room_id) diff --git a/synapse/storage/databases/main/monthly_active_users.py b/synapse/storage/databases/main/monthly_active_users.py index efd136a864..db9a24db5e 100644 --- a/synapse/storage/databases/main/monthly_active_users.py +++ b/synapse/storage/databases/main/monthly_active_users.py @@ -217,7 +217,7 @@ class MonthlyActiveUsersWorkerStore(RegistrationWorkerStore): def _reap_users(txn: LoggingTransaction, reserved_users: List[str]) -> None: """ Args: - reserved_users (tuple): reserved users to preserve + reserved_users: reserved users to preserve """ thirty_days_ago = int(self._clock.time_msec()) - (1000 * 60 * 60 * 24 * 30) @@ -370,8 +370,8 @@ class MonthlyActiveUsersWorkerStore(RegistrationWorkerStore): should not appear in the MAU stats). Args: - txn (cursor): - user_id (str): user to add/update + txn: + user_id: user to add/update """ assert ( self._update_on_this_worker @@ -401,7 +401,7 @@ class MonthlyActiveUsersWorkerStore(RegistrationWorkerStore): add the user to the monthly active tables Args: - user_id(str): the user_id to query + user_id: the user_id to query """ assert ( self._update_on_this_worker diff --git a/synapse/storage/databases/main/registration.py b/synapse/storage/databases/main/registration.py index 5167089e03..31f0f2bd3d 100644 --- a/synapse/storage/databases/main/registration.py +++ b/synapse/storage/databases/main/registration.py @@ -953,7 +953,7 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore): """Returns user id from threepid Args: - txn (cursor): + txn: medium: threepid medium e.g. email address: threepid address e.g. me@example.com @@ -1283,8 +1283,8 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore): """Sets an expiration date to the account with the given user ID. Args: - user_id (str): User ID to set an expiration date for. - use_delta (bool): If set to False, the expiration date for the user will be + user_id: User ID to set an expiration date for. + use_delta: If set to False, the expiration date for the user will be now + validity period. If set to True, this expiration date will be a random value in the [now + period - d ; now + period] range, d being a delta equal to 10% of the validity period. diff --git a/synapse/storage/databases/main/room.py b/synapse/storage/databases/main/room.py index 7d97f8f60e..4fbaefad73 100644 --- a/synapse/storage/databases/main/room.py +++ b/synapse/storage/databases/main/room.py @@ -2057,7 +2057,8 @@ class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore): Args: report_id: ID of reported event in database Returns: - event_report: json list of information from event report + JSON dict of information from an event report or None if the + report does not exist. """ def _get_event_report_txn( @@ -2130,8 +2131,9 @@ class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore): user_id: search for user_id. Ignored if user_id is None room_id: search for room_id. Ignored if room_id is None Returns: - event_reports: json list of event reports - count: total number of event reports matching the filter criteria + Tuple of: + json list of event reports + total number of event reports matching the filter criteria """ def _get_event_reports_paginate_txn( diff --git a/synapse/storage/databases/main/user_directory.py b/synapse/storage/databases/main/user_directory.py index ddb25b5cea..698d6f7515 100644 --- a/synapse/storage/databases/main/user_directory.py +++ b/synapse/storage/databases/main/user_directory.py @@ -185,9 +185,8 @@ class UserDirectoryBackgroundUpdateStore(StateDeltasStore): - who should be in the user_directory. Args: - progress (dict) - batch_size (int): Maximum number of state events to process - per cycle. + progress + batch_size: Maximum number of state events to process per cycle. Returns: number of events processed. @@ -708,10 +707,10 @@ class UserDirectoryStore(UserDirectoryBackgroundUpdateStore): Returns the rooms that a user is in. Args: - user_id(str): Must be a local user + user_id: Must be a local user Returns: - list: user_id + List of room IDs """ rows = await self.db_pool.simple_select_onecol( table="users_who_share_private_rooms", diff --git a/synapse/types.py b/synapse/types.py index 773f0438d5..f2d436ddc3 100644 --- a/synapse/types.py +++ b/synapse/types.py @@ -143,8 +143,8 @@ class Requester: Requester. Args: - store (DataStore): Used to convert AS ID to AS object - input (dict): A dict produced by `serialize` + store: Used to convert AS ID to AS object + input: A dict produced by `serialize` Returns: Requester diff --git a/synapse/util/async_helpers.py b/synapse/util/async_helpers.py index 7f1d41eb3c..d24c4f68c4 100644 --- a/synapse/util/async_helpers.py +++ b/synapse/util/async_helpers.py @@ -217,7 +217,8 @@ async def concurrently_execute( limit: Maximum number of conccurent executions. Returns: - Deferred: Resolved when all function invocations have finished. + None, when all function invocations have finished. The return values + from those functions are discarded. """ it = iter(args) diff --git a/synapse/util/caches/__init__.py b/synapse/util/caches/__init__.py index f7c3a6794e..9387632d0d 100644 --- a/synapse/util/caches/__init__.py +++ b/synapse/util/caches/__init__.py @@ -197,7 +197,7 @@ def register_cache( resize_callback: A function which can be called to resize the cache. Returns: - CacheMetric: an object which provides inc_{hits,misses,evictions} methods + an object which provides inc_{hits,misses,evictions} methods """ if resizable: if not resize_callback: diff --git a/synapse/util/caches/deferred_cache.py b/synapse/util/caches/deferred_cache.py index bcb1cba362..bf7bd351e0 100644 --- a/synapse/util/caches/deferred_cache.py +++ b/synapse/util/caches/deferred_cache.py @@ -153,7 +153,7 @@ class DeferredCache(Generic[KT, VT]): Args: key: callback: Gets called when the entry in the cache is invalidated - update_metrics (bool): whether to update the cache hit rate metrics + update_metrics: whether to update the cache hit rate metrics Returns: A Deferred which completes with the result. Note that this may later fail diff --git a/synapse/util/caches/dictionary_cache.py b/synapse/util/caches/dictionary_cache.py index fa91479c97..5eaf70c7ab 100644 --- a/synapse/util/caches/dictionary_cache.py +++ b/synapse/util/caches/dictionary_cache.py @@ -169,10 +169,11 @@ class DictionaryCache(Generic[KT, DKT, DV]): if it is in the cache. Returns: - DictionaryEntry: If `dict_keys` is not None then `DictionaryEntry` - will contain include the keys that are in the cache. If None then - will either return the full dict if in the cache, or the empty - dict (with `full` set to False) if it isn't. + If `dict_keys` is not None then `DictionaryEntry` will contain include + the keys that are in the cache. + + If None then will either return the full dict if in the cache, or the + empty dict (with `full` set to False) if it isn't. """ if dict_keys is None: # The caller wants the full set of dictionary keys for this cache key diff --git a/synapse/util/caches/expiringcache.py b/synapse/util/caches/expiringcache.py index c6a5d0dfc0..01ad02af67 100644 --- a/synapse/util/caches/expiringcache.py +++ b/synapse/util/caches/expiringcache.py @@ -207,7 +207,7 @@ class ExpiringCache(Generic[KT, VT]): items from the cache. Returns: - bool: Whether the cache changed size or not. + Whether the cache changed size or not. """ new_size = int(self._original_max_size * factor) if new_size != self._max_size: diff --git a/synapse/util/caches/lrucache.py b/synapse/util/caches/lrucache.py index aa93109d13..dcf0eac3bf 100644 --- a/synapse/util/caches/lrucache.py +++ b/synapse/util/caches/lrucache.py @@ -389,11 +389,11 @@ class LruCache(Generic[KT, VT]): cache_name: The name of this cache, for the prometheus metrics. If unset, no metrics will be reported on this cache. - cache_type (type): + cache_type: type of underlying cache to be used. Typically one of dict or TreeCache. - size_callback (func(V) -> int | None): + size_callback: metrics_collection_callback: metrics collection callback. This is called early in the metrics @@ -403,7 +403,7 @@ class LruCache(Generic[KT, VT]): Ignored if cache_name is None. - apply_cache_factor_from_config (bool): If true, `max_size` will be + apply_cache_factor_from_config: If true, `max_size` will be multiplied by a cache factor derived from the homeserver config clock: @@ -796,7 +796,7 @@ class LruCache(Generic[KT, VT]): items from the cache. Returns: - bool: Whether the cache changed size or not. + Whether the cache changed size or not. """ if not self.apply_cache_factor_from_config: return False diff --git a/synapse/util/ratelimitutils.py b/synapse/util/ratelimitutils.py index 9f64fed0d7..2aceb1a47f 100644 --- a/synapse/util/ratelimitutils.py +++ b/synapse/util/ratelimitutils.py @@ -183,7 +183,7 @@ class FederationRateLimiter: # Handle request ... Args: - host (str): Origin of incoming request. + host: Origin of incoming request. Returns: context manager which returns a deferred. diff --git a/synapse/util/threepids.py b/synapse/util/threepids.py index 1e9c2faa64..54bc7589fd 100644 --- a/synapse/util/threepids.py +++ b/synapse/util/threepids.py @@ -48,7 +48,7 @@ async def check_3pid_allowed( registration: whether we want to bind the 3PID as part of registering a new user. Returns: - bool: whether the 3PID medium/address is allowed to be added to this HS + whether the 3PID medium/address is allowed to be added to this HS """ if not await hs.get_password_auth_provider().is_3pid_allowed( medium, address, registration diff --git a/synapse/util/wheel_timer.py b/synapse/util/wheel_timer.py index 177e198e7e..b1ec7f4bd8 100644 --- a/synapse/util/wheel_timer.py +++ b/synapse/util/wheel_timer.py @@ -90,10 +90,10 @@ class WheelTimer(Generic[T]): """Fetch any objects that have timed out Args: - now (ms): Current time in msec + now: Current time in msec Returns: - list: List of objects that have timed out + List of objects that have timed out """ now_key = int(now / self.bucket_size) diff --git a/tests/http/__init__.py b/tests/http/__init__.py index e74f7f5b48..093537adef 100644 --- a/tests/http/__init__.py +++ b/tests/http/__init__.py @@ -13,6 +13,7 @@ # limitations under the License. import os.path import subprocess +from typing import List from zope.interface import implementer @@ -70,14 +71,14 @@ subjectAltName = %(sanentries)s """ -def create_test_cert_file(sanlist): +def create_test_cert_file(sanlist: List[bytes]) -> str: """build an x509 certificate file Args: - sanlist: list[bytes]: a list of subjectAltName values for the cert + sanlist: a list of subjectAltName values for the cert Returns: - str: the path to the file + The path to the file """ global cert_file_count csr_filename = "server.csr" diff --git a/tests/replication/slave/storage/test_events.py b/tests/replication/slave/storage/test_events.py index 96f3880923..dce71f7334 100644 --- a/tests/replication/slave/storage/test_events.py +++ b/tests/replication/slave/storage/test_events.py @@ -143,6 +143,7 @@ class EventsWorkerStoreTestCase(BaseSlavedStoreTestCase): self.persist(type="m.room.create", key="", creator=USER_ID) self.check("get_invited_rooms_for_local_user", [USER_ID_2], []) event = self.persist(type="m.room.member", key=USER_ID_2, membership="invite") + assert event.internal_metadata.stream_ordering is not None self.replicate() @@ -230,6 +231,7 @@ class EventsWorkerStoreTestCase(BaseSlavedStoreTestCase): j2 = self.persist( type="m.room.member", sender=USER_ID_2, key=USER_ID_2, membership="join" ) + assert j2.internal_metadata.stream_ordering is not None self.replicate() expected_pos = PersistedEventPosition( @@ -287,6 +289,7 @@ class EventsWorkerStoreTestCase(BaseSlavedStoreTestCase): ) ) self.replicate() + assert j2.internal_metadata.stream_ordering is not None event_source = RoomEventSource(self.hs) event_source.store = self.slaved_store @@ -336,10 +339,10 @@ class EventsWorkerStoreTestCase(BaseSlavedStoreTestCase): event_id = 0 - def persist(self, backfill=False, **kwargs): + def persist(self, backfill=False, **kwargs) -> FrozenEvent: """ Returns: - synapse.events.FrozenEvent: The event that was persisted. + The event that was persisted. """ event, context = self.build_event(**kwargs) diff --git a/tests/replication/test_multi_media_repo.py b/tests/replication/test_multi_media_repo.py index 13aa5eb51a..96cdf2c45b 100644 --- a/tests/replication/test_multi_media_repo.py +++ b/tests/replication/test_multi_media_repo.py @@ -15,8 +15,9 @@ import logging import os from typing import Optional, Tuple +from twisted.internet.interfaces import IOpenSSLServerConnectionCreator from twisted.internet.protocol import Factory -from twisted.protocols.tls import TLSMemoryBIOFactory +from twisted.protocols.tls import TLSMemoryBIOFactory, TLSMemoryBIOProtocol from twisted.web.http import HTTPChannel from twisted.web.server import Request @@ -102,7 +103,7 @@ class MediaRepoShardTestCase(BaseMultiWorkerStreamTestCase): ) # fish the test server back out of the server-side TLS protocol. - http_server = server_tls_protocol.wrappedProtocol + http_server: HTTPChannel = server_tls_protocol.wrappedProtocol # type: ignore[assignment] # give the reactor a pump to get the TLS juices flowing. self.reactor.pump((0.1,)) @@ -238,16 +239,15 @@ def get_connection_factory(): return test_server_connection_factory -def _build_test_server(connection_creator): +def _build_test_server( + connection_creator: IOpenSSLServerConnectionCreator, +) -> TLSMemoryBIOProtocol: """Construct a test server This builds an HTTP channel, wrapped with a TLSMemoryBIOProtocol Args: - connection_creator (IOpenSSLServerConnectionCreator): thing to build - SSL connections - sanlist (list[bytes]): list of the SAN entries for the cert returned - by the server + connection_creator: thing to build SSL connections Returns: TLSMemoryBIOProtocol diff --git a/tests/server_notices/test_resource_limits_server_notices.py b/tests/server_notices/test_resource_limits_server_notices.py index bf403045e9..7cbc40736c 100644 --- a/tests/server_notices/test_resource_limits_server_notices.py +++ b/tests/server_notices/test_resource_limits_server_notices.py @@ -11,6 +11,7 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. +from typing import Tuple from unittest.mock import Mock from twisted.test.proto_helpers import MemoryReactor @@ -350,14 +351,15 @@ class TestResourceLimitsServerNoticesWithRealRooms(unittest.HomeserverTestCase): self.assertTrue(notice_in_room, "No server notice in room") - def _trigger_notice_and_join(self): + def _trigger_notice_and_join(self) -> Tuple[str, str, str]: """Creates enough active users to hit the MAU limit and trigger a system notice about it, then joins the system notices room with one of the users created. Returns: - user_id (str): The ID of the user that joined the room. - tok (str): The access token of the user that joined the room. - room_id (str): The ID of the room that's been joined. + A tuple of: + user_id: The ID of the user that joined the room. + tok: The access token of the user that joined the room. + room_id: The ID of the room that's been joined. """ user_id = None tok = None diff --git a/tests/unittest.py b/tests/unittest.py index 5116be338e..a120c2976c 100644 --- a/tests/unittest.py +++ b/tests/unittest.py @@ -360,13 +360,13 @@ class HomeserverTestCase(TestCase): store.db_pool.updates.do_next_background_update(False), by=0.1 ) - def make_homeserver(self, reactor, clock): + def make_homeserver(self, reactor: MemoryReactor, clock: Clock): """ Make and return a homeserver. Args: reactor: A Twisted Reactor, or something that pretends to be one. - clock (synapse.util.Clock): The Clock, associated with the reactor. + clock: The Clock, associated with the reactor. Returns: A homeserver suitable for testing. @@ -426,9 +426,8 @@ class HomeserverTestCase(TestCase): Args: reactor: A Twisted Reactor, or something that pretends to be one. - clock (synapse.util.Clock): The Clock, associated with the reactor. - homeserver (synapse.server.HomeServer): The HomeServer to test - against. + clock: The Clock, associated with the reactor. + homeserver: The HomeServer to test against. Function to optionally be overridden in subclasses. """ @@ -452,11 +451,10 @@ class HomeserverTestCase(TestCase): given content. Args: - method (bytes/unicode): The HTTP request method ("verb"). - path (bytes/unicode): The HTTP path, suitably URL encoded (e.g. - escaped UTF-8 & spaces and such). - content (bytes or dict): The body of the request. JSON-encoded, if - a dict. + method: The HTTP request method ("verb"). + path: The HTTP path, suitably URL encoded (e.g. escaped UTF-8 & spaces + and such). content (bytes or dict): The body of the request. + JSON-encoded, if a dict. shorthand: Whether to try and be helpful and prefix the given URL with the usual REST API path, if it doesn't contain it. federation_auth_origin: if set to not-None, we will add a fake -- cgit 1.5.1 From 618e4ab81b70e37bdb8e9224bd84fcfe4b15bdea Mon Sep 17 00:00:00 2001 From: Andrew Morgan <1342360+anoadragon453@users.noreply.github.com> Date: Wed, 16 Nov 2022 15:25:35 +0000 Subject: Fix an invalid comparison of `UserPresenceState` to `str` (#14393) --- changelog.d/14393.bugfix | 1 + synapse/handlers/presence.py | 2 +- tests/handlers/test_presence.py | 41 +++++++++++++++++++++++++++++++++++------ tests/module_api/test_api.py | 3 +++ tests/replication/_base.py | 7 ++++++- 5 files changed, 46 insertions(+), 8 deletions(-) create mode 100644 changelog.d/14393.bugfix (limited to 'synapse/handlers') diff --git a/changelog.d/14393.bugfix b/changelog.d/14393.bugfix new file mode 100644 index 0000000000..97177bc62f --- /dev/null +++ b/changelog.d/14393.bugfix @@ -0,0 +1 @@ +Fix a bug introduced in 1.58.0 where a user with presence state 'org.matrix.msc3026.busy' would mistakenly be set to 'online' when calling `/sync` or `/events` on a worker process. \ No newline at end of file diff --git a/synapse/handlers/presence.py b/synapse/handlers/presence.py index b7bc787636..cf08737d11 100644 --- a/synapse/handlers/presence.py +++ b/synapse/handlers/presence.py @@ -478,7 +478,7 @@ class WorkerPresenceHandler(BasePresenceHandler): return _NullContextManager() prev_state = await self.current_state_for_user(user_id) - if prev_state != PresenceState.BUSY: + if prev_state.state != PresenceState.BUSY: # We set state here but pass ignore_status_msg = True as we don't want to # cause the status message to be cleared. # Note that this causes last_active_ts to be incremented which is not diff --git a/tests/handlers/test_presence.py b/tests/handlers/test_presence.py index c96dc6caf2..c5981ff965 100644 --- a/tests/handlers/test_presence.py +++ b/tests/handlers/test_presence.py @@ -15,6 +15,7 @@ from typing import Optional from unittest.mock import Mock, call +from parameterized import parameterized from signedjson.key import generate_signing_key from synapse.api.constants import EventTypes, Membership, PresenceState @@ -37,6 +38,7 @@ from synapse.rest.client import room from synapse.types import UserID, get_domain_from_id from tests import unittest +from tests.replication._base import BaseMultiWorkerStreamTestCase class PresenceUpdateTestCase(unittest.HomeserverTestCase): @@ -505,7 +507,7 @@ class PresenceTimeoutTestCase(unittest.TestCase): self.assertEqual(state, new_state) -class PresenceHandlerTestCase(unittest.HomeserverTestCase): +class PresenceHandlerTestCase(BaseMultiWorkerStreamTestCase): def prepare(self, reactor, clock, hs): self.presence_handler = hs.get_presence_handler() self.clock = hs.get_clock() @@ -716,20 +718,47 @@ class PresenceHandlerTestCase(unittest.HomeserverTestCase): # our status message should be the same as it was before self.assertEqual(state.status_msg, status_msg) - def test_set_presence_from_syncing_keeps_busy(self): - """Test that presence set by syncing doesn't affect busy status""" - # while this isn't the default - self.presence_handler._busy_presence_enabled = True + @parameterized.expand([(False,), (True,)]) + @unittest.override_config( + { + "experimental_features": { + "msc3026_enabled": True, + }, + } + ) + def test_set_presence_from_syncing_keeps_busy(self, test_with_workers: bool): + """Test that presence set by syncing doesn't affect busy status + Args: + test_with_workers: If True, check the presence state of the user by calling + /sync against a worker, rather than the main process. + """ user_id = "@test:server" status_msg = "I'm busy!" + # By default, we call /sync against the main process. + worker_to_sync_against = self.hs + if test_with_workers: + # Create a worker and use it to handle /sync traffic instead. + # This is used to test that presence changes get replicated from workers + # to the main process correctly. + worker_to_sync_against = self.make_worker_hs( + "synapse.app.generic_worker", {"worker_name": "presence_writer"} + ) + + # Set presence to BUSY self._set_presencestate_with_status_msg(user_id, PresenceState.BUSY, status_msg) + # Perform a sync with a presence state other than busy. This should NOT change + # our presence status; we only change from busy if we explicitly set it via + # /presence/*. self.get_success( - self.presence_handler.user_syncing(user_id, True, PresenceState.ONLINE) + worker_to_sync_against.get_presence_handler().user_syncing( + user_id, True, PresenceState.ONLINE + ) ) + # Check against the main process that the user's presence did not change. state = self.get_success( self.presence_handler.get_state(UserID.from_string(user_id)) ) diff --git a/tests/module_api/test_api.py b/tests/module_api/test_api.py index 02cef6f876..058ca57e55 100644 --- a/tests/module_api/test_api.py +++ b/tests/module_api/test_api.py @@ -778,8 +778,11 @@ def _test_sending_local_online_presence_to_local_user( worker process. The test users will still sync with the main process. The purpose of testing with a worker is to check whether a Synapse module running on a worker can inform other workers/ the main process that they should include additional presence when a user next syncs. + If this argument is True, `test_case` MUST be an instance of BaseMultiWorkerStreamTestCase. """ if test_with_workers: + assert isinstance(test_case, BaseMultiWorkerStreamTestCase) + # Create a worker process to make module_api calls against worker_hs = test_case.make_worker_hs( "synapse.app.generic_worker", {"worker_name": "presence_writer"} diff --git a/tests/replication/_base.py b/tests/replication/_base.py index 121f3d8d65..3029a16dda 100644 --- a/tests/replication/_base.py +++ b/tests/replication/_base.py @@ -542,8 +542,13 @@ class FakeRedisPubSubProtocol(Protocol): self.send("OK") elif command == b"GET": self.send(None) + + # Connection keep-alives. + elif command == b"PING": + self.send("PONG") + else: - raise Exception("Unknown command") + raise Exception(f"Unknown command: {command}") def send(self, msg): """Send a message back to the client.""" -- cgit 1.5.1 From 1526ff389f02d14d0df729bd6ea35836e758c449 Mon Sep 17 00:00:00 2001 From: Mathieu Velten Date: Mon, 21 Nov 2022 16:46:14 +0100 Subject: Faster joins: filter out non local events when a room doesn't have its full state (#14404) Signed-off-by: Mathieu Velten --- changelog.d/14404.misc | 1 + synapse/federation/sender/per_destination_queue.py | 1 + synapse/handlers/federation.py | 15 +++++++---- synapse/visibility.py | 29 +++++++++++++++++++--- tests/test_visibility.py | 10 ++++---- 5 files changed, 43 insertions(+), 13 deletions(-) create mode 100644 changelog.d/14404.misc (limited to 'synapse/handlers') diff --git a/changelog.d/14404.misc b/changelog.d/14404.misc new file mode 100644 index 0000000000..b9ab525f2b --- /dev/null +++ b/changelog.d/14404.misc @@ -0,0 +1 @@ +Faster joins: filter out non local events when a room doesn't have its full state. diff --git a/synapse/federation/sender/per_destination_queue.py b/synapse/federation/sender/per_destination_queue.py index 084c45a95c..3ae5e8634c 100644 --- a/synapse/federation/sender/per_destination_queue.py +++ b/synapse/federation/sender/per_destination_queue.py @@ -505,6 +505,7 @@ class PerDestinationQueue: new_pdus = await filter_events_for_server( self._storage_controllers, self._destination, + self._server_name, new_pdus, redact=False, ) diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index 188f0956ef..d92582fd5c 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -379,6 +379,7 @@ class FederationHandler: filtered_extremities = await filter_events_for_server( self._storage_controllers, self.server_name, + self.server_name, events_to_check, redact=False, check_history_visibility_only=True, @@ -1231,7 +1232,9 @@ class FederationHandler: async def on_backfill_request( self, origin: str, room_id: str, pdu_list: List[str], limit: int ) -> List[EventBase]: - await self._event_auth_handler.assert_host_in_room(room_id, origin) + # We allow partially joined rooms since in this case we are filtering out + # non-local events in `filter_events_for_server`. + await self._event_auth_handler.assert_host_in_room(room_id, origin, True) # Synapse asks for 100 events per backfill request. Do not allow more. limit = min(limit, 100) @@ -1252,7 +1255,7 @@ class FederationHandler: ) events = await filter_events_for_server( - self._storage_controllers, origin, events + self._storage_controllers, origin, self.server_name, events ) return events @@ -1283,7 +1286,7 @@ class FederationHandler: await self._event_auth_handler.assert_host_in_room(event.room_id, origin) events = await filter_events_for_server( - self._storage_controllers, origin, [event] + self._storage_controllers, origin, self.server_name, [event] ) event = events[0] return event @@ -1296,7 +1299,9 @@ class FederationHandler: latest_events: List[str], limit: int, ) -> List[EventBase]: - await self._event_auth_handler.assert_host_in_room(room_id, origin) + # We allow partially joined rooms since in this case we are filtering out + # non-local events in `filter_events_for_server`. + await self._event_auth_handler.assert_host_in_room(room_id, origin, True) # Only allow up to 20 events to be retrieved per request. limit = min(limit, 20) @@ -1309,7 +1314,7 @@ class FederationHandler: ) missing_events = await filter_events_for_server( - self._storage_controllers, origin, missing_events + self._storage_controllers, origin, self.server_name, missing_events ) return missing_events diff --git a/synapse/visibility.py b/synapse/visibility.py index 40a9c5b53f..b443857571 100644 --- a/synapse/visibility.py +++ b/synapse/visibility.py @@ -563,7 +563,8 @@ def get_effective_room_visibility_from_state(state: StateMap[EventBase]) -> str: async def filter_events_for_server( storage: StorageControllers, - server_name: str, + target_server_name: str, + local_server_name: str, events: List[EventBase], redact: bool = True, check_history_visibility_only: bool = False, @@ -603,7 +604,7 @@ async def filter_events_for_server( # if the server is either in the room or has been invited # into the room. for ev in memberships.values(): - assert get_domain_from_id(ev.state_key) == server_name + assert get_domain_from_id(ev.state_key) == target_server_name memtype = ev.membership if memtype == Membership.JOIN: @@ -622,6 +623,24 @@ async def filter_events_for_server( # to no users having been erased. erased_senders = {} + # Filter out non-local events when we are in the middle of a partial join, since our servers + # list can be out of date and we could leak events to servers not in the room anymore. + # This can also be true for local events but we consider it to be an acceptable risk. + + # We do this check as a first step and before retrieving membership events because + # otherwise a room could be fully joined after we retrieve those, which would then bypass + # this check but would base the filtering on an outdated view of the membership events. + + partial_state_invisible_events = set() + if not check_history_visibility_only: + for e in events: + sender_domain = get_domain_from_id(e.sender) + if ( + sender_domain != local_server_name + and await storage.main.is_partial_state_room(e.room_id) + ): + partial_state_invisible_events.add(e) + # Let's check to see if all the events have a history visibility # of "shared" or "world_readable". If that's the case then we don't # need to check membership (as we know the server is in the room). @@ -636,7 +655,7 @@ async def filter_events_for_server( if event_to_history_vis[e.event_id] not in (HistoryVisibility.SHARED, HistoryVisibility.WORLD_READABLE) ], - server_name, + target_server_name, ) to_return = [] @@ -645,6 +664,10 @@ async def filter_events_for_server( visible = check_event_is_visible( event_to_history_vis[e.event_id], event_to_memberships.get(e.event_id, {}) ) + + if e in partial_state_invisible_events: + visible = False + if visible and not erased: to_return.append(e) elif redact: diff --git a/tests/test_visibility.py b/tests/test_visibility.py index c385b2f8d4..d0b9ad5454 100644 --- a/tests/test_visibility.py +++ b/tests/test_visibility.py @@ -61,7 +61,7 @@ class FilterEventsForServerTestCase(unittest.HomeserverTestCase): filtered = self.get_success( filter_events_for_server( - self._storage_controllers, "test_server", events_to_filter + self._storage_controllers, "test_server", "hs", events_to_filter ) ) @@ -83,7 +83,7 @@ class FilterEventsForServerTestCase(unittest.HomeserverTestCase): self.assertEqual( self.get_success( filter_events_for_server( - self._storage_controllers, "remote_hs", [outlier] + self._storage_controllers, "remote_hs", "hs", [outlier] ) ), [outlier], @@ -94,7 +94,7 @@ class FilterEventsForServerTestCase(unittest.HomeserverTestCase): filtered = self.get_success( filter_events_for_server( - self._storage_controllers, "remote_hs", [outlier, evt] + self._storage_controllers, "remote_hs", "local_hs", [outlier, evt] ) ) self.assertEqual(len(filtered), 2, f"expected 2 results, got: {filtered}") @@ -106,7 +106,7 @@ class FilterEventsForServerTestCase(unittest.HomeserverTestCase): # be redacted) filtered = self.get_success( filter_events_for_server( - self._storage_controllers, "other_server", [outlier, evt] + self._storage_controllers, "other_server", "local_hs", [outlier, evt] ) ) self.assertEqual(filtered[0], outlier) @@ -141,7 +141,7 @@ class FilterEventsForServerTestCase(unittest.HomeserverTestCase): # ... and the filtering happens. filtered = self.get_success( filter_events_for_server( - self._storage_controllers, "test_server", events_to_filter + self._storage_controllers, "test_server", "local_hs", events_to_filter ) ) -- cgit 1.5.1 From 1799a54a545618782840a60950ef4b64da9ee24d Mon Sep 17 00:00:00 2001 From: Patrick Cloke Date: Tue, 22 Nov 2022 07:26:11 -0500 Subject: Batch fetch bundled annotations (#14491) Avoid an n+1 query problem and fetch the bundled aggregations for m.annotation relations in a single query instead of a query per event. This applies similar logic for as was previously done for edits in 8b309adb436c162510ed1402f33b8741d71fc058 (#11660) and threads in b65acead428653b988351ae8d7b22127a22039cd (#11752). --- changelog.d/14491.feature | 1 + synapse/handlers/relations.py | 197 ++++++++++++++++------------ synapse/storage/databases/main/relations.py | 139 ++++++++++++-------- synapse/util/caches/descriptors.py | 2 +- tests/rest/client/test_relations.py | 4 +- 5 files changed, 202 insertions(+), 141 deletions(-) create mode 100644 changelog.d/14491.feature (limited to 'synapse/handlers') diff --git a/changelog.d/14491.feature b/changelog.d/14491.feature new file mode 100644 index 0000000000..4fca7282f7 --- /dev/null +++ b/changelog.d/14491.feature @@ -0,0 +1 @@ +Reduce database load of [Client-Server endpoints](https://spec.matrix.org/v1.4/client-server-api/#aggregations) which return bundled aggregations. diff --git a/synapse/handlers/relations.py b/synapse/handlers/relations.py index 8e71dda970..ca94239f61 100644 --- a/synapse/handlers/relations.py +++ b/synapse/handlers/relations.py @@ -13,7 +13,16 @@ # limitations under the License. import enum import logging -from typing import TYPE_CHECKING, Dict, FrozenSet, Iterable, List, Optional, Tuple +from typing import ( + TYPE_CHECKING, + Collection, + Dict, + FrozenSet, + Iterable, + List, + Optional, + Tuple, +) import attr @@ -259,48 +268,64 @@ class RelationsHandler: e.msg, ) - async def get_annotations_for_event( - self, - event_id: str, - room_id: str, - limit: int = 5, - ignored_users: FrozenSet[str] = frozenset(), - ) -> List[JsonDict]: - """Get a list of annotations on the event, grouped by event type and + async def get_annotations_for_events( + self, event_ids: Collection[str], ignored_users: FrozenSet[str] = frozenset() + ) -> Dict[str, List[JsonDict]]: + """Get a list of annotations to the given events, grouped by event type and aggregation key, sorted by count. - This is used e.g. to get the what and how many reactions have happend + This is used e.g. to get the what and how many reactions have happened on an event. Args: - event_id: Fetch events that relate to this event ID. - room_id: The room the event belongs to. - limit: Only fetch the `limit` groups. + event_ids: Fetch events that relate to these event IDs. ignored_users: The users ignored by the requesting user. Returns: - List of groups of annotations that match. Each row is a dict with - `type`, `key` and `count` fields. + A map of event IDs to a list of groups of annotations that match. + Each entry is a dict with `type`, `key` and `count` fields. """ # Get the base results for all users. - full_results = await self._main_store.get_aggregation_groups_for_event( - event_id, room_id, limit + full_results = await self._main_store.get_aggregation_groups_for_events( + event_ids ) + # Avoid additional logic if there are no ignored users. + if not ignored_users: + return { + event_id: results + for event_id, results in full_results.items() + if results + } + # Then subtract off the results for any ignored users. ignored_results = await self._main_store.get_aggregation_groups_for_users( - event_id, room_id, limit, ignored_users + [event_id for event_id, results in full_results.items() if results], + ignored_users, ) - filtered_results = [] - for result in full_results: - key = (result["type"], result["key"]) - if key in ignored_results: - result = result.copy() - result["count"] -= ignored_results[key] - if result["count"] <= 0: - continue - filtered_results.append(result) + filtered_results = {} + for event_id, results in full_results.items(): + # If no annotations, skip. + if not results: + continue + + # If there are not ignored results for this event, copy verbatim. + if event_id not in ignored_results: + filtered_results[event_id] = results + continue + + # Otherwise, subtract out the ignored results. + event_ignored_results = ignored_results[event_id] + for result in results: + key = (result["type"], result["key"]) + if key in event_ignored_results: + # Ensure to not modify the cache. + result = result.copy() + result["count"] -= event_ignored_results[key] + if result["count"] <= 0: + continue + filtered_results.setdefault(event_id, []).append(result) return filtered_results @@ -366,59 +391,62 @@ class RelationsHandler: results = {} for event_id, summary in summaries.items(): - if summary: - thread_count, latest_thread_event = summary - - # Subtract off the count of any ignored users. - for ignored_user in ignored_users: - thread_count -= ignored_results.get((event_id, ignored_user), 0) - - # This is gnarly, but if the latest event is from an ignored user, - # attempt to find one that isn't from an ignored user. - if latest_thread_event.sender in ignored_users: - room_id = latest_thread_event.room_id - - # If the root event is not found, something went wrong, do - # not include a summary of the thread. - event = await self._event_handler.get_event(user, room_id, event_id) - if event is None: - continue + # If no thread, skip. + if not summary: + continue - potential_events, _ = await self.get_relations_for_event( - event_id, - event, - room_id, - RelationTypes.THREAD, - ignored_users, - ) + thread_count, latest_thread_event = summary - # If all found events are from ignored users, do not include - # a summary of the thread. - if not potential_events: - continue + # Subtract off the count of any ignored users. + for ignored_user in ignored_users: + thread_count -= ignored_results.get((event_id, ignored_user), 0) - # The *last* event returned is the one that is cared about. - event = await self._event_handler.get_event( - user, room_id, potential_events[-1].event_id - ) - # It is unexpected that the event will not exist. - if event is None: - logger.warning( - "Unable to fetch latest event in a thread with event ID: %s", - potential_events[-1].event_id, - ) - continue - latest_thread_event = event - - results[event_id] = _ThreadAggregation( - latest_event=latest_thread_event, - count=thread_count, - # If there's a thread summary it must also exist in the - # participated dictionary. - current_user_participated=events_by_id[event_id].sender == user_id - or participated[event_id], + # This is gnarly, but if the latest event is from an ignored user, + # attempt to find one that isn't from an ignored user. + if latest_thread_event.sender in ignored_users: + room_id = latest_thread_event.room_id + + # If the root event is not found, something went wrong, do + # not include a summary of the thread. + event = await self._event_handler.get_event(user, room_id, event_id) + if event is None: + continue + + potential_events, _ = await self.get_relations_for_event( + event_id, + event, + room_id, + RelationTypes.THREAD, + ignored_users, ) + # If all found events are from ignored users, do not include + # a summary of the thread. + if not potential_events: + continue + + # The *last* event returned is the one that is cared about. + event = await self._event_handler.get_event( + user, room_id, potential_events[-1].event_id + ) + # It is unexpected that the event will not exist. + if event is None: + logger.warning( + "Unable to fetch latest event in a thread with event ID: %s", + potential_events[-1].event_id, + ) + continue + latest_thread_event = event + + results[event_id] = _ThreadAggregation( + latest_event=latest_thread_event, + count=thread_count, + # If there's a thread summary it must also exist in the + # participated dictionary. + current_user_participated=events_by_id[event_id].sender == user_id + or participated[event_id], + ) + return results @trace @@ -496,17 +524,18 @@ class RelationsHandler: # (as that is what makes it part of the thread). relations_by_id[latest_thread_event.event_id] = RelationTypes.THREAD - # Fetch other relations per event. - for event in events_by_id.values(): - # Fetch any annotations (ie, reactions) to bundle with this event. - annotations = await self.get_annotations_for_event( - event.event_id, event.room_id, ignored_users=ignored_users - ) + # Fetch any annotations (ie, reactions) to bundle with this event. + annotations_by_event_id = await self.get_annotations_for_events( + events_by_id.keys(), ignored_users=ignored_users + ) + for event_id, annotations in annotations_by_event_id.items(): if annotations: - results.setdefault( - event.event_id, BundledAggregations() - ).annotations = {"chunk": annotations} + results.setdefault(event_id, BundledAggregations()).annotations = { + "chunk": annotations + } + # Fetch other relations per event. + for event in events_by_id.values(): # Fetch any references to bundle with this event. references, next_token = await self.get_relations_for_event( event.event_id, diff --git a/synapse/storage/databases/main/relations.py b/synapse/storage/databases/main/relations.py index ca431002c8..f96a16956a 100644 --- a/synapse/storage/databases/main/relations.py +++ b/synapse/storage/databases/main/relations.py @@ -20,6 +20,7 @@ from typing import ( FrozenSet, Iterable, List, + Mapping, Optional, Set, Tuple, @@ -394,106 +395,136 @@ class RelationsWorkerStore(SQLBaseStore): ) return result is not None - @cached(tree=True) - async def get_aggregation_groups_for_event( - self, event_id: str, room_id: str, limit: int = 5 - ) -> List[JsonDict]: - """Get a list of annotations on the event, grouped by event type and + @cached() + async def get_aggregation_groups_for_event(self, event_id: str) -> List[JsonDict]: + raise NotImplementedError() + + @cachedList( + cached_method_name="get_aggregation_groups_for_event", list_name="event_ids" + ) + async def get_aggregation_groups_for_events( + self, event_ids: Collection[str] + ) -> Mapping[str, Optional[List[JsonDict]]]: + """Get a list of annotations on the given events, grouped by event type and aggregation key, sorted by count. This is used e.g. to get the what and how many reactions have happend on an event. Args: - event_id: Fetch events that relate to this event ID. - room_id: The room the event belongs to. - limit: Only fetch the `limit` groups. + event_ids: Fetch events that relate to these event IDs. Returns: - List of groups of annotations that match. Each row is a dict with - `type`, `key` and `count` fields. + A map of event IDs to a list of groups of annotations that match. + Each entry is a dict with `type`, `key` and `count` fields. """ + # The number of entries to return per event ID. + limit = 5 - args = [ - event_id, - room_id, - RelationTypes.ANNOTATION, - limit, - ] + clause, args = make_in_list_sql_clause( + self.database_engine, "relates_to_id", event_ids + ) + args.append(RelationTypes.ANNOTATION) - sql = """ - SELECT type, aggregation_key, COUNT(DISTINCT sender) - FROM event_relations - INNER JOIN events USING (event_id) - WHERE relates_to_id = ? AND room_id = ? AND relation_type = ? - GROUP BY relation_type, type, aggregation_key - ORDER BY COUNT(*) DESC - LIMIT ? + sql = f""" + SELECT + relates_to_id, + annotation.type, + aggregation_key, + COUNT(DISTINCT annotation.sender) + FROM events AS annotation + INNER JOIN event_relations USING (event_id) + INNER JOIN events AS parent ON + parent.event_id = relates_to_id + AND parent.room_id = annotation.room_id + WHERE + {clause} + AND relation_type = ? + GROUP BY relates_to_id, annotation.type, aggregation_key + ORDER BY relates_to_id, COUNT(*) DESC """ - def _get_aggregation_groups_for_event_txn( + def _get_aggregation_groups_for_events_txn( txn: LoggingTransaction, - ) -> List[JsonDict]: + ) -> Mapping[str, List[JsonDict]]: txn.execute(sql, args) - return [{"type": row[0], "key": row[1], "count": row[2]} for row in txn] + result: Dict[str, List[JsonDict]] = {} + for event_id, type, key, count in cast( + List[Tuple[str, str, str, int]], txn + ): + event_results = result.setdefault(event_id, []) + + # Limit the number of results per event ID. + if len(event_results) == limit: + continue + + event_results.append({"type": type, "key": key, "count": count}) + + return result return await self.db_pool.runInteraction( - "get_aggregation_groups_for_event", _get_aggregation_groups_for_event_txn + "get_aggregation_groups_for_events", _get_aggregation_groups_for_events_txn ) async def get_aggregation_groups_for_users( - self, - event_id: str, - room_id: str, - limit: int, - users: FrozenSet[str] = frozenset(), - ) -> Dict[Tuple[str, str], int]: + self, event_ids: Collection[str], users: FrozenSet[str] + ) -> Dict[str, Dict[Tuple[str, str], int]]: """Fetch the partial aggregations for an event for specific users. This is used, in conjunction with get_aggregation_groups_for_event, to remove information from the results for ignored users. Args: - event_id: Fetch events that relate to this event ID. - room_id: The room the event belongs to. - limit: Only fetch the `limit` groups. + event_ids: Fetch events that relate to these event IDs. users: The users to fetch information for. Returns: - A map of (event type, aggregation key) to a count of users. + A map of event ID to a map of (event type, aggregation key) to a + count of users. """ if not users: return {} - args: List[Union[str, int]] = [ - event_id, - room_id, - RelationTypes.ANNOTATION, - ] + events_sql, args = make_in_list_sql_clause( + self.database_engine, "relates_to_id", event_ids + ) users_sql, users_args = make_in_list_sql_clause( - self.database_engine, "sender", users + self.database_engine, "annotation.sender", users ) args.extend(users_args) + args.append(RelationTypes.ANNOTATION) sql = f""" - SELECT type, aggregation_key, COUNT(DISTINCT sender) - FROM event_relations - INNER JOIN events USING (event_id) - WHERE relates_to_id = ? AND room_id = ? AND relation_type = ? AND {users_sql} - GROUP BY relation_type, type, aggregation_key - ORDER BY COUNT(*) DESC - LIMIT ? + SELECT + relates_to_id, + annotation.type, + aggregation_key, + COUNT(DISTINCT annotation.sender) + FROM events AS annotation + INNER JOIN event_relations USING (event_id) + INNER JOIN events AS parent ON + parent.event_id = relates_to_id + AND parent.room_id = annotation.room_id + WHERE {events_sql} AND {users_sql} AND relation_type = ? + GROUP BY relates_to_id, annotation.type, aggregation_key + ORDER BY relates_to_id, COUNT(*) DESC """ def _get_aggregation_groups_for_users_txn( txn: LoggingTransaction, - ) -> Dict[Tuple[str, str], int]: - txn.execute(sql, args + [limit]) + ) -> Dict[str, Dict[Tuple[str, str], int]]: + txn.execute(sql, args) - return {(row[0], row[1]): row[2] for row in txn} + result: Dict[str, Dict[Tuple[str, str], int]] = {} + for event_id, type, key, count in cast( + List[Tuple[str, str, str, int]], txn + ): + result.setdefault(event_id, {})[(type, key)] = count + + return result return await self.db_pool.runInteraction( "get_aggregation_groups_for_users", _get_aggregation_groups_for_users_txn diff --git a/synapse/util/caches/descriptors.py b/synapse/util/caches/descriptors.py index 75428d19ba..72227359b9 100644 --- a/synapse/util/caches/descriptors.py +++ b/synapse/util/caches/descriptors.py @@ -503,7 +503,7 @@ def cachedList( is specified as a list that is iterated through to lookup keys in the original cache. A new tuple consisting of the (deduplicated) keys that weren't in the cache gets passed to the original function, which is expected to results - in a map of key to value for each passed value. THe new results are stored in the + in a map of key to value for each passed value. The new results are stored in the original cache. Note that any missing values are cached as None. Args: diff --git a/tests/rest/client/test_relations.py b/tests/rest/client/test_relations.py index e3d801f7a8..2d2b683548 100644 --- a/tests/rest/client/test_relations.py +++ b/tests/rest/client/test_relations.py @@ -1108,7 +1108,7 @@ class BundledAggregationsTestCase(BaseRelationsTestCase): # The "user" sent the root event and is making queries for the bundled # aggregations: they have participated. - self._test_bundled_aggregations(RelationTypes.THREAD, _gen_assert(True), 9) + self._test_bundled_aggregations(RelationTypes.THREAD, _gen_assert(True), 8) # The "user2" sent replies in the thread and is making queries for the # bundled aggregations: they have participated. # @@ -1170,7 +1170,7 @@ class BundledAggregationsTestCase(BaseRelationsTestCase): bundled_aggregations["latest_event"].get("unsigned"), ) - self._test_bundled_aggregations(RelationTypes.THREAD, assert_thread, 9) + self._test_bundled_aggregations(RelationTypes.THREAD, assert_thread, 8) def test_nested_thread(self) -> None: """ -- cgit 1.5.1 From 6d7523ef1484ec56f4a6dffdd2ea3d8736b4cc98 Mon Sep 17 00:00:00 2001 From: Patrick Cloke Date: Tue, 22 Nov 2022 09:41:09 -0500 Subject: Batch fetch bundled references (#14508) Avoid an n+1 query problem and fetch the bundled aggregations for m.reference relations in a single query instead of a query per event. This applies similar logic for as was previously done for edits in 8b309adb436c162510ed1402f33b8741d71fc058 (#11660; threads in b65acead428653b988351ae8d7b22127a22039cd (#11752); and annotations in 1799a54a545618782840a60950ef4b64da9ee24d (#14491). --- changelog.d/14508.feature | 1 + synapse/handlers/relations.py | 128 +++++++++++++--------------- synapse/storage/databases/main/cache.py | 1 + synapse/storage/databases/main/events.py | 4 + synapse/storage/databases/main/relations.py | 74 ++++++++++++++-- tests/rest/client/test_relations.py | 4 +- 6 files changed, 133 insertions(+), 79 deletions(-) create mode 100644 changelog.d/14508.feature (limited to 'synapse/handlers') diff --git a/changelog.d/14508.feature b/changelog.d/14508.feature new file mode 100644 index 0000000000..4fca7282f7 --- /dev/null +++ b/changelog.d/14508.feature @@ -0,0 +1 @@ +Reduce database load of [Client-Server endpoints](https://spec.matrix.org/v1.4/client-server-api/#aggregations) which return bundled aggregations. diff --git a/synapse/handlers/relations.py b/synapse/handlers/relations.py index ca94239f61..8414be5879 100644 --- a/synapse/handlers/relations.py +++ b/synapse/handlers/relations.py @@ -13,16 +13,7 @@ # limitations under the License. import enum import logging -from typing import ( - TYPE_CHECKING, - Collection, - Dict, - FrozenSet, - Iterable, - List, - Optional, - Tuple, -) +from typing import TYPE_CHECKING, Collection, Dict, FrozenSet, Iterable, List, Optional import attr @@ -32,7 +23,7 @@ from synapse.events import EventBase, relation_from_event from synapse.logging.opentracing import trace from synapse.storage.databases.main.relations import ThreadsNextBatch, _RelatedEvent from synapse.streams.config import PaginationConfig -from synapse.types import JsonDict, Requester, StreamToken, UserID +from synapse.types import JsonDict, Requester, UserID from synapse.visibility import filter_events_for_client if TYPE_CHECKING: @@ -181,40 +172,6 @@ class RelationsHandler: return return_value - async def get_relations_for_event( - self, - event_id: str, - event: EventBase, - room_id: str, - relation_type: str, - ignored_users: FrozenSet[str] = frozenset(), - ) -> Tuple[List[_RelatedEvent], Optional[StreamToken]]: - """Get a list of events which relate to an event, ordered by topological ordering. - - Args: - event_id: Fetch events that relate to this event ID. - event: The matching EventBase to event_id. - room_id: The room the event belongs to. - relation_type: The type of relation. - ignored_users: The users ignored by the requesting user. - - Returns: - List of event IDs that match relations requested. The rows are of - the form `{"event_id": "..."}`. - """ - - # Call the underlying storage method, which is cached. - related_events, next_token = await self._main_store.get_relations_for_event( - event_id, event, room_id, relation_type, direction="f" - ) - - # Filter out ignored users and convert to the expected format. - related_events = [ - event for event in related_events if event.sender not in ignored_users - ] - - return related_events, next_token - async def redact_events_related_to( self, requester: Requester, @@ -329,6 +286,46 @@ class RelationsHandler: return filtered_results + async def get_references_for_events( + self, event_ids: Collection[str], ignored_users: FrozenSet[str] = frozenset() + ) -> Dict[str, List[_RelatedEvent]]: + """Get a list of references to the given events. + + Args: + event_ids: Fetch events that relate to this event ID. + ignored_users: The users ignored by the requesting user. + + Returns: + A map of event IDs to a list related events. + """ + + related_events = await self._main_store.get_references_for_events(event_ids) + + # Avoid additional logic if there are no ignored users. + if not ignored_users: + return { + event_id: results + for event_id, results in related_events.items() + if results + } + + # Filter out ignored users. + results = {} + for event_id, events in related_events.items(): + # If no references, skip. + if not events: + continue + + # Filter ignored users out. + events = [event for event in events if event.sender not in ignored_users] + # If there are no events left, skip this event. + if not events: + continue + + results[event_id] = events + + return results + async def _get_threads_for_events( self, events_by_id: Dict[str, EventBase], @@ -412,14 +409,18 @@ class RelationsHandler: if event is None: continue - potential_events, _ = await self.get_relations_for_event( - event_id, - event, - room_id, - RelationTypes.THREAD, - ignored_users, + # Attempt to find another event to use as the latest event. + potential_events, _ = await self._main_store.get_relations_for_event( + event_id, event, room_id, RelationTypes.THREAD, direction="f" ) + # Filter out ignored users. + potential_events = [ + event + for event in potential_events + if event.sender not in ignored_users + ] + # If all found events are from ignored users, do not include # a summary of the thread. if not potential_events: @@ -534,27 +535,16 @@ class RelationsHandler: "chunk": annotations } - # Fetch other relations per event. - for event in events_by_id.values(): - # Fetch any references to bundle with this event. - references, next_token = await self.get_relations_for_event( - event.event_id, - event, - event.room_id, - RelationTypes.REFERENCE, - ignored_users=ignored_users, - ) + # Fetch any references to bundle with this event. + references_by_event_id = await self.get_references_for_events( + events_by_id.keys(), ignored_users=ignored_users + ) + for event_id, references in references_by_event_id.items(): if references: - aggregations = results.setdefault(event.event_id, BundledAggregations()) - aggregations.references = { + results.setdefault(event_id, BundledAggregations()).references = { "chunk": [{"event_id": ev.event_id} for ev in references] } - if next_token: - aggregations.references["next_batch"] = await next_token.to_string( - self._main_store - ) - # Fetch any edits (but not for redacted events). # # Note that there is no use in limiting edits by ignored users since the @@ -600,7 +590,7 @@ class RelationsHandler: room_id, requester, allow_departed_users=True ) - # Note that ignored users are not passed into get_relations_for_event + # Note that ignored users are not passed into get_threads # below. Ignored users are handled in filter_events_for_client (and by # not passing them in here we should get a better cache hit rate). thread_roots, next_batch = await self._main_store.get_threads( diff --git a/synapse/storage/databases/main/cache.py b/synapse/storage/databases/main/cache.py index ddb7397714..a58668a380 100644 --- a/synapse/storage/databases/main/cache.py +++ b/synapse/storage/databases/main/cache.py @@ -259,6 +259,7 @@ class CacheInvalidationWorkerStore(SQLBaseStore): if relates_to: self._attempt_to_invalidate_cache("get_relations_for_event", (relates_to,)) + self._attempt_to_invalidate_cache("get_references_for_event", (relates_to,)) self._attempt_to_invalidate_cache( "get_aggregation_groups_for_event", (relates_to,) ) diff --git a/synapse/storage/databases/main/events.py b/synapse/storage/databases/main/events.py index d68f127f9b..0f097a2927 100644 --- a/synapse/storage/databases/main/events.py +++ b/synapse/storage/databases/main/events.py @@ -2049,6 +2049,10 @@ class PersistEventsStore: self.store._invalidate_cache_and_stream( txn, self.store.get_aggregation_groups_for_event, (redacted_relates_to,) ) + if rel_type == RelationTypes.REFERENCE: + self.store._invalidate_cache_and_stream( + txn, self.store.get_references_for_event, (redacted_relates_to,) + ) if rel_type == RelationTypes.REPLACE: self.store._invalidate_cache_and_stream( txn, self.store.get_applicable_edit, (redacted_relates_to,) diff --git a/synapse/storage/databases/main/relations.py b/synapse/storage/databases/main/relations.py index f96a16956a..aea96e9d24 100644 --- a/synapse/storage/databases/main/relations.py +++ b/synapse/storage/databases/main/relations.py @@ -82,8 +82,6 @@ class _RelatedEvent: event_id: str # The sender of the related event. sender: str - topological_ordering: Optional[int] - stream_ordering: int class RelationsWorkerStore(SQLBaseStore): @@ -246,13 +244,17 @@ class RelationsWorkerStore(SQLBaseStore): txn.execute(sql, where_args + [limit + 1]) events = [] - for event_id, relation_type, sender, topo_ordering, stream_ordering in txn: + topo_orderings: List[int] = [] + stream_orderings: List[int] = [] + for event_id, relation_type, sender, topo_ordering, stream_ordering in cast( + List[Tuple[str, str, str, int, int]], txn + ): # Do not include edits for redacted events as they leak event # content. if not is_redacted or relation_type != RelationTypes.REPLACE: - events.append( - _RelatedEvent(event_id, sender, topo_ordering, stream_ordering) - ) + events.append(_RelatedEvent(event_id, sender)) + topo_orderings.append(topo_ordering) + stream_orderings.append(stream_ordering) # If there are more events, generate the next pagination key from the # last event returned. @@ -261,9 +263,11 @@ class RelationsWorkerStore(SQLBaseStore): # Instead of using the last row (which tells us there is more # data), use the last row to be returned. events = events[:limit] + topo_orderings = topo_orderings[:limit] + stream_orderings = stream_orderings[:limit] - topo = events[-1].topological_ordering - token = events[-1].stream_ordering + topo = topo_orderings[-1] + token = stream_orderings[-1] if direction == "b": # Tokens are positions between events. # This token points *after* the last event in the chunk. @@ -530,6 +534,60 @@ class RelationsWorkerStore(SQLBaseStore): "get_aggregation_groups_for_users", _get_aggregation_groups_for_users_txn ) + @cached() + async def get_references_for_event(self, event_id: str) -> List[JsonDict]: + raise NotImplementedError() + + @cachedList(cached_method_name="get_references_for_event", list_name="event_ids") + async def get_references_for_events( + self, event_ids: Collection[str] + ) -> Mapping[str, Optional[List[_RelatedEvent]]]: + """Get a list of references to the given events. + + Args: + event_ids: Fetch events that relate to these event IDs. + + Returns: + A map of event IDs to a list of related event IDs (and their senders). + """ + + clause, args = make_in_list_sql_clause( + self.database_engine, "relates_to_id", event_ids + ) + args.append(RelationTypes.REFERENCE) + + sql = f""" + SELECT relates_to_id, ref.event_id, ref.sender + FROM events AS ref + INNER JOIN event_relations USING (event_id) + INNER JOIN events AS parent ON + parent.event_id = relates_to_id + AND parent.room_id = ref.room_id + WHERE + {clause} + AND relation_type = ? + ORDER BY ref.topological_ordering, ref.stream_ordering + """ + + def _get_references_for_events_txn( + txn: LoggingTransaction, + ) -> Mapping[str, List[_RelatedEvent]]: + txn.execute(sql, args) + + result: Dict[str, List[_RelatedEvent]] = {} + for relates_to_id, event_id, sender in cast( + List[Tuple[str, str, str]], txn + ): + result.setdefault(relates_to_id, []).append( + _RelatedEvent(event_id, sender) + ) + + return result + + return await self.db_pool.runInteraction( + "_get_references_for_events_txn", _get_references_for_events_txn + ) + @cached() def get_applicable_edit(self, event_id: str) -> Optional[EventBase]: raise NotImplementedError() diff --git a/tests/rest/client/test_relations.py b/tests/rest/client/test_relations.py index 2d2b683548..b86f341ff5 100644 --- a/tests/rest/client/test_relations.py +++ b/tests/rest/client/test_relations.py @@ -1108,7 +1108,7 @@ class BundledAggregationsTestCase(BaseRelationsTestCase): # The "user" sent the root event and is making queries for the bundled # aggregations: they have participated. - self._test_bundled_aggregations(RelationTypes.THREAD, _gen_assert(True), 8) + self._test_bundled_aggregations(RelationTypes.THREAD, _gen_assert(True), 7) # The "user2" sent replies in the thread and is making queries for the # bundled aggregations: they have participated. # @@ -1170,7 +1170,7 @@ class BundledAggregationsTestCase(BaseRelationsTestCase): bundled_aggregations["latest_event"].get("unsigned"), ) - self._test_bundled_aggregations(RelationTypes.THREAD, assert_thread, 8) + self._test_bundled_aggregations(RelationTypes.THREAD, assert_thread, 7) def test_nested_thread(self) -> None: """ -- cgit 1.5.1 From 7eb74600423e00c6982493eed18551d7f294140d Mon Sep 17 00:00:00 2001 From: Patrick Cloke Date: Tue, 22 Nov 2022 09:47:32 -0500 Subject: Parallelize calls to fetch bundled aggregations. (#14510) The bundled aggregations for annotations, references, and edits can be parallelized. --- changelog.d/14510.feature | 1 + synapse/handlers/relations.py | 83 ++++++++++++++++++++++++++----------------- 2 files changed, 52 insertions(+), 32 deletions(-) create mode 100644 changelog.d/14510.feature (limited to 'synapse/handlers') diff --git a/changelog.d/14510.feature b/changelog.d/14510.feature new file mode 100644 index 0000000000..4fca7282f7 --- /dev/null +++ b/changelog.d/14510.feature @@ -0,0 +1 @@ +Reduce database load of [Client-Server endpoints](https://spec.matrix.org/v1.4/client-server-api/#aggregations) which return bundled aggregations. diff --git a/synapse/handlers/relations.py b/synapse/handlers/relations.py index 8414be5879..e96f9999a8 100644 --- a/synapse/handlers/relations.py +++ b/synapse/handlers/relations.py @@ -20,10 +20,12 @@ import attr from synapse.api.constants import EventTypes, RelationTypes from synapse.api.errors import SynapseError from synapse.events import EventBase, relation_from_event +from synapse.logging.context import make_deferred_yieldable, run_in_background from synapse.logging.opentracing import trace from synapse.storage.databases.main.relations import ThreadsNextBatch, _RelatedEvent from synapse.streams.config import PaginationConfig from synapse.types import JsonDict, Requester, UserID +from synapse.util.async_helpers import gather_results from synapse.visibility import filter_events_for_client if TYPE_CHECKING: @@ -525,39 +527,56 @@ class RelationsHandler: # (as that is what makes it part of the thread). relations_by_id[latest_thread_event.event_id] = RelationTypes.THREAD - # Fetch any annotations (ie, reactions) to bundle with this event. - annotations_by_event_id = await self.get_annotations_for_events( - events_by_id.keys(), ignored_users=ignored_users - ) - for event_id, annotations in annotations_by_event_id.items(): - if annotations: - results.setdefault(event_id, BundledAggregations()).annotations = { - "chunk": annotations - } - - # Fetch any references to bundle with this event. - references_by_event_id = await self.get_references_for_events( - events_by_id.keys(), ignored_users=ignored_users - ) - for event_id, references in references_by_event_id.items(): - if references: - results.setdefault(event_id, BundledAggregations()).references = { - "chunk": [{"event_id": ev.event_id} for ev in references] - } - - # Fetch any edits (but not for redacted events). - # - # Note that there is no use in limiting edits by ignored users since the - # parent event should be ignored in the first place if the user is ignored. - edits = await self._main_store.get_applicable_edits( - [ - event_id - for event_id, event in events_by_id.items() - if not event.internal_metadata.is_redacted() - ] + async def _fetch_annotations() -> None: + """Fetch any annotations (ie, reactions) to bundle with this event.""" + annotations_by_event_id = await self.get_annotations_for_events( + events_by_id.keys(), ignored_users=ignored_users + ) + for event_id, annotations in annotations_by_event_id.items(): + if annotations: + results.setdefault(event_id, BundledAggregations()).annotations = { + "chunk": annotations + } + + async def _fetch_references() -> None: + """Fetch any references to bundle with this event.""" + references_by_event_id = await self.get_references_for_events( + events_by_id.keys(), ignored_users=ignored_users + ) + for event_id, references in references_by_event_id.items(): + if references: + results.setdefault(event_id, BundledAggregations()).references = { + "chunk": [{"event_id": ev.event_id} for ev in references] + } + + async def _fetch_edits() -> None: + """ + Fetch any edits (but not for redacted events). + + Note that there is no use in limiting edits by ignored users since the + parent event should be ignored in the first place if the user is ignored. + """ + edits = await self._main_store.get_applicable_edits( + [ + event_id + for event_id, event in events_by_id.items() + if not event.internal_metadata.is_redacted() + ] + ) + for event_id, edit in edits.items(): + results.setdefault(event_id, BundledAggregations()).replace = edit + + # Parallelize the calls for annotations, references, and edits since they + # are unrelated. + await make_deferred_yieldable( + gather_results( + ( + run_in_background(_fetch_annotations), + run_in_background(_fetch_references), + run_in_background(_fetch_edits), + ) + ) ) - for event_id, edit in edits.items(): - results.setdefault(event_id, BundledAggregations()).replace = edit return results -- cgit 1.5.1 From 9cae44f49e6bf4f6b8a20ab11a65da417bb1565f Mon Sep 17 00:00:00 2001 From: Sean Quah <8349537+squahtx@users.noreply.github.com> Date: Tue, 22 Nov 2022 16:46:52 +0000 Subject: Track unconverted device list outbound pokes using a position instead (#14516) When a local device list change is added to `device_lists_changes_in_room`, the `converted_to_destinations` flag is set to `FALSE` and the `_handle_new_device_update_async` background process is started. This background process looks for unconverted rows in `device_lists_changes_in_room`, copies them to `device_lists_outbound_pokes` and updates the flag. To update the `converted_to_destinations` flag, the database performs a `DELETE` and `INSERT` internally, which fragments the table. To avoid this, track unconverted rows using a `(stream ID, room ID)` position instead of the flag. From now on, the `converted_to_destinations` column indicates rows that need converting to outbound pokes, but does not indicate whether the conversion has already taken place. Closes #14037. Signed-off-by: Sean Quah --- changelog.d/14516.misc | 1 + synapse/handlers/device.py | 30 +++++- synapse/storage/database.py | 13 +-- synapse/storage/databases/main/devices.py | 107 +++++++++++++-------- .../73/12refactor_device_list_outbound_pokes.sql | 53 ++++++++++ tests/storage/test_devices.py | 3 +- 6 files changed, 158 insertions(+), 49 deletions(-) create mode 100644 changelog.d/14516.misc create mode 100644 synapse/storage/schema/main/delta/73/12refactor_device_list_outbound_pokes.sql (limited to 'synapse/handlers') diff --git a/changelog.d/14516.misc b/changelog.d/14516.misc new file mode 100644 index 0000000000..51666c6ffc --- /dev/null +++ b/changelog.d/14516.misc @@ -0,0 +1 @@ +Refactor conversion of device list changes in room to outbound pokes to track unconverted rows using a `(stream ID, room ID)` position instead of updating the `converted_to_destinations` flag on every row. diff --git a/synapse/handlers/device.py b/synapse/handlers/device.py index c597639a7f..da3ddafeae 100644 --- a/synapse/handlers/device.py +++ b/synapse/handlers/device.py @@ -682,13 +682,33 @@ class DeviceHandler(DeviceWorkerHandler): hosts_already_sent_to: Set[str] = set() try: + stream_id, room_id = await self.store.get_device_change_last_converted_pos() + while True: self._handle_new_device_update_new_data = False - rows = await self.store.get_uncoverted_outbound_room_pokes() + max_stream_id = self.store.get_device_stream_token() + rows = await self.store.get_uncoverted_outbound_room_pokes( + stream_id, room_id + ) if not rows: # If the DB returned nothing then there is nothing left to # do, *unless* a new device list update happened during the # DB query. + + # Advance `(stream_id, room_id)`. + # `max_stream_id` comes from *before* the query for unconverted + # rows, which means that any unconverted rows must have a larger + # stream ID. + if max_stream_id > stream_id: + stream_id, room_id = max_stream_id, "" + await self.store.set_device_change_last_converted_pos( + stream_id, room_id + ) + else: + assert max_stream_id == stream_id + # Avoid moving `room_id` backwards. + pass + if self._handle_new_device_update_new_data: continue else: @@ -718,7 +738,6 @@ class DeviceHandler(DeviceWorkerHandler): user_id=user_id, device_id=device_id, room_id=room_id, - stream_id=stream_id, hosts=hosts, context=opentracing_context, ) @@ -752,6 +771,12 @@ class DeviceHandler(DeviceWorkerHandler): hosts_already_sent_to.update(hosts) current_stream_id = stream_id + # Advance `(stream_id, room_id)`. + _, _, room_id, stream_id, _ = rows[-1] + await self.store.set_device_change_last_converted_pos( + stream_id, room_id + ) + finally: self._handle_new_device_update_is_processing = False @@ -834,7 +859,6 @@ class DeviceHandler(DeviceWorkerHandler): user_id=user_id, device_id=device_id, room_id=room_id, - stream_id=None, hosts=potentially_changed_hosts, context=None, ) diff --git a/synapse/storage/database.py b/synapse/storage/database.py index 0dc44b246c..a14b13aec8 100644 --- a/synapse/storage/database.py +++ b/synapse/storage/database.py @@ -2075,13 +2075,14 @@ class DatabasePool: retcols: Collection[str], allow_none: bool = False, ) -> Optional[Dict[str, Any]]: - select_sql = "SELECT %s FROM %s WHERE %s" % ( - ", ".join(retcols), - table, - " AND ".join("%s = ?" % (k,) for k in keyvalues), - ) + select_sql = "SELECT %s FROM %s" % (", ".join(retcols), table) + + if keyvalues: + select_sql += " WHERE %s" % (" AND ".join("%s = ?" % k for k in keyvalues),) + txn.execute(select_sql, list(keyvalues.values())) + else: + txn.execute(select_sql) - txn.execute(select_sql, list(keyvalues.values())) row = txn.fetchone() if not row: diff --git a/synapse/storage/databases/main/devices.py b/synapse/storage/databases/main/devices.py index 57230df5ae..37629115ab 100644 --- a/synapse/storage/databases/main/devices.py +++ b/synapse/storage/databases/main/devices.py @@ -2008,27 +2008,48 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): ) async def get_uncoverted_outbound_room_pokes( - self, limit: int = 10 + self, start_stream_id: int, start_room_id: str, limit: int = 10 ) -> List[Tuple[str, str, str, int, Optional[Dict[str, str]]]]: """Get device list changes by room that have not yet been handled and written to `device_lists_outbound_pokes`. + Args: + start_stream_id: Together with `start_room_id`, indicates the position after + which to return device list changes. + start_room_id: Together with `start_stream_id`, indicates the position after + which to return device list changes. + limit: The maximum number of device list changes to return. + Returns: - A list of user ID, device ID, room ID, stream ID and optional opentracing context. + A list of user ID, device ID, room ID, stream ID and optional opentracing + context, in order of ascending (stream ID, room ID). """ sql = """ SELECT user_id, device_id, room_id, stream_id, opentracing_context FROM device_lists_changes_in_room - WHERE NOT converted_to_destinations - ORDER BY stream_id + WHERE + (stream_id, room_id) > (?, ?) AND + stream_id <= ? AND + NOT converted_to_destinations + ORDER BY stream_id ASC, room_id ASC LIMIT ? """ def get_uncoverted_outbound_room_pokes_txn( txn: LoggingTransaction, ) -> List[Tuple[str, str, str, int, Optional[Dict[str, str]]]]: - txn.execute(sql, (limit,)) + txn.execute( + sql, + ( + start_stream_id, + start_room_id, + # Avoid returning rows if there may be uncommitted device list + # changes with smaller stream IDs. + self._device_list_id_gen.get_current_token(), + limit, + ), + ) return [ ( @@ -2050,49 +2071,25 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): user_id: str, device_id: str, room_id: str, - stream_id: Optional[int], hosts: Collection[str], context: Optional[Dict[str, str]], ) -> None: """Queue the device update to be sent to the given set of hosts, calculated from the room ID. - - Marks the associated row in `device_lists_changes_in_room` as handled, - if `stream_id` is provided. """ + if not hosts: + return def add_device_list_outbound_pokes_txn( txn: LoggingTransaction, stream_ids: List[int] ) -> None: - if hosts: - self._add_device_outbound_poke_to_stream_txn( - txn, - user_id=user_id, - device_id=device_id, - hosts=hosts, - stream_ids=stream_ids, - context=context, - ) - - if stream_id: - self.db_pool.simple_update_txn( - txn, - table="device_lists_changes_in_room", - keyvalues={ - "user_id": user_id, - "device_id": device_id, - "stream_id": stream_id, - "room_id": room_id, - }, - updatevalues={"converted_to_destinations": True}, - ) - - if not hosts: - # If there are no hosts then we don't try and generate stream IDs. - return await self.db_pool.runInteraction( - "add_device_list_outbound_pokes", - add_device_list_outbound_pokes_txn, - [], + self._add_device_outbound_poke_to_stream_txn( + txn, + user_id=user_id, + device_id=device_id, + hosts=hosts, + stream_ids=stream_ids, + context=context, ) async with self._device_list_id_gen.get_next_mult(len(hosts)) as stream_ids: @@ -2156,3 +2153,37 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): "get_pending_remote_device_list_updates_for_room", get_pending_remote_device_list_updates_for_room_txn, ) + + async def get_device_change_last_converted_pos(self) -> Tuple[int, str]: + """ + Get the position of the last row in `device_list_changes_in_room` that has been + converted to `device_lists_outbound_pokes`. + + Rows with a strictly greater position where `converted_to_destinations` is + `FALSE` have not been converted. + """ + + row = await self.db_pool.simple_select_one( + table="device_lists_changes_converted_stream_position", + keyvalues={}, + retcols=["stream_id", "room_id"], + desc="get_device_change_last_converted_pos", + ) + return row["stream_id"], row["room_id"] + + async def set_device_change_last_converted_pos( + self, + stream_id: int, + room_id: str, + ) -> None: + """ + Set the position of the last row in `device_list_changes_in_room` that has been + converted to `device_lists_outbound_pokes`. + """ + + await self.db_pool.simple_update_one( + table="device_lists_changes_converted_stream_position", + keyvalues={}, + updatevalues={"stream_id": stream_id, "room_id": room_id}, + desc="set_device_change_last_converted_pos", + ) diff --git a/synapse/storage/schema/main/delta/73/12refactor_device_list_outbound_pokes.sql b/synapse/storage/schema/main/delta/73/12refactor_device_list_outbound_pokes.sql new file mode 100644 index 0000000000..93d7fcb79b --- /dev/null +++ b/synapse/storage/schema/main/delta/73/12refactor_device_list_outbound_pokes.sql @@ -0,0 +1,53 @@ +/* Copyright 2022 The Matrix.org Foundation C.I.C + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +-- Prior to this schema delta, we tracked the set of unconverted rows in +-- `device_lists_changes_in_room` using the `converted_to_destinations` flag. When rows +-- were converted to `device_lists_outbound_pokes`, the `converted_to_destinations` flag +-- would be set. +-- +-- After this schema delta, the `converted_to_destinations` is still populated like +-- before, but the set of unconverted rows is determined by the `stream_id` in the new +-- `device_lists_changes_converted_stream_position` table. +-- +-- If rolled back, Synapse will re-send all device list changes that happened since the +-- schema delta. + +CREATE TABLE IF NOT EXISTS device_lists_changes_converted_stream_position( + Lock CHAR(1) NOT NULL DEFAULT 'X' UNIQUE, -- Makes sure this table only has one row. + -- The (stream id, room id) of the last row in `device_lists_changes_in_room` that + -- has been converted to `device_lists_outbound_pokes`. Rows with a strictly larger + -- (stream id, room id) where `converted_to_destinations` is `FALSE` have not been + -- converted. + stream_id BIGINT NOT NULL, + -- `room_id` may be an empty string, which compares less than all valid room IDs. + room_id TEXT NOT NULL, + CHECK (Lock='X') +); + +INSERT INTO device_lists_changes_converted_stream_position (stream_id, room_id) VALUES ( + ( + SELECT COALESCE( + -- The last converted stream id is the smallest unconverted stream id minus + -- one. + MIN(stream_id) - 1, + -- If there is no unconverted stream id, the last converted stream id is the + -- largest stream id. + -- Otherwise, pick 1, since stream ids start at 2. + (SELECT COALESCE(MAX(stream_id), 1) FROM device_lists_changes_in_room) + ) FROM device_lists_changes_in_room WHERE NOT converted_to_destinations + ), + '' +); diff --git a/tests/storage/test_devices.py b/tests/storage/test_devices.py index f37505b6cf..8e7db2c4ec 100644 --- a/tests/storage/test_devices.py +++ b/tests/storage/test_devices.py @@ -28,7 +28,7 @@ class DeviceStoreTestCase(HomeserverTestCase): """ for device_id in device_ids: - stream_id = self.get_success( + self.get_success( self.store.add_device_change_to_streams( user_id, [device_id], ["!some:room"] ) @@ -39,7 +39,6 @@ class DeviceStoreTestCase(HomeserverTestCase): user_id=user_id, device_id=device_id, room_id="!some:room", - stream_id=stream_id, hosts=[host], context={}, ) -- cgit 1.5.1 From 6d47b7e32589e816eb766446cc1ff19ea73fc7c1 Mon Sep 17 00:00:00 2001 From: Patrick Cloke Date: Tue, 22 Nov 2022 14:08:04 -0500 Subject: Add a type hint for `get_device_handler()` and fix incorrect types. (#14055) This was the last untyped handler from the HomeServer object. Since it was being treated as Any (and thus unchecked) it was being used incorrectly in a few places. --- changelog.d/14055.misc | 1 + synapse/handlers/deactivate_account.py | 4 +++ synapse/handlers/device.py | 65 ++++++++++++++++++++++++++-------- synapse/handlers/e2e_keys.py | 61 ++++++++++++++++--------------- synapse/handlers/register.py | 4 +++ synapse/handlers/set_password.py | 6 +++- synapse/handlers/sso.py | 9 +++++ synapse/module_api/__init__.py | 10 +++++- synapse/replication/http/devices.py | 11 ++++-- synapse/rest/admin/__init__.py | 26 ++++++++------ synapse/rest/admin/devices.py | 13 +++++-- synapse/rest/client/devices.py | 17 ++++++--- synapse/rest/client/logout.py | 9 +++-- synapse/server.py | 2 +- tests/handlers/test_device.py | 19 ++++++---- tests/rest/admin/test_device.py | 5 ++- 16 files changed, 185 insertions(+), 77 deletions(-) create mode 100644 changelog.d/14055.misc (limited to 'synapse/handlers') diff --git a/changelog.d/14055.misc b/changelog.d/14055.misc new file mode 100644 index 0000000000..02980bc528 --- /dev/null +++ b/changelog.d/14055.misc @@ -0,0 +1 @@ +Add missing type hints to `HomeServer`. diff --git a/synapse/handlers/deactivate_account.py b/synapse/handlers/deactivate_account.py index 816e1a6d79..d74d135c0c 100644 --- a/synapse/handlers/deactivate_account.py +++ b/synapse/handlers/deactivate_account.py @@ -16,6 +16,7 @@ import logging from typing import TYPE_CHECKING, Optional from synapse.api.errors import SynapseError +from synapse.handlers.device import DeviceHandler from synapse.metrics.background_process_metrics import run_as_background_process from synapse.types import Codes, Requester, UserID, create_requester @@ -76,6 +77,9 @@ class DeactivateAccountHandler: True if identity server supports removing threepids, otherwise False. """ + # This can only be called on the main process. + assert isinstance(self._device_handler, DeviceHandler) + # Check if this user can be deactivated if not await self._third_party_rules.check_can_deactivate_user( user_id, by_admin diff --git a/synapse/handlers/device.py b/synapse/handlers/device.py index da3ddafeae..b1e55e1b9e 100644 --- a/synapse/handlers/device.py +++ b/synapse/handlers/device.py @@ -65,6 +65,8 @@ DELETE_STALE_DEVICES_INTERVAL_MS = 24 * 60 * 60 * 1000 class DeviceWorkerHandler: + device_list_updater: "DeviceListWorkerUpdater" + def __init__(self, hs: "HomeServer"): self.clock = hs.get_clock() self.hs = hs @@ -76,6 +78,8 @@ class DeviceWorkerHandler: self.server_name = hs.hostname self._msc3852_enabled = hs.config.experimental.msc3852_enabled + self.device_list_updater = DeviceListWorkerUpdater(hs) + @trace async def get_devices_by_user(self, user_id: str) -> List[JsonDict]: """ @@ -99,6 +103,19 @@ class DeviceWorkerHandler: log_kv(device_map) return devices + async def get_dehydrated_device( + self, user_id: str + ) -> Optional[Tuple[str, JsonDict]]: + """Retrieve the information for a dehydrated device. + + Args: + user_id: the user whose dehydrated device we are looking for + Returns: + a tuple whose first item is the device ID, and the second item is + the dehydrated device information + """ + return await self.store.get_dehydrated_device(user_id) + @trace async def get_device(self, user_id: str, device_id: str) -> JsonDict: """Retrieve the given device @@ -127,7 +144,7 @@ class DeviceWorkerHandler: @cancellable async def get_device_changes_in_shared_rooms( self, user_id: str, room_ids: Collection[str], from_token: StreamToken - ) -> Collection[str]: + ) -> Set[str]: """Get the set of users whose devices have changed who share a room with the given user. """ @@ -320,6 +337,8 @@ class DeviceWorkerHandler: class DeviceHandler(DeviceWorkerHandler): + device_list_updater: "DeviceListUpdater" + def __init__(self, hs: "HomeServer"): super().__init__(hs) @@ -606,19 +625,6 @@ class DeviceHandler(DeviceWorkerHandler): await self.delete_devices(user_id, [old_device_id]) return device_id - async def get_dehydrated_device( - self, user_id: str - ) -> Optional[Tuple[str, JsonDict]]: - """Retrieve the information for a dehydrated device. - - Args: - user_id: the user whose dehydrated device we are looking for - Returns: - a tuple whose first item is the device ID, and the second item is - the dehydrated device information - """ - return await self.store.get_dehydrated_device(user_id) - async def rehydrate_device( self, user_id: str, access_token: str, device_id: str ) -> dict: @@ -882,7 +888,36 @@ def _update_device_from_client_ips( ) -class DeviceListUpdater: +class DeviceListWorkerUpdater: + "Handles incoming device list updates from federation and contacts the main process over replication" + + def __init__(self, hs: "HomeServer"): + from synapse.replication.http.devices import ( + ReplicationUserDevicesResyncRestServlet, + ) + + self._user_device_resync_client = ( + ReplicationUserDevicesResyncRestServlet.make_client(hs) + ) + + async def user_device_resync( + self, user_id: str, mark_failed_as_stale: bool = True + ) -> Optional[JsonDict]: + """Fetches all devices for a user and updates the device cache with them. + + Args: + user_id: The user's id whose device_list will be updated. + mark_failed_as_stale: Whether to mark the user's device list as stale + if the attempt to resync failed. + Returns: + A dict with device info as under the "devices" in the result of this + request: + https://matrix.org/docs/spec/server_server/r0.1.2#get-matrix-federation-v1-user-devices-userid + """ + return await self._user_device_resync_client(user_id=user_id) + + +class DeviceListUpdater(DeviceListWorkerUpdater): "Handles incoming device list updates from federation and updates the DB" def __init__(self, hs: "HomeServer", device_handler: DeviceHandler): diff --git a/synapse/handlers/e2e_keys.py b/synapse/handlers/e2e_keys.py index bf1221f523..5fe102e2f2 100644 --- a/synapse/handlers/e2e_keys.py +++ b/synapse/handlers/e2e_keys.py @@ -27,9 +27,9 @@ from twisted.internet import defer from synapse.api.constants import EduTypes from synapse.api.errors import CodeMessageException, Codes, NotFoundError, SynapseError +from synapse.handlers.device import DeviceHandler from synapse.logging.context import make_deferred_yieldable, run_in_background from synapse.logging.opentracing import log_kv, set_tag, tag_args, trace -from synapse.replication.http.devices import ReplicationUserDevicesResyncRestServlet from synapse.types import ( JsonDict, UserID, @@ -56,27 +56,23 @@ class E2eKeysHandler: self.is_mine = hs.is_mine self.clock = hs.get_clock() - self._edu_updater = SigningKeyEduUpdater(hs, self) - federation_registry = hs.get_federation_registry() - self._is_master = hs.config.worker.worker_app is None - if not self._is_master: - self._user_device_resync_client = ( - ReplicationUserDevicesResyncRestServlet.make_client(hs) - ) - else: + is_master = hs.config.worker.worker_app is None + if is_master: + edu_updater = SigningKeyEduUpdater(hs) + # Only register this edu handler on master as it requires writing # device updates to the db federation_registry.register_edu_handler( EduTypes.SIGNING_KEY_UPDATE, - self._edu_updater.incoming_signing_key_update, + edu_updater.incoming_signing_key_update, ) # also handle the unstable version # FIXME: remove this when enough servers have upgraded federation_registry.register_edu_handler( EduTypes.UNSTABLE_SIGNING_KEY_UPDATE, - self._edu_updater.incoming_signing_key_update, + edu_updater.incoming_signing_key_update, ) # doesn't really work as part of the generic query API, because the @@ -319,14 +315,13 @@ class E2eKeysHandler: # probably be tracking their device lists. However, we haven't # done an initial sync on the device list so we do it now. try: - if self._is_master: - resync_results = await self.device_handler.device_list_updater.user_device_resync( + resync_results = ( + await self.device_handler.device_list_updater.user_device_resync( user_id ) - else: - resync_results = await self._user_device_resync_client( - user_id=user_id - ) + ) + if resync_results is None: + raise ValueError("Device resync failed") # Add the device keys to the results. user_devices = resync_results["devices"] @@ -605,6 +600,8 @@ class E2eKeysHandler: async def upload_keys_for_user( self, user_id: str, device_id: str, keys: JsonDict ) -> JsonDict: + # This can only be called from the main process. + assert isinstance(self.device_handler, DeviceHandler) time_now = self.clock.time_msec() @@ -732,6 +729,8 @@ class E2eKeysHandler: user_id: the user uploading the keys keys: the signing keys """ + # This can only be called from the main process. + assert isinstance(self.device_handler, DeviceHandler) # if a master key is uploaded, then check it. Otherwise, load the # stored master key, to check signatures on other keys @@ -823,6 +822,9 @@ class E2eKeysHandler: Raises: SynapseError: if the signatures dict is not valid. """ + # This can only be called from the main process. + assert isinstance(self.device_handler, DeviceHandler) + failures = {} # signatures to be stored. Each item will be a SignatureListItem @@ -1200,6 +1202,9 @@ class E2eKeysHandler: A tuple of the retrieved key content, the key's ID and the matching VerifyKey. If the key cannot be retrieved, all values in the tuple will instead be None. """ + # This can only be called from the main process. + assert isinstance(self.device_handler, DeviceHandler) + try: remote_result = await self.federation.query_user_devices( user.domain, user.to_string() @@ -1396,11 +1401,14 @@ class SignatureListItem: class SigningKeyEduUpdater: """Handles incoming signing key updates from federation and updates the DB""" - def __init__(self, hs: "HomeServer", e2e_keys_handler: E2eKeysHandler): + def __init__(self, hs: "HomeServer"): self.store = hs.get_datastores().main self.federation = hs.get_federation_client() self.clock = hs.get_clock() - self.e2e_keys_handler = e2e_keys_handler + + device_handler = hs.get_device_handler() + assert isinstance(device_handler, DeviceHandler) + self._device_handler = device_handler self._remote_edu_linearizer = Linearizer(name="remote_signing_key") @@ -1445,9 +1453,6 @@ class SigningKeyEduUpdater: user_id: the user whose updates we are processing """ - device_handler = self.e2e_keys_handler.device_handler - device_list_updater = device_handler.device_list_updater - async with self._remote_edu_linearizer.queue(user_id): pending_updates = self._pending_updates.pop(user_id, []) if not pending_updates: @@ -1459,13 +1464,11 @@ class SigningKeyEduUpdater: logger.info("pending updates: %r", pending_updates) for master_key, self_signing_key in pending_updates: - new_device_ids = ( - await device_list_updater.process_cross_signing_key_update( - user_id, - master_key, - self_signing_key, - ) + new_device_ids = await self._device_handler.device_list_updater.process_cross_signing_key_update( + user_id, + master_key, + self_signing_key, ) device_ids = device_ids + new_device_ids - await device_handler.notify_device_update(user_id, device_ids) + await self._device_handler.notify_device_update(user_id, device_ids) diff --git a/synapse/handlers/register.py b/synapse/handlers/register.py index ca1c7a1866..6307fa9c5d 100644 --- a/synapse/handlers/register.py +++ b/synapse/handlers/register.py @@ -38,6 +38,7 @@ from synapse.api.errors import ( ) from synapse.appservice import ApplicationService from synapse.config.server import is_threepid_reserved +from synapse.handlers.device import DeviceHandler from synapse.http.servlet import assert_params_in_dict from synapse.replication.http.login import RegisterDeviceReplicationServlet from synapse.replication.http.register import ( @@ -841,6 +842,9 @@ class RegistrationHandler: refresh_token = None refresh_token_id = None + # This can only run on the main process. + assert isinstance(self.device_handler, DeviceHandler) + registered_device_id = await self.device_handler.check_device_registered( user_id, device_id, diff --git a/synapse/handlers/set_password.py b/synapse/handlers/set_password.py index 73861bbd40..bd9d0bb34b 100644 --- a/synapse/handlers/set_password.py +++ b/synapse/handlers/set_password.py @@ -15,6 +15,7 @@ import logging from typing import TYPE_CHECKING, Optional from synapse.api.errors import Codes, StoreError, SynapseError +from synapse.handlers.device import DeviceHandler from synapse.types import Requester if TYPE_CHECKING: @@ -29,7 +30,10 @@ class SetPasswordHandler: def __init__(self, hs: "HomeServer"): self.store = hs.get_datastores().main self._auth_handler = hs.get_auth_handler() - self._device_handler = hs.get_device_handler() + # This can only be instantiated on the main process. + device_handler = hs.get_device_handler() + assert isinstance(device_handler, DeviceHandler) + self._device_handler = device_handler async def set_password( self, diff --git a/synapse/handlers/sso.py b/synapse/handlers/sso.py index 749d7e93b0..e1c0bff1b2 100644 --- a/synapse/handlers/sso.py +++ b/synapse/handlers/sso.py @@ -37,6 +37,7 @@ from twisted.web.server import Request from synapse.api.constants import LoginType from synapse.api.errors import Codes, NotFoundError, RedirectException, SynapseError from synapse.config.sso import SsoAttributeRequirement +from synapse.handlers.device import DeviceHandler from synapse.handlers.register import init_counters_for_auth_provider from synapse.handlers.ui_auth import UIAuthSessionDataConstants from synapse.http import get_request_user_agent @@ -1035,6 +1036,8 @@ class SsoHandler: ) -> None: """Revoke any devices and in-flight logins tied to a provider session. + Can only be called from the main process. + Args: auth_provider_id: A unique identifier for this SSO provider, e.g. "oidc" or "saml". @@ -1042,6 +1045,12 @@ class SsoHandler: expected_user_id: The user we're expecting to logout. If set, it will ignore sessions belonging to other users and log an error. """ + + # It is expected that this is the main process. + assert isinstance( + self._device_handler, DeviceHandler + ), "revoking SSO sessions can only be called on the main process" + # Invalidate any running user-mapping sessions to_delete = [] for session_id, session in self._username_mapping_sessions.items(): diff --git a/synapse/module_api/__init__.py b/synapse/module_api/__init__.py index 1adc1fd64f..96a661177a 100644 --- a/synapse/module_api/__init__.py +++ b/synapse/module_api/__init__.py @@ -86,6 +86,7 @@ from synapse.handlers.auth import ( ON_LOGGED_OUT_CALLBACK, AuthHandler, ) +from synapse.handlers.device import DeviceHandler from synapse.handlers.push_rules import RuleSpec, check_actions from synapse.http.client import SimpleHttpClient from synapse.http.server import ( @@ -207,6 +208,7 @@ class ModuleApi: self._registration_handler = hs.get_registration_handler() self._send_email_handler = hs.get_send_email_handler() self._push_rules_handler = hs.get_push_rules_handler() + self._device_handler = hs.get_device_handler() self.custom_template_dir = hs.config.server.custom_template_directory try: @@ -784,6 +786,8 @@ class ModuleApi: ) -> Generator["defer.Deferred[Any]", Any, None]: """Invalidate an access token for a user + Can only be called from the main process. + Added in Synapse v0.25.0. Args: @@ -796,6 +800,10 @@ class ModuleApi: Raises: synapse.api.errors.AuthError: the access token is invalid """ + assert isinstance( + self._device_handler, DeviceHandler + ), "invalidate_access_token can only be called on the main process" + # see if the access token corresponds to a device user_info = yield defer.ensureDeferred( self._auth.get_user_by_access_token(access_token) @@ -805,7 +813,7 @@ class ModuleApi: if device_id: # delete the device, which will also delete its access tokens yield defer.ensureDeferred( - self._hs.get_device_handler().delete_devices(user_id, [device_id]) + self._device_handler.delete_devices(user_id, [device_id]) ) else: # no associated device. Just delete the access token. diff --git a/synapse/replication/http/devices.py b/synapse/replication/http/devices.py index c21629def8..7c4941c3d3 100644 --- a/synapse/replication/http/devices.py +++ b/synapse/replication/http/devices.py @@ -13,7 +13,7 @@ # limitations under the License. import logging -from typing import TYPE_CHECKING, Tuple +from typing import TYPE_CHECKING, Optional, Tuple from twisted.web.server import Request @@ -63,7 +63,12 @@ class ReplicationUserDevicesResyncRestServlet(ReplicationEndpoint): def __init__(self, hs: "HomeServer"): super().__init__(hs) - self.device_list_updater = hs.get_device_handler().device_list_updater + from synapse.handlers.device import DeviceHandler + + handler = hs.get_device_handler() + assert isinstance(handler, DeviceHandler) + self.device_list_updater = handler.device_list_updater + self.store = hs.get_datastores().main self.clock = hs.get_clock() @@ -73,7 +78,7 @@ class ReplicationUserDevicesResyncRestServlet(ReplicationEndpoint): async def _handle_request( # type: ignore[override] self, request: Request, user_id: str - ) -> Tuple[int, JsonDict]: + ) -> Tuple[int, Optional[JsonDict]]: user_devices = await self.device_list_updater.user_device_resync(user_id) return 200, user_devices diff --git a/synapse/rest/admin/__init__.py b/synapse/rest/admin/__init__.py index c62ea22116..fb73886df0 100644 --- a/synapse/rest/admin/__init__.py +++ b/synapse/rest/admin/__init__.py @@ -238,6 +238,10 @@ def register_servlets(hs: "HomeServer", http_server: HttpServer) -> None: """ Register all the admin servlets. """ + # Admin servlets aren't registered on workers. + if hs.config.worker.worker_app is not None: + return + register_servlets_for_client_rest_resource(hs, http_server) BlockRoomRestServlet(hs).register(http_server) ListRoomRestServlet(hs).register(http_server) @@ -254,9 +258,6 @@ def register_servlets(hs: "HomeServer", http_server: HttpServer) -> None: UserTokenRestServlet(hs).register(http_server) UserRestServletV2(hs).register(http_server) UsersRestServletV2(hs).register(http_server) - DeviceRestServlet(hs).register(http_server) - DevicesRestServlet(hs).register(http_server) - DeleteDevicesRestServlet(hs).register(http_server) UserMediaStatisticsRestServlet(hs).register(http_server) EventReportDetailRestServlet(hs).register(http_server) EventReportsRestServlet(hs).register(http_server) @@ -280,12 +281,13 @@ def register_servlets(hs: "HomeServer", http_server: HttpServer) -> None: UserByExternalId(hs).register(http_server) UserByThreePid(hs).register(http_server) - # Some servlets only get registered for the main process. - if hs.config.worker.worker_app is None: - SendServerNoticeServlet(hs).register(http_server) - BackgroundUpdateEnabledRestServlet(hs).register(http_server) - BackgroundUpdateRestServlet(hs).register(http_server) - BackgroundUpdateStartJobRestServlet(hs).register(http_server) + DeviceRestServlet(hs).register(http_server) + DevicesRestServlet(hs).register(http_server) + DeleteDevicesRestServlet(hs).register(http_server) + SendServerNoticeServlet(hs).register(http_server) + BackgroundUpdateEnabledRestServlet(hs).register(http_server) + BackgroundUpdateRestServlet(hs).register(http_server) + BackgroundUpdateStartJobRestServlet(hs).register(http_server) def register_servlets_for_client_rest_resource( @@ -294,9 +296,11 @@ def register_servlets_for_client_rest_resource( """Register only the servlets which need to be exposed on /_matrix/client/xxx""" WhoisRestServlet(hs).register(http_server) PurgeHistoryStatusRestServlet(hs).register(http_server) - DeactivateAccountRestServlet(hs).register(http_server) PurgeHistoryRestServlet(hs).register(http_server) - ResetPasswordRestServlet(hs).register(http_server) + # The following resources can only be run on the main process. + if hs.config.worker.worker_app is None: + DeactivateAccountRestServlet(hs).register(http_server) + ResetPasswordRestServlet(hs).register(http_server) SearchUsersRestServlet(hs).register(http_server) UserRegisterServlet(hs).register(http_server) AccountValidityRenewServlet(hs).register(http_server) diff --git a/synapse/rest/admin/devices.py b/synapse/rest/admin/devices.py index d934880102..3b2f2d9abb 100644 --- a/synapse/rest/admin/devices.py +++ b/synapse/rest/admin/devices.py @@ -16,6 +16,7 @@ from http import HTTPStatus from typing import TYPE_CHECKING, Tuple from synapse.api.errors import NotFoundError, SynapseError +from synapse.handlers.device import DeviceHandler from synapse.http.servlet import ( RestServlet, assert_params_in_dict, @@ -43,7 +44,9 @@ class DeviceRestServlet(RestServlet): def __init__(self, hs: "HomeServer"): super().__init__() self.auth = hs.get_auth() - self.device_handler = hs.get_device_handler() + handler = hs.get_device_handler() + assert isinstance(handler, DeviceHandler) + self.device_handler = handler self.store = hs.get_datastores().main self.is_mine = hs.is_mine @@ -112,7 +115,9 @@ class DevicesRestServlet(RestServlet): def __init__(self, hs: "HomeServer"): self.auth = hs.get_auth() - self.device_handler = hs.get_device_handler() + handler = hs.get_device_handler() + assert isinstance(handler, DeviceHandler) + self.device_handler = handler self.store = hs.get_datastores().main self.is_mine = hs.is_mine @@ -143,7 +148,9 @@ class DeleteDevicesRestServlet(RestServlet): def __init__(self, hs: "HomeServer"): self.auth = hs.get_auth() - self.device_handler = hs.get_device_handler() + handler = hs.get_device_handler() + assert isinstance(handler, DeviceHandler) + self.device_handler = handler self.store = hs.get_datastores().main self.is_mine = hs.is_mine diff --git a/synapse/rest/client/devices.py b/synapse/rest/client/devices.py index 8f3cbd4ea2..69b803f9f8 100644 --- a/synapse/rest/client/devices.py +++ b/synapse/rest/client/devices.py @@ -20,6 +20,7 @@ from pydantic import Extra, StrictStr from synapse.api import errors from synapse.api.errors import NotFoundError +from synapse.handlers.device import DeviceHandler from synapse.http.server import HttpServer from synapse.http.servlet import ( RestServlet, @@ -80,7 +81,9 @@ class DeleteDevicesRestServlet(RestServlet): super().__init__() self.hs = hs self.auth = hs.get_auth() - self.device_handler = hs.get_device_handler() + handler = hs.get_device_handler() + assert isinstance(handler, DeviceHandler) + self.device_handler = handler self.auth_handler = hs.get_auth_handler() class PostBody(RequestBodyModel): @@ -125,7 +128,9 @@ class DeviceRestServlet(RestServlet): super().__init__() self.hs = hs self.auth = hs.get_auth() - self.device_handler = hs.get_device_handler() + handler = hs.get_device_handler() + assert isinstance(handler, DeviceHandler) + self.device_handler = handler self.auth_handler = hs.get_auth_handler() self._msc3852_enabled = hs.config.experimental.msc3852_enabled @@ -256,7 +261,9 @@ class DehydratedDeviceServlet(RestServlet): super().__init__() self.hs = hs self.auth = hs.get_auth() - self.device_handler = hs.get_device_handler() + handler = hs.get_device_handler() + assert isinstance(handler, DeviceHandler) + self.device_handler = handler async def on_GET(self, request: SynapseRequest) -> Tuple[int, JsonDict]: requester = await self.auth.get_user_by_req(request) @@ -313,7 +320,9 @@ class ClaimDehydratedDeviceServlet(RestServlet): super().__init__() self.hs = hs self.auth = hs.get_auth() - self.device_handler = hs.get_device_handler() + handler = hs.get_device_handler() + assert isinstance(handler, DeviceHandler) + self.device_handler = handler class PostBody(RequestBodyModel): device_id: StrictStr diff --git a/synapse/rest/client/logout.py b/synapse/rest/client/logout.py index 23dfa4518f..6d34625ad5 100644 --- a/synapse/rest/client/logout.py +++ b/synapse/rest/client/logout.py @@ -15,6 +15,7 @@ import logging from typing import TYPE_CHECKING, Tuple +from synapse.handlers.device import DeviceHandler from synapse.http.server import HttpServer from synapse.http.servlet import RestServlet from synapse.http.site import SynapseRequest @@ -34,7 +35,9 @@ class LogoutRestServlet(RestServlet): super().__init__() self.auth = hs.get_auth() self._auth_handler = hs.get_auth_handler() - self._device_handler = hs.get_device_handler() + handler = hs.get_device_handler() + assert isinstance(handler, DeviceHandler) + self._device_handler = handler async def on_POST(self, request: SynapseRequest) -> Tuple[int, JsonDict]: requester = await self.auth.get_user_by_req(request, allow_expired=True) @@ -59,7 +62,9 @@ class LogoutAllRestServlet(RestServlet): super().__init__() self.auth = hs.get_auth() self._auth_handler = hs.get_auth_handler() - self._device_handler = hs.get_device_handler() + handler = hs.get_device_handler() + assert isinstance(handler, DeviceHandler) + self._device_handler = handler async def on_POST(self, request: SynapseRequest) -> Tuple[int, JsonDict]: requester = await self.auth.get_user_by_req(request, allow_expired=True) diff --git a/synapse/server.py b/synapse/server.py index f0a60d0056..5baae2325e 100644 --- a/synapse/server.py +++ b/synapse/server.py @@ -510,7 +510,7 @@ class HomeServer(metaclass=abc.ABCMeta): ) @cache_in_self - def get_device_handler(self): + def get_device_handler(self) -> DeviceWorkerHandler: if self.config.worker.worker_app: return DeviceWorkerHandler(self) else: diff --git a/tests/handlers/test_device.py b/tests/handlers/test_device.py index b8b465d35b..ce7525e29c 100644 --- a/tests/handlers/test_device.py +++ b/tests/handlers/test_device.py @@ -19,7 +19,7 @@ from typing import Optional from twisted.test.proto_helpers import MemoryReactor from synapse.api.errors import NotFoundError, SynapseError -from synapse.handlers.device import MAX_DEVICE_DISPLAY_NAME_LEN +from synapse.handlers.device import MAX_DEVICE_DISPLAY_NAME_LEN, DeviceHandler from synapse.server import HomeServer from synapse.util import Clock @@ -32,7 +32,9 @@ user2 = "@theresa:bbb" class DeviceTestCase(unittest.HomeserverTestCase): def make_homeserver(self, reactor: MemoryReactor, clock: Clock) -> HomeServer: hs = self.setup_test_homeserver("server", federation_http_client=None) - self.handler = hs.get_device_handler() + handler = hs.get_device_handler() + assert isinstance(handler, DeviceHandler) + self.handler = handler self.store = hs.get_datastores().main return hs @@ -61,6 +63,7 @@ class DeviceTestCase(unittest.HomeserverTestCase): self.assertEqual(res, "fco") dev = self.get_success(self.handler.store.get_device("@boris:foo", "fco")) + assert dev is not None self.assertEqual(dev["display_name"], "display name") def test_device_is_preserved_if_exists(self) -> None: @@ -83,6 +86,7 @@ class DeviceTestCase(unittest.HomeserverTestCase): self.assertEqual(res2, "fco") dev = self.get_success(self.handler.store.get_device("@boris:foo", "fco")) + assert dev is not None self.assertEqual(dev["display_name"], "display name") def test_device_id_is_made_up_if_unspecified(self) -> None: @@ -95,6 +99,7 @@ class DeviceTestCase(unittest.HomeserverTestCase): ) dev = self.get_success(self.handler.store.get_device("@theresa:foo", device_id)) + assert dev is not None self.assertEqual(dev["display_name"], "display") def test_get_devices_by_user(self) -> None: @@ -264,7 +269,9 @@ class DeviceTestCase(unittest.HomeserverTestCase): class DehydrationTestCase(unittest.HomeserverTestCase): def make_homeserver(self, reactor: MemoryReactor, clock: Clock) -> HomeServer: hs = self.setup_test_homeserver("server", federation_http_client=None) - self.handler = hs.get_device_handler() + handler = hs.get_device_handler() + assert isinstance(handler, DeviceHandler) + self.handler = handler self.registration = hs.get_registration_handler() self.auth = hs.get_auth() self.store = hs.get_datastores().main @@ -284,9 +291,9 @@ class DehydrationTestCase(unittest.HomeserverTestCase): ) ) - retrieved_device_id, device_data = self.get_success( - self.handler.get_dehydrated_device(user_id=user_id) - ) + result = self.get_success(self.handler.get_dehydrated_device(user_id=user_id)) + assert result is not None + retrieved_device_id, device_data = result self.assertEqual(retrieved_device_id, stored_dehydrated_device_id) self.assertEqual(device_data, {"device_data": {"foo": "bar"}}) diff --git a/tests/rest/admin/test_device.py b/tests/rest/admin/test_device.py index d52aee8f92..03f2112b07 100644 --- a/tests/rest/admin/test_device.py +++ b/tests/rest/admin/test_device.py @@ -19,6 +19,7 @@ from twisted.test.proto_helpers import MemoryReactor import synapse.rest.admin from synapse.api.errors import Codes +from synapse.handlers.device import DeviceHandler from synapse.rest.client import login from synapse.server import HomeServer from synapse.util import Clock @@ -34,7 +35,9 @@ class DeviceRestTestCase(unittest.HomeserverTestCase): ] def prepare(self, reactor: MemoryReactor, clock: Clock, hs: HomeServer) -> None: - self.handler = hs.get_device_handler() + handler = hs.get_device_handler() + assert isinstance(handler, DeviceHandler) + self.handler = handler self.admin_user = self.register_user("admin", "pass", admin=True) self.admin_user_tok = self.login("admin", "pass") -- cgit 1.5.1 From 3b4e1508689cc09eba30509249459a64431558fc Mon Sep 17 00:00:00 2001 From: Mathieu Velten Date: Thu, 24 Nov 2022 09:10:47 +0100 Subject: Faster joins: use servers list approximation in `assert_host_in_room` (#14515) Signed-off-by: Mathieu Velten --- changelog.d/14515.misc | 1 + synapse/handlers/event_auth.py | 28 +++++++++++++++++----------- 2 files changed, 18 insertions(+), 11 deletions(-) create mode 100644 changelog.d/14515.misc (limited to 'synapse/handlers') diff --git a/changelog.d/14515.misc b/changelog.d/14515.misc new file mode 100644 index 0000000000..a0effb4dbe --- /dev/null +++ b/changelog.d/14515.misc @@ -0,0 +1 @@ +Faster joins: use servers list approximation received during `send_join` (potentially updated with received membership events) in `assert_host_in_room`. \ No newline at end of file diff --git a/synapse/handlers/event_auth.py b/synapse/handlers/event_auth.py index 3bbad0271b..f91dbbecb7 100644 --- a/synapse/handlers/event_auth.py +++ b/synapse/handlers/event_auth.py @@ -45,6 +45,7 @@ class EventAuthHandler: def __init__(self, hs: "HomeServer"): self._clock = hs.get_clock() self._store = hs.get_datastores().main + self._state_storage_controller = hs.get_storage_controllers().state self._server_name = hs.hostname async def check_auth_rules_from_context( @@ -179,17 +180,22 @@ class EventAuthHandler: this function may return an incorrect result as we are not able to fully track server membership in a room without full state. """ - if not allow_partial_state_rooms and await self._store.is_partial_state_room( - room_id - ): - raise AuthError( - 403, - "Unable to authorise you right now; room is partial-stated here.", - errcode=Codes.UNABLE_DUE_TO_PARTIAL_STATE, - ) - - if not await self.is_host_in_room(room_id, host): - raise AuthError(403, "Host not in room.") + if await self._store.is_partial_state_room(room_id): + if allow_partial_state_rooms: + current_hosts = await self._state_storage_controller.get_current_hosts_in_room_or_partial_state_approximation( + room_id + ) + if host not in current_hosts: + raise AuthError(403, "Host not in room (partial-state approx).") + else: + raise AuthError( + 403, + "Unable to authorise you right now; room is partial-stated here.", + errcode=Codes.UNABLE_DUE_TO_PARTIAL_STATE, + ) + else: + if not await self.is_host_in_room(room_id, host): + raise AuthError(403, "Host not in room.") async def check_restricted_join_rules( self, -- cgit 1.5.1 From f6c74d1cb2ed966802b01a2b037f09ce7a842c18 Mon Sep 17 00:00:00 2001 From: Benjamin Kampmann Date: Thu, 24 Nov 2022 09:10:51 +0000 Subject: Implement message forward pagination from start when no from is given, fixes #12383 (#14149) Fixes https://github.com/matrix-org/synapse/issues/12383 --- changelog.d/14149.bugfix | 1 + synapse/handlers/pagination.py | 6 ++++++ synapse/streams/events.py | 13 +++++++++++++ tests/rest/admin/test_room.py | 40 ++++++++++++++++++++++++++++++++++++++++ 4 files changed, 60 insertions(+) create mode 100644 changelog.d/14149.bugfix (limited to 'synapse/handlers') diff --git a/changelog.d/14149.bugfix b/changelog.d/14149.bugfix new file mode 100644 index 0000000000..b31c658266 --- /dev/null +++ b/changelog.d/14149.bugfix @@ -0,0 +1 @@ +Fix #12383: paginate room messages from the start if no from is given. Contributed by @gnunicorn . \ No newline at end of file diff --git a/synapse/handlers/pagination.py b/synapse/handlers/pagination.py index a4ca9cb8b4..c572508a02 100644 --- a/synapse/handlers/pagination.py +++ b/synapse/handlers/pagination.py @@ -448,6 +448,12 @@ class PaginationHandler: if pagin_config.from_token: from_token = pagin_config.from_token + elif pagin_config.direction == "f": + from_token = ( + await self.hs.get_event_sources().get_start_token_for_pagination( + room_id + ) + ) else: from_token = ( await self.hs.get_event_sources().get_current_token_for_pagination( diff --git a/synapse/streams/events.py b/synapse/streams/events.py index f331e1af16..619eb7f601 100644 --- a/synapse/streams/events.py +++ b/synapse/streams/events.py @@ -73,6 +73,19 @@ class EventSources: ) return token + @trace + async def get_start_token_for_pagination(self, room_id: str) -> StreamToken: + """Get the start token for a given room to be used to paginate + events. + + The returned token does not have the current values for fields other + than `room`, since they are not used during pagination. + + Returns: + The start token for pagination. + """ + return StreamToken.START + @trace async def get_current_token_for_pagination(self, room_id: str) -> StreamToken: """Get the current token for a given room to be used to paginate diff --git a/tests/rest/admin/test_room.py b/tests/rest/admin/test_room.py index d156be82b0..e0f5d54aba 100644 --- a/tests/rest/admin/test_room.py +++ b/tests/rest/admin/test_room.py @@ -1857,6 +1857,46 @@ class RoomMessagesTestCase(unittest.HomeserverTestCase): self.assertIn("chunk", channel.json_body) self.assertIn("end", channel.json_body) + def test_room_messages_backward(self) -> None: + """Test room messages can be retrieved by an admin that isn't in the room.""" + latest_event_id = self.helper.send( + self.room_id, body="message 1", tok=self.user_tok + )["event_id"] + + # Check that we get the first and second message when querying /messages. + channel = self.make_request( + "GET", + "/_synapse/admin/v1/rooms/%s/messages?dir=b" % (self.room_id,), + access_token=self.admin_user_tok, + ) + self.assertEqual(channel.code, 200, channel.json_body) + + chunk = channel.json_body["chunk"] + self.assertEqual(len(chunk), 6, [event["content"] for event in chunk]) + + # in backwards, this is the first event + self.assertEqual(chunk[0]["event_id"], latest_event_id) + + def test_room_messages_forward(self) -> None: + """Test room messages can be retrieved by an admin that isn't in the room.""" + latest_event_id = self.helper.send( + self.room_id, body="message 1", tok=self.user_tok + )["event_id"] + + # Check that we get the first and second message when querying /messages. + channel = self.make_request( + "GET", + "/_synapse/admin/v1/rooms/%s/messages?dir=f" % (self.room_id,), + access_token=self.admin_user_tok, + ) + self.assertEqual(channel.code, 200, channel.json_body) + + chunk = channel.json_body["chunk"] + self.assertEqual(len(chunk), 6, [event["content"] for event in chunk]) + + # in forward, this is the last event + self.assertEqual(chunk[5]["event_id"], latest_event_id) + def test_room_messages_purge(self) -> None: """Test room messages can be retrieved by an admin that isn't in the room.""" store = self.hs.get_datastores().main -- cgit 1.5.1 From 09de2aecb05cb46e0513396e2675b24c8beedb68 Mon Sep 17 00:00:00 2001 From: Ashish Kumar Date: Fri, 25 Nov 2022 19:16:50 +0400 Subject: Add support for handling avatar with SSO login (#13917) This commit adds support for handling a provided avatar picture URL when logging in via SSO. Signed-off-by: Ashish Kumar Fixes #9357. --- changelog.d/13917.feature | 1 + docs/usage/configuration/config_documentation.md | 9 +- mypy.ini | 4 +- synapse/handlers/oidc.py | 7 ++ synapse/handlers/sso.py | 111 +++++++++++++++++ tests/handlers/test_sso.py | 145 +++++++++++++++++++++++ 6 files changed, 275 insertions(+), 2 deletions(-) create mode 100644 changelog.d/13917.feature create mode 100644 tests/handlers/test_sso.py (limited to 'synapse/handlers') diff --git a/changelog.d/13917.feature b/changelog.d/13917.feature new file mode 100644 index 0000000000..4eb942ab38 --- /dev/null +++ b/changelog.d/13917.feature @@ -0,0 +1 @@ +Adds support for handling avatar in SSO login. Contributed by @ashfame. diff --git a/docs/usage/configuration/config_documentation.md b/docs/usage/configuration/config_documentation.md index fae2771fad..749af12aac 100644 --- a/docs/usage/configuration/config_documentation.md +++ b/docs/usage/configuration/config_documentation.md @@ -2968,10 +2968,17 @@ Options for each entry include: For the default provider, the following settings are available: - * subject_claim: name of the claim containing a unique identifier + * `subject_claim`: name of the claim containing a unique identifier for the user. Defaults to 'sub', which OpenID Connect compliant providers should provide. + * `picture_claim`: name of the claim containing an url for the user's profile picture. + Defaults to 'picture', which OpenID Connect compliant providers should provide + and has to refer to a direct image file such as PNG, JPEG, or GIF image file. + + Currently only supported in monolithic (single-process) server configurations + where the media repository runs within the Synapse process. + * `localpart_template`: Jinja2 template for the localpart of the MXID. If this is not set, the user will be prompted to choose their own username (see the documentation for the `sso_auth_account_details.html` diff --git a/mypy.ini b/mypy.ini index 25b3c93748..0b6e7df267 100644 --- a/mypy.ini +++ b/mypy.ini @@ -119,6 +119,9 @@ disallow_untyped_defs = True [mypy-tests.storage.test_profile] disallow_untyped_defs = True +[mypy-tests.handlers.test_sso] +disallow_untyped_defs = True + [mypy-tests.storage.test_user_directory] disallow_untyped_defs = True @@ -137,7 +140,6 @@ disallow_untyped_defs = False [mypy-tests.utils] disallow_untyped_defs = True - ;; Dependencies without annotations ;; Before ignoring a module, check to see if type stubs are available. ;; The `typeshed` project maintains stubs here: diff --git a/synapse/handlers/oidc.py b/synapse/handlers/oidc.py index 41c675f408..03de6a4ba6 100644 --- a/synapse/handlers/oidc.py +++ b/synapse/handlers/oidc.py @@ -1435,6 +1435,7 @@ class UserAttributeDict(TypedDict): localpart: Optional[str] confirm_localpart: bool display_name: Optional[str] + picture: Optional[str] # may be omitted by older `OidcMappingProviders` emails: List[str] @@ -1520,6 +1521,7 @@ env.filters.update( @attr.s(slots=True, frozen=True, auto_attribs=True) class JinjaOidcMappingConfig: subject_claim: str + picture_claim: str localpart_template: Optional[Template] display_name_template: Optional[Template] email_template: Optional[Template] @@ -1539,6 +1541,7 @@ class JinjaOidcMappingProvider(OidcMappingProvider[JinjaOidcMappingConfig]): @staticmethod def parse_config(config: dict) -> JinjaOidcMappingConfig: subject_claim = config.get("subject_claim", "sub") + picture_claim = config.get("picture_claim", "picture") def parse_template_config(option_name: str) -> Optional[Template]: if option_name not in config: @@ -1572,6 +1575,7 @@ class JinjaOidcMappingProvider(OidcMappingProvider[JinjaOidcMappingConfig]): return JinjaOidcMappingConfig( subject_claim=subject_claim, + picture_claim=picture_claim, localpart_template=localpart_template, display_name_template=display_name_template, email_template=email_template, @@ -1611,10 +1615,13 @@ class JinjaOidcMappingProvider(OidcMappingProvider[JinjaOidcMappingConfig]): if email: emails.append(email) + picture = userinfo.get("picture") + return UserAttributeDict( localpart=localpart, display_name=display_name, emails=emails, + picture=picture, confirm_localpart=self._config.confirm_localpart, ) diff --git a/synapse/handlers/sso.py b/synapse/handlers/sso.py index e1c0bff1b2..44e70fc4b8 100644 --- a/synapse/handlers/sso.py +++ b/synapse/handlers/sso.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. import abc +import hashlib +import io import logging from typing import ( TYPE_CHECKING, @@ -138,6 +140,7 @@ class UserAttributes: localpart: Optional[str] confirm_localpart: bool = False display_name: Optional[str] = None + picture: Optional[str] = None emails: Collection[str] = attr.Factory(list) @@ -196,6 +199,10 @@ class SsoHandler: self._error_template = hs.config.sso.sso_error_template self._bad_user_template = hs.config.sso.sso_auth_bad_user_template self._profile_handler = hs.get_profile_handler() + self._media_repo = ( + hs.get_media_repository() if hs.config.media.can_load_media_repo else None + ) + self._http_client = hs.get_proxied_blacklisted_http_client() # The following template is shown after a successful user interactive # authentication session. It tells the user they can close the window. @@ -495,6 +502,8 @@ class SsoHandler: await self._profile_handler.set_displayname( user_id_obj, requester, attributes.display_name, True ) + if attributes.picture: + await self.set_avatar(user_id, attributes.picture) await self._auth_handler.complete_sso_login( user_id, @@ -703,8 +712,110 @@ class SsoHandler: await self._store.record_user_external_id( auth_provider_id, remote_user_id, registered_user_id ) + + # Set avatar, if available + if attributes.picture: + await self.set_avatar(registered_user_id, attributes.picture) + return registered_user_id + async def set_avatar(self, user_id: str, picture_https_url: str) -> bool: + """Set avatar of the user. + + This downloads the image file from the URL provided, stores that in + the media repository and then sets the avatar on the user's profile. + + It can detect if the same image is being saved again and bails early by storing + the hash of the file in the `upload_name` of the avatar image. + + Currently, it only supports server configurations which run the media repository + within the same process. + + It silently fails and logs a warning by raising an exception and catching it + internally if: + * it is unable to fetch the image itself (non 200 status code) or + * the image supplied is bigger than max allowed size or + * the image type is not one of the allowed image types. + + Args: + user_id: matrix user ID in the form @localpart:domain as a string. + + picture_https_url: HTTPS url for the picture image file. + + Returns: `True` if the user's avatar has been successfully set to the image at + `picture_https_url`. + """ + if self._media_repo is None: + logger.info( + "failed to set user avatar because out-of-process media repositories " + "are not supported yet " + ) + return False + + try: + uid = UserID.from_string(user_id) + + def is_allowed_mime_type(content_type: str) -> bool: + if ( + self._profile_handler.allowed_avatar_mimetypes + and content_type + not in self._profile_handler.allowed_avatar_mimetypes + ): + return False + return True + + # download picture, enforcing size limit & mime type check + picture = io.BytesIO() + + content_length, headers, uri, code = await self._http_client.get_file( + url=picture_https_url, + output_stream=picture, + max_size=self._profile_handler.max_avatar_size, + is_allowed_content_type=is_allowed_mime_type, + ) + + if code != 200: + raise Exception( + "GET request to download sso avatar image returned {}".format(code) + ) + + # upload name includes hash of the image file's content so that we can + # easily check if it requires an update or not, the next time user logs in + upload_name = "sso_avatar_" + hashlib.sha256(picture.read()).hexdigest() + + # bail if user already has the same avatar + profile = await self._profile_handler.get_profile(user_id) + if profile["avatar_url"] is not None: + server_name = profile["avatar_url"].split("/")[-2] + media_id = profile["avatar_url"].split("/")[-1] + if server_name == self._server_name: + media = await self._media_repo.store.get_local_media(media_id) + if media is not None and upload_name == media["upload_name"]: + logger.info("skipping saving the user avatar") + return True + + # store it in media repository + avatar_mxc_url = await self._media_repo.create_content( + media_type=headers[b"Content-Type"][0].decode("utf-8"), + upload_name=upload_name, + content=picture, + content_length=content_length, + auth_user=uid, + ) + + # save it as user avatar + await self._profile_handler.set_avatar_url( + uid, + create_requester(uid), + str(avatar_mxc_url), + ) + + logger.info("successfully saved the user avatar") + return True + except Exception: + logger.warning("failed to save the user avatar") + return False + async def complete_sso_ui_auth_request( self, auth_provider_id: str, diff --git a/tests/handlers/test_sso.py b/tests/handlers/test_sso.py new file mode 100644 index 0000000000..137deab138 --- /dev/null +++ b/tests/handlers/test_sso.py @@ -0,0 +1,145 @@ +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +from http import HTTPStatus +from typing import BinaryIO, Callable, Dict, List, Optional, Tuple +from unittest.mock import Mock + +from twisted.test.proto_helpers import MemoryReactor +from twisted.web.http_headers import Headers + +from synapse.api.errors import Codes, SynapseError +from synapse.http.client import RawHeaders +from synapse.server import HomeServer +from synapse.util import Clock + +from tests import unittest +from tests.test_utils import SMALL_PNG, FakeResponse + + +class TestSSOHandler(unittest.HomeserverTestCase): + def make_homeserver(self, reactor: MemoryReactor, clock: Clock) -> HomeServer: + self.http_client = Mock(spec=["get_file"]) + self.http_client.get_file.side_effect = mock_get_file + self.http_client.user_agent = b"Synapse Test" + hs = self.setup_test_homeserver( + proxied_blacklisted_http_client=self.http_client + ) + return hs + + async def test_set_avatar(self) -> None: + """Tests successfully setting the avatar of a newly created user""" + handler = self.hs.get_sso_handler() + + # Create a new user to set avatar for + reg_handler = self.hs.get_registration_handler() + user_id = self.get_success(reg_handler.register_user(approved=True)) + + self.assertTrue( + self.get_success(handler.set_avatar(user_id, "http://my.server/me.png")) + ) + + # Ensure avatar is set on this newly created user, + # so no need to compare for the exact image + profile_handler = self.hs.get_profile_handler() + profile = self.get_success(profile_handler.get_profile(user_id)) + self.assertIsNot(profile["avatar_url"], None) + + @unittest.override_config({"max_avatar_size": 1}) + async def test_set_avatar_too_big_image(self) -> None: + """Tests that saving an avatar fails when it is too big""" + handler = self.hs.get_sso_handler() + + # any random user works since image check is supposed to fail + user_id = "@sso-user:test" + + self.assertFalse( + self.get_success(handler.set_avatar(user_id, "http://my.server/me.png")) + ) + + @unittest.override_config({"allowed_avatar_mimetypes": ["image/jpeg"]}) + async def test_set_avatar_incorrect_mime_type(self) -> None: + """Tests that saving an avatar fails when its mime type is not allowed""" + handler = self.hs.get_sso_handler() + + # any random user works since image check is supposed to fail + user_id = "@sso-user:test" + + self.assertFalse( + self.get_success(handler.set_avatar(user_id, "http://my.server/me.png")) + ) + + async def test_skip_saving_avatar_when_not_changed(self) -> None: + """Tests whether saving of avatar correctly skips if the avatar hasn't + changed""" + handler = self.hs.get_sso_handler() + + # Create a new user to set avatar for + reg_handler = self.hs.get_registration_handler() + user_id = self.get_success(reg_handler.register_user(approved=True)) + + # set avatar for the first time, should be a success + self.assertTrue( + self.get_success(handler.set_avatar(user_id, "http://my.server/me.png")) + ) + + # get avatar picture for comparison after another attempt + profile_handler = self.hs.get_profile_handler() + profile = self.get_success(profile_handler.get_profile(user_id)) + url_to_match = profile["avatar_url"] + + # set same avatar for the second time, should be a success + self.assertTrue( + self.get_success(handler.set_avatar(user_id, "http://my.server/me.png")) + ) + + # compare avatar picture's url from previous step + profile = self.get_success(profile_handler.get_profile(user_id)) + self.assertEqual(profile["avatar_url"], url_to_match) + + +async def mock_get_file( + url: str, + output_stream: BinaryIO, + max_size: Optional[int] = None, + headers: Optional[RawHeaders] = None, + is_allowed_content_type: Optional[Callable[[str], bool]] = None, +) -> Tuple[int, Dict[bytes, List[bytes]], str, int]: + + fake_response = FakeResponse(code=404) + if url == "http://my.server/me.png": + fake_response = FakeResponse( + code=200, + headers=Headers( + {"Content-Type": ["image/png"], "Content-Length": [str(len(SMALL_PNG))]} + ), + body=SMALL_PNG, + ) + + if max_size is not None and max_size < len(SMALL_PNG): + raise SynapseError( + HTTPStatus.BAD_GATEWAY, + "Requested file is too large > %r bytes" % (max_size,), + Codes.TOO_LARGE, + ) + + if is_allowed_content_type and not is_allowed_content_type("image/png"): + raise SynapseError( + HTTPStatus.BAD_GATEWAY, + ( + "Requested file's content type not allowed for this operation: %s" + % "image/png" + ), + ) + + output_stream.write(fake_response.body) + + return len(SMALL_PNG), {b"Content-Type": [b"image/png"]}, "", 200 -- cgit 1.5.1 From d748bbc8f8268d2e8457374d529adafb20b9f5f4 Mon Sep 17 00:00:00 2001 From: Patrick Cloke Date: Mon, 28 Nov 2022 09:40:17 -0500 Subject: Include thread information when sending receipts over federation. (#14466) Include the thread_id field when sending read receipts over federation. This might result in the same user having multiple read receipts per-room, meaning multiple EDUs must be sent to encapsulate those receipts. This restructures the PerDestinationQueue APIs to support multiple receipt EDUs, queue_read_receipt now becomes linear time in the number of queued threaded receipts in the room for the given user, it is expected this is a small number since receipt EDUs are sent as filler in transactions. --- changelog.d/14466.bugfix | 1 + synapse/federation/sender/per_destination_queue.py | 183 ++++++++++++++------- synapse/handlers/receipts.py | 1 - tests/federation/test_federation_sender.py | 77 +++++++++ 4 files changed, 198 insertions(+), 64 deletions(-) create mode 100644 changelog.d/14466.bugfix (limited to 'synapse/handlers') diff --git a/changelog.d/14466.bugfix b/changelog.d/14466.bugfix new file mode 100644 index 0000000000..82f6e6b68e --- /dev/null +++ b/changelog.d/14466.bugfix @@ -0,0 +1 @@ +Fix a bug introduced in Synapse 1.70.0 where a receipt's thread ID was not sent over federation. diff --git a/synapse/federation/sender/per_destination_queue.py b/synapse/federation/sender/per_destination_queue.py index 3ae5e8634c..5af2784f1e 100644 --- a/synapse/federation/sender/per_destination_queue.py +++ b/synapse/federation/sender/per_destination_queue.py @@ -35,7 +35,7 @@ from synapse.logging import issue9533_logger from synapse.logging.opentracing import SynapseTags, set_tag from synapse.metrics import sent_transactions_counter from synapse.metrics.background_process_metrics import run_as_background_process -from synapse.types import ReadReceipt +from synapse.types import JsonDict, ReadReceipt from synapse.util.retryutils import NotRetryingDestination, get_retry_limiter from synapse.visibility import filter_events_for_server @@ -136,8 +136,11 @@ class PerDestinationQueue: # destination self._pending_presence: Dict[str, UserPresenceState] = {} - # room_id -> receipt_type -> user_id -> receipt_dict - self._pending_rrs: Dict[str, Dict[str, Dict[str, dict]]] = {} + # List of room_id -> receipt_type -> user_id -> receipt_dict, + # + # Each receipt can only have a single receipt per + # (room ID, receipt type, user ID, thread ID) tuple. + self._pending_receipt_edus: List[Dict[str, Dict[str, Dict[str, dict]]]] = [] self._rrs_pending_flush = False # stream_id of last successfully sent to-device message. @@ -202,17 +205,53 @@ class PerDestinationQueue: Args: receipt: receipt to be queued """ - self._pending_rrs.setdefault(receipt.room_id, {}).setdefault( - receipt.receipt_type, {} - )[receipt.user_id] = {"event_ids": receipt.event_ids, "data": receipt.data} + serialized_receipt: JsonDict = { + "event_ids": receipt.event_ids, + "data": receipt.data, + } + if receipt.thread_id is not None: + serialized_receipt["data"]["thread_id"] = receipt.thread_id + + # Find which EDU to add this receipt to. There's three situations depending + # on the (room ID, receipt type, user, thread ID) tuple: + # + # 1. If it fully matches, clobber the information. + # 2. If it is missing, add the information. + # 3. If the subset tuple of (room ID, receipt type, user) matches, check + # the next EDU (or add a new EDU). + for edu in self._pending_receipt_edus: + receipt_content = edu.setdefault(receipt.room_id, {}).setdefault( + receipt.receipt_type, {} + ) + # If this room ID, receipt type, user ID is not in this EDU, OR if + # the full tuple matches, use the current EDU. + if ( + receipt.user_id not in receipt_content + or receipt_content[receipt.user_id].get("thread_id") + == receipt.thread_id + ): + receipt_content[receipt.user_id] = serialized_receipt + break + + # If no matching EDU was found, create a new one. + else: + self._pending_receipt_edus.append( + { + receipt.room_id: { + receipt.receipt_type: {receipt.user_id: serialized_receipt} + } + } + ) def flush_read_receipts_for_room(self, room_id: str) -> None: - # if we don't have any read-receipts for this room, it may be that we've already - # sent them out, so we don't need to flush. - if room_id not in self._pending_rrs: - return - self._rrs_pending_flush = True - self.attempt_new_transaction() + # If there are any pending receipts for this room then force-flush them + # in a new transaction. + for edu in self._pending_receipt_edus: + if room_id in edu: + self._rrs_pending_flush = True + self.attempt_new_transaction() + # No use in checking remaining EDUs if the room was found. + break def send_keyed_edu(self, edu: Edu, key: Hashable) -> None: self._pending_edus_keyed[(edu.edu_type, key)] = edu @@ -351,7 +390,7 @@ class PerDestinationQueue: self._pending_edus = [] self._pending_edus_keyed = {} self._pending_presence = {} - self._pending_rrs = {} + self._pending_receipt_edus = [] self._start_catching_up() except FederationDeniedError as e: @@ -543,22 +582,27 @@ class PerDestinationQueue: self._destination, last_successful_stream_ordering ) - def _get_rr_edus(self, force_flush: bool) -> Iterable[Edu]: - if not self._pending_rrs: + def _get_receipt_edus(self, force_flush: bool, limit: int) -> Iterable[Edu]: + if not self._pending_receipt_edus: return if not force_flush and not self._rrs_pending_flush: # not yet time for this lot return - edu = Edu( - origin=self._server_name, - destination=self._destination, - edu_type=EduTypes.RECEIPT, - content=self._pending_rrs, - ) - self._pending_rrs = {} - self._rrs_pending_flush = False - yield edu + # Send at most limit EDUs for receipts. + for content in self._pending_receipt_edus[:limit]: + yield Edu( + origin=self._server_name, + destination=self._destination, + edu_type=EduTypes.RECEIPT, + content=content, + ) + self._pending_receipt_edus = self._pending_receipt_edus[limit:] + + # If there are still pending read-receipts, don't reset the pending flush + # flag. + if not self._pending_receipt_edus: + self._rrs_pending_flush = False def _pop_pending_edus(self, limit: int) -> List[Edu]: pending_edus = self._pending_edus @@ -645,27 +689,61 @@ class _TransactionQueueManager: async def __aenter__(self) -> Tuple[List[EventBase], List[Edu]]: # First we calculate the EDUs we want to send, if any. - # We start by fetching device related EDUs, i.e device updates and to - # device messages. We have to keep 2 free slots for presence and rr_edus. - device_edu_limit = MAX_EDUS_PER_TRANSACTION - 2 + # There's a maximum number of EDUs that can be sent with a transaction, + # generally device updates and to-device messages get priority, but we + # want to ensure that there's room for some other EDUs as well. + # + # This is done by: + # + # * Add a presence EDU, if one exists. + # * Add up-to a small limit of read receipt EDUs. + # * Add to-device EDUs, but leave some space for device list updates. + # * Add device list updates EDUs. + # * If there's any remaining room, add other EDUs. + pending_edus = [] + + # Add presence EDU. + if self.queue._pending_presence: + pending_edus.append( + Edu( + origin=self.queue._server_name, + destination=self.queue._destination, + edu_type=EduTypes.PRESENCE, + content={ + "push": [ + format_user_presence_state( + presence, self.queue._clock.time_msec() + ) + for presence in self.queue._pending_presence.values() + ] + }, + ) + ) + self.queue._pending_presence = {} - # We prioritize to-device messages so that existing encryption channels + # Add read receipt EDUs. + pending_edus.extend(self.queue._get_receipt_edus(force_flush=False, limit=5)) + edu_limit = MAX_EDUS_PER_TRANSACTION - len(pending_edus) + + # Next, prioritize to-device messages so that existing encryption channels # work. We also keep a few slots spare (by reducing the limit) so that # we can still trickle out some device list updates. ( to_device_edus, device_stream_id, - ) = await self.queue._get_to_device_message_edus(device_edu_limit - 10) + ) = await self.queue._get_to_device_message_edus(edu_limit - 10) if to_device_edus: self._device_stream_id = device_stream_id else: self.queue._last_device_stream_id = device_stream_id - device_edu_limit -= len(to_device_edus) + pending_edus.extend(to_device_edus) + edu_limit -= len(to_device_edus) + # Add device list update EDUs. device_update_edus, dev_list_id = await self.queue._get_device_update_edus( - device_edu_limit + edu_limit ) if device_update_edus: @@ -673,40 +751,17 @@ class _TransactionQueueManager: else: self.queue._last_device_list_stream_id = dev_list_id - pending_edus = device_update_edus + to_device_edus - - # Now add the read receipt EDU. - pending_edus.extend(self.queue._get_rr_edus(force_flush=False)) - - # And presence EDU. - if self.queue._pending_presence: - pending_edus.append( - Edu( - origin=self.queue._server_name, - destination=self.queue._destination, - edu_type=EduTypes.PRESENCE, - content={ - "push": [ - format_user_presence_state( - presence, self.queue._clock.time_msec() - ) - for presence in self.queue._pending_presence.values() - ] - }, - ) - ) - self.queue._pending_presence = {} + pending_edus.extend(device_update_edus) + edu_limit -= len(device_update_edus) # Finally add any other types of EDUs if there is room. - pending_edus.extend( - self.queue._pop_pending_edus(MAX_EDUS_PER_TRANSACTION - len(pending_edus)) - ) - while ( - len(pending_edus) < MAX_EDUS_PER_TRANSACTION - and self.queue._pending_edus_keyed - ): + other_edus = self.queue._pop_pending_edus(edu_limit) + pending_edus.extend(other_edus) + edu_limit -= len(other_edus) + while edu_limit > 0 and self.queue._pending_edus_keyed: _, val = self.queue._pending_edus_keyed.popitem() pending_edus.append(val) + edu_limit -= 1 # Now we look for any PDUs to send, by getting up to 50 PDUs from the # queue @@ -717,8 +772,10 @@ class _TransactionQueueManager: # if we've decided to send a transaction anyway, and we have room, we # may as well send any pending RRs - if len(pending_edus) < MAX_EDUS_PER_TRANSACTION: - pending_edus.extend(self.queue._get_rr_edus(force_flush=True)) + if edu_limit: + pending_edus.extend( + self.queue._get_receipt_edus(force_flush=True, limit=edu_limit) + ) if self._pdus: self._last_stream_ordering = self._pdus[ diff --git a/synapse/handlers/receipts.py b/synapse/handlers/receipts.py index ac01582442..6a4fed1156 100644 --- a/synapse/handlers/receipts.py +++ b/synapse/handlers/receipts.py @@ -92,7 +92,6 @@ class ReceiptsHandler: continue # Check if these receipts apply to a thread. - thread_id = None data = user_values.get("data", {}) thread_id = data.get("thread_id") # If the thread ID is invalid, consider it missing. diff --git a/tests/federation/test_federation_sender.py b/tests/federation/test_federation_sender.py index f1e357764f..01f147418b 100644 --- a/tests/federation/test_federation_sender.py +++ b/tests/federation/test_federation_sender.py @@ -83,6 +83,83 @@ class FederationSenderReceiptsTestCases(HomeserverTestCase): ], ) + @override_config({"send_federation": True}) + def test_send_receipts_thread(self): + mock_send_transaction = ( + self.hs.get_federation_transport_client().send_transaction + ) + mock_send_transaction.return_value = make_awaitable({}) + + # Create receipts for: + # + # * The same room / user on multiple threads. + # * A different user in the same room. + sender = self.hs.get_federation_sender() + for user, thread in ( + ("alice", None), + ("alice", "thread"), + ("bob", None), + ("bob", "diff-thread"), + ): + receipt = ReadReceipt( + "room_id", + "m.read", + user, + ["event_id"], + thread_id=thread, + data={"ts": 1234}, + ) + self.successResultOf( + defer.ensureDeferred(sender.send_read_receipt(receipt)) + ) + + self.pump() + + # expect a call to send_transaction with two EDUs to separate threads. + mock_send_transaction.assert_called_once() + json_cb = mock_send_transaction.call_args[0][1] + data = json_cb() + # Note that the ordering of the EDUs doesn't matter. + self.assertCountEqual( + data["edus"], + [ + { + "edu_type": EduTypes.RECEIPT, + "content": { + "room_id": { + "m.read": { + "alice": { + "event_ids": ["event_id"], + "data": {"ts": 1234, "thread_id": "thread"}, + }, + "bob": { + "event_ids": ["event_id"], + "data": {"ts": 1234, "thread_id": "diff-thread"}, + }, + } + } + }, + }, + { + "edu_type": EduTypes.RECEIPT, + "content": { + "room_id": { + "m.read": { + "alice": { + "event_ids": ["event_id"], + "data": {"ts": 1234}, + }, + "bob": { + "event_ids": ["event_id"], + "data": {"ts": 1234}, + }, + } + } + }, + }, + ], + ) + @override_config({"send_federation": True}) def test_send_receipts_with_backoff(self): """Send two receipts in quick succession; the second should be flushed, but -- cgit 1.5.1 From 1183c372fa9da01b2667f1b83dab958dad432c68 Mon Sep 17 00:00:00 2001 From: Andrew Ferrazzutti Date: Mon, 28 Nov 2022 11:17:29 -0500 Subject: Use `device_one_time_keys_count` to match MSC3202 (#14565) * Use `device_one_time_keys_count` to match MSC3202 Rename the `device_one_time_key_counts` key in responses to `device_one_time_keys_count` to match the name specified by MSC3202. Also change related variable/class names for consistency. Signed-off-by: Andrew Ferrazzutti * Update changelog.d/14565.misc * Revert name change for `one_time_key_counts` key as this is a different key altogether from `device_one_time_keys_count`, which is used for `/sync` instead of appservice transactions. Signed-off-by: Andrew Ferrazzutti --- changelog.d/14565.misc | 1 + synapse/appservice/__init__.py | 10 +++++----- synapse/appservice/api.py | 11 +++++++---- synapse/appservice/scheduler.py | 16 ++++++++-------- synapse/handlers/sync.py | 6 +++--- synapse/storage/databases/main/appservice.py | 10 +++++----- synapse/storage/databases/main/end_to_end_keys.py | 8 ++++---- tests/appservice/test_scheduler.py | 6 +++--- tests/handlers/test_appservice.py | 4 ++-- 9 files changed, 38 insertions(+), 34 deletions(-) create mode 100644 changelog.d/14565.misc (limited to 'synapse/handlers') diff --git a/changelog.d/14565.misc b/changelog.d/14565.misc new file mode 100644 index 0000000000..19a62b036c --- /dev/null +++ b/changelog.d/14565.misc @@ -0,0 +1 @@ +In application service transactions that include the experimental `org.matrix.msc3202.device_one_time_key_counts` key, include a duplicate key of `org.matrix.msc3202.device_one_time_keys_count` to match the name proposed by [MSC3202](https://github.com/matrix-org/matrix-spec-proposals/blob/travis/msc/otk-dl-appservice/proposals/3202-encrypted-appservices.md). diff --git a/synapse/appservice/__init__.py b/synapse/appservice/__init__.py index 500bdde3a9..bf4e6c629b 100644 --- a/synapse/appservice/__init__.py +++ b/synapse/appservice/__init__.py @@ -32,9 +32,9 @@ if TYPE_CHECKING: logger = logging.getLogger(__name__) -# Type for the `device_one_time_key_counts` field in an appservice transaction +# Type for the `device_one_time_keys_count` field in an appservice transaction # user ID -> {device ID -> {algorithm -> count}} -TransactionOneTimeKeyCounts = Dict[str, Dict[str, Dict[str, int]]] +TransactionOneTimeKeysCount = Dict[str, Dict[str, Dict[str, int]]] # Type for the `device_unused_fallback_key_types` field in an appservice transaction # user ID -> {device ID -> [algorithm]} @@ -376,7 +376,7 @@ class AppServiceTransaction: events: List[EventBase], ephemeral: List[JsonDict], to_device_messages: List[JsonDict], - one_time_key_counts: TransactionOneTimeKeyCounts, + one_time_keys_count: TransactionOneTimeKeysCount, unused_fallback_keys: TransactionUnusedFallbackKeys, device_list_summary: DeviceListUpdates, ): @@ -385,7 +385,7 @@ class AppServiceTransaction: self.events = events self.ephemeral = ephemeral self.to_device_messages = to_device_messages - self.one_time_key_counts = one_time_key_counts + self.one_time_keys_count = one_time_keys_count self.unused_fallback_keys = unused_fallback_keys self.device_list_summary = device_list_summary @@ -402,7 +402,7 @@ class AppServiceTransaction: events=self.events, ephemeral=self.ephemeral, to_device_messages=self.to_device_messages, - one_time_key_counts=self.one_time_key_counts, + one_time_keys_count=self.one_time_keys_count, unused_fallback_keys=self.unused_fallback_keys, device_list_summary=self.device_list_summary, txn_id=self.id, diff --git a/synapse/appservice/api.py b/synapse/appservice/api.py index 60774b240d..edafd433cd 100644 --- a/synapse/appservice/api.py +++ b/synapse/appservice/api.py @@ -23,7 +23,7 @@ from synapse.api.constants import EventTypes, Membership, ThirdPartyEntityKind from synapse.api.errors import CodeMessageException from synapse.appservice import ( ApplicationService, - TransactionOneTimeKeyCounts, + TransactionOneTimeKeysCount, TransactionUnusedFallbackKeys, ) from synapse.events import EventBase @@ -262,7 +262,7 @@ class ApplicationServiceApi(SimpleHttpClient): events: List[EventBase], ephemeral: List[JsonDict], to_device_messages: List[JsonDict], - one_time_key_counts: TransactionOneTimeKeyCounts, + one_time_keys_count: TransactionOneTimeKeysCount, unused_fallback_keys: TransactionUnusedFallbackKeys, device_list_summary: DeviceListUpdates, txn_id: Optional[int] = None, @@ -310,10 +310,13 @@ class ApplicationServiceApi(SimpleHttpClient): # TODO: Update to stable prefixes once MSC3202 completes FCP merge if service.msc3202_transaction_extensions: - if one_time_key_counts: + if one_time_keys_count: body[ "org.matrix.msc3202.device_one_time_key_counts" - ] = one_time_key_counts + ] = one_time_keys_count + body[ + "org.matrix.msc3202.device_one_time_keys_count" + ] = one_time_keys_count if unused_fallback_keys: body[ "org.matrix.msc3202.device_unused_fallback_key_types" diff --git a/synapse/appservice/scheduler.py b/synapse/appservice/scheduler.py index 430ffbcd1f..7b562795a3 100644 --- a/synapse/appservice/scheduler.py +++ b/synapse/appservice/scheduler.py @@ -64,7 +64,7 @@ from typing import ( from synapse.appservice import ( ApplicationService, ApplicationServiceState, - TransactionOneTimeKeyCounts, + TransactionOneTimeKeysCount, TransactionUnusedFallbackKeys, ) from synapse.appservice.api import ApplicationServiceApi @@ -258,7 +258,7 @@ class _ServiceQueuer: ): return - one_time_key_counts: Optional[TransactionOneTimeKeyCounts] = None + one_time_keys_count: Optional[TransactionOneTimeKeysCount] = None unused_fallback_keys: Optional[TransactionUnusedFallbackKeys] = None if ( @@ -269,7 +269,7 @@ class _ServiceQueuer: # for the users which are mentioned in this transaction, # as well as the appservice's sender. ( - one_time_key_counts, + one_time_keys_count, unused_fallback_keys, ) = await self._compute_msc3202_otk_counts_and_fallback_keys( service, events, ephemeral, to_device_messages_to_send @@ -281,7 +281,7 @@ class _ServiceQueuer: events, ephemeral, to_device_messages_to_send, - one_time_key_counts, + one_time_keys_count, unused_fallback_keys, device_list_summary, ) @@ -296,7 +296,7 @@ class _ServiceQueuer: events: Iterable[EventBase], ephemerals: Iterable[JsonDict], to_device_messages: Iterable[JsonDict], - ) -> Tuple[TransactionOneTimeKeyCounts, TransactionUnusedFallbackKeys]: + ) -> Tuple[TransactionOneTimeKeysCount, TransactionUnusedFallbackKeys]: """ Given a list of the events, ephemeral messages and to-device messages, - first computes a list of application services users that may have @@ -367,7 +367,7 @@ class _TransactionController: events: List[EventBase], ephemeral: Optional[List[JsonDict]] = None, to_device_messages: Optional[List[JsonDict]] = None, - one_time_key_counts: Optional[TransactionOneTimeKeyCounts] = None, + one_time_keys_count: Optional[TransactionOneTimeKeysCount] = None, unused_fallback_keys: Optional[TransactionUnusedFallbackKeys] = None, device_list_summary: Optional[DeviceListUpdates] = None, ) -> None: @@ -380,7 +380,7 @@ class _TransactionController: events: The persistent events to include in the transaction. ephemeral: The ephemeral events to include in the transaction. to_device_messages: The to-device messages to include in the transaction. - one_time_key_counts: Counts of remaining one-time keys for relevant + one_time_keys_count: Counts of remaining one-time keys for relevant appservice devices in the transaction. unused_fallback_keys: Lists of unused fallback keys for relevant appservice devices in the transaction. @@ -397,7 +397,7 @@ class _TransactionController: events=events, ephemeral=ephemeral or [], to_device_messages=to_device_messages or [], - one_time_key_counts=one_time_key_counts or {}, + one_time_keys_count=one_time_keys_count or {}, unused_fallback_keys=unused_fallback_keys or {}, device_list_summary=device_list_summary or DeviceListUpdates(), ) diff --git a/synapse/handlers/sync.py b/synapse/handlers/sync.py index 259456b55d..c8858b22dd 100644 --- a/synapse/handlers/sync.py +++ b/synapse/handlers/sync.py @@ -1426,14 +1426,14 @@ class SyncHandler: logger.debug("Fetching OTK data") device_id = sync_config.device_id - one_time_key_counts: JsonDict = {} + one_time_keys_count: JsonDict = {} unused_fallback_key_types: List[str] = [] if device_id: # TODO: We should have a way to let clients differentiate between the states of: # * no change in OTK count since the provided since token # * the server has zero OTKs left for this device # Spec issue: https://github.com/matrix-org/matrix-doc/issues/3298 - one_time_key_counts = await self.store.count_e2e_one_time_keys( + one_time_keys_count = await self.store.count_e2e_one_time_keys( user_id, device_id ) unused_fallback_key_types = ( @@ -1463,7 +1463,7 @@ class SyncHandler: archived=sync_result_builder.archived, to_device=sync_result_builder.to_device, device_lists=device_lists, - device_one_time_keys_count=one_time_key_counts, + device_one_time_keys_count=one_time_keys_count, device_unused_fallback_key_types=unused_fallback_key_types, next_batch=sync_result_builder.now_token, ) diff --git a/synapse/storage/databases/main/appservice.py b/synapse/storage/databases/main/appservice.py index 25da0c56c5..c2c8018ee2 100644 --- a/synapse/storage/databases/main/appservice.py +++ b/synapse/storage/databases/main/appservice.py @@ -20,7 +20,7 @@ from synapse.appservice import ( ApplicationService, ApplicationServiceState, AppServiceTransaction, - TransactionOneTimeKeyCounts, + TransactionOneTimeKeysCount, TransactionUnusedFallbackKeys, ) from synapse.config.appservice import load_appservices @@ -260,7 +260,7 @@ class ApplicationServiceTransactionWorkerStore( events: List[EventBase], ephemeral: List[JsonDict], to_device_messages: List[JsonDict], - one_time_key_counts: TransactionOneTimeKeyCounts, + one_time_keys_count: TransactionOneTimeKeysCount, unused_fallback_keys: TransactionUnusedFallbackKeys, device_list_summary: DeviceListUpdates, ) -> AppServiceTransaction: @@ -273,7 +273,7 @@ class ApplicationServiceTransactionWorkerStore( events: A list of persistent events to put in the transaction. ephemeral: A list of ephemeral events to put in the transaction. to_device_messages: A list of to-device messages to put in the transaction. - one_time_key_counts: Counts of remaining one-time keys for relevant + one_time_keys_count: Counts of remaining one-time keys for relevant appservice devices in the transaction. unused_fallback_keys: Lists of unused fallback keys for relevant appservice devices in the transaction. @@ -299,7 +299,7 @@ class ApplicationServiceTransactionWorkerStore( events=events, ephemeral=ephemeral, to_device_messages=to_device_messages, - one_time_key_counts=one_time_key_counts, + one_time_keys_count=one_time_keys_count, unused_fallback_keys=unused_fallback_keys, device_list_summary=device_list_summary, ) @@ -379,7 +379,7 @@ class ApplicationServiceTransactionWorkerStore( events=events, ephemeral=[], to_device_messages=[], - one_time_key_counts={}, + one_time_keys_count={}, unused_fallback_keys={}, device_list_summary=DeviceListUpdates(), ) diff --git a/synapse/storage/databases/main/end_to_end_keys.py b/synapse/storage/databases/main/end_to_end_keys.py index cf33e73e2b..643c47d608 100644 --- a/synapse/storage/databases/main/end_to_end_keys.py +++ b/synapse/storage/databases/main/end_to_end_keys.py @@ -33,7 +33,7 @@ from typing_extensions import Literal from synapse.api.constants import DeviceKeyAlgorithms from synapse.appservice import ( - TransactionOneTimeKeyCounts, + TransactionOneTimeKeysCount, TransactionUnusedFallbackKeys, ) from synapse.logging.opentracing import log_kv, set_tag, trace @@ -514,7 +514,7 @@ class EndToEndKeyWorkerStore(EndToEndKeyBackgroundStore, CacheInvalidationWorker async def count_bulk_e2e_one_time_keys_for_as( self, user_ids: Collection[str] - ) -> TransactionOneTimeKeyCounts: + ) -> TransactionOneTimeKeysCount: """ Counts, in bulk, the one-time keys for all the users specified. Intended to be used by application services for populating OTK counts in @@ -528,7 +528,7 @@ class EndToEndKeyWorkerStore(EndToEndKeyBackgroundStore, CacheInvalidationWorker def _count_bulk_e2e_one_time_keys_txn( txn: LoggingTransaction, - ) -> TransactionOneTimeKeyCounts: + ) -> TransactionOneTimeKeysCount: user_in_where_clause, user_parameters = make_in_list_sql_clause( self.database_engine, "user_id", user_ids ) @@ -541,7 +541,7 @@ class EndToEndKeyWorkerStore(EndToEndKeyBackgroundStore, CacheInvalidationWorker """ txn.execute(sql, user_parameters) - result: TransactionOneTimeKeyCounts = {} + result: TransactionOneTimeKeysCount = {} for user_id, device_id, algorithm, count in txn: # We deliberately construct empty dictionaries for diff --git a/tests/appservice/test_scheduler.py b/tests/appservice/test_scheduler.py index 0b22afdc75..0a1ae83a2b 100644 --- a/tests/appservice/test_scheduler.py +++ b/tests/appservice/test_scheduler.py @@ -69,7 +69,7 @@ class ApplicationServiceSchedulerTransactionCtrlTestCase(unittest.TestCase): events=events, ephemeral=[], to_device_messages=[], # txn made and saved - one_time_key_counts={}, + one_time_keys_count={}, unused_fallback_keys={}, device_list_summary=DeviceListUpdates(), ) @@ -96,7 +96,7 @@ class ApplicationServiceSchedulerTransactionCtrlTestCase(unittest.TestCase): events=events, ephemeral=[], to_device_messages=[], # txn made and saved - one_time_key_counts={}, + one_time_keys_count={}, unused_fallback_keys={}, device_list_summary=DeviceListUpdates(), ) @@ -125,7 +125,7 @@ class ApplicationServiceSchedulerTransactionCtrlTestCase(unittest.TestCase): events=events, ephemeral=[], to_device_messages=[], - one_time_key_counts={}, + one_time_keys_count={}, unused_fallback_keys={}, device_list_summary=DeviceListUpdates(), ) diff --git a/tests/handlers/test_appservice.py b/tests/handlers/test_appservice.py index 144e49d0fd..9ed26d87a7 100644 --- a/tests/handlers/test_appservice.py +++ b/tests/handlers/test_appservice.py @@ -25,7 +25,7 @@ import synapse.storage from synapse.api.constants import EduTypes, EventTypes from synapse.appservice import ( ApplicationService, - TransactionOneTimeKeyCounts, + TransactionOneTimeKeysCount, TransactionUnusedFallbackKeys, ) from synapse.handlers.appservice import ApplicationServicesHandler @@ -1123,7 +1123,7 @@ class ApplicationServicesHandlerOtkCountsTestCase(unittest.HomeserverTestCase): # Capture what was sent as an AS transaction. self.send_mock.assert_called() last_args, _last_kwargs = self.send_mock.call_args - otks: Optional[TransactionOneTimeKeyCounts] = last_args[self.ARG_OTK_COUNTS] + otks: Optional[TransactionOneTimeKeysCount] = last_args[self.ARG_OTK_COUNTS] unused_fallbacks: Optional[TransactionUnusedFallbackKeys] = last_args[ self.ARG_FALLBACK_KEYS ] -- cgit 1.5.1 From 72f3e381375ba10d576a23025ca312397114de6b Mon Sep 17 00:00:00 2001 From: Shay Date: Mon, 28 Nov 2022 19:18:12 -0800 Subject: Fix possible variable shadow in `create_new_client_event` (#14575) --- changelog.d/14575.misc | 1 + synapse/handlers/message.py | 6 ++++-- 2 files changed, 5 insertions(+), 2 deletions(-) create mode 100644 changelog.d/14575.misc (limited to 'synapse/handlers') diff --git a/changelog.d/14575.misc b/changelog.d/14575.misc new file mode 100644 index 0000000000..f6fa54eaa2 --- /dev/null +++ b/changelog.d/14575.misc @@ -0,0 +1 @@ +Fix a possible variable shadow in `create_new_client_event`. \ No newline at end of file diff --git a/synapse/handlers/message.py b/synapse/handlers/message.py index 4cf593cfdc..5cbe89f4fd 100644 --- a/synapse/handlers/message.py +++ b/synapse/handlers/message.py @@ -1135,11 +1135,13 @@ class EventCreationHandler: ) state_events = await self.store.get_events_as_list(state_event_ids) # Create a StateMap[str] - state_map = {(e.type, e.state_key): e.event_id for e in state_events} + current_state_ids = { + (e.type, e.state_key): e.event_id for e in state_events + } # Actually strip down and only use the necessary auth events auth_event_ids = self._event_auth_handler.compute_auth_events( event=temp_event, - current_state_ids=state_map, + current_state_ids=current_state_ids, for_verification=False, ) -- cgit 1.5.1 From c7e29ca277cf60bfdc488b93f4321b046fa6b46f Mon Sep 17 00:00:00 2001 From: Erik Johnston Date: Tue, 29 Nov 2022 10:36:41 +0000 Subject: POC delete stale non-e2e devices for users (#14038) This should help reduce the number of devices e.g. simple bots the repeatedly login rack up. We only delete non-e2e devices as they should be safe to delete, whereas if we delete e2e devices for a user we may accidentally break their ability to receive e2e keys for a message. Co-authored-by: Patrick Cloke Co-authored-by: Sean Quah <8349537+squahtx@users.noreply.github.com> --- changelog.d/14038.misc | 1 + synapse/handlers/device.py | 13 +++++- synapse/storage/databases/main/devices.py | 67 ++++++++++++++++++++++++++++++- tests/handlers/test_device.py | 2 +- tests/storage/test_client_ips.py | 4 +- 5 files changed, 83 insertions(+), 4 deletions(-) create mode 100644 changelog.d/14038.misc (limited to 'synapse/handlers') diff --git a/changelog.d/14038.misc b/changelog.d/14038.misc new file mode 100644 index 0000000000..f9bfc581ad --- /dev/null +++ b/changelog.d/14038.misc @@ -0,0 +1 @@ +Prune user's old devices on login if they have too many. diff --git a/synapse/handlers/device.py b/synapse/handlers/device.py index b1e55e1b9e..7c4dd8cf5a 100644 --- a/synapse/handlers/device.py +++ b/synapse/handlers/device.py @@ -421,6 +421,9 @@ class DeviceHandler(DeviceWorkerHandler): self._check_device_name_length(initial_device_display_name) + # Prune the user's device list if they already have a lot of devices. + await self._prune_too_many_devices(user_id) + if device_id is not None: new_device = await self.store.store_device( user_id=user_id, @@ -452,6 +455,14 @@ class DeviceHandler(DeviceWorkerHandler): raise errors.StoreError(500, "Couldn't generate a device ID.") + async def _prune_too_many_devices(self, user_id: str) -> None: + """Delete any excess old devices this user may have.""" + device_ids = await self.store.check_too_many_devices_for_user(user_id) + if not device_ids: + return + + await self.delete_devices(user_id, device_ids) + async def _delete_stale_devices(self) -> None: """Background task that deletes devices which haven't been accessed for more than a configured time period. @@ -481,7 +492,7 @@ class DeviceHandler(DeviceWorkerHandler): device_ids = [d for d in device_ids if d != except_device_id] await self.delete_devices(user_id, device_ids) - async def delete_devices(self, user_id: str, device_ids: List[str]) -> None: + async def delete_devices(self, user_id: str, device_ids: Collection[str]) -> None: """Delete several devices Args: diff --git a/synapse/storage/databases/main/devices.py b/synapse/storage/databases/main/devices.py index 534f7fc04a..1e83c62753 100644 --- a/synapse/storage/databases/main/devices.py +++ b/synapse/storage/databases/main/devices.py @@ -1533,6 +1533,70 @@ class DeviceBackgroundUpdateStore(SQLBaseStore): return rows + async def check_too_many_devices_for_user(self, user_id: str) -> Collection[str]: + """Check if the user has a lot of devices, and if so return the set of + devices we can prune. + + This does *not* return hidden devices or devices with E2E keys. + """ + + num_devices = await self.db_pool.simple_select_one_onecol( + table="devices", + keyvalues={"user_id": user_id, "hidden": False}, + retcol="COALESCE(COUNT(*), 0)", + desc="count_devices", + ) + + # We let users have up to ten devices without pruning. + if num_devices <= 10: + return () + + # We prune everything older than N days. + max_last_seen = self._clock.time_msec() - 14 * 24 * 60 * 60 * 1000 + + if num_devices > 50: + # If the user has more than 50 devices, then we chose a last seen + # that ensures we keep at most 50 devices. + sql = """ + SELECT last_seen FROM devices + WHERE + user_id = ? + AND NOT hidden + AND last_seen IS NOT NULL + AND key_json IS NULL + ORDER BY last_seen DESC + LIMIT 1 + OFFSET 50 + """ + + rows = await self.db_pool.execute( + "check_too_many_devices_for_user_last_seen", None, sql, (user_id,) + ) + if rows: + max_last_seen = max(rows[0][0], max_last_seen) + + # Now fetch the devices to delete. + sql = """ + SELECT DISTINCT device_id FROM devices + LEFT JOIN e2e_device_keys_json USING (user_id, device_id) + WHERE + user_id = ? + AND NOT hidden + AND last_seen < ? + AND key_json IS NULL + """ + + def check_too_many_devices_for_user_txn( + txn: LoggingTransaction, + ) -> Collection[str]: + txn.execute(sql, (user_id, max_last_seen)) + return {device_id for device_id, in txn} + + return await self.db_pool.runInteraction( + "check_too_many_devices_for_user", + check_too_many_devices_for_user_txn, + ) + class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): # Because we have write access, this will be a StreamIdGenerator @@ -1591,6 +1655,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): values={}, insertion_values={ "display_name": initial_device_display_name, + "last_seen": self._clock.time_msec(), "hidden": False, }, desc="store_device", @@ -1636,7 +1701,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): ) raise StoreError(500, "Problem storing device.") - async def delete_devices(self, user_id: str, device_ids: List[str]) -> None: + async def delete_devices(self, user_id: str, device_ids: Collection[str]) -> None: """Deletes several devices. Args: diff --git a/tests/handlers/test_device.py b/tests/handlers/test_device.py index ce7525e29c..a456bffd63 100644 --- a/tests/handlers/test_device.py +++ b/tests/handlers/test_device.py @@ -115,7 +115,7 @@ class DeviceTestCase(unittest.HomeserverTestCase): "device_id": "xyz", "display_name": "display 0", "last_seen_ip": None, - "last_seen_ts": None, + "last_seen_ts": 1000000, }, device_map["xyz"], ) diff --git a/tests/storage/test_client_ips.py b/tests/storage/test_client_ips.py index 49ad3c1324..a9af1babed 100644 --- a/tests/storage/test_client_ips.py +++ b/tests/storage/test_client_ips.py @@ -169,6 +169,8 @@ class ClientIpStoreTestCase(unittest.HomeserverTestCase): ) ) + last_seen = self.clock.time_msec() + if after_persisting: # Trigger the storage loop self.reactor.advance(10) @@ -189,7 +191,7 @@ class ClientIpStoreTestCase(unittest.HomeserverTestCase): "device_id": device_id, "ip": None, "user_agent": None, - "last_seen": None, + "last_seen": last_seen, }, ], ) -- cgit 1.5.1 From c29e2c630624beb0b5557aa0f7ccdcedbe62def1 Mon Sep 17 00:00:00 2001 From: David Robertson Date: Tue, 29 Nov 2022 17:48:48 +0000 Subject: Revert "POC delete stale non-e2e devices for users (#14038)" (#14582) --- changelog.d/14582.bugfix | 1 + synapse/handlers/device.py | 13 +----- synapse/storage/databases/main/devices.py | 68 +------------------------------ tests/handlers/test_device.py | 2 +- tests/storage/test_client_ips.py | 4 +- 5 files changed, 5 insertions(+), 83 deletions(-) create mode 100644 changelog.d/14582.bugfix (limited to 'synapse/handlers') diff --git a/changelog.d/14582.bugfix b/changelog.d/14582.bugfix new file mode 100644 index 0000000000..caad468e70 --- /dev/null +++ b/changelog.d/14582.bugfix @@ -0,0 +1 @@ +Fix a regression in Synapse 1.73.0rc1 where Synapse's main process would stop responding to HTTP requests when a user with a large number of devices logs in. diff --git a/synapse/handlers/device.py b/synapse/handlers/device.py index 7c4dd8cf5a..b1e55e1b9e 100644 --- a/synapse/handlers/device.py +++ b/synapse/handlers/device.py @@ -421,9 +421,6 @@ class DeviceHandler(DeviceWorkerHandler): self._check_device_name_length(initial_device_display_name) - # Prune the user's device list if they already have a lot of devices. - await self._prune_too_many_devices(user_id) - if device_id is not None: new_device = await self.store.store_device( user_id=user_id, @@ -455,14 +452,6 @@ class DeviceHandler(DeviceWorkerHandler): raise errors.StoreError(500, "Couldn't generate a device ID.") - async def _prune_too_many_devices(self, user_id: str) -> None: - """Delete any excess old devices this user may have.""" - device_ids = await self.store.check_too_many_devices_for_user(user_id) - if not device_ids: - return - - await self.delete_devices(user_id, device_ids) - async def _delete_stale_devices(self) -> None: """Background task that deletes devices which haven't been accessed for more than a configured time period. @@ -492,7 +481,7 @@ class DeviceHandler(DeviceWorkerHandler): device_ids = [d for d in device_ids if d != except_device_id] await self.delete_devices(user_id, device_ids) - async def delete_devices(self, user_id: str, device_ids: Collection[str]) -> None: + async def delete_devices(self, user_id: str, device_ids: List[str]) -> None: """Delete several devices Args: diff --git a/synapse/storage/databases/main/devices.py b/synapse/storage/databases/main/devices.py index 0378035cff..534f7fc04a 100644 --- a/synapse/storage/databases/main/devices.py +++ b/synapse/storage/databases/main/devices.py @@ -1533,71 +1533,6 @@ class DeviceBackgroundUpdateStore(SQLBaseStore): return rows - async def check_too_many_devices_for_user(self, user_id: str) -> Collection[str]: - """Check if the user has a lot of devices, and if so return the set of - devices we can prune. - - This does *not* return hidden devices or devices with E2E keys. - """ - - num_devices = await self.db_pool.simple_select_one_onecol( - table="devices", - keyvalues={"user_id": user_id, "hidden": False}, - retcol="COALESCE(COUNT(*), 0)", - desc="count_devices", - ) - - # We let users have up to ten devices without pruning. - if num_devices <= 10: - return () - - # We prune everything older than N days. - max_last_seen = self._clock.time_msec() - 14 * 24 * 60 * 60 * 1000 - - if num_devices > 50: - # If the user has more than 50 devices, then we chose a last seen - # that ensures we keep at most 50 devices. - sql = """ - SELECT last_seen FROM devices - LEFT JOIN e2e_device_keys_json USING (user_id, device_id) - WHERE - user_id = ? - AND NOT hidden - AND last_seen IS NOT NULL - AND key_json IS NULL - ORDER BY last_seen DESC - LIMIT 1 - OFFSET 50 - """ - - rows = await self.db_pool.execute( - "check_too_many_devices_for_user_last_seen", None, sql, (user_id,) - ) - if rows: - max_last_seen = max(rows[0][0], max_last_seen) - - # Now fetch the devices to delete. - sql = """ - SELECT DISTINCT device_id FROM devices - LEFT JOIN e2e_device_keys_json USING (user_id, device_id) - WHERE - user_id = ? - AND NOT hidden - AND last_seen < ? - AND key_json IS NULL - """ - - def check_too_many_devices_for_user_txn( - txn: LoggingTransaction, - ) -> Collection[str]: - txn.execute(sql, (user_id, max_last_seen)) - return {device_id for device_id, in txn} - - return await self.db_pool.runInteraction( - "check_too_many_devices_for_user", - check_too_many_devices_for_user_txn, - ) - class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): # Because we have write access, this will be a StreamIdGenerator @@ -1656,7 +1591,6 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): values={}, insertion_values={ "display_name": initial_device_display_name, - "last_seen": self._clock.time_msec(), "hidden": False, }, desc="store_device", @@ -1702,7 +1636,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): ) raise StoreError(500, "Problem storing device.") - async def delete_devices(self, user_id: str, device_ids: Collection[str]) -> None: + async def delete_devices(self, user_id: str, device_ids: List[str]) -> None: """Deletes several devices. Args: diff --git a/tests/handlers/test_device.py b/tests/handlers/test_device.py index a456bffd63..ce7525e29c 100644 --- a/tests/handlers/test_device.py +++ b/tests/handlers/test_device.py @@ -115,7 +115,7 @@ class DeviceTestCase(unittest.HomeserverTestCase): "device_id": "xyz", "display_name": "display 0", "last_seen_ip": None, - "last_seen_ts": 1000000, + "last_seen_ts": None, }, device_map["xyz"], ) diff --git a/tests/storage/test_client_ips.py b/tests/storage/test_client_ips.py index a9af1babed..49ad3c1324 100644 --- a/tests/storage/test_client_ips.py +++ b/tests/storage/test_client_ips.py @@ -169,8 +169,6 @@ class ClientIpStoreTestCase(unittest.HomeserverTestCase): ) ) - last_seen = self.clock.time_msec() - if after_persisting: # Trigger the storage loop self.reactor.advance(10) @@ -191,7 +189,7 @@ class ClientIpStoreTestCase(unittest.HomeserverTestCase): "device_id": device_id, "ip": None, "user_agent": None, - "last_seen": last_seen, + "last_seen": None, }, ], ) -- cgit 1.5.1 From fac8a38525387e344e3595a092578e0ffedd49ae Mon Sep 17 00:00:00 2001 From: Patrick Cloke Date: Fri, 2 Dec 2022 10:28:41 -0500 Subject: Properly handle unknown results for the stream change cache. (#14592) StreamChangeCache.get_all_changed_entities can return None to signify it does not have information at the given stream position. Two callers (related to device lists and presence) were treating this response the same as an empty list (i.e. there being no updates). --- changelog.d/14592.bugfix | 1 + synapse/handlers/presence.py | 4 ++-- synapse/storage/databases/main/devices.py | 33 ++++++++++++++++++------------- 3 files changed, 22 insertions(+), 16 deletions(-) create mode 100644 changelog.d/14592.bugfix (limited to 'synapse/handlers') diff --git a/changelog.d/14592.bugfix b/changelog.d/14592.bugfix new file mode 100644 index 0000000000..149ee99dd7 --- /dev/null +++ b/changelog.d/14592.bugfix @@ -0,0 +1 @@ +Fix a long-standing bug where a device list update might not be sent to clients in certain circumstances. diff --git a/synapse/handlers/presence.py b/synapse/handlers/presence.py index cf08737d11..1799174c2f 100644 --- a/synapse/handlers/presence.py +++ b/synapse/handlers/presence.py @@ -1764,14 +1764,14 @@ class PresenceEventSource(EventSource[int, UserPresenceState]): Returns: A list of presence states for the given user to receive. """ + updated_users = None if from_key: # Only return updates since the last sync updated_users = self.store.presence_stream_cache.get_all_entities_changed( from_key ) - if not updated_users: - updated_users = [] + if updated_users is not None: # Get the actual presence update for each change users_to_state = await self.get_presence_handler().current_state_for_users( updated_users diff --git a/synapse/storage/databases/main/devices.py b/synapse/storage/databases/main/devices.py index 534f7fc04a..8ba995df3b 100644 --- a/synapse/storage/databases/main/devices.py +++ b/synapse/storage/databases/main/devices.py @@ -842,12 +842,11 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): user_ids, from_key ) - if not user_ids_to_check: + # If an empty set was returned, there's nothing to do. + if user_ids_to_check is not None and not user_ids_to_check: return set() def _get_users_whose_devices_changed_txn(txn: LoggingTransaction) -> Set[str]: - changes: Set[str] = set() - stream_id_where_clause = "stream_id > ?" sql_args = [from_key] @@ -858,19 +857,25 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): sql = f""" SELECT DISTINCT user_id FROM device_lists_stream WHERE {stream_id_where_clause} - AND """ - # Query device changes with a batch of users at a time - # Assertion for mypy's benefit; see also - # https://mypy.readthedocs.io/en/stable/common_issues.html#narrowing-and-inner-functions - assert user_ids_to_check is not None - for chunk in batch_iter(user_ids_to_check, 100): - clause, args = make_in_list_sql_clause( - txn.database_engine, "user_id", chunk - ) - txn.execute(sql + clause, sql_args + args) - changes.update(user_id for user_id, in txn) + # If the stream change cache gave us no information, fetch *all* + # users between the stream IDs. + if user_ids_to_check is None: + txn.execute(sql, sql_args) + return {user_id for user_id, in txn} + + # Otherwise, fetch changes for the given users. + else: + changes: Set[str] = set() + + # Query device changes with a batch of users at a time + for chunk in batch_iter(user_ids_to_check, 100): + clause, args = make_in_list_sql_clause( + txn.database_engine, "user_id", chunk + ) + txn.execute(sql + " AND " + clause, sql_args + args) + changes.update(user_id for user_id, in txn) return changes -- cgit 1.5.1 From 501f62d1a62296f79e46e1bd60dc5d1a8b28847d Mon Sep 17 00:00:00 2001 From: reivilibre Date: Mon, 5 Dec 2022 13:07:55 +0000 Subject: Faster remote room joins: stream the un-partial-stating of rooms over replication. [rei:frrj/streams/unpsr] (#14473) --- changelog.d/14473.misc | 1 + synapse/handlers/device.py | 2 +- synapse/handlers/federation.py | 4 + synapse/replication/tcp/streams/__init__.py | 3 + synapse/replication/tcp/streams/partial_state.py | 48 +++++ synapse/storage/databases/main/room.py | 237 +++++++++++++++------ .../delta/73/20_un_partial_stated_room_stream.sql | 32 +++ ..._un_partial_stated_room_stream_seq.sql.postgres | 20 ++ 8 files changed, 280 insertions(+), 67 deletions(-) create mode 100644 changelog.d/14473.misc create mode 100644 synapse/replication/tcp/streams/partial_state.py create mode 100644 synapse/storage/schema/main/delta/73/20_un_partial_stated_room_stream.sql create mode 100644 synapse/storage/schema/main/delta/73/21_un_partial_stated_room_stream_seq.sql.postgres (limited to 'synapse/handlers') diff --git a/changelog.d/14473.misc b/changelog.d/14473.misc new file mode 100644 index 0000000000..deccd4e91a --- /dev/null +++ b/changelog.d/14473.misc @@ -0,0 +1 @@ +Faster remote room joins: stream the un-partial-stating of rooms over replication. \ No newline at end of file diff --git a/synapse/handlers/device.py b/synapse/handlers/device.py index b1e55e1b9e..d4750a32e6 100644 --- a/synapse/handlers/device.py +++ b/synapse/handlers/device.py @@ -996,7 +996,7 @@ class DeviceListUpdater(DeviceListWorkerUpdater): # Check if we are partially joining any rooms. If so we need to store # all device list updates so that we can handle them correctly once we # know who is in the room. - # TODO(faster joins): this fetches and processes a bunch of data that we don't + # TODO(faster_joins): this fetches and processes a bunch of data that we don't # use. Could be replaced by a tighter query e.g. # SELECT EXISTS(SELECT 1 FROM partial_state_rooms) partial_rooms = await self.store.get_partial_state_room_resync_info() diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index d92582fd5c..3398fcaf7d 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -152,6 +152,7 @@ class FederationHandler: self._federation_event_handler = hs.get_federation_event_handler() self._device_handler = hs.get_device_handler() self._bulk_push_rule_evaluator = hs.get_bulk_push_rule_evaluator() + self._notifier = hs.get_notifier() self._clean_room_for_join_client = ReplicationCleanRoomRestServlet.make_client( hs @@ -1692,6 +1693,9 @@ class FederationHandler: self._storage_controllers.state.notify_room_un_partial_stated( room_id ) + # Poke the notifier so that other workers see the write to + # the un-partial-stated rooms stream. + self._notifier.notify_replication() # TODO(faster_joins) update room stats and user directory? # https://github.com/matrix-org/synapse/issues/12814 diff --git a/synapse/replication/tcp/streams/__init__.py b/synapse/replication/tcp/streams/__init__.py index b1cd55bf6f..8575666d9c 100644 --- a/synapse/replication/tcp/streams/__init__.py +++ b/synapse/replication/tcp/streams/__init__.py @@ -42,6 +42,7 @@ from synapse.replication.tcp.streams._base import ( ) from synapse.replication.tcp.streams.events import EventsStream from synapse.replication.tcp.streams.federation import FederationStream +from synapse.replication.tcp.streams.partial_state import UnPartialStatedRoomStream STREAMS_MAP = { stream.NAME: stream @@ -61,6 +62,7 @@ STREAMS_MAP = { TagAccountDataStream, AccountDataStream, UserSignatureStream, + UnPartialStatedRoomStream, ) } @@ -80,4 +82,5 @@ __all__ = [ "TagAccountDataStream", "AccountDataStream", "UserSignatureStream", + "UnPartialStatedRoomStream", ] diff --git a/synapse/replication/tcp/streams/partial_state.py b/synapse/replication/tcp/streams/partial_state.py new file mode 100644 index 0000000000..18f087ffa2 --- /dev/null +++ b/synapse/replication/tcp/streams/partial_state.py @@ -0,0 +1,48 @@ +# Copyright 2022 The Matrix.org Foundation C.I.C. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +from typing import TYPE_CHECKING + +import attr + +from synapse.replication.tcp.streams import Stream +from synapse.replication.tcp.streams._base import current_token_without_instance + +if TYPE_CHECKING: + from synapse.server import HomeServer + + +@attr.s(slots=True, frozen=True, auto_attribs=True) +class UnPartialStatedRoomStreamRow: + # ID of the room that has been un-partial-stated. + room_id: str + + +class UnPartialStatedRoomStream(Stream): + """ + Stream to notify about rooms becoming un-partial-stated; + that is, when the background sync finishes such that we now have full state for + the room. + """ + + NAME = "un_partial_stated_room" + ROW_TYPE = UnPartialStatedRoomStreamRow + + def __init__(self, hs: "HomeServer"): + store = hs.get_datastores().main + super().__init__( + hs.get_instance_name(), + # TODO(faster_joins, multiple writers): we need to account for instance names + current_token_without_instance(store.get_un_partial_stated_rooms_token), + store.get_un_partial_stated_rooms_from_stream, + ) diff --git a/synapse/storage/databases/main/room.py b/synapse/storage/databases/main/room.py index 1309bfd374..78906a5e1d 100644 --- a/synapse/storage/databases/main/room.py +++ b/synapse/storage/databases/main/room.py @@ -1,5 +1,5 @@ # Copyright 2014-2016 OpenMarket Ltd -# Copyright 2019 The Matrix.org Foundation C.I.C. +# Copyright 2019, 2022 The Matrix.org Foundation C.I.C. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. @@ -50,8 +50,14 @@ from synapse.storage.database import ( LoggingTransaction, ) from synapse.storage.databases.main.cache import CacheInvalidationWorkerStore +from synapse.storage.engines import PostgresEngine from synapse.storage.types import Cursor -from synapse.storage.util.id_generators import IdGenerator +from synapse.storage.util.id_generators import ( + AbstractStreamIdGenerator, + IdGenerator, + MultiWriterIdGenerator, + StreamIdGenerator, +) from synapse.types import JsonDict, RetentionPolicy, ThirdPartyInstanceID from synapse.util import json_encoder from synapse.util.caches.descriptors import cached @@ -114,6 +120,26 @@ class RoomWorkerStore(CacheInvalidationWorkerStore): self.config: HomeServerConfig = hs.config + self._un_partial_stated_rooms_stream_id_gen: AbstractStreamIdGenerator + + if isinstance(database.engine, PostgresEngine): + self._un_partial_stated_rooms_stream_id_gen = MultiWriterIdGenerator( + db_conn=db_conn, + db=database, + stream_name="un_partial_stated_room_stream", + instance_name=self._instance_name, + tables=[ + ("un_partial_stated_room_stream", "instance_name", "stream_id") + ], + sequence_name="un_partial_stated_room_stream_sequence", + # TODO(faster_joins, multiple writers) Support multiple writers. + writers=["master"], + ) + else: + self._un_partial_stated_rooms_stream_id_gen = StreamIdGenerator( + db_conn, "un_partial_stated_room_stream", "stream_id" + ) + async def store_room( self, room_id: str, @@ -1216,70 +1242,6 @@ class RoomWorkerStore(CacheInvalidationWorkerStore): return room_servers - async def clear_partial_state_room(self, room_id: str) -> bool: - """Clears the partial state flag for a room. - - Args: - room_id: The room whose partial state flag is to be cleared. - - Returns: - `True` if the partial state flag has been cleared successfully. - - `False` if the partial state flag could not be cleared because the room - still contains events with partial state. - """ - try: - await self.db_pool.runInteraction( - "clear_partial_state_room", self._clear_partial_state_room_txn, room_id - ) - return True - except self.db_pool.engine.module.IntegrityError as e: - # Assume that any `IntegrityError`s are due to partial state events. - logger.info( - "Exception while clearing lazy partial-state-room %s, retrying: %s", - room_id, - e, - ) - return False - - def _clear_partial_state_room_txn( - self, txn: LoggingTransaction, room_id: str - ) -> None: - DatabasePool.simple_delete_txn( - txn, - table="partial_state_rooms_servers", - keyvalues={"room_id": room_id}, - ) - DatabasePool.simple_delete_one_txn( - txn, - table="partial_state_rooms", - keyvalues={"room_id": room_id}, - ) - self._invalidate_cache_and_stream(txn, self.is_partial_state_room, (room_id,)) - self._invalidate_cache_and_stream( - txn, self.get_partial_state_servers_at_join, (room_id,) - ) - - # We now delete anything from `device_lists_remote_pending` with a - # stream ID less than the minimum - # `partial_state_rooms.device_lists_stream_id`, as we no longer need them. - device_lists_stream_id = DatabasePool.simple_select_one_onecol_txn( - txn, - table="partial_state_rooms", - keyvalues={}, - retcol="MIN(device_lists_stream_id)", - allow_none=True, - ) - if device_lists_stream_id is None: - # There are no rooms being currently partially joined, so we delete everything. - txn.execute("DELETE FROM device_lists_remote_pending") - else: - sql = """ - DELETE FROM device_lists_remote_pending - WHERE stream_id <= ? - """ - txn.execute(sql, (device_lists_stream_id,)) - @cached() async def is_partial_state_room(self, room_id: str) -> bool: """Checks if this room has partial state. @@ -1315,6 +1277,66 @@ class RoomWorkerStore(CacheInvalidationWorkerStore): ) return result["join_event_id"], result["device_lists_stream_id"] + def get_un_partial_stated_rooms_token(self) -> int: + # TODO(faster_joins, multiple writers): This is inappropriate if there + # are multiple writers because workers that don't write often will + # hold all readers up. + # (See `MultiWriterIdGenerator.get_persisted_upto_position` for an + # explanation.) + return self._un_partial_stated_rooms_stream_id_gen.get_current_token() + + async def get_un_partial_stated_rooms_from_stream( + self, instance_name: str, last_id: int, current_id: int, limit: int + ) -> Tuple[List[Tuple[int, Tuple[str]]], int, bool]: + """Get updates for caches replication stream. + + Args: + instance_name: The writer we want to fetch updates from. Unused + here since there is only ever one writer. + last_id: The token to fetch updates from. Exclusive. + current_id: The token to fetch updates up to. Inclusive. + limit: The requested limit for the number of rows to return. The + function may return more or fewer rows. + + Returns: + A tuple consisting of: the updates, a token to use to fetch + subsequent updates, and whether we returned fewer rows than exists + between the requested tokens due to the limit. + + The token returned can be used in a subsequent call to this + function to get further updatees. + + The updates are a list of 2-tuples of stream ID and the row data + """ + + if last_id == current_id: + return [], current_id, False + + def get_un_partial_stated_rooms_from_stream_txn( + txn: LoggingTransaction, + ) -> Tuple[List[Tuple[int, Tuple[str]]], int, bool]: + sql = """ + SELECT stream_id, room_id + FROM un_partial_stated_room_stream + WHERE ? < stream_id AND stream_id <= ? AND instance_name = ? + ORDER BY stream_id ASC + LIMIT ? + """ + txn.execute(sql, (last_id, current_id, instance_name, limit)) + updates = [(row[0], (row[1],)) for row in txn] + limited = False + upto_token = current_id + if len(updates) >= limit: + upto_token = updates[-1][0] + limited = True + + return updates, upto_token, limited + + return await self.db_pool.runInteraction( + "get_un_partial_stated_rooms_from_stream", + get_un_partial_stated_rooms_from_stream_txn, + ) + class _BackgroundUpdates: REMOVE_TOMESTONED_ROOMS_BG_UPDATE = "remove_tombstoned_rooms_from_directory" @@ -1806,6 +1828,8 @@ class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore): self._event_reports_id_gen = IdGenerator(db_conn, "event_reports", "id") + self._instance_name = hs.get_instance_name() + async def upsert_room_on_join( self, room_id: str, room_version: RoomVersion, state_events: List[EventBase] ) -> None: @@ -2270,3 +2294,84 @@ class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore): self.is_room_blocked, (room_id,), ) + + async def clear_partial_state_room(self, room_id: str) -> bool: + """Clears the partial state flag for a room. + + Args: + room_id: The room whose partial state flag is to be cleared. + + Returns: + `True` if the partial state flag has been cleared successfully. + + `False` if the partial state flag could not be cleared because the room + still contains events with partial state. + """ + try: + async with self._un_partial_stated_rooms_stream_id_gen.get_next() as un_partial_state_room_stream_id: + await self.db_pool.runInteraction( + "clear_partial_state_room", + self._clear_partial_state_room_txn, + room_id, + un_partial_state_room_stream_id, + ) + return True + except self.db_pool.engine.module.IntegrityError as e: + # Assume that any `IntegrityError`s are due to partial state events. + logger.info( + "Exception while clearing lazy partial-state-room %s, retrying: %s", + room_id, + e, + ) + return False + + def _clear_partial_state_room_txn( + self, + txn: LoggingTransaction, + room_id: str, + un_partial_state_room_stream_id: int, + ) -> None: + DatabasePool.simple_delete_txn( + txn, + table="partial_state_rooms_servers", + keyvalues={"room_id": room_id}, + ) + DatabasePool.simple_delete_one_txn( + txn, + table="partial_state_rooms", + keyvalues={"room_id": room_id}, + ) + self._invalidate_cache_and_stream(txn, self.is_partial_state_room, (room_id,)) + self._invalidate_cache_and_stream( + txn, self.get_partial_state_servers_at_join, (room_id,) + ) + + DatabasePool.simple_insert_txn( + txn, + "un_partial_stated_room_stream", + { + "stream_id": un_partial_state_room_stream_id, + "instance_name": self._instance_name, + "room_id": room_id, + }, + ) + + # We now delete anything from `device_lists_remote_pending` with a + # stream ID less than the minimum + # `partial_state_rooms.device_lists_stream_id`, as we no longer need them. + device_lists_stream_id = DatabasePool.simple_select_one_onecol_txn( + txn, + table="partial_state_rooms", + keyvalues={}, + retcol="MIN(device_lists_stream_id)", + allow_none=True, + ) + if device_lists_stream_id is None: + # There are no rooms being currently partially joined, so we delete everything. + txn.execute("DELETE FROM device_lists_remote_pending") + else: + sql = """ + DELETE FROM device_lists_remote_pending + WHERE stream_id <= ? + """ + txn.execute(sql, (device_lists_stream_id,)) diff --git a/synapse/storage/schema/main/delta/73/20_un_partial_stated_room_stream.sql b/synapse/storage/schema/main/delta/73/20_un_partial_stated_room_stream.sql new file mode 100644 index 0000000000..743196cfe3 --- /dev/null +++ b/synapse/storage/schema/main/delta/73/20_un_partial_stated_room_stream.sql @@ -0,0 +1,32 @@ +/* Copyright 2022 The Matrix.org Foundation C.I.C + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +-- Stream for notifying that a room has become un-partial-stated. +CREATE TABLE un_partial_stated_room_stream( + -- Position in the stream + stream_id BIGINT PRIMARY KEY NOT NULL, + + -- Which instance wrote this entry. + instance_name TEXT NOT NULL, + + -- Which room has been un-partial-stated. + room_id TEXT NOT NULL REFERENCES rooms(room_id) ON DELETE CASCADE +); + +-- We want an index here because of the foreign key constraint: +-- upon deleting a room, the database needs to be able to check here. +-- This index is not unique because we can join a room multiple times in a server's lifetime, +-- so the same room could be un-partial-stated multiple times! +CREATE INDEX un_partial_stated_room_stream_room_id ON un_partial_stated_room_stream (room_id); diff --git a/synapse/storage/schema/main/delta/73/21_un_partial_stated_room_stream_seq.sql.postgres b/synapse/storage/schema/main/delta/73/21_un_partial_stated_room_stream_seq.sql.postgres new file mode 100644 index 0000000000..c1aac0b385 --- /dev/null +++ b/synapse/storage/schema/main/delta/73/21_un_partial_stated_room_stream_seq.sql.postgres @@ -0,0 +1,20 @@ +/* Copyright 2022 The Matrix.org Foundation C.I.C + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +CREATE SEQUENCE IF NOT EXISTS un_partial_stated_room_stream_sequence; + +SELECT setval('un_partial_stated_room_stream_sequence', ( + SELECT COALESCE(MAX(stream_id), 1) FROM un_partial_stated_room_stream +)); -- cgit 1.5.1 From cee9445884eb62c070fb0b03a112a862e8dea7c4 Mon Sep 17 00:00:00 2001 From: Erik Johnston Date: Mon, 5 Dec 2022 20:19:14 +0000 Subject: Better return type for `get_all_entities_changed` (#14604) Help callers from using the return value incorrectly by ensuring that callers explicitly check if there was a cache hit or not. --- changelog.d/14604.bugfix | 1 + synapse/handlers/appservice.py | 4 +- synapse/handlers/presence.py | 12 ++-- synapse/handlers/sync.py | 6 +- synapse/handlers/typing.py | 8 +-- synapse/storage/databases/main/devices.py | 111 ++++++++++++++++++----------- synapse/util/caches/stream_change_cache.py | 52 ++++++++++---- tests/util/test_stream_change_cache.py | 20 +++--- 8 files changed, 138 insertions(+), 76 deletions(-) create mode 100644 changelog.d/14604.bugfix (limited to 'synapse/handlers') diff --git a/changelog.d/14604.bugfix b/changelog.d/14604.bugfix new file mode 100644 index 0000000000..149ee99dd7 --- /dev/null +++ b/changelog.d/14604.bugfix @@ -0,0 +1 @@ +Fix a long-standing bug where a device list update might not be sent to clients in certain circumstances. diff --git a/synapse/handlers/appservice.py b/synapse/handlers/appservice.py index 66f5b8d108..f68027aaed 100644 --- a/synapse/handlers/appservice.py +++ b/synapse/handlers/appservice.py @@ -615,8 +615,8 @@ class ApplicationServicesHandler: ) # Fetch the users who have modified their device list since then. - users_with_changed_device_lists = ( - await self.store.get_users_whose_devices_changed(from_key, to_key=new_key) + users_with_changed_device_lists = await self.store.get_all_devices_changed( + from_key, to_key=new_key ) # Filter out any users the application service is not interested in diff --git a/synapse/handlers/presence.py b/synapse/handlers/presence.py index 1799174c2f..2af90b25a3 100644 --- a/synapse/handlers/presence.py +++ b/synapse/handlers/presence.py @@ -1692,10 +1692,12 @@ class PresenceEventSource(EventSource[int, UserPresenceState]): if from_key is not None: # First get all users that have had a presence update - updated_users = stream_change_cache.get_all_entities_changed(from_key) + result = stream_change_cache.get_all_entities_changed(from_key) # Cross-reference users we're interested in with those that have had updates. - if updated_users is not None: + if result.hit: + updated_users = result.entities + # If we have the full list of changes for presence we can # simply check which ones share a room with the user. get_updates_counter.labels("stream").inc() @@ -1767,9 +1769,9 @@ class PresenceEventSource(EventSource[int, UserPresenceState]): updated_users = None if from_key: # Only return updates since the last sync - updated_users = self.store.presence_stream_cache.get_all_entities_changed( - from_key - ) + result = self.store.presence_stream_cache.get_all_entities_changed(from_key) + if result.hit: + updated_users = result.entities if updated_users is not None: # Get the actual presence update for each change diff --git a/synapse/handlers/sync.py b/synapse/handlers/sync.py index c8858b22dd..0b395a104d 100644 --- a/synapse/handlers/sync.py +++ b/synapse/handlers/sync.py @@ -1528,10 +1528,12 @@ class SyncHandler: # # If we don't have that info cached then we get all the users that # share a room with our user and check if those users have changed. - changed_users = self.store.get_cached_device_list_changes( + cache_result = self.store.get_cached_device_list_changes( since_token.device_list_key ) - if changed_users is not None: + if cache_result.hit: + changed_users = cache_result.entities + result = await self.store.get_rooms_for_users(changed_users) for changed_user_id, entries in result.items(): diff --git a/synapse/handlers/typing.py b/synapse/handlers/typing.py index a0ea719430..3f656ea4f5 100644 --- a/synapse/handlers/typing.py +++ b/synapse/handlers/typing.py @@ -420,11 +420,11 @@ class TypingWriterHandler(FollowerTypingHandler): if last_id == current_id: return [], current_id, False - changed_rooms: Optional[ - Iterable[str] - ] = self._typing_stream_change_cache.get_all_entities_changed(last_id) + result = self._typing_stream_change_cache.get_all_entities_changed(last_id) - if changed_rooms is None: + if result.hit: + changed_rooms: Iterable[str] = result.entities + else: changed_rooms = self._room_serials rows = [] diff --git a/synapse/storage/databases/main/devices.py b/synapse/storage/databases/main/devices.py index 8ba995df3b..a5bb4d404e 100644 --- a/synapse/storage/databases/main/devices.py +++ b/synapse/storage/databases/main/devices.py @@ -58,7 +58,10 @@ from synapse.types import JsonDict, get_verify_key_from_cross_signing_key from synapse.util import json_decoder, json_encoder from synapse.util.caches.descriptors import cached, cachedList from synapse.util.caches.lrucache import LruCache -from synapse.util.caches.stream_change_cache import StreamChangeCache +from synapse.util.caches.stream_change_cache import ( + AllEntitiesChangedResult, + StreamChangeCache, +) from synapse.util.cancellation import cancellable from synapse.util.iterutils import batch_iter from synapse.util.stringutils import shortstr @@ -799,18 +802,66 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): def get_cached_device_list_changes( self, from_key: int, - ) -> Optional[List[str]]: + ) -> AllEntitiesChangedResult: """Get set of users whose devices have changed since `from_key`, or None if that information is not in our cache. """ return self._device_list_stream_cache.get_all_entities_changed(from_key) + @cancellable + async def get_all_devices_changed( + self, + from_key: int, + to_key: int, + ) -> Set[str]: + """Get all users whose devices have changed in the given range. + + Args: + from_key: The minimum device lists stream token to query device list + changes for, exclusive. + to_key: The maximum device lists stream token to query device list + changes for, inclusive. + + Returns: + The set of user_ids whose devices have changed since `from_key` + (exclusive) until `to_key` (inclusive). + """ + + result = self._device_list_stream_cache.get_all_entities_changed(from_key) + + if result.hit: + # We know which users might have changed devices. + if not result.entities: + # If no users then we can return early. + return set() + + # Otherwise we need to filter down the list + return await self.get_users_whose_devices_changed( + from_key, result.entities, to_key + ) + + # If the cache didn't tell us anything, we just need to query the full + # range. + sql = """ + SELECT DISTINCT user_id FROM device_lists_stream + WHERE ? < stream_id AND stream_id <= ? + """ + + rows = await self.db_pool.execute( + "get_all_devices_changed", + None, + sql, + from_key, + to_key, + ) + return {u for u, in rows} + @cancellable async def get_users_whose_devices_changed( self, from_key: int, - user_ids: Optional[Collection[str]] = None, + user_ids: Collection[str], to_key: Optional[int] = None, ) -> Set[str]: """Get set of users whose devices have changed since `from_key` that @@ -830,52 +881,32 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): """ # Get set of users who *may* have changed. Users not in the returned # list have definitely not changed. - user_ids_to_check: Optional[Collection[str]] - if user_ids is None: - # Get set of all users that have had device list changes since 'from_key' - user_ids_to_check = self._device_list_stream_cache.get_all_entities_changed( - from_key - ) - else: - # The same as above, but filter results to only those users in 'user_ids' - user_ids_to_check = self._device_list_stream_cache.get_entities_changed( - user_ids, from_key - ) + user_ids_to_check = self._device_list_stream_cache.get_entities_changed( + user_ids, from_key + ) # If an empty set was returned, there's nothing to do. - if user_ids_to_check is not None and not user_ids_to_check: + if not user_ids_to_check: return set() - def _get_users_whose_devices_changed_txn(txn: LoggingTransaction) -> Set[str]: - stream_id_where_clause = "stream_id > ?" - sql_args = [from_key] - - if to_key: - stream_id_where_clause += " AND stream_id <= ?" - sql_args.append(to_key) + if to_key is None: + to_key = self._device_list_id_gen.get_current_token() - sql = f""" + def _get_users_whose_devices_changed_txn(txn: LoggingTransaction) -> Set[str]: + sql = """ SELECT DISTINCT user_id FROM device_lists_stream - WHERE {stream_id_where_clause} + WHERE ? < stream_id AND stream_id <= ? AND %s """ - # If the stream change cache gave us no information, fetch *all* - # users between the stream IDs. - if user_ids_to_check is None: - txn.execute(sql, sql_args) - return {user_id for user_id, in txn} + changes: Set[str] = set() - # Otherwise, fetch changes for the given users. - else: - changes: Set[str] = set() - - # Query device changes with a batch of users at a time - for chunk in batch_iter(user_ids_to_check, 100): - clause, args = make_in_list_sql_clause( - txn.database_engine, "user_id", chunk - ) - txn.execute(sql + " AND " + clause, sql_args + args) - changes.update(user_id for user_id, in txn) + # Query device changes with a batch of users at a time + for chunk in batch_iter(user_ids_to_check, 100): + clause, args = make_in_list_sql_clause( + txn.database_engine, "user_id", chunk + ) + txn.execute(sql % (clause,), [from_key, to_key] + args) + changes.update(user_id for user_id, in txn) return changes diff --git a/synapse/util/caches/stream_change_cache.py b/synapse/util/caches/stream_change_cache.py index 042de8d7c8..c8b17acb59 100644 --- a/synapse/util/caches/stream_change_cache.py +++ b/synapse/util/caches/stream_change_cache.py @@ -16,6 +16,7 @@ import logging import math from typing import Collection, Dict, FrozenSet, List, Mapping, Optional, Set, Union +import attr from sortedcontainers import SortedDict from synapse.util import caches @@ -26,6 +27,29 @@ logger = logging.getLogger(__name__) EntityType = str +@attr.s(auto_attribs=True, frozen=True, slots=True) +class AllEntitiesChangedResult: + """Return type of `get_all_entities_changed`. + + Callers must check that there was a cache hit, via `result.hit`, before + using the entities in `result.entities`. + + This specifically does *not* implement helpers such as `__bool__` to ensure + that callers do the correct checks. + """ + + _entities: Optional[List[EntityType]] + + @property + def hit(self) -> bool: + return self._entities is not None + + @property + def entities(self) -> List[EntityType]: + assert self._entities is not None + return self._entities + + class StreamChangeCache: """ Keeps track of the stream positions of the latest change in a set of entities. @@ -153,19 +177,19 @@ class StreamChangeCache: This will be all entities if the given stream position is at or earlier than the earliest known stream position. """ - changed_entities = self.get_all_entities_changed(stream_pos) - if changed_entities is not None: + cache_result = self.get_all_entities_changed(stream_pos) + if cache_result.hit: # We now do an intersection, trying to do so in the most efficient # way possible (some of these sets are *large*). First check in the # given iterable is already a set that we can reuse, otherwise we # create a set of the *smallest* of the two iterables and call # `intersection(..)` on it (this can be twice as fast as the reverse). if isinstance(entities, (set, frozenset)): - result = entities.intersection(changed_entities) - elif len(changed_entities) < len(entities): - result = set(changed_entities).intersection(entities) + result = entities.intersection(cache_result.entities) + elif len(cache_result.entities) < len(entities): + result = set(cache_result.entities).intersection(entities) else: - result = set(entities).intersection(changed_entities) + result = set(entities).intersection(cache_result.entities) self.metrics.inc_hits() else: result = set(entities) @@ -202,12 +226,12 @@ class StreamChangeCache: self.metrics.inc_hits() return stream_pos < self._cache.peekitem()[0] - def get_all_entities_changed(self, stream_pos: int) -> Optional[List[EntityType]]: + def get_all_entities_changed(self, stream_pos: int) -> AllEntitiesChangedResult: """ Returns all entities that have had changes after the given position. - If the stream change cache does not go far enough back, i.e. the position - is too old, it will return None. + If the stream change cache does not go far enough back, i.e. the + position is too old, it will return None. Returns the entities in the order that they were changed. @@ -215,23 +239,21 @@ class StreamChangeCache: stream_pos: The stream position to check for changes after. Return: - Entities which have changed after the given stream position. - - None if the given stream position is at or earlier than the earliest - known stream position. + A class indicating if we have the requested data cached, and if so + includes the entities in the order they were changed. """ assert isinstance(stream_pos, int) # _cache is not valid at or before the earliest known stream position, so # return None to mark that it is unknown if an entity has changed. if stream_pos <= self._earliest_known_stream_pos: - return None + return AllEntitiesChangedResult(None) changed_entities: List[EntityType] = [] for k in self._cache.islice(start=self._cache.bisect_right(stream_pos)): changed_entities.extend(self._cache[k]) - return changed_entities + return AllEntitiesChangedResult(changed_entities) def entity_has_changed(self, entity: EntityType, stream_pos: int) -> None: """ diff --git a/tests/util/test_stream_change_cache.py b/tests/util/test_stream_change_cache.py index a29cc872f9..0305741c99 100644 --- a/tests/util/test_stream_change_cache.py +++ b/tests/util/test_stream_change_cache.py @@ -73,8 +73,10 @@ class StreamChangeCacheTests(unittest.HomeserverTestCase): # The oldest item has been popped off self.assertTrue("user@foo.com" not in cache._entity_to_key) - self.assertEqual(cache.get_all_entities_changed(3), ["user@elsewhere.org"]) - self.assertIsNone(cache.get_all_entities_changed(2)) + self.assertEqual( + cache.get_all_entities_changed(3).entities, ["user@elsewhere.org"] + ) + self.assertFalse(cache.get_all_entities_changed(2).hit) # If we update an existing entity, it keeps the two existing entities cache.entity_has_changed("bar@baz.net", 5) @@ -82,10 +84,10 @@ class StreamChangeCacheTests(unittest.HomeserverTestCase): {"bar@baz.net", "user@elsewhere.org"}, set(cache._entity_to_key) ) self.assertEqual( - cache.get_all_entities_changed(3), + cache.get_all_entities_changed(3).entities, ["user@elsewhere.org", "bar@baz.net"], ) - self.assertIsNone(cache.get_all_entities_changed(2)) + self.assertFalse(cache.get_all_entities_changed(2).hit) def test_get_all_entities_changed(self) -> None: """ @@ -105,10 +107,12 @@ class StreamChangeCacheTests(unittest.HomeserverTestCase): # Results are ordered so either of these are valid. ok1 = ["bar@baz.net", "anotheruser@foo.com", "user@elsewhere.org"] ok2 = ["anotheruser@foo.com", "bar@baz.net", "user@elsewhere.org"] - self.assertTrue(r == ok1 or r == ok2) + self.assertTrue(r.entities == ok1 or r.entities == ok2) - self.assertEqual(cache.get_all_entities_changed(3), ["user@elsewhere.org"]) - self.assertEqual(cache.get_all_entities_changed(1), None) + self.assertEqual( + cache.get_all_entities_changed(3).entities, ["user@elsewhere.org"] + ) + self.assertFalse(cache.get_all_entities_changed(1).hit) # ... later, things gest more updates cache.entity_has_changed("user@foo.com", 5) @@ -128,7 +132,7 @@ class StreamChangeCacheTests(unittest.HomeserverTestCase): "anotheruser@foo.com", ] r = cache.get_all_entities_changed(3) - self.assertTrue(r == ok1 or r == ok2) + self.assertTrue(r.entities == ok1 or r.entities == ok2) def test_has_any_entity_changed(self) -> None: """ -- cgit 1.5.1 From cb59e080627745d089d073d9dac276362d9abaf6 Mon Sep 17 00:00:00 2001 From: Richard van der Hoff <1389908+richvdh@users.noreply.github.com> Date: Tue, 6 Dec 2022 09:52:55 +0000 Subject: Improve logging and opentracing for to-device message handling (#14598) A batch of changes intended to make it easier to trace to-device messages through the system. The intention here is that a client can set a property org.matrix.msgid in any to-device message it sends. That ID is then included in any tracing or logging related to the message. (Suggestions as to where this field should be documented welcome. I'm not enthusiastic about speccing it - it's very much an optional extra to help with debugging.) I've also generally improved the data we send to opentracing for these messages. --- changelog.d/14598.feature | 1 + synapse/api/constants.py | 3 + synapse/federation/sender/per_destination_queue.py | 2 +- synapse/handlers/appservice.py | 3 - synapse/handlers/devicemessage.py | 36 +++++---- synapse/handlers/sync.py | 26 ++++-- synapse/logging/opentracing.py | 11 ++- synapse/rest/client/sendtodevice.py | 1 - synapse/storage/databases/main/deviceinbox.py | 92 ++++++++++++++++++---- tests/handlers/test_appservice.py | 7 +- 10 files changed, 136 insertions(+), 46 deletions(-) create mode 100644 changelog.d/14598.feature (limited to 'synapse/handlers') diff --git a/changelog.d/14598.feature b/changelog.d/14598.feature new file mode 100644 index 0000000000..88d561e286 --- /dev/null +++ b/changelog.d/14598.feature @@ -0,0 +1 @@ +Improve opentracing and logging for to-device message handling. \ No newline at end of file diff --git a/synapse/api/constants.py b/synapse/api/constants.py index bc04a0755b..89723d24fa 100644 --- a/synapse/api/constants.py +++ b/synapse/api/constants.py @@ -230,6 +230,9 @@ class EventContentFields: # The authorising user for joining a restricted room. AUTHORISING_USER: Final = "join_authorised_via_users_server" + # an unspecced field added to to-device messages to identify them uniquely-ish + TO_DEVICE_MSGID: Final = "org.matrix.msgid" + class RoomTypes: """Understood values of the room_type field of m.room.create events.""" diff --git a/synapse/federation/sender/per_destination_queue.py b/synapse/federation/sender/per_destination_queue.py index 5af2784f1e..ffc9d95ee7 100644 --- a/synapse/federation/sender/per_destination_queue.py +++ b/synapse/federation/sender/per_destination_queue.py @@ -641,7 +641,7 @@ class PerDestinationQueue: if not message_id: continue - set_tag(SynapseTags.TO_DEVICE_MESSAGE_ID, message_id) + set_tag(SynapseTags.TO_DEVICE_EDU_ID, message_id) edus = [ Edu( diff --git a/synapse/handlers/appservice.py b/synapse/handlers/appservice.py index f68027aaed..5d1d21cdc8 100644 --- a/synapse/handlers/appservice.py +++ b/synapse/handlers/appservice.py @@ -578,9 +578,6 @@ class ApplicationServicesHandler: device_id, ), messages in recipient_device_to_messages.items(): for message_json in messages: - # Remove 'message_id' from the to-device message, as it's an internal ID - message_json.pop("message_id", None) - message_payload.append( { "to_user_id": user_id, diff --git a/synapse/handlers/devicemessage.py b/synapse/handlers/devicemessage.py index 444c08bc2e..75e89850f5 100644 --- a/synapse/handlers/devicemessage.py +++ b/synapse/handlers/devicemessage.py @@ -15,7 +15,7 @@ import logging from typing import TYPE_CHECKING, Any, Dict -from synapse.api.constants import EduTypes, ToDeviceEventTypes +from synapse.api.constants import EduTypes, EventContentFields, ToDeviceEventTypes from synapse.api.errors import SynapseError from synapse.api.ratelimiting import Ratelimiter from synapse.logging.context import run_in_background @@ -216,14 +216,24 @@ class DeviceMessageHandler: """ sender_user_id = requester.user.to_string() - message_id = random_string(16) - set_tag(SynapseTags.TO_DEVICE_MESSAGE_ID, message_id) - - log_kv({"number_of_to_device_messages": len(messages)}) - set_tag("sender", sender_user_id) + set_tag(SynapseTags.TO_DEVICE_TYPE, message_type) + set_tag(SynapseTags.TO_DEVICE_SENDER, sender_user_id) local_messages = {} remote_messages: Dict[str, Dict[str, Dict[str, JsonDict]]] = {} for user_id, by_device in messages.items(): + # add an opentracing log entry for each message + for device_id, message_content in by_device.items(): + log_kv( + { + "event": "send_to_device_message", + "user_id": user_id, + "device_id": device_id, + EventContentFields.TO_DEVICE_MSGID: message_content.get( + EventContentFields.TO_DEVICE_MSGID + ), + } + ) + # Ratelimit local cross-user key requests by the sending device. if ( message_type == ToDeviceEventTypes.RoomKeyRequest @@ -233,6 +243,7 @@ class DeviceMessageHandler: requester, (sender_user_id, requester.device_id) ) if not allowed: + log_kv({"message": f"dropping key requests to {user_id}"}) logger.info( "Dropping room_key_request from %s to %s due to rate limit", sender_user_id, @@ -247,18 +258,11 @@ class DeviceMessageHandler: "content": message_content, "type": message_type, "sender": sender_user_id, - "message_id": message_id, } for device_id, message_content in by_device.items() } if messages_by_device: local_messages[user_id] = messages_by_device - log_kv( - { - "user_id": user_id, - "device_id": list(messages_by_device), - } - ) else: destination = get_domain_from_id(user_id) remote_messages.setdefault(destination, {})[user_id] = by_device @@ -267,7 +271,11 @@ class DeviceMessageHandler: remote_edu_contents = {} for destination, messages in remote_messages.items(): - log_kv({"destination": destination}) + # The EDU contains a "message_id" property which is used for + # idempotence. Make up a random one. + message_id = random_string(16) + log_kv({"destination": destination, "message_id": message_id}) + remote_edu_contents[destination] = { "messages": messages, "sender": sender_user_id, diff --git a/synapse/handlers/sync.py b/synapse/handlers/sync.py index 0b395a104d..dace9b606f 100644 --- a/synapse/handlers/sync.py +++ b/synapse/handlers/sync.py @@ -31,14 +31,20 @@ from typing import ( import attr from prometheus_client import Counter -from synapse.api.constants import EventTypes, Membership +from synapse.api.constants import EventContentFields, EventTypes, Membership from synapse.api.filtering import FilterCollection from synapse.api.presence import UserPresenceState from synapse.api.room_versions import KNOWN_ROOM_VERSIONS from synapse.events import EventBase from synapse.handlers.relations import BundledAggregations from synapse.logging.context import current_context -from synapse.logging.opentracing import SynapseTags, log_kv, set_tag, start_active_span +from synapse.logging.opentracing import ( + SynapseTags, + log_kv, + set_tag, + start_active_span, + trace, +) from synapse.push.clientformat import format_push_rules_for_user from synapse.storage.databases.main.event_push_actions import RoomNotifCounts from synapse.storage.databases.main.roommember import extract_heroes_from_room_summary @@ -1586,6 +1592,7 @@ class SyncHandler: else: return DeviceListUpdates() + @trace async def _generate_sync_entry_for_to_device( self, sync_result_builder: "SyncResultBuilder" ) -> None: @@ -1605,11 +1612,16 @@ class SyncHandler: ) for message in messages: - # We pop here as we shouldn't be sending the message ID down - # `/sync` - message_id = message.pop("message_id", None) - if message_id: - set_tag(SynapseTags.TO_DEVICE_MESSAGE_ID, message_id) + log_kv( + { + "event": "to_device_message", + "sender": message["sender"], + "type": message["type"], + EventContentFields.TO_DEVICE_MSGID: message["content"].get( + EventContentFields.TO_DEVICE_MSGID + ), + } + ) logger.debug( "Returning %d to-device messages between %d and %d (current token: %d)", diff --git a/synapse/logging/opentracing.py b/synapse/logging/opentracing.py index b69060854f..a705af8356 100644 --- a/synapse/logging/opentracing.py +++ b/synapse/logging/opentracing.py @@ -292,8 +292,15 @@ logger = logging.getLogger(__name__) class SynapseTags: - # The message ID of any to_device message processed - TO_DEVICE_MESSAGE_ID = "to_device.message_id" + # The message ID of any to_device EDU processed + TO_DEVICE_EDU_ID = "to_device.edu_id" + + # Details about to-device messages + TO_DEVICE_TYPE = "to_device.type" + TO_DEVICE_SENDER = "to_device.sender" + TO_DEVICE_RECIPIENT = "to_device.recipient" + TO_DEVICE_RECIPIENT_DEVICE = "to_device.recipient_device" + TO_DEVICE_MSGID = "to_device.msgid" # client-generated ID # Whether the sync response has new data to be returned to the client. SYNC_RESULT = "sync.new_data" diff --git a/synapse/rest/client/sendtodevice.py b/synapse/rest/client/sendtodevice.py index 46a8b03829..55d52f0b28 100644 --- a/synapse/rest/client/sendtodevice.py +++ b/synapse/rest/client/sendtodevice.py @@ -46,7 +46,6 @@ class SendToDeviceRestServlet(servlet.RestServlet): def on_PUT( self, request: SynapseRequest, message_type: str, txn_id: str ) -> Awaitable[Tuple[int, JsonDict]]: - set_tag("message_type", message_type) set_tag("txn_id", txn_id) return self.txns.fetch_or_execute_request( request, self._put, request, message_type, txn_id diff --git a/synapse/storage/databases/main/deviceinbox.py b/synapse/storage/databases/main/deviceinbox.py index 73c95ffb6f..48a54d9cb8 100644 --- a/synapse/storage/databases/main/deviceinbox.py +++ b/synapse/storage/databases/main/deviceinbox.py @@ -26,8 +26,15 @@ from typing import ( cast, ) +from synapse.api.constants import EventContentFields from synapse.logging import issue9533_logger -from synapse.logging.opentracing import log_kv, set_tag, trace +from synapse.logging.opentracing import ( + SynapseTags, + log_kv, + set_tag, + start_active_span, + trace, +) from synapse.replication.tcp.streams import ToDeviceStream from synapse.storage._base import SQLBaseStore, db_to_json from synapse.storage.database import ( @@ -397,6 +404,17 @@ class DeviceInboxWorkerStore(SQLBaseStore): (recipient_user_id, recipient_device_id), [] ).append(message_dict) + # start a new span for each message, so that we can tag each separately + with start_active_span("get_to_device_message"): + set_tag(SynapseTags.TO_DEVICE_TYPE, message_dict["type"]) + set_tag(SynapseTags.TO_DEVICE_SENDER, message_dict["sender"]) + set_tag(SynapseTags.TO_DEVICE_RECIPIENT, recipient_user_id) + set_tag(SynapseTags.TO_DEVICE_RECIPIENT_DEVICE, recipient_device_id) + set_tag( + SynapseTags.TO_DEVICE_MSGID, + message_dict["content"].get(EventContentFields.TO_DEVICE_MSGID), + ) + if limit is not None and rowcount == limit: # We ended up bumping up against the message limit. There may be more messages # to retrieve. Return what we have, as well as the last stream position that @@ -678,12 +696,35 @@ class DeviceInboxWorkerStore(SQLBaseStore): ], ) - if remote_messages_by_destination: - issue9533_logger.debug( - "Queued outgoing to-device messages with stream_id %i for %s", - stream_id, - list(remote_messages_by_destination.keys()), - ) + for destination, edu in remote_messages_by_destination.items(): + if issue9533_logger.isEnabledFor(logging.DEBUG): + issue9533_logger.debug( + "Queued outgoing to-device messages with " + "stream_id %i, EDU message_id %s, type %s for %s: %s", + stream_id, + edu["message_id"], + edu["type"], + destination, + [ + f"{user_id}/{device_id} (msgid " + f"{msg.get(EventContentFields.TO_DEVICE_MSGID)})" + for (user_id, messages_by_device) in edu["messages"].items() + for (device_id, msg) in messages_by_device.items() + ], + ) + + for (user_id, messages_by_device) in edu["messages"].items(): + for (device_id, msg) in messages_by_device.items(): + with start_active_span("store_outgoing_to_device_message"): + set_tag(SynapseTags.TO_DEVICE_EDU_ID, edu["sender"]) + set_tag(SynapseTags.TO_DEVICE_EDU_ID, edu["message_id"]) + set_tag(SynapseTags.TO_DEVICE_TYPE, edu["type"]) + set_tag(SynapseTags.TO_DEVICE_RECIPIENT, user_id) + set_tag(SynapseTags.TO_DEVICE_RECIPIENT_DEVICE, device_id) + set_tag( + SynapseTags.TO_DEVICE_MSGID, + msg.get(EventContentFields.TO_DEVICE_MSGID), + ) async with self._device_inbox_id_gen.get_next() as stream_id: now_ms = self._clock.time_msec() @@ -801,7 +842,19 @@ class DeviceInboxWorkerStore(SQLBaseStore): # Only insert into the local inbox if the device exists on # this server device_id = row["device_id"] - message_json = json_encoder.encode(messages_by_device[device_id]) + + with start_active_span("serialise_to_device_message"): + msg = messages_by_device[device_id] + set_tag(SynapseTags.TO_DEVICE_TYPE, msg["type"]) + set_tag(SynapseTags.TO_DEVICE_SENDER, msg["sender"]) + set_tag(SynapseTags.TO_DEVICE_RECIPIENT, user_id) + set_tag(SynapseTags.TO_DEVICE_RECIPIENT_DEVICE, device_id) + set_tag( + SynapseTags.TO_DEVICE_MSGID, + msg["content"].get(EventContentFields.TO_DEVICE_MSGID), + ) + message_json = json_encoder.encode(msg) + messages_json_for_user[device_id] = message_json if messages_json_for_user: @@ -821,15 +874,20 @@ class DeviceInboxWorkerStore(SQLBaseStore): ], ) - issue9533_logger.debug( - "Stored to-device messages with stream_id %i for %s", - stream_id, - [ - (user_id, device_id) - for (user_id, messages_by_device) in local_by_user_then_device.items() - for device_id in messages_by_device.keys() - ], - ) + if issue9533_logger.isEnabledFor(logging.DEBUG): + issue9533_logger.debug( + "Stored to-device messages with stream_id %i: %s", + stream_id, + [ + f"{user_id}/{device_id} (msgid " + f"{msg['content'].get(EventContentFields.TO_DEVICE_MSGID)})" + for ( + user_id, + messages_by_device, + ) in messages_by_user_then_device.items() + for (device_id, msg) in messages_by_device.items() + ], + ) class DeviceInboxBackgroundUpdateStore(SQLBaseStore): diff --git a/tests/handlers/test_appservice.py b/tests/handlers/test_appservice.py index 9ed26d87a7..57bfbd7734 100644 --- a/tests/handlers/test_appservice.py +++ b/tests/handlers/test_appservice.py @@ -765,7 +765,12 @@ class ApplicationServicesHandlerSendEventsTestCase(unittest.HomeserverTestCase): fake_device_ids = [f"device_{num}" for num in range(number_of_messages - 1)] messages = { self.exclusive_as_user: { - device_id: to_device_message_content for device_id in fake_device_ids + device_id: { + "type": "test_to_device_message", + "sender": "@some:sender", + "content": to_device_message_content, + } + for device_id in fake_device_ids } } -- cgit 1.5.1 From c2de2ca63060324cf2f80ddf3289b0fd7a4d861b Mon Sep 17 00:00:00 2001 From: Erik Johnston Date: Fri, 9 Dec 2022 09:37:07 +0000 Subject: Delete stale non-e2e devices for users, take 2 (#14595) This should help reduce the number of devices e.g. simple bots the repeatedly login rack up. We only delete non-e2e devices as they should be safe to delete, whereas if we delete e2e devices for a user we may accidentally break their ability to receive e2e keys for a message. --- changelog.d/14595.misc | 1 + synapse/handlers/device.py | 31 +++++++++++- synapse/storage/databases/main/devices.py | 79 ++++++++++++++++++++++++++++++- tests/handlers/test_device.py | 2 +- tests/storage/test_client_ips.py | 4 +- 5 files changed, 113 insertions(+), 4 deletions(-) create mode 100644 changelog.d/14595.misc (limited to 'synapse/handlers') diff --git a/changelog.d/14595.misc b/changelog.d/14595.misc new file mode 100644 index 0000000000..f9bfc581ad --- /dev/null +++ b/changelog.d/14595.misc @@ -0,0 +1 @@ +Prune user's old devices on login if they have too many. diff --git a/synapse/handlers/device.py b/synapse/handlers/device.py index d4750a32e6..7674c187ef 100644 --- a/synapse/handlers/device.py +++ b/synapse/handlers/device.py @@ -52,6 +52,7 @@ from synapse.util import stringutils from synapse.util.async_helpers import Linearizer from synapse.util.caches.expiringcache import ExpiringCache from synapse.util.cancellation import cancellable +from synapse.util.iterutils import batch_iter from synapse.util.metrics import measure_func from synapse.util.retryutils import NotRetryingDestination @@ -421,6 +422,9 @@ class DeviceHandler(DeviceWorkerHandler): self._check_device_name_length(initial_device_display_name) + # Prune the user's device list if they already have a lot of devices. + await self._prune_too_many_devices(user_id) + if device_id is not None: new_device = await self.store.store_device( user_id=user_id, @@ -452,6 +456,31 @@ class DeviceHandler(DeviceWorkerHandler): raise errors.StoreError(500, "Couldn't generate a device ID.") + async def _prune_too_many_devices(self, user_id: str) -> None: + """Delete any excess old devices this user may have.""" + device_ids = await self.store.check_too_many_devices_for_user(user_id) + if not device_ids: + return + + # We don't want to block and try and delete tonnes of devices at once, + # so we cap the number of devices we delete synchronously. + first_batch, remaining_device_ids = device_ids[:10], device_ids[10:] + await self.delete_devices(user_id, first_batch) + + if not remaining_device_ids: + return + + # Now spawn a background loop that deletes the rest. + async def _prune_too_many_devices_loop() -> None: + for batch in batch_iter(remaining_device_ids, 10): + await self.delete_devices(user_id, batch) + + await self.clock.sleep(1) + + run_as_background_process( + "_prune_too_many_devices_loop", _prune_too_many_devices_loop + ) + async def _delete_stale_devices(self) -> None: """Background task that deletes devices which haven't been accessed for more than a configured time period. @@ -481,7 +510,7 @@ class DeviceHandler(DeviceWorkerHandler): device_ids = [d for d in device_ids if d != except_device_id] await self.delete_devices(user_id, device_ids) - async def delete_devices(self, user_id: str, device_ids: List[str]) -> None: + async def delete_devices(self, user_id: str, device_ids: Collection[str]) -> None: """Delete several devices Args: diff --git a/synapse/storage/databases/main/devices.py b/synapse/storage/databases/main/devices.py index a5bb4d404e..08ccd46a2b 100644 --- a/synapse/storage/databases/main/devices.py +++ b/synapse/storage/databases/main/devices.py @@ -1569,6 +1569,72 @@ class DeviceBackgroundUpdateStore(SQLBaseStore): return rows + async def check_too_many_devices_for_user(self, user_id: str) -> List[str]: + """Check if the user has a lot of devices, and if so return the set of + devices we can prune. + + This does *not* return hidden devices or devices with E2E keys. + """ + + num_devices = await self.db_pool.simple_select_one_onecol( + table="devices", + keyvalues={"user_id": user_id, "hidden": False}, + retcol="COALESCE(COUNT(*), 0)", + desc="count_devices", + ) + + # We let users have up to ten devices without pruning. + if num_devices <= 10: + return [] + + # We prune everything older than N days. + max_last_seen = self._clock.time_msec() - 14 * 24 * 60 * 60 * 1000 + + if num_devices > 50: + # If the user has more than 50 devices, then we chose a last seen + # that ensures we keep at most 50 devices. + sql = """ + SELECT last_seen FROM devices + LEFT JOIN e2e_device_keys_json USING (user_id, device_id) + WHERE + user_id = ? + AND NOT hidden + AND last_seen IS NOT NULL + AND key_json IS NULL + ORDER BY last_seen DESC + LIMIT 1 + OFFSET 50 + """ + + rows = await self.db_pool.execute( + "check_too_many_devices_for_user_last_seen", None, sql, (user_id,) + ) + if rows: + max_last_seen = max(rows[0][0], max_last_seen) + + # Now fetch the devices to delete. + sql = """ + SELECT DISTINCT device_id FROM devices + LEFT JOIN e2e_device_keys_json USING (user_id, device_id) + WHERE + user_id = ? + AND NOT hidden + AND last_seen < ? + AND key_json IS NULL + ORDER BY last_seen + """ + + def check_too_many_devices_for_user_txn( + txn: LoggingTransaction, + ) -> List[str]: + txn.execute(sql, (user_id, max_last_seen)) + return [device_id for device_id, in txn] + + return await self.db_pool.runInteraction( + "check_too_many_devices_for_user", + check_too_many_devices_for_user_txn, + ) + class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): # Because we have write access, this will be a StreamIdGenerator @@ -1627,6 +1693,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): values={}, insertion_values={ "display_name": initial_device_display_name, + "last_seen": self._clock.time_msec(), "hidden": False, }, desc="store_device", @@ -1672,7 +1739,15 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): ) raise StoreError(500, "Problem storing device.") - async def delete_devices(self, user_id: str, device_ids: List[str]) -> None: + @cached(max_entries=0) + async def delete_device(self, user_id: str, device_id: str) -> None: + raise NotImplementedError() + + # Note: sometimes deleting rows out of `device_inbox` can take a long time, + # so we use a cache so that we deduplicate in flight requests to delete + # devices. + @cachedList(cached_method_name="delete_device", list_name="device_ids") + async def delete_devices(self, user_id: str, device_ids: Collection[str]) -> dict: """Deletes several devices. Args: @@ -1709,6 +1784,8 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): for device_id in device_ids: self.device_id_exists_cache.invalidate((user_id, device_id)) + return {} + async def update_device( self, user_id: str, device_id: str, new_display_name: Optional[str] = None ) -> None: diff --git a/tests/handlers/test_device.py b/tests/handlers/test_device.py index ce7525e29c..a456bffd63 100644 --- a/tests/handlers/test_device.py +++ b/tests/handlers/test_device.py @@ -115,7 +115,7 @@ class DeviceTestCase(unittest.HomeserverTestCase): "device_id": "xyz", "display_name": "display 0", "last_seen_ip": None, - "last_seen_ts": None, + "last_seen_ts": 1000000, }, device_map["xyz"], ) diff --git a/tests/storage/test_client_ips.py b/tests/storage/test_client_ips.py index 49ad3c1324..a9af1babed 100644 --- a/tests/storage/test_client_ips.py +++ b/tests/storage/test_client_ips.py @@ -169,6 +169,8 @@ class ClientIpStoreTestCase(unittest.HomeserverTestCase): ) ) + last_seen = self.clock.time_msec() + if after_persisting: # Trigger the storage loop self.reactor.advance(10) @@ -189,7 +191,7 @@ class ClientIpStoreTestCase(unittest.HomeserverTestCase): "device_id": device_id, "ip": None, "user_agent": None, - "last_seen": None, + "last_seen": last_seen, }, ], ) -- cgit 1.5.1 From 94bc21e69f89ad873ad7a0deb6d9c4ff3cb480ef Mon Sep 17 00:00:00 2001 From: Erik Johnston Date: Fri, 9 Dec 2022 13:31:32 +0000 Subject: Limit the number of devices we delete at once (#14649) --- changelog.d/14649.misc | 1 + synapse/handlers/device.py | 4 +++- synapse/storage/databases/main/devices.py | 11 ++++++++--- tests/handlers/test_device.py | 31 +++++++++++++++++++++++++++++++ 4 files changed, 43 insertions(+), 4 deletions(-) create mode 100644 changelog.d/14649.misc (limited to 'synapse/handlers') diff --git a/changelog.d/14649.misc b/changelog.d/14649.misc new file mode 100644 index 0000000000..f9bfc581ad --- /dev/null +++ b/changelog.d/14649.misc @@ -0,0 +1 @@ +Prune user's old devices on login if they have too many. diff --git a/synapse/handlers/device.py b/synapse/handlers/device.py index 7674c187ef..c935c7be90 100644 --- a/synapse/handlers/device.py +++ b/synapse/handlers/device.py @@ -458,10 +458,12 @@ class DeviceHandler(DeviceWorkerHandler): async def _prune_too_many_devices(self, user_id: str) -> None: """Delete any excess old devices this user may have.""" - device_ids = await self.store.check_too_many_devices_for_user(user_id) + device_ids = await self.store.check_too_many_devices_for_user(user_id, 100) if not device_ids: return + logger.info("Pruning %d old devices for user %s", len(device_ids), user_id) + # We don't want to block and try and delete tonnes of devices at once, # so we cap the number of devices we delete synchronously. first_batch, remaining_device_ids = device_ids[:10], device_ids[10:] diff --git a/synapse/storage/databases/main/devices.py b/synapse/storage/databases/main/devices.py index 08ccd46a2b..95d4c0622d 100644 --- a/synapse/storage/databases/main/devices.py +++ b/synapse/storage/databases/main/devices.py @@ -1569,11 +1569,15 @@ class DeviceBackgroundUpdateStore(SQLBaseStore): return rows - async def check_too_many_devices_for_user(self, user_id: str) -> List[str]: + async def check_too_many_devices_for_user( + self, user_id: str, limit: int + ) -> List[str]: """Check if the user has a lot of devices, and if so return the set of devices we can prune. This does *not* return hidden devices or devices with E2E keys. + + Returns at most `limit` number of devices, ordered by last seen. """ num_devices = await self.db_pool.simple_select_one_onecol( @@ -1614,7 +1618,7 @@ class DeviceBackgroundUpdateStore(SQLBaseStore): # Now fetch the devices to delete. sql = """ - SELECT DISTINCT device_id FROM devices + SELECT device_id FROM devices LEFT JOIN e2e_device_keys_json USING (user_id, device_id) WHERE user_id = ? @@ -1622,12 +1626,13 @@ class DeviceBackgroundUpdateStore(SQLBaseStore): AND last_seen < ? AND key_json IS NULL ORDER BY last_seen + LIMIT ? """ def check_too_many_devices_for_user_txn( txn: LoggingTransaction, ) -> List[str]: - txn.execute(sql, (user_id, max_last_seen)) + txn.execute(sql, (user_id, max_last_seen, limit)) return [device_id for device_id, in txn] return await self.db_pool.runInteraction( diff --git a/tests/handlers/test_device.py b/tests/handlers/test_device.py index a456bffd63..e51cac9b33 100644 --- a/tests/handlers/test_device.py +++ b/tests/handlers/test_device.py @@ -20,6 +20,8 @@ from twisted.test.proto_helpers import MemoryReactor from synapse.api.errors import NotFoundError, SynapseError from synapse.handlers.device import MAX_DEVICE_DISPLAY_NAME_LEN, DeviceHandler +from synapse.rest import admin +from synapse.rest.client import account, login from synapse.server import HomeServer from synapse.util import Clock @@ -30,6 +32,12 @@ user2 = "@theresa:bbb" class DeviceTestCase(unittest.HomeserverTestCase): + servlets = [ + login.register_servlets, + admin.register_servlets, + account.register_servlets, + ] + def make_homeserver(self, reactor: MemoryReactor, clock: Clock) -> HomeServer: hs = self.setup_test_homeserver("server", federation_http_client=None) handler = hs.get_device_handler() @@ -229,6 +237,29 @@ class DeviceTestCase(unittest.HomeserverTestCase): NotFoundError, ) + def test_login_delete_old_devices(self) -> None: + """Delete old devices if the user already has too many.""" + + user_id = self.register_user("user", "pass") + + # Create a bunch of devices + for _ in range(50): + self.login("user", "pass") + self.reactor.advance(1) + + # Advance the clock for ages (as we only delete old devices) + self.reactor.advance(60 * 60 * 24 * 300) + + # Log in again to start the pruning + self.login("user", "pass") + + # Give the background job time to do its thing + self.reactor.pump([1.0] * 100) + + # We should now only have the most recent device. + devices = self.get_success(self.handler.get_devices_by_user(user_id)) + self.assertEqual(len(devices), 1) + def _record_users(self) -> None: # check this works for both devices which have a recorded client_ip, # and those which don't. -- cgit 1.5.1 From 74b89c27613a34ec9b291ad3066db7ce0adff1db Mon Sep 17 00:00:00 2001 From: reivilibre Date: Mon, 12 Dec 2022 13:55:23 +0000 Subject: Revert the deletion of stale devices due to performance issues. (#14662) --- changelog.d/14595.misc | 1 - changelog.d/14649.misc | 1 - changelog.d/14662.removal | 1 + synapse/handlers/device.py | 33 +----------- synapse/storage/databases/main/devices.py | 84 +------------------------------ tests/handlers/test_device.py | 33 +----------- tests/storage/test_client_ips.py | 4 +- 7 files changed, 5 insertions(+), 152 deletions(-) delete mode 100644 changelog.d/14595.misc delete mode 100644 changelog.d/14649.misc create mode 100644 changelog.d/14662.removal (limited to 'synapse/handlers') diff --git a/changelog.d/14595.misc b/changelog.d/14595.misc deleted file mode 100644 index f9bfc581ad..0000000000 --- a/changelog.d/14595.misc +++ /dev/null @@ -1 +0,0 @@ -Prune user's old devices on login if they have too many. diff --git a/changelog.d/14649.misc b/changelog.d/14649.misc deleted file mode 100644 index f9bfc581ad..0000000000 --- a/changelog.d/14649.misc +++ /dev/null @@ -1 +0,0 @@ -Prune user's old devices on login if they have too many. diff --git a/changelog.d/14662.removal b/changelog.d/14662.removal new file mode 100644 index 0000000000..19a387bbb4 --- /dev/null +++ b/changelog.d/14662.removal @@ -0,0 +1 @@ +(remove from changelog: unreleased) Revert the deletion of stale devices due to performance issues. \ No newline at end of file diff --git a/synapse/handlers/device.py b/synapse/handlers/device.py index c935c7be90..d4750a32e6 100644 --- a/synapse/handlers/device.py +++ b/synapse/handlers/device.py @@ -52,7 +52,6 @@ from synapse.util import stringutils from synapse.util.async_helpers import Linearizer from synapse.util.caches.expiringcache import ExpiringCache from synapse.util.cancellation import cancellable -from synapse.util.iterutils import batch_iter from synapse.util.metrics import measure_func from synapse.util.retryutils import NotRetryingDestination @@ -422,9 +421,6 @@ class DeviceHandler(DeviceWorkerHandler): self._check_device_name_length(initial_device_display_name) - # Prune the user's device list if they already have a lot of devices. - await self._prune_too_many_devices(user_id) - if device_id is not None: new_device = await self.store.store_device( user_id=user_id, @@ -456,33 +452,6 @@ class DeviceHandler(DeviceWorkerHandler): raise errors.StoreError(500, "Couldn't generate a device ID.") - async def _prune_too_many_devices(self, user_id: str) -> None: - """Delete any excess old devices this user may have.""" - device_ids = await self.store.check_too_many_devices_for_user(user_id, 100) - if not device_ids: - return - - logger.info("Pruning %d old devices for user %s", len(device_ids), user_id) - - # We don't want to block and try and delete tonnes of devices at once, - # so we cap the number of devices we delete synchronously. - first_batch, remaining_device_ids = device_ids[:10], device_ids[10:] - await self.delete_devices(user_id, first_batch) - - if not remaining_device_ids: - return - - # Now spawn a background loop that deletes the rest. - async def _prune_too_many_devices_loop() -> None: - for batch in batch_iter(remaining_device_ids, 10): - await self.delete_devices(user_id, batch) - - await self.clock.sleep(1) - - run_as_background_process( - "_prune_too_many_devices_loop", _prune_too_many_devices_loop - ) - async def _delete_stale_devices(self) -> None: """Background task that deletes devices which haven't been accessed for more than a configured time period. @@ -512,7 +481,7 @@ class DeviceHandler(DeviceWorkerHandler): device_ids = [d for d in device_ids if d != except_device_id] await self.delete_devices(user_id, device_ids) - async def delete_devices(self, user_id: str, device_ids: Collection[str]) -> None: + async def delete_devices(self, user_id: str, device_ids: List[str]) -> None: """Delete several devices Args: diff --git a/synapse/storage/databases/main/devices.py b/synapse/storage/databases/main/devices.py index 95d4c0622d..a5bb4d404e 100644 --- a/synapse/storage/databases/main/devices.py +++ b/synapse/storage/databases/main/devices.py @@ -1569,77 +1569,6 @@ class DeviceBackgroundUpdateStore(SQLBaseStore): return rows - async def check_too_many_devices_for_user( - self, user_id: str, limit: int - ) -> List[str]: - """Check if the user has a lot of devices, and if so return the set of - devices we can prune. - - This does *not* return hidden devices or devices with E2E keys. - - Returns at most `limit` number of devices, ordered by last seen. - """ - - num_devices = await self.db_pool.simple_select_one_onecol( - table="devices", - keyvalues={"user_id": user_id, "hidden": False}, - retcol="COALESCE(COUNT(*), 0)", - desc="count_devices", - ) - - # We let users have up to ten devices without pruning. - if num_devices <= 10: - return [] - - # We prune everything older than N days. - max_last_seen = self._clock.time_msec() - 14 * 24 * 60 * 60 * 1000 - - if num_devices > 50: - # If the user has more than 50 devices, then we chose a last seen - # that ensures we keep at most 50 devices. - sql = """ - SELECT last_seen FROM devices - LEFT JOIN e2e_device_keys_json USING (user_id, device_id) - WHERE - user_id = ? - AND NOT hidden - AND last_seen IS NOT NULL - AND key_json IS NULL - ORDER BY last_seen DESC - LIMIT 1 - OFFSET 50 - """ - - rows = await self.db_pool.execute( - "check_too_many_devices_for_user_last_seen", None, sql, (user_id,) - ) - if rows: - max_last_seen = max(rows[0][0], max_last_seen) - - # Now fetch the devices to delete. - sql = """ - SELECT device_id FROM devices - LEFT JOIN e2e_device_keys_json USING (user_id, device_id) - WHERE - user_id = ? - AND NOT hidden - AND last_seen < ? - AND key_json IS NULL - ORDER BY last_seen - LIMIT ? - """ - - def check_too_many_devices_for_user_txn( - txn: LoggingTransaction, - ) -> List[str]: - txn.execute(sql, (user_id, max_last_seen, limit)) - return [device_id for device_id, in txn] - - return await self.db_pool.runInteraction( - "check_too_many_devices_for_user", - check_too_many_devices_for_user_txn, - ) - class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): # Because we have write access, this will be a StreamIdGenerator @@ -1698,7 +1627,6 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): values={}, insertion_values={ "display_name": initial_device_display_name, - "last_seen": self._clock.time_msec(), "hidden": False, }, desc="store_device", @@ -1744,15 +1672,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): ) raise StoreError(500, "Problem storing device.") - @cached(max_entries=0) - async def delete_device(self, user_id: str, device_id: str) -> None: - raise NotImplementedError() - - # Note: sometimes deleting rows out of `device_inbox` can take a long time, - # so we use a cache so that we deduplicate in flight requests to delete - # devices. - @cachedList(cached_method_name="delete_device", list_name="device_ids") - async def delete_devices(self, user_id: str, device_ids: Collection[str]) -> dict: + async def delete_devices(self, user_id: str, device_ids: List[str]) -> None: """Deletes several devices. Args: @@ -1789,8 +1709,6 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): for device_id in device_ids: self.device_id_exists_cache.invalidate((user_id, device_id)) - return {} - async def update_device( self, user_id: str, device_id: str, new_display_name: Optional[str] = None ) -> None: diff --git a/tests/handlers/test_device.py b/tests/handlers/test_device.py index e51cac9b33..ce7525e29c 100644 --- a/tests/handlers/test_device.py +++ b/tests/handlers/test_device.py @@ -20,8 +20,6 @@ from twisted.test.proto_helpers import MemoryReactor from synapse.api.errors import NotFoundError, SynapseError from synapse.handlers.device import MAX_DEVICE_DISPLAY_NAME_LEN, DeviceHandler -from synapse.rest import admin -from synapse.rest.client import account, login from synapse.server import HomeServer from synapse.util import Clock @@ -32,12 +30,6 @@ user2 = "@theresa:bbb" class DeviceTestCase(unittest.HomeserverTestCase): - servlets = [ - login.register_servlets, - admin.register_servlets, - account.register_servlets, - ] - def make_homeserver(self, reactor: MemoryReactor, clock: Clock) -> HomeServer: hs = self.setup_test_homeserver("server", federation_http_client=None) handler = hs.get_device_handler() @@ -123,7 +115,7 @@ class DeviceTestCase(unittest.HomeserverTestCase): "device_id": "xyz", "display_name": "display 0", "last_seen_ip": None, - "last_seen_ts": 1000000, + "last_seen_ts": None, }, device_map["xyz"], ) @@ -237,29 +229,6 @@ class DeviceTestCase(unittest.HomeserverTestCase): NotFoundError, ) - def test_login_delete_old_devices(self) -> None: - """Delete old devices if the user already has too many.""" - - user_id = self.register_user("user", "pass") - - # Create a bunch of devices - for _ in range(50): - self.login("user", "pass") - self.reactor.advance(1) - - # Advance the clock for ages (as we only delete old devices) - self.reactor.advance(60 * 60 * 24 * 300) - - # Log in again to start the pruning - self.login("user", "pass") - - # Give the background job time to do its thing - self.reactor.pump([1.0] * 100) - - # We should now only have the most recent device. - devices = self.get_success(self.handler.get_devices_by_user(user_id)) - self.assertEqual(len(devices), 1) - def _record_users(self) -> None: # check this works for both devices which have a recorded client_ip, # and those which don't. diff --git a/tests/storage/test_client_ips.py b/tests/storage/test_client_ips.py index 81e4e596e4..7f7f4ef892 100644 --- a/tests/storage/test_client_ips.py +++ b/tests/storage/test_client_ips.py @@ -170,8 +170,6 @@ class ClientIpStoreTestCase(unittest.HomeserverTestCase): ) ) - last_seen = self.clock.time_msec() - if after_persisting: # Trigger the storage loop self.reactor.advance(10) @@ -192,7 +190,7 @@ class ClientIpStoreTestCase(unittest.HomeserverTestCase): "device_id": device_id, "ip": None, "user_agent": None, - "last_seen": last_seen, + "last_seen": None, }, ], ) -- cgit 1.5.1 From b5b5f6608462a988b05502a3b70b6a57ca3846d2 Mon Sep 17 00:00:00 2001 From: David Robertson Date: Mon, 12 Dec 2022 16:19:30 +0000 Subject: Move `StateFilter` to `synapse.types` (#14668) * Move `StateFilter` to `synapse.types` * Changelog --- changelog.d/14668.misc | 1 + synapse/events/builder.py | 2 +- synapse/events/snapshot.py | 2 +- synapse/handlers/federation.py | 2 +- synapse/handlers/federation_event.py | 2 +- synapse/handlers/message.py | 2 +- synapse/handlers/pagination.py | 2 +- synapse/handlers/register.py | 2 +- synapse/handlers/room.py | 2 +- synapse/handlers/room_member.py | 2 +- synapse/handlers/search.py | 2 +- synapse/handlers/sync.py | 2 +- synapse/module_api/__init__.py | 2 +- synapse/push/bulk_push_rule_evaluator.py | 2 +- synapse/push/mailer.py | 2 +- synapse/rest/admin/rooms.py | 2 +- synapse/rest/client/room.py | 2 +- synapse/state/__init__.py | 2 +- synapse/storage/controllers/persist_events.py | 2 +- synapse/storage/controllers/state.py | 2 +- synapse/storage/databases/main/state.py | 2 +- synapse/storage/databases/state/bg_updates.py | 2 +- synapse/storage/databases/state/store.py | 2 +- synapse/storage/state.py | 567 ---------------- synapse/types.py | 928 -------------------------- synapse/types/__init__.py | 928 ++++++++++++++++++++++++++ synapse/types/state.py | 567 ++++++++++++++++ synapse/visibility.py | 2 +- tests/storage/test_state.py | 2 +- 29 files changed, 1520 insertions(+), 1519 deletions(-) create mode 100644 changelog.d/14668.misc delete mode 100644 synapse/storage/state.py delete mode 100644 synapse/types.py create mode 100644 synapse/types/__init__.py create mode 100644 synapse/types/state.py (limited to 'synapse/handlers') diff --git a/changelog.d/14668.misc b/changelog.d/14668.misc new file mode 100644 index 0000000000..5269d8a97d --- /dev/null +++ b/changelog.d/14668.misc @@ -0,0 +1 @@ +Move `StateFilter` to `synapse.types`. diff --git a/synapse/events/builder.py b/synapse/events/builder.py index d62906043f..94dd1298e1 100644 --- a/synapse/events/builder.py +++ b/synapse/events/builder.py @@ -28,8 +28,8 @@ from synapse.event_auth import auth_types_for_event from synapse.events import EventBase, _EventInternalMetadata, make_event_from_dict from synapse.state import StateHandler from synapse.storage.databases.main import DataStore -from synapse.storage.state import StateFilter from synapse.types import EventID, JsonDict +from synapse.types.state import StateFilter from synapse.util import Clock from synapse.util.stringutils import random_string diff --git a/synapse/events/snapshot.py b/synapse/events/snapshot.py index 1c0e96bec7..6eaef8b57a 100644 --- a/synapse/events/snapshot.py +++ b/synapse/events/snapshot.py @@ -23,7 +23,7 @@ from synapse.types import JsonDict, StateMap if TYPE_CHECKING: from synapse.storage.controllers import StorageControllers from synapse.storage.databases.main import DataStore - from synapse.storage.state import StateFilter + from synapse.types.state import StateFilter @attr.s(slots=True, auto_attribs=True) diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index 3398fcaf7d..b2784d7333 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -70,8 +70,8 @@ from synapse.replication.http.federation import ( ) from synapse.storage.databases.main.events import PartialStateConflictError from synapse.storage.databases.main.events_worker import EventRedactBehaviour -from synapse.storage.state import StateFilter from synapse.types import JsonDict, get_domain_from_id +from synapse.types.state import StateFilter from synapse.util.async_helpers import Linearizer from synapse.util.retryutils import NotRetryingDestination from synapse.visibility import filter_events_for_server diff --git a/synapse/handlers/federation_event.py b/synapse/handlers/federation_event.py index f7223b03c3..d2facdab60 100644 --- a/synapse/handlers/federation_event.py +++ b/synapse/handlers/federation_event.py @@ -75,7 +75,6 @@ from synapse.replication.http.federation import ( from synapse.state import StateResolutionStore from synapse.storage.databases.main.events import PartialStateConflictError from synapse.storage.databases.main.events_worker import EventRedactBehaviour -from synapse.storage.state import StateFilter from synapse.types import ( PersistedEventPosition, RoomStreamToken, @@ -83,6 +82,7 @@ from synapse.types import ( UserID, get_domain_from_id, ) +from synapse.types.state import StateFilter from synapse.util.async_helpers import Linearizer, concurrently_execute from synapse.util.iterutils import batch_iter from synapse.util.retryutils import NotRetryingDestination diff --git a/synapse/handlers/message.py b/synapse/handlers/message.py index 5cbe89f4fd..d6e90ef259 100644 --- a/synapse/handlers/message.py +++ b/synapse/handlers/message.py @@ -59,7 +59,6 @@ from synapse.replication.http.send_event import ReplicationSendEventRestServlet from synapse.replication.http.send_events import ReplicationSendEventsRestServlet from synapse.storage.databases.main.events import PartialStateConflictError from synapse.storage.databases.main.events_worker import EventRedactBehaviour -from synapse.storage.state import StateFilter from synapse.types import ( MutableStateMap, PersistedEventPosition, @@ -70,6 +69,7 @@ from synapse.types import ( UserID, create_requester, ) +from synapse.types.state import StateFilter from synapse.util import json_decoder, json_encoder, log_failure, unwrapFirstError from synapse.util.async_helpers import Linearizer, gather_results from synapse.util.caches.expiringcache import ExpiringCache diff --git a/synapse/handlers/pagination.py b/synapse/handlers/pagination.py index c572508a02..8c8ff18a1a 100644 --- a/synapse/handlers/pagination.py +++ b/synapse/handlers/pagination.py @@ -27,9 +27,9 @@ from synapse.handlers.room import ShutdownRoomResponse from synapse.logging.opentracing import trace from synapse.metrics.background_process_metrics import run_as_background_process from synapse.rest.admin._base import assert_user_is_admin -from synapse.storage.state import StateFilter from synapse.streams.config import PaginationConfig from synapse.types import JsonDict, Requester, StreamKeyType +from synapse.types.state import StateFilter from synapse.util.async_helpers import ReadWriteLock from synapse.util.stringutils import random_string from synapse.visibility import filter_events_for_client diff --git a/synapse/handlers/register.py b/synapse/handlers/register.py index 6307fa9c5d..c611efb760 100644 --- a/synapse/handlers/register.py +++ b/synapse/handlers/register.py @@ -46,8 +46,8 @@ from synapse.replication.http.register import ( ReplicationRegisterServlet, ) from synapse.spam_checker_api import RegistrationBehaviour -from synapse.storage.state import StateFilter from synapse.types import RoomAlias, UserID, create_requester +from synapse.types.state import StateFilter if TYPE_CHECKING: from synapse.server import HomeServer diff --git a/synapse/handlers/room.py b/synapse/handlers/room.py index 6dcfd86fdf..f81241c2b3 100644 --- a/synapse/handlers/room.py +++ b/synapse/handlers/room.py @@ -62,7 +62,6 @@ from synapse.events.utils import copy_and_fixup_power_levels_contents from synapse.handlers.relations import BundledAggregations from synapse.module_api import NOT_SPAM from synapse.rest.admin._base import assert_user_is_admin -from synapse.storage.state import StateFilter from synapse.streams import EventSource from synapse.types import ( JsonDict, @@ -77,6 +76,7 @@ from synapse.types import ( UserID, create_requester, ) +from synapse.types.state import StateFilter from synapse.util import stringutils from synapse.util.caches.response_cache import ResponseCache from synapse.util.stringutils import parse_and_validate_server_name diff --git a/synapse/handlers/room_member.py b/synapse/handlers/room_member.py index 6ad2b38b8f..0c39e852a1 100644 --- a/synapse/handlers/room_member.py +++ b/synapse/handlers/room_member.py @@ -34,7 +34,6 @@ from synapse.events.snapshot import EventContext from synapse.handlers.profile import MAX_AVATAR_URL_LEN, MAX_DISPLAYNAME_LEN from synapse.logging import opentracing from synapse.module_api import NOT_SPAM -from synapse.storage.state import StateFilter from synapse.types import ( JsonDict, Requester, @@ -45,6 +44,7 @@ from synapse.types import ( create_requester, get_domain_from_id, ) +from synapse.types.state import StateFilter from synapse.util.async_helpers import Linearizer from synapse.util.distributor import user_left_room diff --git a/synapse/handlers/search.py b/synapse/handlers/search.py index bcab98c6d5..33115ce488 100644 --- a/synapse/handlers/search.py +++ b/synapse/handlers/search.py @@ -23,8 +23,8 @@ from synapse.api.constants import EventTypes, Membership from synapse.api.errors import NotFoundError, SynapseError from synapse.api.filtering import Filter from synapse.events import EventBase -from synapse.storage.state import StateFilter from synapse.types import JsonDict, StreamKeyType, UserID +from synapse.types.state import StateFilter from synapse.visibility import filter_events_for_client if TYPE_CHECKING: diff --git a/synapse/handlers/sync.py b/synapse/handlers/sync.py index dace9b606f..7d6a653747 100644 --- a/synapse/handlers/sync.py +++ b/synapse/handlers/sync.py @@ -49,7 +49,6 @@ from synapse.push.clientformat import format_push_rules_for_user from synapse.storage.databases.main.event_push_actions import RoomNotifCounts from synapse.storage.databases.main.roommember import extract_heroes_from_room_summary from synapse.storage.roommember import MemberSummary -from synapse.storage.state import StateFilter from synapse.types import ( DeviceListUpdates, JsonDict, @@ -61,6 +60,7 @@ from synapse.types import ( StreamToken, UserID, ) +from synapse.types.state import StateFilter from synapse.util.async_helpers import concurrently_execute from synapse.util.caches.expiringcache import ExpiringCache from synapse.util.caches.lrucache import LruCache diff --git a/synapse/module_api/__init__.py b/synapse/module_api/__init__.py index 96a661177a..0092a03c59 100644 --- a/synapse/module_api/__init__.py +++ b/synapse/module_api/__init__.py @@ -111,7 +111,6 @@ from synapse.storage.background_updates import ( ) from synapse.storage.database import DatabasePool, LoggingTransaction from synapse.storage.databases.main.roommember import ProfileInfo -from synapse.storage.state import StateFilter from synapse.types import ( DomainSpecificString, JsonDict, @@ -124,6 +123,7 @@ from synapse.types import ( UserProfile, create_requester, ) +from synapse.types.state import StateFilter from synapse.util import Clock from synapse.util.async_helpers import maybe_awaitable from synapse.util.caches.descriptors import CachedFunction, cached diff --git a/synapse/push/bulk_push_rule_evaluator.py b/synapse/push/bulk_push_rule_evaluator.py index 9ed35d8461..36e5b327ef 100644 --- a/synapse/push/bulk_push_rule_evaluator.py +++ b/synapse/push/bulk_push_rule_evaluator.py @@ -35,8 +35,8 @@ from synapse.events import EventBase, relation_from_event from synapse.events.snapshot import EventContext from synapse.state import POWER_KEY from synapse.storage.databases.main.roommember import EventIdMembership -from synapse.storage.state import StateFilter from synapse.synapse_rust.push import FilteredPushRules, PushRuleEvaluator +from synapse.types.state import StateFilter from synapse.util.caches import register_cache from synapse.util.metrics import measure_func from synapse.visibility import filter_event_for_clients_with_state diff --git a/synapse/push/mailer.py b/synapse/push/mailer.py index c2575ba3d9..93b255ced5 100644 --- a/synapse/push/mailer.py +++ b/synapse/push/mailer.py @@ -37,8 +37,8 @@ from synapse.push.push_types import ( TemplateVars, ) from synapse.storage.databases.main.event_push_actions import EmailPushAction -from synapse.storage.state import StateFilter from synapse.types import StateMap, UserID +from synapse.types.state import StateFilter from synapse.util.async_helpers import concurrently_execute from synapse.visibility import filter_events_for_client diff --git a/synapse/rest/admin/rooms.py b/synapse/rest/admin/rooms.py index 747e6fda83..e957aa28ca 100644 --- a/synapse/rest/admin/rooms.py +++ b/synapse/rest/admin/rooms.py @@ -34,9 +34,9 @@ from synapse.rest.admin._base import ( assert_user_is_admin, ) from synapse.storage.databases.main.room import RoomSortOrder -from synapse.storage.state import StateFilter from synapse.streams.config import PaginationConfig from synapse.types import JsonDict, RoomID, UserID, create_requester +from synapse.types.state import StateFilter from synapse.util import json_decoder if TYPE_CHECKING: diff --git a/synapse/rest/client/room.py b/synapse/rest/client/room.py index 514eb6afc8..790614d721 100644 --- a/synapse/rest/client/room.py +++ b/synapse/rest/client/room.py @@ -55,9 +55,9 @@ from synapse.logging.opentracing import set_tag from synapse.metrics.background_process_metrics import run_as_background_process from synapse.rest.client._base import client_patterns from synapse.rest.client.transactions import HttpTransactionCache -from synapse.storage.state import StateFilter from synapse.streams.config import PaginationConfig from synapse.types import JsonDict, StreamToken, ThirdPartyInstanceID, UserID +from synapse.types.state import StateFilter from synapse.util import json_decoder from synapse.util.cancellation import cancellable from synapse.util.stringutils import parse_and_validate_server_name, random_string diff --git a/synapse/state/__init__.py b/synapse/state/__init__.py index 833ffec3de..ee5469d5a8 100644 --- a/synapse/state/__init__.py +++ b/synapse/state/__init__.py @@ -44,8 +44,8 @@ from synapse.logging.context import ContextResourceUsage from synapse.replication.http.state import ReplicationUpdateCurrentStateRestServlet from synapse.state import v1, v2 from synapse.storage.databases.main.events_worker import EventRedactBehaviour -from synapse.storage.state import StateFilter from synapse.types import StateMap +from synapse.types.state import StateFilter from synapse.util.async_helpers import Linearizer from synapse.util.caches.expiringcache import ExpiringCache from synapse.util.metrics import Measure, measure_func diff --git a/synapse/storage/controllers/persist_events.py b/synapse/storage/controllers/persist_events.py index 33ffef521b..f1d2c71c91 100644 --- a/synapse/storage/controllers/persist_events.py +++ b/synapse/storage/controllers/persist_events.py @@ -58,13 +58,13 @@ from synapse.storage.controllers.state import StateStorageController from synapse.storage.databases import Databases from synapse.storage.databases.main.events import DeltaState from synapse.storage.databases.main.events_worker import EventRedactBehaviour -from synapse.storage.state import StateFilter from synapse.types import ( PersistedEventPosition, RoomStreamToken, StateMap, get_domain_from_id, ) +from synapse.types.state import StateFilter from synapse.util.async_helpers import ObservableDeferred, yieldable_gather_results from synapse.util.metrics import Measure diff --git a/synapse/storage/controllers/state.py b/synapse/storage/controllers/state.py index 2b31ce54bb..26d79c6e62 100644 --- a/synapse/storage/controllers/state.py +++ b/synapse/storage/controllers/state.py @@ -31,12 +31,12 @@ from synapse.api.constants import EventTypes from synapse.events import EventBase from synapse.logging.opentracing import tag_args, trace from synapse.storage.roommember import ProfileInfo -from synapse.storage.state import StateFilter from synapse.storage.util.partial_state_events_tracker import ( PartialCurrentStateTracker, PartialStateEventsTracker, ) from synapse.types import MutableStateMap, StateMap +from synapse.types.state import StateFilter from synapse.util.cancellation import cancellable if TYPE_CHECKING: diff --git a/synapse/storage/databases/main/state.py b/synapse/storage/databases/main/state.py index af7bebee80..c801a93b5b 100644 --- a/synapse/storage/databases/main/state.py +++ b/synapse/storage/databases/main/state.py @@ -33,8 +33,8 @@ from synapse.storage.database import ( ) from synapse.storage.databases.main.events_worker import EventsWorkerStore from synapse.storage.databases.main.roommember import RoomMemberWorkerStore -from synapse.storage.state import StateFilter from synapse.types import JsonDict, JsonMapping, StateMap +from synapse.types.state import StateFilter from synapse.util.caches import intern_string from synapse.util.caches.descriptors import cached, cachedList from synapse.util.cancellation import cancellable diff --git a/synapse/storage/databases/state/bg_updates.py b/synapse/storage/databases/state/bg_updates.py index 4a4ad0f492..d743282f13 100644 --- a/synapse/storage/databases/state/bg_updates.py +++ b/synapse/storage/databases/state/bg_updates.py @@ -22,8 +22,8 @@ from synapse.storage.database import ( LoggingTransaction, ) from synapse.storage.engines import PostgresEngine -from synapse.storage.state import StateFilter from synapse.types import MutableStateMap, StateMap +from synapse.types.state import StateFilter from synapse.util.caches import intern_string if TYPE_CHECKING: diff --git a/synapse/storage/databases/state/store.py b/synapse/storage/databases/state/store.py index f8cfcaca83..1a7232b276 100644 --- a/synapse/storage/databases/state/store.py +++ b/synapse/storage/databases/state/store.py @@ -25,10 +25,10 @@ from synapse.storage.database import ( LoggingTransaction, ) from synapse.storage.databases.state.bg_updates import StateBackgroundUpdateStore -from synapse.storage.state import StateFilter from synapse.storage.types import Cursor from synapse.storage.util.sequence import build_sequence_generator from synapse.types import MutableStateMap, StateKey, StateMap +from synapse.types.state import StateFilter from synapse.util.caches.descriptors import cached from synapse.util.caches.dictionary_cache import DictionaryCache from synapse.util.cancellation import cancellable diff --git a/synapse/storage/state.py b/synapse/storage/state.py deleted file mode 100644 index 0004d955b4..0000000000 --- a/synapse/storage/state.py +++ /dev/null @@ -1,567 +0,0 @@ -# Copyright 2014-2016 OpenMarket Ltd -# Copyright 2022 The Matrix.org Foundation C.I.C. -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. -import logging -from typing import ( - TYPE_CHECKING, - Callable, - Collection, - Dict, - Iterable, - List, - Mapping, - Optional, - Set, - Tuple, - TypeVar, -) - -import attr -from frozendict import frozendict - -from synapse.api.constants import EventTypes -from synapse.types import MutableStateMap, StateKey, StateMap - -if TYPE_CHECKING: - from typing import FrozenSet # noqa: used within quoted type hint; flake8 sad - - -logger = logging.getLogger(__name__) - -# Used for generic functions below -T = TypeVar("T") - - -@attr.s(slots=True, frozen=True, auto_attribs=True) -class StateFilter: - """A filter used when querying for state. - - Attributes: - types: Map from type to set of state keys (or None). This specifies - which state_keys for the given type to fetch from the DB. If None - then all events with that type are fetched. If the set is empty - then no events with that type are fetched. - include_others: Whether to fetch events with types that do not - appear in `types`. - """ - - types: "frozendict[str, Optional[FrozenSet[str]]]" - include_others: bool = False - - def __attrs_post_init__(self) -> None: - # If `include_others` is set we canonicalise the filter by removing - # wildcards from the types dictionary - if self.include_others: - # this is needed to work around the fact that StateFilter is frozen - object.__setattr__( - self, - "types", - frozendict({k: v for k, v in self.types.items() if v is not None}), - ) - - @staticmethod - def all() -> "StateFilter": - """Returns a filter that fetches everything. - - Returns: - The state filter. - """ - return _ALL_STATE_FILTER - - @staticmethod - def none() -> "StateFilter": - """Returns a filter that fetches nothing. - - Returns: - The new state filter. - """ - return _NONE_STATE_FILTER - - @staticmethod - def from_types(types: Iterable[Tuple[str, Optional[str]]]) -> "StateFilter": - """Creates a filter that only fetches the given types - - Args: - types: A list of type and state keys to fetch. A state_key of None - fetches everything for that type - - Returns: - The new state filter. - """ - type_dict: Dict[str, Optional[Set[str]]] = {} - for typ, s in types: - if typ in type_dict: - if type_dict[typ] is None: - continue - - if s is None: - type_dict[typ] = None - continue - - type_dict.setdefault(typ, set()).add(s) # type: ignore - - return StateFilter( - types=frozendict( - (k, frozenset(v) if v is not None else None) - for k, v in type_dict.items() - ) - ) - - @staticmethod - def from_lazy_load_member_list(members: Iterable[str]) -> "StateFilter": - """Creates a filter that returns all non-member events, plus the member - events for the given users - - Args: - members: Set of user IDs - - Returns: - The new state filter - """ - return StateFilter( - types=frozendict({EventTypes.Member: frozenset(members)}), - include_others=True, - ) - - @staticmethod - def freeze( - types: Mapping[str, Optional[Collection[str]]], include_others: bool - ) -> "StateFilter": - """ - Returns a (frozen) StateFilter with the same contents as the parameters - specified here, which can be made of mutable types. - """ - types_with_frozen_values: Dict[str, Optional[FrozenSet[str]]] = {} - for state_types, state_keys in types.items(): - if state_keys is not None: - types_with_frozen_values[state_types] = frozenset(state_keys) - else: - types_with_frozen_values[state_types] = None - - return StateFilter( - frozendict(types_with_frozen_values), include_others=include_others - ) - - def return_expanded(self) -> "StateFilter": - """Creates a new StateFilter where type wild cards have been removed - (except for memberships). The returned filter is a superset of the - current one, i.e. anything that passes the current filter will pass - the returned filter. - - This helps the caching as the DictionaryCache knows if it has *all* the - state, but does not know if it has all of the keys of a particular type, - which makes wildcard lookups expensive unless we have a complete cache. - Hence, if we are doing a wildcard lookup, populate the cache fully so - that we can do an efficient lookup next time. - - Note that since we have two caches, one for membership events and one for - other events, we can be a bit more clever than simply returning - `StateFilter.all()` if `has_wildcards()` is True. - - We return a StateFilter where: - 1. the list of membership events to return is the same - 2. if there is a wildcard that matches non-member events we - return all non-member events - - Returns: - The new state filter. - """ - - if self.is_full(): - # If we're going to return everything then there's nothing to do - return self - - if not self.has_wildcards(): - # If there are no wild cards, there's nothing to do - return self - - if EventTypes.Member in self.types: - get_all_members = self.types[EventTypes.Member] is None - else: - get_all_members = self.include_others - - has_non_member_wildcard = self.include_others or any( - state_keys is None - for t, state_keys in self.types.items() - if t != EventTypes.Member - ) - - if not has_non_member_wildcard: - # If there are no non-member wild cards we can just return ourselves - return self - - if get_all_members: - # We want to return everything. - return StateFilter.all() - elif EventTypes.Member in self.types: - # We want to return all non-members, but only particular - # memberships - return StateFilter( - types=frozendict({EventTypes.Member: self.types[EventTypes.Member]}), - include_others=True, - ) - else: - # We want to return all non-members - return _ALL_NON_MEMBER_STATE_FILTER - - def make_sql_filter_clause(self) -> Tuple[str, List[str]]: - """Converts the filter to an SQL clause. - - For example: - - f = StateFilter.from_types([("m.room.create", "")]) - clause, args = f.make_sql_filter_clause() - clause == "(type = ? AND state_key = ?)" - args == ['m.room.create', ''] - - - Returns: - The SQL string (may be empty) and arguments. An empty SQL string is - returned when the filter matches everything (i.e. is "full"). - """ - - where_clause = "" - where_args: List[str] = [] - - if self.is_full(): - return where_clause, where_args - - if not self.include_others and not self.types: - # i.e. this is an empty filter, so we need to return a clause that - # will match nothing - return "1 = 2", [] - - # First we build up a lost of clauses for each type/state_key combo - clauses = [] - for etype, state_keys in self.types.items(): - if state_keys is None: - clauses.append("(type = ?)") - where_args.append(etype) - continue - - for state_key in state_keys: - clauses.append("(type = ? AND state_key = ?)") - where_args.extend((etype, state_key)) - - # This will match anything that appears in `self.types` - where_clause = " OR ".join(clauses) - - # If we want to include stuff that's not in the types dict then we add - # a `OR type NOT IN (...)` clause to the end. - if self.include_others: - if where_clause: - where_clause += " OR " - - where_clause += "type NOT IN (%s)" % (",".join(["?"] * len(self.types)),) - where_args.extend(self.types) - - return where_clause, where_args - - def max_entries_returned(self) -> Optional[int]: - """Returns the maximum number of entries this filter will return if - known, otherwise returns None. - - For example a simple state filter asking for `("m.room.create", "")` - will return 1, whereas the default state filter will return None. - - This is used to bail out early if the right number of entries have been - fetched. - """ - if self.has_wildcards(): - return None - - return len(self.concrete_types()) - - def filter_state(self, state_dict: StateMap[T]) -> MutableStateMap[T]: - """Returns the state filtered with by this StateFilter. - - Args: - state: The state map to filter - - Returns: - The filtered state map. - This is a copy, so it's safe to mutate. - """ - if self.is_full(): - return dict(state_dict) - - filtered_state = {} - for k, v in state_dict.items(): - typ, state_key = k - if typ in self.types: - state_keys = self.types[typ] - if state_keys is None or state_key in state_keys: - filtered_state[k] = v - elif self.include_others: - filtered_state[k] = v - - return filtered_state - - def is_full(self) -> bool: - """Whether this filter fetches everything or not - - Returns: - True if the filter fetches everything. - """ - return self.include_others and not self.types - - def has_wildcards(self) -> bool: - """Whether the filter includes wildcards or is attempting to fetch - specific state. - - Returns: - True if the filter includes wildcards. - """ - - return self.include_others or any( - state_keys is None for state_keys in self.types.values() - ) - - def concrete_types(self) -> List[Tuple[str, str]]: - """Returns a list of concrete type/state_keys (i.e. not None) that - will be fetched. This will be a complete list if `has_wildcards` - returns False, but otherwise will be a subset (or even empty). - - Returns: - A list of type/state_keys tuples. - """ - return [ - (t, s) - for t, state_keys in self.types.items() - if state_keys is not None - for s in state_keys - ] - - def get_member_split(self) -> Tuple["StateFilter", "StateFilter"]: - """Return the filter split into two: one which assumes it's exclusively - matching against member state, and one which assumes it's matching - against non member state. - - This is useful due to the returned filters giving correct results for - `is_full()`, `has_wildcards()`, etc, when operating against maps that - either exclusively contain member events or only contain non-member - events. (Which is the case when dealing with the member vs non-member - state caches). - - Returns: - The member and non member filters - """ - - if EventTypes.Member in self.types: - state_keys = self.types[EventTypes.Member] - if state_keys is None: - member_filter = StateFilter.all() - else: - member_filter = StateFilter(frozendict({EventTypes.Member: state_keys})) - elif self.include_others: - member_filter = StateFilter.all() - else: - member_filter = StateFilter.none() - - non_member_filter = StateFilter( - types=frozendict( - {k: v for k, v in self.types.items() if k != EventTypes.Member} - ), - include_others=self.include_others, - ) - - return member_filter, non_member_filter - - def _decompose_into_four_parts( - self, - ) -> Tuple[Tuple[bool, Set[str]], Tuple[Set[str], Set[StateKey]]]: - """ - Decomposes this state filter into 4 constituent parts, which can be - thought of as this: - all? - minus_wildcards + plus_wildcards + plus_state_keys - - where - * all represents ALL state - * minus_wildcards represents entire state types to remove - * plus_wildcards represents entire state types to add - * plus_state_keys represents individual state keys to add - - See `recompose_from_four_parts` for the other direction of this - correspondence. - """ - is_all = self.include_others - excluded_types: Set[str] = {t for t in self.types if is_all} - wildcard_types: Set[str] = {t for t, s in self.types.items() if s is None} - concrete_keys: Set[StateKey] = set(self.concrete_types()) - - return (is_all, excluded_types), (wildcard_types, concrete_keys) - - @staticmethod - def _recompose_from_four_parts( - all_part: bool, - minus_wildcards: Set[str], - plus_wildcards: Set[str], - plus_state_keys: Set[StateKey], - ) -> "StateFilter": - """ - Recomposes a state filter from 4 parts. - - See `decompose_into_four_parts` (the other direction of this - correspondence) for descriptions on each of the parts. - """ - - # {state type -> set of state keys OR None for wildcard} - # (The same structure as that of a StateFilter.) - new_types: Dict[str, Optional[Set[str]]] = {} - - # if we start with all, insert the excluded statetypes as empty sets - # to prevent them from being included - if all_part: - new_types.update({state_type: set() for state_type in minus_wildcards}) - - # insert the plus wildcards - new_types.update({state_type: None for state_type in plus_wildcards}) - - # insert the specific state keys - for state_type, state_key in plus_state_keys: - if state_type in new_types: - entry = new_types[state_type] - if entry is not None: - entry.add(state_key) - elif not all_part: - # don't insert if the entire type is already included by - # include_others as this would actually shrink the state allowed - # by this filter. - new_types[state_type] = {state_key} - - return StateFilter.freeze(new_types, include_others=all_part) - - def approx_difference(self, other: "StateFilter") -> "StateFilter": - """ - Returns a state filter which represents `self - other`. - - This is useful for determining what state remains to be pulled out of the - database if we want the state included by `self` but already have the state - included by `other`. - - The returned state filter - - MUST include all state events that are included by this filter (`self`) - unless they are included by `other`; - - MUST NOT include state events not included by this filter (`self`); and - - MAY be an over-approximation: the returned state filter - MAY additionally include some state events from `other`. - - This implementation attempts to return the narrowest such state filter. - In the case that `self` contains wildcards for state types where - `other` contains specific state keys, an approximation must be made: - the returned state filter keeps the wildcard, as state filters are not - able to express 'all state keys except some given examples'. - e.g. - StateFilter(m.room.member -> None (wildcard)) - minus - StateFilter(m.room.member -> {'@wombat:example.org'}) - is approximated as - StateFilter(m.room.member -> None (wildcard)) - """ - - # We first transform self and other into an alternative representation: - # - whether or not they include all events to begin with ('all') - # - if so, which event types are excluded? ('excludes') - # - which entire event types to include ('wildcards') - # - which concrete state keys to include ('concrete state keys') - (self_all, self_excludes), ( - self_wildcards, - self_concrete_keys, - ) = self._decompose_into_four_parts() - (other_all, other_excludes), ( - other_wildcards, - other_concrete_keys, - ) = other._decompose_into_four_parts() - - # Start with an estimate of the difference based on self - new_all = self_all - # Wildcards from the other can be added to the exclusion filter - new_excludes = self_excludes | other_wildcards - # We remove wildcards that appeared as wildcards in the other - new_wildcards = self_wildcards - other_wildcards - # We filter out the concrete state keys that appear in the other - # as wildcards or concrete state keys. - new_concrete_keys = { - (state_type, state_key) - for (state_type, state_key) in self_concrete_keys - if state_type not in other_wildcards - } - other_concrete_keys - - if other_all: - if self_all: - # If self starts with all, then we add as wildcards any - # types which appear in the other's exclusion filter (but - # aren't in the self exclusion filter). This is as the other - # filter will return everything BUT the types in its exclusion, so - # we need to add those excluded types that also match the self - # filter as wildcard types in the new filter. - new_wildcards |= other_excludes.difference(self_excludes) - - # If other is an `include_others` then the difference isn't. - new_all = False - # (We have no need for excludes when we don't start with all, as there - # is nothing to exclude.) - new_excludes = set() - - # We also filter out all state types that aren't in the exclusion - # list of the other. - new_wildcards &= other_excludes - new_concrete_keys = { - (state_type, state_key) - for (state_type, state_key) in new_concrete_keys - if state_type in other_excludes - } - - # Transform our newly-constructed state filter from the alternative - # representation back into the normal StateFilter representation. - return StateFilter._recompose_from_four_parts( - new_all, new_excludes, new_wildcards, new_concrete_keys - ) - - def must_await_full_state(self, is_mine_id: Callable[[str], bool]) -> bool: - """Check if we need to wait for full state to complete to calculate this state - - If we have a state filter which is completely satisfied even with partial - state, then we don't need to await_full_state before we can return it. - - Args: - is_mine_id: a callable which confirms if a given state_key matches a mxid - of a local user - """ - # if we haven't requested membership events, then it depends on the value of - # 'include_others' - if EventTypes.Member not in self.types: - return self.include_others - - # if we're looking for *all* membership events, then we have to wait - member_state_keys = self.types[EventTypes.Member] - if member_state_keys is None: - return True - - # otherwise, consider whose membership we are looking for. If it's entirely - # local users, then we don't need to wait. - for state_key in member_state_keys: - if not is_mine_id(state_key): - # remote user - return True - - # local users only - return False - - -_ALL_STATE_FILTER = StateFilter(types=frozendict(), include_others=True) -_ALL_NON_MEMBER_STATE_FILTER = StateFilter( - types=frozendict({EventTypes.Member: frozenset()}), include_others=True -) -_NONE_STATE_FILTER = StateFilter(types=frozendict(), include_others=False) diff --git a/synapse/types.py b/synapse/types.py deleted file mode 100644 index f2d436ddc3..0000000000 --- a/synapse/types.py +++ /dev/null @@ -1,928 +0,0 @@ -# Copyright 2014-2016 OpenMarket Ltd -# Copyright 2019 The Matrix.org Foundation C.I.C. -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. -import abc -import re -import string -from typing import ( - TYPE_CHECKING, - Any, - ClassVar, - Dict, - List, - Mapping, - Match, - MutableMapping, - NoReturn, - Optional, - Set, - Tuple, - Type, - TypeVar, - Union, -) - -import attr -from frozendict import frozendict -from signedjson.key import decode_verify_key_bytes -from signedjson.types import VerifyKey -from typing_extensions import Final, TypedDict -from unpaddedbase64 import decode_base64 -from zope.interface import Interface - -from twisted.internet.defer import CancelledError -from twisted.internet.interfaces import ( - IReactorCore, - IReactorPluggableNameResolver, - IReactorSSL, - IReactorTCP, - IReactorThreads, - IReactorTime, -) - -from synapse.api.errors import Codes, SynapseError -from synapse.util.cancellation import cancellable -from synapse.util.stringutils import parse_and_validate_server_name - -if TYPE_CHECKING: - from synapse.appservice.api import ApplicationService - from synapse.storage.databases.main import DataStore, PurgeEventsStore - from synapse.storage.databases.main.appservice import ApplicationServiceWorkerStore - -# Define a state map type from type/state_key to T (usually an event ID or -# event) -T = TypeVar("T") -StateKey = Tuple[str, str] -StateMap = Mapping[StateKey, T] -MutableStateMap = MutableMapping[StateKey, T] - -# JSON types. These could be made stronger, but will do for now. -# A JSON-serialisable dict. -JsonDict = Dict[str, Any] -# A JSON-serialisable mapping; roughly speaking an immutable JSONDict. -# Useful when you have a TypedDict which isn't going to be mutated and you don't want -# to cast to JsonDict everywhere. -JsonMapping = Mapping[str, Any] -# A JSON-serialisable object. -JsonSerializable = object - - -# Note that this seems to require inheriting *directly* from Interface in order -# for mypy-zope to realize it is an interface. -class ISynapseReactor( - IReactorTCP, - IReactorSSL, - IReactorPluggableNameResolver, - IReactorTime, - IReactorCore, - IReactorThreads, - Interface, -): - """The interfaces necessary for Synapse to function.""" - - -@attr.s(frozen=True, slots=True, auto_attribs=True) -class Requester: - """ - Represents the user making a request - - Attributes: - user: id of the user making the request - access_token_id: *ID* of the access token used for this - request, or None if it came via the appservice API or similar - is_guest: True if the user making this request is a guest user - shadow_banned: True if the user making this request has been shadow-banned. - device_id: device_id which was set at authentication time - app_service: the AS requesting on behalf of the user - authenticated_entity: The entity that authenticated when making the request. - This is different to the user_id when an admin user or the server is - "puppeting" the user. - """ - - user: "UserID" - access_token_id: Optional[int] - is_guest: bool - shadow_banned: bool - device_id: Optional[str] - app_service: Optional["ApplicationService"] - authenticated_entity: str - - def serialize(self) -> Dict[str, Any]: - """Converts self to a type that can be serialized as JSON, and then - deserialized by `deserialize` - - Returns: - dict - """ - return { - "user_id": self.user.to_string(), - "access_token_id": self.access_token_id, - "is_guest": self.is_guest, - "shadow_banned": self.shadow_banned, - "device_id": self.device_id, - "app_server_id": self.app_service.id if self.app_service else None, - "authenticated_entity": self.authenticated_entity, - } - - @staticmethod - def deserialize( - store: "ApplicationServiceWorkerStore", input: Dict[str, Any] - ) -> "Requester": - """Converts a dict that was produced by `serialize` back into a - Requester. - - Args: - store: Used to convert AS ID to AS object - input: A dict produced by `serialize` - - Returns: - Requester - """ - appservice = None - if input["app_server_id"]: - appservice = store.get_app_service_by_id(input["app_server_id"]) - - return Requester( - user=UserID.from_string(input["user_id"]), - access_token_id=input["access_token_id"], - is_guest=input["is_guest"], - shadow_banned=input["shadow_banned"], - device_id=input["device_id"], - app_service=appservice, - authenticated_entity=input["authenticated_entity"], - ) - - -def create_requester( - user_id: Union[str, "UserID"], - access_token_id: Optional[int] = None, - is_guest: bool = False, - shadow_banned: bool = False, - device_id: Optional[str] = None, - app_service: Optional["ApplicationService"] = None, - authenticated_entity: Optional[str] = None, -) -> Requester: - """ - Create a new ``Requester`` object - - Args: - user_id: id of the user making the request - access_token_id: *ID* of the access token used for this - request, or None if it came via the appservice API or similar - is_guest: True if the user making this request is a guest user - shadow_banned: True if the user making this request is shadow-banned. - device_id: device_id which was set at authentication time - app_service: the AS requesting on behalf of the user - authenticated_entity: The entity that authenticated when making the request. - This is different to the user_id when an admin user or the server is - "puppeting" the user. - - Returns: - Requester - """ - if not isinstance(user_id, UserID): - user_id = UserID.from_string(user_id) - - if authenticated_entity is None: - authenticated_entity = user_id.to_string() - - return Requester( - user_id, - access_token_id, - is_guest, - shadow_banned, - device_id, - app_service, - authenticated_entity, - ) - - -def get_domain_from_id(string: str) -> str: - idx = string.find(":") - if idx == -1: - raise SynapseError(400, "Invalid ID: %r" % (string,)) - return string[idx + 1 :] - - -def get_localpart_from_id(string: str) -> str: - idx = string.find(":") - if idx == -1: - raise SynapseError(400, "Invalid ID: %r" % (string,)) - return string[1:idx] - - -DS = TypeVar("DS", bound="DomainSpecificString") - - -@attr.s(slots=True, frozen=True, repr=False, auto_attribs=True) -class DomainSpecificString(metaclass=abc.ABCMeta): - """Common base class among ID/name strings that have a local part and a - domain name, prefixed with a sigil. - - Has the fields: - - 'localpart' : The local part of the name (without the leading sigil) - 'domain' : The domain part of the name - """ - - SIGIL: ClassVar[str] = abc.abstractproperty() # type: ignore - - localpart: str - domain: str - - # Because this is a frozen class, it is deeply immutable. - def __copy__(self: DS) -> DS: - return self - - def __deepcopy__(self: DS, memo: Dict[str, object]) -> DS: - return self - - @classmethod - def from_string(cls: Type[DS], s: str) -> DS: - """Parse the string given by 's' into a structure object.""" - if len(s) < 1 or s[0:1] != cls.SIGIL: - raise SynapseError( - 400, - "Expected %s string to start with '%s'" % (cls.__name__, cls.SIGIL), - Codes.INVALID_PARAM, - ) - - parts = s[1:].split(":", 1) - if len(parts) != 2: - raise SynapseError( - 400, - "Expected %s of the form '%slocalname:domain'" - % (cls.__name__, cls.SIGIL), - Codes.INVALID_PARAM, - ) - - domain = parts[1] - # This code will need changing if we want to support multiple domain - # names on one HS - return cls(localpart=parts[0], domain=domain) - - def to_string(self) -> str: - """Return a string encoding the fields of the structure object.""" - return "%s%s:%s" % (self.SIGIL, self.localpart, self.domain) - - @classmethod - def is_valid(cls: Type[DS], s: str) -> bool: - """Parses the input string and attempts to ensure it is valid.""" - # TODO: this does not reject an empty localpart or an overly-long string. - # See https://spec.matrix.org/v1.2/appendices/#identifier-grammar - try: - obj = cls.from_string(s) - # Apply additional validation to the domain. This is only done - # during is_valid (and not part of from_string) since it is - # possible for invalid data to exist in room-state, etc. - parse_and_validate_server_name(obj.domain) - return True - except Exception: - return False - - __repr__ = to_string - - -@attr.s(slots=True, frozen=True, repr=False) -class UserID(DomainSpecificString): - """Structure representing a user ID.""" - - SIGIL = "@" - - -@attr.s(slots=True, frozen=True, repr=False) -class RoomAlias(DomainSpecificString): - """Structure representing a room name.""" - - SIGIL = "#" - - -@attr.s(slots=True, frozen=True, repr=False) -class RoomID(DomainSpecificString): - """Structure representing a room id.""" - - SIGIL = "!" - - -@attr.s(slots=True, frozen=True, repr=False) -class EventID(DomainSpecificString): - """Structure representing an event id.""" - - SIGIL = "$" - - -mxid_localpart_allowed_characters = set( - "_-./=" + string.ascii_lowercase + string.digits -) - - -def contains_invalid_mxid_characters(localpart: str) -> bool: - """Check for characters not allowed in an mxid or groupid localpart - - Args: - localpart: the localpart to be checked - - Returns: - True if there are any naughty characters - """ - return any(c not in mxid_localpart_allowed_characters for c in localpart) - - -UPPER_CASE_PATTERN = re.compile(b"[A-Z_]") - -# the following is a pattern which matches '=', and bytes which are not allowed in a mxid -# localpart. -# -# It works by: -# * building a string containing the allowed characters (excluding '=') -# * escaping every special character with a backslash (to stop '-' being interpreted as a -# range operator) -# * wrapping it in a '[^...]' regex -# * converting the whole lot to a 'bytes' sequence, so that we can use it to match -# bytes rather than strings -# -NON_MXID_CHARACTER_PATTERN = re.compile( - ("[^%s]" % (re.escape("".join(mxid_localpart_allowed_characters - {"="})),)).encode( - "ascii" - ) -) - - -def map_username_to_mxid_localpart( - username: Union[str, bytes], case_sensitive: bool = False -) -> str: - """Map a username onto a string suitable for a MXID - - This follows the algorithm laid out at - https://matrix.org/docs/spec/appendices.html#mapping-from-other-character-sets. - - Args: - username: username to be mapped - case_sensitive: true if TEST and test should be mapped - onto different mxids - - Returns: - string suitable for a mxid localpart - """ - if not isinstance(username, bytes): - username = username.encode("utf-8") - - # first we sort out upper-case characters - if case_sensitive: - - def f1(m: Match[bytes]) -> bytes: - return b"_" + m.group().lower() - - username = UPPER_CASE_PATTERN.sub(f1, username) - else: - username = username.lower() - - # then we sort out non-ascii characters by converting to the hex equivalent. - def f2(m: Match[bytes]) -> bytes: - return b"=%02x" % (m.group()[0],) - - username = NON_MXID_CHARACTER_PATTERN.sub(f2, username) - - # we also do the =-escaping to mxids starting with an underscore. - username = re.sub(b"^_", b"=5f", username) - - # we should now only have ascii bytes left, so can decode back to a string. - return username.decode("ascii") - - -@attr.s(frozen=True, slots=True, order=False) -class RoomStreamToken: - """Tokens are positions between events. The token "s1" comes after event 1. - - s0 s1 - | | - [0] ▼ [1] ▼ [2] - - Tokens can either be a point in the live event stream or a cursor going - through historic events. - - When traversing the live event stream, events are ordered by - `stream_ordering` (when they arrived at the homeserver). - - When traversing historic events, events are first ordered by their `depth` - (`topological_ordering` in the event graph) and tie-broken by - `stream_ordering` (when the event arrived at the homeserver). - - If you're looking for more info about what a token with all of the - underscores means, ex. - `s2633508_17_338_6732159_1082514_541479_274711_265584_1`, see the docstring - for `StreamToken` below. - - --- - - Live tokens start with an "s" followed by the `stream_ordering` of the event - that comes before the position of the token. Said another way: - `stream_ordering` uniquely identifies a persisted event. The live token - means "the position just after the event identified by `stream_ordering`". - An example token is: - - s2633508 - - --- - - Historic tokens start with a "t" followed by the `depth` - (`topological_ordering` in the event graph) of the event that comes before - the position of the token, followed by "-", followed by the - `stream_ordering` of the event that comes before the position of the token. - An example token is: - - t426-2633508 - - --- - - There is also a third mode for live tokens where the token starts with "m", - which is sometimes used when using sharded event persisters. In this case - the events stream is considered to be a set of streams (one for each writer) - and the token encodes the vector clock of positions of each writer in their - respective streams. - - The format of the token in such case is an initial integer min position, - followed by the mapping of instance ID to position separated by '.' and '~': - - m{min_pos}~{writer1}.{pos1}~{writer2}.{pos2}. ... - - The `min_pos` corresponds to the minimum position all writers have persisted - up to, and then only writers that are ahead of that position need to be - encoded. An example token is: - - m56~2.58~3.59 - - Which corresponds to a set of three (or more writers) where instances 2 and - 3 (these are instance IDs that can be looked up in the DB to fetch the more - commonly used instance names) are at positions 58 and 59 respectively, and - all other instances are at position 56. - - Note: The `RoomStreamToken` cannot have both a topological part and an - instance map. - - --- - - For caching purposes, `RoomStreamToken`s and by extension, all their - attributes, must be hashable. - """ - - topological: Optional[int] = attr.ib( - validator=attr.validators.optional(attr.validators.instance_of(int)), - ) - stream: int = attr.ib(validator=attr.validators.instance_of(int)) - - instance_map: "frozendict[str, int]" = attr.ib( - factory=frozendict, - validator=attr.validators.deep_mapping( - key_validator=attr.validators.instance_of(str), - value_validator=attr.validators.instance_of(int), - mapping_validator=attr.validators.instance_of(frozendict), - ), - ) - - def __attrs_post_init__(self) -> None: - """Validates that both `topological` and `instance_map` aren't set.""" - - if self.instance_map and self.topological: - raise ValueError( - "Cannot set both 'topological' and 'instance_map' on 'RoomStreamToken'." - ) - - @classmethod - async def parse(cls, store: "PurgeEventsStore", string: str) -> "RoomStreamToken": - try: - if string[0] == "s": - return cls(topological=None, stream=int(string[1:])) - if string[0] == "t": - parts = string[1:].split("-", 1) - return cls(topological=int(parts[0]), stream=int(parts[1])) - if string[0] == "m": - parts = string[1:].split("~") - stream = int(parts[0]) - - instance_map = {} - for part in parts[1:]: - key, value = part.split(".") - instance_id = int(key) - pos = int(value) - - instance_name = await store.get_name_from_instance_id(instance_id) # type: ignore[attr-defined] - instance_map[instance_name] = pos - - return cls( - topological=None, - stream=stream, - instance_map=frozendict(instance_map), - ) - except CancelledError: - raise - except Exception: - pass - raise SynapseError(400, "Invalid room stream token %r" % (string,)) - - @classmethod - def parse_stream_token(cls, string: str) -> "RoomStreamToken": - try: - if string[0] == "s": - return cls(topological=None, stream=int(string[1:])) - except Exception: - pass - raise SynapseError(400, "Invalid room stream token %r" % (string,)) - - def copy_and_advance(self, other: "RoomStreamToken") -> "RoomStreamToken": - """Return a new token such that if an event is after both this token and - the other token, then its after the returned token too. - """ - - if self.topological or other.topological: - raise Exception("Can't advance topological tokens") - - max_stream = max(self.stream, other.stream) - - instance_map = { - instance: max( - self.instance_map.get(instance, self.stream), - other.instance_map.get(instance, other.stream), - ) - for instance in set(self.instance_map).union(other.instance_map) - } - - return RoomStreamToken(None, max_stream, frozendict(instance_map)) - - def as_historical_tuple(self) -> Tuple[int, int]: - """Returns a tuple of `(topological, stream)` for historical tokens. - - Raises if not an historical token (i.e. doesn't have a topological part). - """ - if self.topological is None: - raise Exception( - "Cannot call `RoomStreamToken.as_historical_tuple` on live token" - ) - - return self.topological, self.stream - - def get_stream_pos_for_instance(self, instance_name: str) -> int: - """Get the stream position that the given writer was at at this token. - - This only makes sense for "live" tokens that may have a vector clock - component, and so asserts that this is a "live" token. - """ - assert self.topological is None - - # If we don't have an entry for the instance we can assume that it was - # at `self.stream`. - return self.instance_map.get(instance_name, self.stream) - - def get_max_stream_pos(self) -> int: - """Get the maximum stream position referenced in this token. - - The corresponding "min" position is, by definition just `self.stream`. - - This is used to handle tokens that have non-empty `instance_map`, and so - reference stream positions after the `self.stream` position. - """ - return max(self.instance_map.values(), default=self.stream) - - async def to_string(self, store: "DataStore") -> str: - if self.topological is not None: - return "t%d-%d" % (self.topological, self.stream) - elif self.instance_map: - entries = [] - for name, pos in self.instance_map.items(): - instance_id = await store.get_id_for_instance(name) - entries.append(f"{instance_id}.{pos}") - - encoded_map = "~".join(entries) - return f"m{self.stream}~{encoded_map}" - else: - return "s%d" % (self.stream,) - - -class StreamKeyType: - """Known stream types. - - A stream is a list of entities ordered by an incrementing "stream token". - """ - - ROOM: Final = "room_key" - PRESENCE: Final = "presence_key" - TYPING: Final = "typing_key" - RECEIPT: Final = "receipt_key" - ACCOUNT_DATA: Final = "account_data_key" - PUSH_RULES: Final = "push_rules_key" - TO_DEVICE: Final = "to_device_key" - DEVICE_LIST: Final = "device_list_key" - - -@attr.s(slots=True, frozen=True, auto_attribs=True) -class StreamToken: - """A collection of keys joined together by underscores in the following - order and which represent the position in their respective streams. - - ex. `s2633508_17_338_6732159_1082514_541479_274711_265584_1` - 1. `room_key`: `s2633508` which is a `RoomStreamToken` - - `RoomStreamToken`'s can also look like `t426-2633508` or `m56~2.58~3.59` - - See the docstring for `RoomStreamToken` for more details. - 2. `presence_key`: `17` - 3. `typing_key`: `338` - 4. `receipt_key`: `6732159` - 5. `account_data_key`: `1082514` - 6. `push_rules_key`: `541479` - 7. `to_device_key`: `274711` - 8. `device_list_key`: `265584` - 9. `groups_key`: `1` (note that this key is now unused) - - You can see how many of these keys correspond to the various - fields in a "/sync" response: - ```json - { - "next_batch": "s12_4_0_1_1_1_1_4_1", - "presence": { - "events": [] - }, - "device_lists": { - "changed": [] - }, - "rooms": { - "join": { - "!QrZlfIDQLNLdZHqTnt:hs1": { - "timeline": { - "events": [], - "prev_batch": "s10_4_0_1_1_1_1_4_1", - "limited": false - }, - "state": { - "events": [] - }, - "account_data": { - "events": [] - }, - "ephemeral": { - "events": [] - } - } - } - } - } - ``` - - --- - - For caching purposes, `StreamToken`s and by extension, all their attributes, - must be hashable. - """ - - room_key: RoomStreamToken = attr.ib( - validator=attr.validators.instance_of(RoomStreamToken) - ) - presence_key: int - typing_key: int - receipt_key: int - account_data_key: int - push_rules_key: int - to_device_key: int - device_list_key: int - # Note that the groups key is no longer used and may have bogus values. - groups_key: int - - _SEPARATOR = "_" - START: ClassVar["StreamToken"] - - @classmethod - @cancellable - async def from_string(cls, store: "DataStore", string: str) -> "StreamToken": - """ - Creates a RoomStreamToken from its textual representation. - """ - try: - keys = string.split(cls._SEPARATOR) - while len(keys) < len(attr.fields(cls)): - # i.e. old token from before receipt_key - keys.append("0") - return cls( - await RoomStreamToken.parse(store, keys[0]), *(int(k) for k in keys[1:]) - ) - except CancelledError: - raise - except Exception: - raise SynapseError(400, "Invalid stream token") - - async def to_string(self, store: "DataStore") -> str: - return self._SEPARATOR.join( - [ - await self.room_key.to_string(store), - str(self.presence_key), - str(self.typing_key), - str(self.receipt_key), - str(self.account_data_key), - str(self.push_rules_key), - str(self.to_device_key), - str(self.device_list_key), - # Note that the groups key is no longer used, but it is still - # serialized so that there will not be confusion in the future - # if additional tokens are added. - str(self.groups_key), - ] - ) - - @property - def room_stream_id(self) -> int: - return self.room_key.stream - - def copy_and_advance(self, key: str, new_value: Any) -> "StreamToken": - """Advance the given key in the token to a new value if and only if the - new value is after the old value. - - :raises TypeError: if `key` is not the one of the keys tracked by a StreamToken. - """ - if key == StreamKeyType.ROOM: - new_token = self.copy_and_replace( - StreamKeyType.ROOM, self.room_key.copy_and_advance(new_value) - ) - return new_token - - new_token = self.copy_and_replace(key, new_value) - new_id = int(getattr(new_token, key)) - old_id = int(getattr(self, key)) - - if old_id < new_id: - return new_token - else: - return self - - def copy_and_replace(self, key: str, new_value: Any) -> "StreamToken": - return attr.evolve(self, **{key: new_value}) - - -StreamToken.START = StreamToken(RoomStreamToken(None, 0), 0, 0, 0, 0, 0, 0, 0, 0) - - -@attr.s(slots=True, frozen=True, auto_attribs=True) -class PersistedEventPosition: - """Position of a newly persisted event with instance that persisted it. - - This can be used to test whether the event is persisted before or after a - RoomStreamToken. - """ - - instance_name: str - stream: int - - def persisted_after(self, token: RoomStreamToken) -> bool: - return token.get_stream_pos_for_instance(self.instance_name) < self.stream - - def to_room_stream_token(self) -> RoomStreamToken: - """Converts the position to a room stream token such that events - persisted in the same room after this position will be after the - returned `RoomStreamToken`. - - Note: no guarantees are made about ordering w.r.t. events in other - rooms. - """ - # Doing the naive thing satisfies the desired properties described in - # the docstring. - return RoomStreamToken(None, self.stream) - - -@attr.s(slots=True, frozen=True, auto_attribs=True) -class ThirdPartyInstanceID: - appservice_id: Optional[str] - network_id: Optional[str] - - # Deny iteration because it will bite you if you try to create a singleton - # set by: - # users = set(user) - def __iter__(self) -> NoReturn: - raise ValueError("Attempted to iterate a %s" % (type(self).__name__,)) - - # Because this class is a frozen class, it is deeply immutable. - def __copy__(self) -> "ThirdPartyInstanceID": - return self - - def __deepcopy__(self, memo: Dict[str, object]) -> "ThirdPartyInstanceID": - return self - - @classmethod - def from_string(cls, s: str) -> "ThirdPartyInstanceID": - bits = s.split("|", 2) - if len(bits) != 2: - raise SynapseError(400, "Invalid ID %r" % (s,)) - - return cls(appservice_id=bits[0], network_id=bits[1]) - - def to_string(self) -> str: - return "%s|%s" % (self.appservice_id, self.network_id) - - __str__ = to_string - - -@attr.s(slots=True, frozen=True, auto_attribs=True) -class ReadReceipt: - """Information about a read-receipt""" - - room_id: str - receipt_type: str - user_id: str - event_ids: List[str] - thread_id: Optional[str] - data: JsonDict - - -@attr.s(slots=True, frozen=True, auto_attribs=True) -class DeviceListUpdates: - """ - An object containing a diff of information regarding other users' device lists, intended for - a recipient to carry out device list tracking. - - Attributes: - changed: A set of users whose device lists have changed recently. - left: A set of users who the recipient no longer needs to track the device lists of. - Typically when those users no longer share any end-to-end encryption enabled rooms. - """ - - # We need to use a factory here, otherwise `set` is not evaluated at - # object instantiation, but instead at class definition instantiation. - # The latter happening only once, thus always giving you the same sets - # across multiple DeviceListUpdates instances. - # Also see: don't define mutable default arguments. - changed: Set[str] = attr.ib(factory=set) - left: Set[str] = attr.ib(factory=set) - - def __bool__(self) -> bool: - return bool(self.changed or self.left) - - -def get_verify_key_from_cross_signing_key( - key_info: Mapping[str, Any] -) -> Tuple[str, VerifyKey]: - """Get the key ID and signedjson verify key from a cross-signing key dict - - Args: - key_info: a cross-signing key dict, which must have a "keys" - property that has exactly one item in it - - Returns: - the key ID and verify key for the cross-signing key - """ - # make sure that a `keys` field is provided - if "keys" not in key_info: - raise ValueError("Invalid key") - keys = key_info["keys"] - # and that it contains exactly one key - if len(keys) == 1: - key_id, key_data = next(iter(keys.items())) - return key_id, decode_verify_key_bytes(key_id, decode_base64(key_data)) - else: - raise ValueError("Invalid key") - - -@attr.s(auto_attribs=True, frozen=True, slots=True) -class UserInfo: - """Holds information about a user. Result of get_userinfo_by_id. - - Attributes: - user_id: ID of the user. - appservice_id: Application service ID that created this user. - consent_server_notice_sent: Version of policy documents the user has been sent. - consent_version: Version of policy documents the user has consented to. - creation_ts: Creation timestamp of the user. - is_admin: True if the user is an admin. - is_deactivated: True if the user has been deactivated. - is_guest: True if the user is a guest user. - is_shadow_banned: True if the user has been shadow-banned. - user_type: User type (None for normal user, 'support' and 'bot' other options). - """ - - user_id: UserID - appservice_id: Optional[int] - consent_server_notice_sent: Optional[str] - consent_version: Optional[str] - user_type: Optional[str] - creation_ts: int - is_admin: bool - is_deactivated: bool - is_guest: bool - is_shadow_banned: bool - - -class UserProfile(TypedDict): - user_id: str - display_name: Optional[str] - avatar_url: Optional[str] - - -@attr.s(auto_attribs=True, frozen=True, slots=True) -class RetentionPolicy: - min_lifetime: Optional[int] = None - max_lifetime: Optional[int] = None diff --git a/synapse/types/__init__.py b/synapse/types/__init__.py new file mode 100644 index 0000000000..f2d436ddc3 --- /dev/null +++ b/synapse/types/__init__.py @@ -0,0 +1,928 @@ +# Copyright 2014-2016 OpenMarket Ltd +# Copyright 2019 The Matrix.org Foundation C.I.C. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +import abc +import re +import string +from typing import ( + TYPE_CHECKING, + Any, + ClassVar, + Dict, + List, + Mapping, + Match, + MutableMapping, + NoReturn, + Optional, + Set, + Tuple, + Type, + TypeVar, + Union, +) + +import attr +from frozendict import frozendict +from signedjson.key import decode_verify_key_bytes +from signedjson.types import VerifyKey +from typing_extensions import Final, TypedDict +from unpaddedbase64 import decode_base64 +from zope.interface import Interface + +from twisted.internet.defer import CancelledError +from twisted.internet.interfaces import ( + IReactorCore, + IReactorPluggableNameResolver, + IReactorSSL, + IReactorTCP, + IReactorThreads, + IReactorTime, +) + +from synapse.api.errors import Codes, SynapseError +from synapse.util.cancellation import cancellable +from synapse.util.stringutils import parse_and_validate_server_name + +if TYPE_CHECKING: + from synapse.appservice.api import ApplicationService + from synapse.storage.databases.main import DataStore, PurgeEventsStore + from synapse.storage.databases.main.appservice import ApplicationServiceWorkerStore + +# Define a state map type from type/state_key to T (usually an event ID or +# event) +T = TypeVar("T") +StateKey = Tuple[str, str] +StateMap = Mapping[StateKey, T] +MutableStateMap = MutableMapping[StateKey, T] + +# JSON types. These could be made stronger, but will do for now. +# A JSON-serialisable dict. +JsonDict = Dict[str, Any] +# A JSON-serialisable mapping; roughly speaking an immutable JSONDict. +# Useful when you have a TypedDict which isn't going to be mutated and you don't want +# to cast to JsonDict everywhere. +JsonMapping = Mapping[str, Any] +# A JSON-serialisable object. +JsonSerializable = object + + +# Note that this seems to require inheriting *directly* from Interface in order +# for mypy-zope to realize it is an interface. +class ISynapseReactor( + IReactorTCP, + IReactorSSL, + IReactorPluggableNameResolver, + IReactorTime, + IReactorCore, + IReactorThreads, + Interface, +): + """The interfaces necessary for Synapse to function.""" + + +@attr.s(frozen=True, slots=True, auto_attribs=True) +class Requester: + """ + Represents the user making a request + + Attributes: + user: id of the user making the request + access_token_id: *ID* of the access token used for this + request, or None if it came via the appservice API or similar + is_guest: True if the user making this request is a guest user + shadow_banned: True if the user making this request has been shadow-banned. + device_id: device_id which was set at authentication time + app_service: the AS requesting on behalf of the user + authenticated_entity: The entity that authenticated when making the request. + This is different to the user_id when an admin user or the server is + "puppeting" the user. + """ + + user: "UserID" + access_token_id: Optional[int] + is_guest: bool + shadow_banned: bool + device_id: Optional[str] + app_service: Optional["ApplicationService"] + authenticated_entity: str + + def serialize(self) -> Dict[str, Any]: + """Converts self to a type that can be serialized as JSON, and then + deserialized by `deserialize` + + Returns: + dict + """ + return { + "user_id": self.user.to_string(), + "access_token_id": self.access_token_id, + "is_guest": self.is_guest, + "shadow_banned": self.shadow_banned, + "device_id": self.device_id, + "app_server_id": self.app_service.id if self.app_service else None, + "authenticated_entity": self.authenticated_entity, + } + + @staticmethod + def deserialize( + store: "ApplicationServiceWorkerStore", input: Dict[str, Any] + ) -> "Requester": + """Converts a dict that was produced by `serialize` back into a + Requester. + + Args: + store: Used to convert AS ID to AS object + input: A dict produced by `serialize` + + Returns: + Requester + """ + appservice = None + if input["app_server_id"]: + appservice = store.get_app_service_by_id(input["app_server_id"]) + + return Requester( + user=UserID.from_string(input["user_id"]), + access_token_id=input["access_token_id"], + is_guest=input["is_guest"], + shadow_banned=input["shadow_banned"], + device_id=input["device_id"], + app_service=appservice, + authenticated_entity=input["authenticated_entity"], + ) + + +def create_requester( + user_id: Union[str, "UserID"], + access_token_id: Optional[int] = None, + is_guest: bool = False, + shadow_banned: bool = False, + device_id: Optional[str] = None, + app_service: Optional["ApplicationService"] = None, + authenticated_entity: Optional[str] = None, +) -> Requester: + """ + Create a new ``Requester`` object + + Args: + user_id: id of the user making the request + access_token_id: *ID* of the access token used for this + request, or None if it came via the appservice API or similar + is_guest: True if the user making this request is a guest user + shadow_banned: True if the user making this request is shadow-banned. + device_id: device_id which was set at authentication time + app_service: the AS requesting on behalf of the user + authenticated_entity: The entity that authenticated when making the request. + This is different to the user_id when an admin user or the server is + "puppeting" the user. + + Returns: + Requester + """ + if not isinstance(user_id, UserID): + user_id = UserID.from_string(user_id) + + if authenticated_entity is None: + authenticated_entity = user_id.to_string() + + return Requester( + user_id, + access_token_id, + is_guest, + shadow_banned, + device_id, + app_service, + authenticated_entity, + ) + + +def get_domain_from_id(string: str) -> str: + idx = string.find(":") + if idx == -1: + raise SynapseError(400, "Invalid ID: %r" % (string,)) + return string[idx + 1 :] + + +def get_localpart_from_id(string: str) -> str: + idx = string.find(":") + if idx == -1: + raise SynapseError(400, "Invalid ID: %r" % (string,)) + return string[1:idx] + + +DS = TypeVar("DS", bound="DomainSpecificString") + + +@attr.s(slots=True, frozen=True, repr=False, auto_attribs=True) +class DomainSpecificString(metaclass=abc.ABCMeta): + """Common base class among ID/name strings that have a local part and a + domain name, prefixed with a sigil. + + Has the fields: + + 'localpart' : The local part of the name (without the leading sigil) + 'domain' : The domain part of the name + """ + + SIGIL: ClassVar[str] = abc.abstractproperty() # type: ignore + + localpart: str + domain: str + + # Because this is a frozen class, it is deeply immutable. + def __copy__(self: DS) -> DS: + return self + + def __deepcopy__(self: DS, memo: Dict[str, object]) -> DS: + return self + + @classmethod + def from_string(cls: Type[DS], s: str) -> DS: + """Parse the string given by 's' into a structure object.""" + if len(s) < 1 or s[0:1] != cls.SIGIL: + raise SynapseError( + 400, + "Expected %s string to start with '%s'" % (cls.__name__, cls.SIGIL), + Codes.INVALID_PARAM, + ) + + parts = s[1:].split(":", 1) + if len(parts) != 2: + raise SynapseError( + 400, + "Expected %s of the form '%slocalname:domain'" + % (cls.__name__, cls.SIGIL), + Codes.INVALID_PARAM, + ) + + domain = parts[1] + # This code will need changing if we want to support multiple domain + # names on one HS + return cls(localpart=parts[0], domain=domain) + + def to_string(self) -> str: + """Return a string encoding the fields of the structure object.""" + return "%s%s:%s" % (self.SIGIL, self.localpart, self.domain) + + @classmethod + def is_valid(cls: Type[DS], s: str) -> bool: + """Parses the input string and attempts to ensure it is valid.""" + # TODO: this does not reject an empty localpart or an overly-long string. + # See https://spec.matrix.org/v1.2/appendices/#identifier-grammar + try: + obj = cls.from_string(s) + # Apply additional validation to the domain. This is only done + # during is_valid (and not part of from_string) since it is + # possible for invalid data to exist in room-state, etc. + parse_and_validate_server_name(obj.domain) + return True + except Exception: + return False + + __repr__ = to_string + + +@attr.s(slots=True, frozen=True, repr=False) +class UserID(DomainSpecificString): + """Structure representing a user ID.""" + + SIGIL = "@" + + +@attr.s(slots=True, frozen=True, repr=False) +class RoomAlias(DomainSpecificString): + """Structure representing a room name.""" + + SIGIL = "#" + + +@attr.s(slots=True, frozen=True, repr=False) +class RoomID(DomainSpecificString): + """Structure representing a room id.""" + + SIGIL = "!" + + +@attr.s(slots=True, frozen=True, repr=False) +class EventID(DomainSpecificString): + """Structure representing an event id.""" + + SIGIL = "$" + + +mxid_localpart_allowed_characters = set( + "_-./=" + string.ascii_lowercase + string.digits +) + + +def contains_invalid_mxid_characters(localpart: str) -> bool: + """Check for characters not allowed in an mxid or groupid localpart + + Args: + localpart: the localpart to be checked + + Returns: + True if there are any naughty characters + """ + return any(c not in mxid_localpart_allowed_characters for c in localpart) + + +UPPER_CASE_PATTERN = re.compile(b"[A-Z_]") + +# the following is a pattern which matches '=', and bytes which are not allowed in a mxid +# localpart. +# +# It works by: +# * building a string containing the allowed characters (excluding '=') +# * escaping every special character with a backslash (to stop '-' being interpreted as a +# range operator) +# * wrapping it in a '[^...]' regex +# * converting the whole lot to a 'bytes' sequence, so that we can use it to match +# bytes rather than strings +# +NON_MXID_CHARACTER_PATTERN = re.compile( + ("[^%s]" % (re.escape("".join(mxid_localpart_allowed_characters - {"="})),)).encode( + "ascii" + ) +) + + +def map_username_to_mxid_localpart( + username: Union[str, bytes], case_sensitive: bool = False +) -> str: + """Map a username onto a string suitable for a MXID + + This follows the algorithm laid out at + https://matrix.org/docs/spec/appendices.html#mapping-from-other-character-sets. + + Args: + username: username to be mapped + case_sensitive: true if TEST and test should be mapped + onto different mxids + + Returns: + string suitable for a mxid localpart + """ + if not isinstance(username, bytes): + username = username.encode("utf-8") + + # first we sort out upper-case characters + if case_sensitive: + + def f1(m: Match[bytes]) -> bytes: + return b"_" + m.group().lower() + + username = UPPER_CASE_PATTERN.sub(f1, username) + else: + username = username.lower() + + # then we sort out non-ascii characters by converting to the hex equivalent. + def f2(m: Match[bytes]) -> bytes: + return b"=%02x" % (m.group()[0],) + + username = NON_MXID_CHARACTER_PATTERN.sub(f2, username) + + # we also do the =-escaping to mxids starting with an underscore. + username = re.sub(b"^_", b"=5f", username) + + # we should now only have ascii bytes left, so can decode back to a string. + return username.decode("ascii") + + +@attr.s(frozen=True, slots=True, order=False) +class RoomStreamToken: + """Tokens are positions between events. The token "s1" comes after event 1. + + s0 s1 + | | + [0] ▼ [1] ▼ [2] + + Tokens can either be a point in the live event stream or a cursor going + through historic events. + + When traversing the live event stream, events are ordered by + `stream_ordering` (when they arrived at the homeserver). + + When traversing historic events, events are first ordered by their `depth` + (`topological_ordering` in the event graph) and tie-broken by + `stream_ordering` (when the event arrived at the homeserver). + + If you're looking for more info about what a token with all of the + underscores means, ex. + `s2633508_17_338_6732159_1082514_541479_274711_265584_1`, see the docstring + for `StreamToken` below. + + --- + + Live tokens start with an "s" followed by the `stream_ordering` of the event + that comes before the position of the token. Said another way: + `stream_ordering` uniquely identifies a persisted event. The live token + means "the position just after the event identified by `stream_ordering`". + An example token is: + + s2633508 + + --- + + Historic tokens start with a "t" followed by the `depth` + (`topological_ordering` in the event graph) of the event that comes before + the position of the token, followed by "-", followed by the + `stream_ordering` of the event that comes before the position of the token. + An example token is: + + t426-2633508 + + --- + + There is also a third mode for live tokens where the token starts with "m", + which is sometimes used when using sharded event persisters. In this case + the events stream is considered to be a set of streams (one for each writer) + and the token encodes the vector clock of positions of each writer in their + respective streams. + + The format of the token in such case is an initial integer min position, + followed by the mapping of instance ID to position separated by '.' and '~': + + m{min_pos}~{writer1}.{pos1}~{writer2}.{pos2}. ... + + The `min_pos` corresponds to the minimum position all writers have persisted + up to, and then only writers that are ahead of that position need to be + encoded. An example token is: + + m56~2.58~3.59 + + Which corresponds to a set of three (or more writers) where instances 2 and + 3 (these are instance IDs that can be looked up in the DB to fetch the more + commonly used instance names) are at positions 58 and 59 respectively, and + all other instances are at position 56. + + Note: The `RoomStreamToken` cannot have both a topological part and an + instance map. + + --- + + For caching purposes, `RoomStreamToken`s and by extension, all their + attributes, must be hashable. + """ + + topological: Optional[int] = attr.ib( + validator=attr.validators.optional(attr.validators.instance_of(int)), + ) + stream: int = attr.ib(validator=attr.validators.instance_of(int)) + + instance_map: "frozendict[str, int]" = attr.ib( + factory=frozendict, + validator=attr.validators.deep_mapping( + key_validator=attr.validators.instance_of(str), + value_validator=attr.validators.instance_of(int), + mapping_validator=attr.validators.instance_of(frozendict), + ), + ) + + def __attrs_post_init__(self) -> None: + """Validates that both `topological` and `instance_map` aren't set.""" + + if self.instance_map and self.topological: + raise ValueError( + "Cannot set both 'topological' and 'instance_map' on 'RoomStreamToken'." + ) + + @classmethod + async def parse(cls, store: "PurgeEventsStore", string: str) -> "RoomStreamToken": + try: + if string[0] == "s": + return cls(topological=None, stream=int(string[1:])) + if string[0] == "t": + parts = string[1:].split("-", 1) + return cls(topological=int(parts[0]), stream=int(parts[1])) + if string[0] == "m": + parts = string[1:].split("~") + stream = int(parts[0]) + + instance_map = {} + for part in parts[1:]: + key, value = part.split(".") + instance_id = int(key) + pos = int(value) + + instance_name = await store.get_name_from_instance_id(instance_id) # type: ignore[attr-defined] + instance_map[instance_name] = pos + + return cls( + topological=None, + stream=stream, + instance_map=frozendict(instance_map), + ) + except CancelledError: + raise + except Exception: + pass + raise SynapseError(400, "Invalid room stream token %r" % (string,)) + + @classmethod + def parse_stream_token(cls, string: str) -> "RoomStreamToken": + try: + if string[0] == "s": + return cls(topological=None, stream=int(string[1:])) + except Exception: + pass + raise SynapseError(400, "Invalid room stream token %r" % (string,)) + + def copy_and_advance(self, other: "RoomStreamToken") -> "RoomStreamToken": + """Return a new token such that if an event is after both this token and + the other token, then its after the returned token too. + """ + + if self.topological or other.topological: + raise Exception("Can't advance topological tokens") + + max_stream = max(self.stream, other.stream) + + instance_map = { + instance: max( + self.instance_map.get(instance, self.stream), + other.instance_map.get(instance, other.stream), + ) + for instance in set(self.instance_map).union(other.instance_map) + } + + return RoomStreamToken(None, max_stream, frozendict(instance_map)) + + def as_historical_tuple(self) -> Tuple[int, int]: + """Returns a tuple of `(topological, stream)` for historical tokens. + + Raises if not an historical token (i.e. doesn't have a topological part). + """ + if self.topological is None: + raise Exception( + "Cannot call `RoomStreamToken.as_historical_tuple` on live token" + ) + + return self.topological, self.stream + + def get_stream_pos_for_instance(self, instance_name: str) -> int: + """Get the stream position that the given writer was at at this token. + + This only makes sense for "live" tokens that may have a vector clock + component, and so asserts that this is a "live" token. + """ + assert self.topological is None + + # If we don't have an entry for the instance we can assume that it was + # at `self.stream`. + return self.instance_map.get(instance_name, self.stream) + + def get_max_stream_pos(self) -> int: + """Get the maximum stream position referenced in this token. + + The corresponding "min" position is, by definition just `self.stream`. + + This is used to handle tokens that have non-empty `instance_map`, and so + reference stream positions after the `self.stream` position. + """ + return max(self.instance_map.values(), default=self.stream) + + async def to_string(self, store: "DataStore") -> str: + if self.topological is not None: + return "t%d-%d" % (self.topological, self.stream) + elif self.instance_map: + entries = [] + for name, pos in self.instance_map.items(): + instance_id = await store.get_id_for_instance(name) + entries.append(f"{instance_id}.{pos}") + + encoded_map = "~".join(entries) + return f"m{self.stream}~{encoded_map}" + else: + return "s%d" % (self.stream,) + + +class StreamKeyType: + """Known stream types. + + A stream is a list of entities ordered by an incrementing "stream token". + """ + + ROOM: Final = "room_key" + PRESENCE: Final = "presence_key" + TYPING: Final = "typing_key" + RECEIPT: Final = "receipt_key" + ACCOUNT_DATA: Final = "account_data_key" + PUSH_RULES: Final = "push_rules_key" + TO_DEVICE: Final = "to_device_key" + DEVICE_LIST: Final = "device_list_key" + + +@attr.s(slots=True, frozen=True, auto_attribs=True) +class StreamToken: + """A collection of keys joined together by underscores in the following + order and which represent the position in their respective streams. + + ex. `s2633508_17_338_6732159_1082514_541479_274711_265584_1` + 1. `room_key`: `s2633508` which is a `RoomStreamToken` + - `RoomStreamToken`'s can also look like `t426-2633508` or `m56~2.58~3.59` + - See the docstring for `RoomStreamToken` for more details. + 2. `presence_key`: `17` + 3. `typing_key`: `338` + 4. `receipt_key`: `6732159` + 5. `account_data_key`: `1082514` + 6. `push_rules_key`: `541479` + 7. `to_device_key`: `274711` + 8. `device_list_key`: `265584` + 9. `groups_key`: `1` (note that this key is now unused) + + You can see how many of these keys correspond to the various + fields in a "/sync" response: + ```json + { + "next_batch": "s12_4_0_1_1_1_1_4_1", + "presence": { + "events": [] + }, + "device_lists": { + "changed": [] + }, + "rooms": { + "join": { + "!QrZlfIDQLNLdZHqTnt:hs1": { + "timeline": { + "events": [], + "prev_batch": "s10_4_0_1_1_1_1_4_1", + "limited": false + }, + "state": { + "events": [] + }, + "account_data": { + "events": [] + }, + "ephemeral": { + "events": [] + } + } + } + } + } + ``` + + --- + + For caching purposes, `StreamToken`s and by extension, all their attributes, + must be hashable. + """ + + room_key: RoomStreamToken = attr.ib( + validator=attr.validators.instance_of(RoomStreamToken) + ) + presence_key: int + typing_key: int + receipt_key: int + account_data_key: int + push_rules_key: int + to_device_key: int + device_list_key: int + # Note that the groups key is no longer used and may have bogus values. + groups_key: int + + _SEPARATOR = "_" + START: ClassVar["StreamToken"] + + @classmethod + @cancellable + async def from_string(cls, store: "DataStore", string: str) -> "StreamToken": + """ + Creates a RoomStreamToken from its textual representation. + """ + try: + keys = string.split(cls._SEPARATOR) + while len(keys) < len(attr.fields(cls)): + # i.e. old token from before receipt_key + keys.append("0") + return cls( + await RoomStreamToken.parse(store, keys[0]), *(int(k) for k in keys[1:]) + ) + except CancelledError: + raise + except Exception: + raise SynapseError(400, "Invalid stream token") + + async def to_string(self, store: "DataStore") -> str: + return self._SEPARATOR.join( + [ + await self.room_key.to_string(store), + str(self.presence_key), + str(self.typing_key), + str(self.receipt_key), + str(self.account_data_key), + str(self.push_rules_key), + str(self.to_device_key), + str(self.device_list_key), + # Note that the groups key is no longer used, but it is still + # serialized so that there will not be confusion in the future + # if additional tokens are added. + str(self.groups_key), + ] + ) + + @property + def room_stream_id(self) -> int: + return self.room_key.stream + + def copy_and_advance(self, key: str, new_value: Any) -> "StreamToken": + """Advance the given key in the token to a new value if and only if the + new value is after the old value. + + :raises TypeError: if `key` is not the one of the keys tracked by a StreamToken. + """ + if key == StreamKeyType.ROOM: + new_token = self.copy_and_replace( + StreamKeyType.ROOM, self.room_key.copy_and_advance(new_value) + ) + return new_token + + new_token = self.copy_and_replace(key, new_value) + new_id = int(getattr(new_token, key)) + old_id = int(getattr(self, key)) + + if old_id < new_id: + return new_token + else: + return self + + def copy_and_replace(self, key: str, new_value: Any) -> "StreamToken": + return attr.evolve(self, **{key: new_value}) + + +StreamToken.START = StreamToken(RoomStreamToken(None, 0), 0, 0, 0, 0, 0, 0, 0, 0) + + +@attr.s(slots=True, frozen=True, auto_attribs=True) +class PersistedEventPosition: + """Position of a newly persisted event with instance that persisted it. + + This can be used to test whether the event is persisted before or after a + RoomStreamToken. + """ + + instance_name: str + stream: int + + def persisted_after(self, token: RoomStreamToken) -> bool: + return token.get_stream_pos_for_instance(self.instance_name) < self.stream + + def to_room_stream_token(self) -> RoomStreamToken: + """Converts the position to a room stream token such that events + persisted in the same room after this position will be after the + returned `RoomStreamToken`. + + Note: no guarantees are made about ordering w.r.t. events in other + rooms. + """ + # Doing the naive thing satisfies the desired properties described in + # the docstring. + return RoomStreamToken(None, self.stream) + + +@attr.s(slots=True, frozen=True, auto_attribs=True) +class ThirdPartyInstanceID: + appservice_id: Optional[str] + network_id: Optional[str] + + # Deny iteration because it will bite you if you try to create a singleton + # set by: + # users = set(user) + def __iter__(self) -> NoReturn: + raise ValueError("Attempted to iterate a %s" % (type(self).__name__,)) + + # Because this class is a frozen class, it is deeply immutable. + def __copy__(self) -> "ThirdPartyInstanceID": + return self + + def __deepcopy__(self, memo: Dict[str, object]) -> "ThirdPartyInstanceID": + return self + + @classmethod + def from_string(cls, s: str) -> "ThirdPartyInstanceID": + bits = s.split("|", 2) + if len(bits) != 2: + raise SynapseError(400, "Invalid ID %r" % (s,)) + + return cls(appservice_id=bits[0], network_id=bits[1]) + + def to_string(self) -> str: + return "%s|%s" % (self.appservice_id, self.network_id) + + __str__ = to_string + + +@attr.s(slots=True, frozen=True, auto_attribs=True) +class ReadReceipt: + """Information about a read-receipt""" + + room_id: str + receipt_type: str + user_id: str + event_ids: List[str] + thread_id: Optional[str] + data: JsonDict + + +@attr.s(slots=True, frozen=True, auto_attribs=True) +class DeviceListUpdates: + """ + An object containing a diff of information regarding other users' device lists, intended for + a recipient to carry out device list tracking. + + Attributes: + changed: A set of users whose device lists have changed recently. + left: A set of users who the recipient no longer needs to track the device lists of. + Typically when those users no longer share any end-to-end encryption enabled rooms. + """ + + # We need to use a factory here, otherwise `set` is not evaluated at + # object instantiation, but instead at class definition instantiation. + # The latter happening only once, thus always giving you the same sets + # across multiple DeviceListUpdates instances. + # Also see: don't define mutable default arguments. + changed: Set[str] = attr.ib(factory=set) + left: Set[str] = attr.ib(factory=set) + + def __bool__(self) -> bool: + return bool(self.changed or self.left) + + +def get_verify_key_from_cross_signing_key( + key_info: Mapping[str, Any] +) -> Tuple[str, VerifyKey]: + """Get the key ID and signedjson verify key from a cross-signing key dict + + Args: + key_info: a cross-signing key dict, which must have a "keys" + property that has exactly one item in it + + Returns: + the key ID and verify key for the cross-signing key + """ + # make sure that a `keys` field is provided + if "keys" not in key_info: + raise ValueError("Invalid key") + keys = key_info["keys"] + # and that it contains exactly one key + if len(keys) == 1: + key_id, key_data = next(iter(keys.items())) + return key_id, decode_verify_key_bytes(key_id, decode_base64(key_data)) + else: + raise ValueError("Invalid key") + + +@attr.s(auto_attribs=True, frozen=True, slots=True) +class UserInfo: + """Holds information about a user. Result of get_userinfo_by_id. + + Attributes: + user_id: ID of the user. + appservice_id: Application service ID that created this user. + consent_server_notice_sent: Version of policy documents the user has been sent. + consent_version: Version of policy documents the user has consented to. + creation_ts: Creation timestamp of the user. + is_admin: True if the user is an admin. + is_deactivated: True if the user has been deactivated. + is_guest: True if the user is a guest user. + is_shadow_banned: True if the user has been shadow-banned. + user_type: User type (None for normal user, 'support' and 'bot' other options). + """ + + user_id: UserID + appservice_id: Optional[int] + consent_server_notice_sent: Optional[str] + consent_version: Optional[str] + user_type: Optional[str] + creation_ts: int + is_admin: bool + is_deactivated: bool + is_guest: bool + is_shadow_banned: bool + + +class UserProfile(TypedDict): + user_id: str + display_name: Optional[str] + avatar_url: Optional[str] + + +@attr.s(auto_attribs=True, frozen=True, slots=True) +class RetentionPolicy: + min_lifetime: Optional[int] = None + max_lifetime: Optional[int] = None diff --git a/synapse/types/state.py b/synapse/types/state.py new file mode 100644 index 0000000000..0004d955b4 --- /dev/null +++ b/synapse/types/state.py @@ -0,0 +1,567 @@ +# Copyright 2014-2016 OpenMarket Ltd +# Copyright 2022 The Matrix.org Foundation C.I.C. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +import logging +from typing import ( + TYPE_CHECKING, + Callable, + Collection, + Dict, + Iterable, + List, + Mapping, + Optional, + Set, + Tuple, + TypeVar, +) + +import attr +from frozendict import frozendict + +from synapse.api.constants import EventTypes +from synapse.types import MutableStateMap, StateKey, StateMap + +if TYPE_CHECKING: + from typing import FrozenSet # noqa: used within quoted type hint; flake8 sad + + +logger = logging.getLogger(__name__) + +# Used for generic functions below +T = TypeVar("T") + + +@attr.s(slots=True, frozen=True, auto_attribs=True) +class StateFilter: + """A filter used when querying for state. + + Attributes: + types: Map from type to set of state keys (or None). This specifies + which state_keys for the given type to fetch from the DB. If None + then all events with that type are fetched. If the set is empty + then no events with that type are fetched. + include_others: Whether to fetch events with types that do not + appear in `types`. + """ + + types: "frozendict[str, Optional[FrozenSet[str]]]" + include_others: bool = False + + def __attrs_post_init__(self) -> None: + # If `include_others` is set we canonicalise the filter by removing + # wildcards from the types dictionary + if self.include_others: + # this is needed to work around the fact that StateFilter is frozen + object.__setattr__( + self, + "types", + frozendict({k: v for k, v in self.types.items() if v is not None}), + ) + + @staticmethod + def all() -> "StateFilter": + """Returns a filter that fetches everything. + + Returns: + The state filter. + """ + return _ALL_STATE_FILTER + + @staticmethod + def none() -> "StateFilter": + """Returns a filter that fetches nothing. + + Returns: + The new state filter. + """ + return _NONE_STATE_FILTER + + @staticmethod + def from_types(types: Iterable[Tuple[str, Optional[str]]]) -> "StateFilter": + """Creates a filter that only fetches the given types + + Args: + types: A list of type and state keys to fetch. A state_key of None + fetches everything for that type + + Returns: + The new state filter. + """ + type_dict: Dict[str, Optional[Set[str]]] = {} + for typ, s in types: + if typ in type_dict: + if type_dict[typ] is None: + continue + + if s is None: + type_dict[typ] = None + continue + + type_dict.setdefault(typ, set()).add(s) # type: ignore + + return StateFilter( + types=frozendict( + (k, frozenset(v) if v is not None else None) + for k, v in type_dict.items() + ) + ) + + @staticmethod + def from_lazy_load_member_list(members: Iterable[str]) -> "StateFilter": + """Creates a filter that returns all non-member events, plus the member + events for the given users + + Args: + members: Set of user IDs + + Returns: + The new state filter + """ + return StateFilter( + types=frozendict({EventTypes.Member: frozenset(members)}), + include_others=True, + ) + + @staticmethod + def freeze( + types: Mapping[str, Optional[Collection[str]]], include_others: bool + ) -> "StateFilter": + """ + Returns a (frozen) StateFilter with the same contents as the parameters + specified here, which can be made of mutable types. + """ + types_with_frozen_values: Dict[str, Optional[FrozenSet[str]]] = {} + for state_types, state_keys in types.items(): + if state_keys is not None: + types_with_frozen_values[state_types] = frozenset(state_keys) + else: + types_with_frozen_values[state_types] = None + + return StateFilter( + frozendict(types_with_frozen_values), include_others=include_others + ) + + def return_expanded(self) -> "StateFilter": + """Creates a new StateFilter where type wild cards have been removed + (except for memberships). The returned filter is a superset of the + current one, i.e. anything that passes the current filter will pass + the returned filter. + + This helps the caching as the DictionaryCache knows if it has *all* the + state, but does not know if it has all of the keys of a particular type, + which makes wildcard lookups expensive unless we have a complete cache. + Hence, if we are doing a wildcard lookup, populate the cache fully so + that we can do an efficient lookup next time. + + Note that since we have two caches, one for membership events and one for + other events, we can be a bit more clever than simply returning + `StateFilter.all()` if `has_wildcards()` is True. + + We return a StateFilter where: + 1. the list of membership events to return is the same + 2. if there is a wildcard that matches non-member events we + return all non-member events + + Returns: + The new state filter. + """ + + if self.is_full(): + # If we're going to return everything then there's nothing to do + return self + + if not self.has_wildcards(): + # If there are no wild cards, there's nothing to do + return self + + if EventTypes.Member in self.types: + get_all_members = self.types[EventTypes.Member] is None + else: + get_all_members = self.include_others + + has_non_member_wildcard = self.include_others or any( + state_keys is None + for t, state_keys in self.types.items() + if t != EventTypes.Member + ) + + if not has_non_member_wildcard: + # If there are no non-member wild cards we can just return ourselves + return self + + if get_all_members: + # We want to return everything. + return StateFilter.all() + elif EventTypes.Member in self.types: + # We want to return all non-members, but only particular + # memberships + return StateFilter( + types=frozendict({EventTypes.Member: self.types[EventTypes.Member]}), + include_others=True, + ) + else: + # We want to return all non-members + return _ALL_NON_MEMBER_STATE_FILTER + + def make_sql_filter_clause(self) -> Tuple[str, List[str]]: + """Converts the filter to an SQL clause. + + For example: + + f = StateFilter.from_types([("m.room.create", "")]) + clause, args = f.make_sql_filter_clause() + clause == "(type = ? AND state_key = ?)" + args == ['m.room.create', ''] + + + Returns: + The SQL string (may be empty) and arguments. An empty SQL string is + returned when the filter matches everything (i.e. is "full"). + """ + + where_clause = "" + where_args: List[str] = [] + + if self.is_full(): + return where_clause, where_args + + if not self.include_others and not self.types: + # i.e. this is an empty filter, so we need to return a clause that + # will match nothing + return "1 = 2", [] + + # First we build up a lost of clauses for each type/state_key combo + clauses = [] + for etype, state_keys in self.types.items(): + if state_keys is None: + clauses.append("(type = ?)") + where_args.append(etype) + continue + + for state_key in state_keys: + clauses.append("(type = ? AND state_key = ?)") + where_args.extend((etype, state_key)) + + # This will match anything that appears in `self.types` + where_clause = " OR ".join(clauses) + + # If we want to include stuff that's not in the types dict then we add + # a `OR type NOT IN (...)` clause to the end. + if self.include_others: + if where_clause: + where_clause += " OR " + + where_clause += "type NOT IN (%s)" % (",".join(["?"] * len(self.types)),) + where_args.extend(self.types) + + return where_clause, where_args + + def max_entries_returned(self) -> Optional[int]: + """Returns the maximum number of entries this filter will return if + known, otherwise returns None. + + For example a simple state filter asking for `("m.room.create", "")` + will return 1, whereas the default state filter will return None. + + This is used to bail out early if the right number of entries have been + fetched. + """ + if self.has_wildcards(): + return None + + return len(self.concrete_types()) + + def filter_state(self, state_dict: StateMap[T]) -> MutableStateMap[T]: + """Returns the state filtered with by this StateFilter. + + Args: + state: The state map to filter + + Returns: + The filtered state map. + This is a copy, so it's safe to mutate. + """ + if self.is_full(): + return dict(state_dict) + + filtered_state = {} + for k, v in state_dict.items(): + typ, state_key = k + if typ in self.types: + state_keys = self.types[typ] + if state_keys is None or state_key in state_keys: + filtered_state[k] = v + elif self.include_others: + filtered_state[k] = v + + return filtered_state + + def is_full(self) -> bool: + """Whether this filter fetches everything or not + + Returns: + True if the filter fetches everything. + """ + return self.include_others and not self.types + + def has_wildcards(self) -> bool: + """Whether the filter includes wildcards or is attempting to fetch + specific state. + + Returns: + True if the filter includes wildcards. + """ + + return self.include_others or any( + state_keys is None for state_keys in self.types.values() + ) + + def concrete_types(self) -> List[Tuple[str, str]]: + """Returns a list of concrete type/state_keys (i.e. not None) that + will be fetched. This will be a complete list if `has_wildcards` + returns False, but otherwise will be a subset (or even empty). + + Returns: + A list of type/state_keys tuples. + """ + return [ + (t, s) + for t, state_keys in self.types.items() + if state_keys is not None + for s in state_keys + ] + + def get_member_split(self) -> Tuple["StateFilter", "StateFilter"]: + """Return the filter split into two: one which assumes it's exclusively + matching against member state, and one which assumes it's matching + against non member state. + + This is useful due to the returned filters giving correct results for + `is_full()`, `has_wildcards()`, etc, when operating against maps that + either exclusively contain member events or only contain non-member + events. (Which is the case when dealing with the member vs non-member + state caches). + + Returns: + The member and non member filters + """ + + if EventTypes.Member in self.types: + state_keys = self.types[EventTypes.Member] + if state_keys is None: + member_filter = StateFilter.all() + else: + member_filter = StateFilter(frozendict({EventTypes.Member: state_keys})) + elif self.include_others: + member_filter = StateFilter.all() + else: + member_filter = StateFilter.none() + + non_member_filter = StateFilter( + types=frozendict( + {k: v for k, v in self.types.items() if k != EventTypes.Member} + ), + include_others=self.include_others, + ) + + return member_filter, non_member_filter + + def _decompose_into_four_parts( + self, + ) -> Tuple[Tuple[bool, Set[str]], Tuple[Set[str], Set[StateKey]]]: + """ + Decomposes this state filter into 4 constituent parts, which can be + thought of as this: + all? - minus_wildcards + plus_wildcards + plus_state_keys + + where + * all represents ALL state + * minus_wildcards represents entire state types to remove + * plus_wildcards represents entire state types to add + * plus_state_keys represents individual state keys to add + + See `recompose_from_four_parts` for the other direction of this + correspondence. + """ + is_all = self.include_others + excluded_types: Set[str] = {t for t in self.types if is_all} + wildcard_types: Set[str] = {t for t, s in self.types.items() if s is None} + concrete_keys: Set[StateKey] = set(self.concrete_types()) + + return (is_all, excluded_types), (wildcard_types, concrete_keys) + + @staticmethod + def _recompose_from_four_parts( + all_part: bool, + minus_wildcards: Set[str], + plus_wildcards: Set[str], + plus_state_keys: Set[StateKey], + ) -> "StateFilter": + """ + Recomposes a state filter from 4 parts. + + See `decompose_into_four_parts` (the other direction of this + correspondence) for descriptions on each of the parts. + """ + + # {state type -> set of state keys OR None for wildcard} + # (The same structure as that of a StateFilter.) + new_types: Dict[str, Optional[Set[str]]] = {} + + # if we start with all, insert the excluded statetypes as empty sets + # to prevent them from being included + if all_part: + new_types.update({state_type: set() for state_type in minus_wildcards}) + + # insert the plus wildcards + new_types.update({state_type: None for state_type in plus_wildcards}) + + # insert the specific state keys + for state_type, state_key in plus_state_keys: + if state_type in new_types: + entry = new_types[state_type] + if entry is not None: + entry.add(state_key) + elif not all_part: + # don't insert if the entire type is already included by + # include_others as this would actually shrink the state allowed + # by this filter. + new_types[state_type] = {state_key} + + return StateFilter.freeze(new_types, include_others=all_part) + + def approx_difference(self, other: "StateFilter") -> "StateFilter": + """ + Returns a state filter which represents `self - other`. + + This is useful for determining what state remains to be pulled out of the + database if we want the state included by `self` but already have the state + included by `other`. + + The returned state filter + - MUST include all state events that are included by this filter (`self`) + unless they are included by `other`; + - MUST NOT include state events not included by this filter (`self`); and + - MAY be an over-approximation: the returned state filter + MAY additionally include some state events from `other`. + + This implementation attempts to return the narrowest such state filter. + In the case that `self` contains wildcards for state types where + `other` contains specific state keys, an approximation must be made: + the returned state filter keeps the wildcard, as state filters are not + able to express 'all state keys except some given examples'. + e.g. + StateFilter(m.room.member -> None (wildcard)) + minus + StateFilter(m.room.member -> {'@wombat:example.org'}) + is approximated as + StateFilter(m.room.member -> None (wildcard)) + """ + + # We first transform self and other into an alternative representation: + # - whether or not they include all events to begin with ('all') + # - if so, which event types are excluded? ('excludes') + # - which entire event types to include ('wildcards') + # - which concrete state keys to include ('concrete state keys') + (self_all, self_excludes), ( + self_wildcards, + self_concrete_keys, + ) = self._decompose_into_four_parts() + (other_all, other_excludes), ( + other_wildcards, + other_concrete_keys, + ) = other._decompose_into_four_parts() + + # Start with an estimate of the difference based on self + new_all = self_all + # Wildcards from the other can be added to the exclusion filter + new_excludes = self_excludes | other_wildcards + # We remove wildcards that appeared as wildcards in the other + new_wildcards = self_wildcards - other_wildcards + # We filter out the concrete state keys that appear in the other + # as wildcards or concrete state keys. + new_concrete_keys = { + (state_type, state_key) + for (state_type, state_key) in self_concrete_keys + if state_type not in other_wildcards + } - other_concrete_keys + + if other_all: + if self_all: + # If self starts with all, then we add as wildcards any + # types which appear in the other's exclusion filter (but + # aren't in the self exclusion filter). This is as the other + # filter will return everything BUT the types in its exclusion, so + # we need to add those excluded types that also match the self + # filter as wildcard types in the new filter. + new_wildcards |= other_excludes.difference(self_excludes) + + # If other is an `include_others` then the difference isn't. + new_all = False + # (We have no need for excludes when we don't start with all, as there + # is nothing to exclude.) + new_excludes = set() + + # We also filter out all state types that aren't in the exclusion + # list of the other. + new_wildcards &= other_excludes + new_concrete_keys = { + (state_type, state_key) + for (state_type, state_key) in new_concrete_keys + if state_type in other_excludes + } + + # Transform our newly-constructed state filter from the alternative + # representation back into the normal StateFilter representation. + return StateFilter._recompose_from_four_parts( + new_all, new_excludes, new_wildcards, new_concrete_keys + ) + + def must_await_full_state(self, is_mine_id: Callable[[str], bool]) -> bool: + """Check if we need to wait for full state to complete to calculate this state + + If we have a state filter which is completely satisfied even with partial + state, then we don't need to await_full_state before we can return it. + + Args: + is_mine_id: a callable which confirms if a given state_key matches a mxid + of a local user + """ + # if we haven't requested membership events, then it depends on the value of + # 'include_others' + if EventTypes.Member not in self.types: + return self.include_others + + # if we're looking for *all* membership events, then we have to wait + member_state_keys = self.types[EventTypes.Member] + if member_state_keys is None: + return True + + # otherwise, consider whose membership we are looking for. If it's entirely + # local users, then we don't need to wait. + for state_key in member_state_keys: + if not is_mine_id(state_key): + # remote user + return True + + # local users only + return False + + +_ALL_STATE_FILTER = StateFilter(types=frozendict(), include_others=True) +_ALL_NON_MEMBER_STATE_FILTER = StateFilter( + types=frozendict({EventTypes.Member: frozenset()}), include_others=True +) +_NONE_STATE_FILTER = StateFilter(types=frozendict(), include_others=False) diff --git a/synapse/visibility.py b/synapse/visibility.py index b443857571..e442de3173 100644 --- a/synapse/visibility.py +++ b/synapse/visibility.py @@ -26,8 +26,8 @@ from synapse.events.utils import prune_event from synapse.logging.opentracing import trace from synapse.storage.controllers import StorageControllers from synapse.storage.databases.main import DataStore -from synapse.storage.state import StateFilter from synapse.types import RetentionPolicy, StateMap, get_domain_from_id +from synapse.types.state import StateFilter from synapse.util import Clock logger = logging.getLogger(__name__) diff --git a/tests/storage/test_state.py b/tests/storage/test_state.py index d4e6d4236c..a433e70870 100644 --- a/tests/storage/test_state.py +++ b/tests/storage/test_state.py @@ -22,8 +22,8 @@ from synapse.api.constants import EventTypes, Membership from synapse.api.room_versions import RoomVersions from synapse.events import EventBase from synapse.server import HomeServer -from synapse.storage.state import StateFilter from synapse.types import JsonDict, RoomID, StateMap, UserID +from synapse.types.state import StateFilter from synapse.util import Clock from tests.unittest import HomeserverTestCase, TestCase -- cgit 1.5.1 From e2a1adbf5d11288f2134ced1f84c6ffdd91a9357 Mon Sep 17 00:00:00 2001 From: David Robertson Date: Tue, 13 Dec 2022 00:54:46 +0000 Subject: Allow selecting "prejoin" events by state keys (#14642) * Declare new config * Parse new config * Read new config * Don't use trial/our TestCase where it's not needed Before: ``` $ time trial tests/events/test_utils.py > /dev/null real 0m2.277s user 0m2.186s sys 0m0.083s ``` After: ``` $ time trial tests/events/test_utils.py > /dev/null real 0m0.566s user 0m0.508s sys 0m0.056s ``` * Helper to upsert to event fields without exceeding size limits. * Use helper when adding invite/knock state Now that we allow admins to include events in prejoin room state with arbitrary state keys, be a good Matrix citizen and ensure they don't accidentally create an oversized event. * Changelog * Move StateFilter tests should have done this in #14668 * Add extra methods to StateFilter * Use StateFilter * Ensure test file enforces typed defs; alphabetise * Workaround surprising get_current_state_ids * Whoops, fix mypy --- changelog.d/14642.feature | 1 + docs/usage/configuration/config_documentation.md | 57 ++- mypy.ini | 12 +- synapse/config/_util.py | 3 + synapse/config/api.py | 63 ++- synapse/events/utils.py | 32 +- synapse/handlers/message.py | 29 +- synapse/storage/databases/main/events_worker.py | 33 +- synapse/types/state.py | 18 + tests/config/test_api.py | 145 ++++++ tests/events/test_utils.py | 35 +- tests/storage/test_state.py | 623 +--------------------- tests/types/__init__.py | 0 tests/types/test_state.py | 627 +++++++++++++++++++++++ 14 files changed, 983 insertions(+), 695 deletions(-) create mode 100644 changelog.d/14642.feature create mode 100644 tests/config/test_api.py create mode 100644 tests/types/__init__.py create mode 100644 tests/types/test_state.py (limited to 'synapse/handlers') diff --git a/changelog.d/14642.feature b/changelog.d/14642.feature new file mode 100644 index 0000000000..cbc9db10c3 --- /dev/null +++ b/changelog.d/14642.feature @@ -0,0 +1 @@ +Allow selecting "prejoin" events by state keys in addition to event types. diff --git a/docs/usage/configuration/config_documentation.md b/docs/usage/configuration/config_documentation.md index dc5e5ac597..4d32902fea 100644 --- a/docs/usage/configuration/config_documentation.md +++ b/docs/usage/configuration/config_documentation.md @@ -2501,32 +2501,53 @@ Config settings related to the client/server API --- ### `room_prejoin_state` -Controls for the state that is shared with users who receive an invite -to a room. By default, the following state event types are shared with users who -receive invites to the room: -- m.room.join_rules -- m.room.canonical_alias -- m.room.avatar -- m.room.encryption -- m.room.name -- m.room.create -- m.room.topic +This setting controls the state that is shared with users upon receiving an +invite to a room, or in reply to a knock on a room. By default, the following +state events are shared with users: + +- `m.room.join_rules` +- `m.room.canonical_alias` +- `m.room.avatar` +- `m.room.encryption` +- `m.room.name` +- `m.room.create` +- `m.room.topic` To change the default behavior, use the following sub-options: -* `disable_default_event_types`: set to true to disable the above defaults. If this - is enabled, only the event types listed in `additional_event_types` are shared. - Defaults to false. -* `additional_event_types`: Additional state event types to share with users when they are invited - to a room. By default, this list is empty (so only the default event types are shared). +* `disable_default_event_types`: boolean. Set to `true` to disable the above + defaults. If this is enabled, only the event types listed in + `additional_event_types` are shared. Defaults to `false`. +* `additional_event_types`: A list of additional state events to include in the + events to be shared. By default, this list is empty (so only the default event + types are shared). + + Each entry in this list should be either a single string or a list of two + strings. + * A standalone string `t` represents all events with type `t` (i.e. + with no restrictions on state keys). + * A pair of strings `[t, s]` represents a single event with type `t` and + state key `s`. The same type can appear in two entries with different state + keys: in this situation, both state keys are included in prejoin state. Example configuration: ```yaml room_prejoin_state: - disable_default_event_types: true + disable_default_event_types: false additional_event_types: - - org.example.custom.event.type - - m.room.join_rules + # Share all events of type `org.example.custom.event.typeA` + - org.example.custom.event.typeA + # Share only events of type `org.example.custom.event.typeB` whose + # state_key is "foo" + - ["org.example.custom.event.typeB", "foo"] + # Share only events of type `org.example.custom.event.typeC` whose + # state_key is "bar" or "baz" + - ["org.example.custom.event.typeC", "bar"] + - ["org.example.custom.event.typeC", "baz"] ``` + +*Changed in Synapse 1.74:* admins can filter the events in prejoin state based +on their state key. + --- ### `track_puppeted_user_ips` diff --git a/mypy.ini b/mypy.ini index 727536df50..37acf589c9 100644 --- a/mypy.ini +++ b/mypy.ini @@ -89,6 +89,12 @@ disallow_untyped_defs = False [mypy-tests.*] disallow_untyped_defs = False +[mypy-tests.config.test_api] +disallow_untyped_defs = True + +[mypy-tests.federation.transport.test_client] +disallow_untyped_defs = True + [mypy-tests.handlers.test_sso] disallow_untyped_defs = True @@ -101,7 +107,7 @@ disallow_untyped_defs = True [mypy-tests.push.test_bulk_push_rule_evaluator] disallow_untyped_defs = True -[mypy-tests.test_server] +[mypy-tests.rest.*] disallow_untyped_defs = True [mypy-tests.state.test_profile] @@ -110,10 +116,10 @@ disallow_untyped_defs = True [mypy-tests.storage.*] disallow_untyped_defs = True -[mypy-tests.rest.*] +[mypy-tests.test_server] disallow_untyped_defs = True -[mypy-tests.federation.transport.test_client] +[mypy-tests.types.*] disallow_untyped_defs = True [mypy-tests.util.caches.*] diff --git a/synapse/config/_util.py b/synapse/config/_util.py index 3edb4b7106..d3a4b484ab 100644 --- a/synapse/config/_util.py +++ b/synapse/config/_util.py @@ -33,6 +33,9 @@ def validate_config( config: the configuration value to be validated config_path: the path within the config file. This will be used as a basis for the error message. + + Raises: + ConfigError, if validation fails. """ try: jsonschema.validate(config, json_schema) diff --git a/synapse/config/api.py b/synapse/config/api.py index e46728e73f..27d50d118f 100644 --- a/synapse/config/api.py +++ b/synapse/config/api.py @@ -13,12 +13,13 @@ # limitations under the License. import logging -from typing import Any, Iterable +from typing import Any, Iterable, Optional, Tuple from synapse.api.constants import EventTypes from synapse.config._base import Config, ConfigError from synapse.config._util import validate_config from synapse.types import JsonDict +from synapse.types.state import StateFilter logger = logging.getLogger(__name__) @@ -26,16 +27,20 @@ logger = logging.getLogger(__name__) class ApiConfig(Config): section = "api" + room_prejoin_state: StateFilter + track_puppetted_users_ips: bool + def read_config(self, config: JsonDict, **kwargs: Any) -> None: validate_config(_MAIN_SCHEMA, config, ()) - self.room_prejoin_state = list(self._get_prejoin_state_types(config)) + self.room_prejoin_state = StateFilter.from_types( + self._get_prejoin_state_entries(config) + ) self.track_puppeted_user_ips = config.get("track_puppeted_user_ips", False) - def _get_prejoin_state_types(self, config: JsonDict) -> Iterable[str]: - """Get the event types to include in the prejoin state - - Parses the config and returns an iterable of the event types to be included. - """ + def _get_prejoin_state_entries( + self, config: JsonDict + ) -> Iterable[Tuple[str, Optional[str]]]: + """Get the event types and state keys to include in the prejoin state.""" room_prejoin_state_config = config.get("room_prejoin_state") or {} # backwards-compatibility support for room_invite_state_types @@ -50,33 +55,39 @@ class ApiConfig(Config): logger.warning(_ROOM_INVITE_STATE_TYPES_WARNING) - yield from config["room_invite_state_types"] + for event_type in config["room_invite_state_types"]: + yield event_type, None return if not room_prejoin_state_config.get("disable_default_event_types"): - yield from _DEFAULT_PREJOIN_STATE_TYPES + yield from _DEFAULT_PREJOIN_STATE_TYPES_AND_STATE_KEYS - yield from room_prejoin_state_config.get("additional_event_types", []) + for entry in room_prejoin_state_config.get("additional_event_types", []): + if isinstance(entry, str): + yield entry, None + else: + yield entry _ROOM_INVITE_STATE_TYPES_WARNING = """\ WARNING: The 'room_invite_state_types' configuration setting is now deprecated, and replaced with 'room_prejoin_state'. New features may not work correctly -unless 'room_invite_state_types' is removed. See the sample configuration file for -details of 'room_prejoin_state'. +unless 'room_invite_state_types' is removed. See the config documentation at + https://matrix-org.github.io/synapse/latest/usage/configuration/config_documentation.html#room_prejoin_state +for details of 'room_prejoin_state'. -------------------------------------------------------------------------------- """ -_DEFAULT_PREJOIN_STATE_TYPES = [ - EventTypes.JoinRules, - EventTypes.CanonicalAlias, - EventTypes.RoomAvatar, - EventTypes.RoomEncryption, - EventTypes.Name, +_DEFAULT_PREJOIN_STATE_TYPES_AND_STATE_KEYS = [ + (EventTypes.JoinRules, ""), + (EventTypes.CanonicalAlias, ""), + (EventTypes.RoomAvatar, ""), + (EventTypes.RoomEncryption, ""), + (EventTypes.Name, ""), # Per MSC1772. - EventTypes.Create, + (EventTypes.Create, ""), # Per MSC3173. - EventTypes.Topic, + (EventTypes.Topic, ""), ] @@ -90,7 +101,17 @@ _ROOM_PREJOIN_STATE_CONFIG_SCHEMA = { "disable_default_event_types": {"type": "boolean"}, "additional_event_types": { "type": "array", - "items": {"type": "string"}, + "items": { + "oneOf": [ + {"type": "string"}, + { + "type": "array", + "items": {"type": "string"}, + "minItems": 2, + "maxItems": 2, + }, + ], + }, }, }, }, diff --git a/synapse/events/utils.py b/synapse/events/utils.py index 71853caad8..13fa93afb8 100644 --- a/synapse/events/utils.py +++ b/synapse/events/utils.py @@ -28,8 +28,14 @@ from typing import ( ) import attr +from canonicaljson import encode_canonical_json -from synapse.api.constants import EventContentFields, EventTypes, RelationTypes +from synapse.api.constants import ( + MAX_PDU_SIZE, + EventContentFields, + EventTypes, + RelationTypes, +) from synapse.api.errors import Codes, SynapseError from synapse.api.room_versions import RoomVersion from synapse.types import JsonDict @@ -674,3 +680,27 @@ def validate_canonicaljson(value: Any) -> None: elif not isinstance(value, (bool, str)) and value is not None: # Other potential JSON values (bool, None, str) are safe. raise SynapseError(400, "Unknown JSON value", Codes.BAD_JSON) + + +def maybe_upsert_event_field( + event: EventBase, container: JsonDict, key: str, value: object +) -> bool: + """Upsert an event field, but only if this doesn't make the event too large. + + Returns true iff the upsert took place. + """ + if key in container: + old_value: object = container[key] + container[key] = value + # NB: here and below, we assume that passing a non-None `time_now` argument to + # get_pdu_json doesn't increase the size of the encoded result. + upsert_okay = len(encode_canonical_json(event.get_pdu_json())) <= MAX_PDU_SIZE + if not upsert_okay: + container[key] = old_value + else: + container[key] = value + upsert_okay = len(encode_canonical_json(event.get_pdu_json())) <= MAX_PDU_SIZE + if not upsert_okay: + del container[key] + + return upsert_okay diff --git a/synapse/handlers/message.py b/synapse/handlers/message.py index d6e90ef259..845f683358 100644 --- a/synapse/handlers/message.py +++ b/synapse/handlers/message.py @@ -50,6 +50,7 @@ from synapse.event_auth import validate_event_for_room_version from synapse.events import EventBase, relation_from_event from synapse.events.builder import EventBuilder from synapse.events.snapshot import EventContext +from synapse.events.utils import maybe_upsert_event_field from synapse.events.validator import EventValidator from synapse.handlers.directory import DirectoryHandler from synapse.logging import opentracing @@ -1739,12 +1740,15 @@ class EventCreationHandler: if event.type == EventTypes.Member: if event.content["membership"] == Membership.INVITE: - event.unsigned[ - "invite_room_state" - ] = await self.store.get_stripped_room_state_from_event_context( - context, - self.room_prejoin_state_types, - membership_user_id=event.sender, + maybe_upsert_event_field( + event, + event.unsigned, + "invite_room_state", + await self.store.get_stripped_room_state_from_event_context( + context, + self.room_prejoin_state_types, + membership_user_id=event.sender, + ), ) invitee = UserID.from_string(event.state_key) @@ -1762,11 +1766,14 @@ class EventCreationHandler: event.signatures.update(returned_invite.signatures) if event.content["membership"] == Membership.KNOCK: - event.unsigned[ - "knock_room_state" - ] = await self.store.get_stripped_room_state_from_event_context( - context, - self.room_prejoin_state_types, + maybe_upsert_event_field( + event, + event.unsigned, + "knock_room_state", + await self.store.get_stripped_room_state_from_event_context( + context, + self.room_prejoin_state_types, + ), ) if event.type == EventTypes.Redaction: diff --git a/synapse/storage/databases/main/events_worker.py b/synapse/storage/databases/main/events_worker.py index 01e935edef..318fd7dc71 100644 --- a/synapse/storage/databases/main/events_worker.py +++ b/synapse/storage/databases/main/events_worker.py @@ -16,11 +16,11 @@ import logging import threading import weakref from enum import Enum, auto +from itertools import chain from typing import ( TYPE_CHECKING, Any, Collection, - Container, Dict, Iterable, List, @@ -76,6 +76,7 @@ from synapse.storage.util.id_generators import ( ) from synapse.storage.util.sequence import build_sequence_generator from synapse.types import JsonDict, get_domain_from_id +from synapse.types.state import StateFilter from synapse.util import unwrapFirstError from synapse.util.async_helpers import ObservableDeferred, delay_cancellation from synapse.util.caches.descriptors import cached, cachedList @@ -879,7 +880,7 @@ class EventsWorkerStore(SQLBaseStore): async def get_stripped_room_state_from_event_context( self, context: EventContext, - state_types_to_include: Container[str], + state_keys_to_include: StateFilter, membership_user_id: Optional[str] = None, ) -> List[JsonDict]: """ @@ -892,7 +893,7 @@ class EventsWorkerStore(SQLBaseStore): Args: context: The event context to retrieve state of the room from. - state_types_to_include: The type of state events to include. + state_keys_to_include: The state events to include, for each event type. membership_user_id: An optional user ID to include the stripped membership state events of. This is useful when generating the stripped state of a room for invites. We want to send membership events of the inviter, so that the @@ -901,21 +902,25 @@ class EventsWorkerStore(SQLBaseStore): Returns: A list of dictionaries, each representing a stripped state event from the room. """ - current_state_ids = await context.get_current_state_ids() + if membership_user_id: + types = chain( + state_keys_to_include.to_types(), + [(EventTypes.Member, membership_user_id)], + ) + filter = StateFilter.from_types(types) + else: + filter = state_keys_to_include + selected_state_ids = await context.get_current_state_ids(filter) # We know this event is not an outlier, so this must be # non-None. - assert current_state_ids is not None - - # The state to include - state_to_include_ids = [ - e_id - for k, e_id in current_state_ids.items() - if k[0] in state_types_to_include - or (membership_user_id and k == (EventTypes.Member, membership_user_id)) - ] + assert selected_state_ids is not None + + # Confusingly, get_current_state_events may return events that are discarded by + # the filter, if they're in context._state_delta_due_to_event. Strip these away. + selected_state_ids = filter.filter_state(selected_state_ids) - state_to_include = await self.get_events(state_to_include_ids) + state_to_include = await self.get_events(selected_state_ids.values()) return [ { diff --git a/synapse/types/state.py b/synapse/types/state.py index 0004d955b4..743a4f9217 100644 --- a/synapse/types/state.py +++ b/synapse/types/state.py @@ -118,6 +118,15 @@ class StateFilter: ) ) + def to_types(self) -> Iterable[Tuple[str, Optional[str]]]: + """The inverse to `from_types`.""" + for (event_type, state_keys) in self.types.items(): + if state_keys is None: + yield event_type, None + else: + for state_key in state_keys: + yield event_type, state_key + @staticmethod def from_lazy_load_member_list(members: Iterable[str]) -> "StateFilter": """Creates a filter that returns all non-member events, plus the member @@ -343,6 +352,15 @@ class StateFilter: for s in state_keys ] + def wildcard_types(self) -> List[str]: + """Returns a list of event types which require us to fetch all state keys. + This will be empty unless `has_wildcards` returns True. + + Returns: + A list of event types. + """ + return [t for t, state_keys in self.types.items() if state_keys is None] + def get_member_split(self) -> Tuple["StateFilter", "StateFilter"]: """Return the filter split into two: one which assumes it's exclusively matching against member state, and one which assumes it's matching diff --git a/tests/config/test_api.py b/tests/config/test_api.py new file mode 100644 index 0000000000..6773c9a277 --- /dev/null +++ b/tests/config/test_api.py @@ -0,0 +1,145 @@ +from unittest import TestCase as StdlibTestCase + +import yaml + +from synapse.config import ConfigError +from synapse.config.api import ApiConfig +from synapse.types.state import StateFilter + +DEFAULT_PREJOIN_STATE_PAIRS = { + ("m.room.join_rules", ""), + ("m.room.canonical_alias", ""), + ("m.room.avatar", ""), + ("m.room.encryption", ""), + ("m.room.name", ""), + ("m.room.create", ""), + ("m.room.topic", ""), +} + + +class TestRoomPrejoinState(StdlibTestCase): + def read_config(self, source: str) -> ApiConfig: + config = ApiConfig() + config.read_config(yaml.safe_load(source)) + return config + + def test_no_prejoin_state(self) -> None: + config = self.read_config("foo: bar") + self.assertFalse(config.room_prejoin_state.has_wildcards()) + self.assertEqual( + set(config.room_prejoin_state.concrete_types()), DEFAULT_PREJOIN_STATE_PAIRS + ) + + def test_disable_default_event_types(self) -> None: + config = self.read_config( + """ +room_prejoin_state: + disable_default_event_types: true + """ + ) + self.assertEqual(config.room_prejoin_state, StateFilter.none()) + + def test_event_without_state_key(self) -> None: + config = self.read_config( + """ +room_prejoin_state: + disable_default_event_types: true + additional_event_types: + - foo + """ + ) + self.assertEqual(config.room_prejoin_state.wildcard_types(), ["foo"]) + self.assertEqual(config.room_prejoin_state.concrete_types(), []) + + def test_event_with_specific_state_key(self) -> None: + config = self.read_config( + """ +room_prejoin_state: + disable_default_event_types: true + additional_event_types: + - [foo, bar] + """ + ) + self.assertFalse(config.room_prejoin_state.has_wildcards()) + self.assertEqual( + set(config.room_prejoin_state.concrete_types()), + {("foo", "bar")}, + ) + + def test_repeated_event_with_specific_state_key(self) -> None: + config = self.read_config( + """ +room_prejoin_state: + disable_default_event_types: true + additional_event_types: + - [foo, bar] + - [foo, baz] + """ + ) + self.assertFalse(config.room_prejoin_state.has_wildcards()) + self.assertEqual( + set(config.room_prejoin_state.concrete_types()), + {("foo", "bar"), ("foo", "baz")}, + ) + + def test_no_specific_state_key_overrides_specific_state_key(self) -> None: + config = self.read_config( + """ +room_prejoin_state: + disable_default_event_types: true + additional_event_types: + - [foo, bar] + - foo + """ + ) + self.assertEqual(config.room_prejoin_state.wildcard_types(), ["foo"]) + self.assertEqual(config.room_prejoin_state.concrete_types(), []) + + config = self.read_config( + """ +room_prejoin_state: + disable_default_event_types: true + additional_event_types: + - foo + - [foo, bar] + """ + ) + self.assertEqual(config.room_prejoin_state.wildcard_types(), ["foo"]) + self.assertEqual(config.room_prejoin_state.concrete_types(), []) + + def test_bad_event_type_entry_raises(self) -> None: + with self.assertRaises(ConfigError): + self.read_config( + """ +room_prejoin_state: + additional_event_types: + - [] + """ + ) + + with self.assertRaises(ConfigError): + self.read_config( + """ +room_prejoin_state: + additional_event_types: + - [a] + """ + ) + + with self.assertRaises(ConfigError): + self.read_config( + """ +room_prejoin_state: + additional_event_types: + - [a, b, c] + """ + ) + + with self.assertRaises(ConfigError): + self.read_config( + """ +room_prejoin_state: + additional_event_types: + - [true, 1.23] + """ + ) diff --git a/tests/events/test_utils.py b/tests/events/test_utils.py index b1c47efac7..a79256846f 100644 --- a/tests/events/test_utils.py +++ b/tests/events/test_utils.py @@ -12,19 +12,20 @@ # See the License for the specific language governing permissions and # limitations under the License. +import unittest as stdlib_unittest + from synapse.api.constants import EventContentFields from synapse.api.room_versions import RoomVersions from synapse.events import make_event_from_dict from synapse.events.utils import ( SerializeEventConfig, copy_and_fixup_power_levels_contents, + maybe_upsert_event_field, prune_event, serialize_event, ) from synapse.util.frozenutils import freeze -from tests import unittest - def MockEvent(**kwargs): if "event_id" not in kwargs: @@ -34,7 +35,31 @@ def MockEvent(**kwargs): return make_event_from_dict(kwargs) -class PruneEventTestCase(unittest.TestCase): +class TestMaybeUpsertEventField(stdlib_unittest.TestCase): + def test_update_okay(self) -> None: + event = make_event_from_dict({"event_id": "$1234"}) + success = maybe_upsert_event_field(event, event.unsigned, "key", "value") + self.assertTrue(success) + self.assertEqual(event.unsigned["key"], "value") + + def test_update_not_okay(self) -> None: + event = make_event_from_dict({"event_id": "$1234"}) + LARGE_STRING = "a" * 100_000 + success = maybe_upsert_event_field(event, event.unsigned, "key", LARGE_STRING) + self.assertFalse(success) + self.assertNotIn("key", event.unsigned) + + def test_update_not_okay_leaves_original_value(self) -> None: + event = make_event_from_dict( + {"event_id": "$1234", "unsigned": {"key": "value"}} + ) + LARGE_STRING = "a" * 100_000 + success = maybe_upsert_event_field(event, event.unsigned, "key", LARGE_STRING) + self.assertFalse(success) + self.assertEqual(event.unsigned["key"], "value") + + +class PruneEventTestCase(stdlib_unittest.TestCase): def run_test(self, evdict, matchdict, **kwargs): """ Asserts that a new event constructed with `evdict` will look like @@ -391,7 +416,7 @@ class PruneEventTestCase(unittest.TestCase): ) -class SerializeEventTestCase(unittest.TestCase): +class SerializeEventTestCase(stdlib_unittest.TestCase): def serialize(self, ev, fields): return serialize_event( ev, 1479807801915, config=SerializeEventConfig(only_event_fields=fields) @@ -513,7 +538,7 @@ class SerializeEventTestCase(unittest.TestCase): ) -class CopyPowerLevelsContentTestCase(unittest.TestCase): +class CopyPowerLevelsContentTestCase(stdlib_unittest.TestCase): def setUp(self) -> None: self.test_content = { "ban": 50, diff --git a/tests/storage/test_state.py b/tests/storage/test_state.py index a433e70870..bad7f0bc60 100644 --- a/tests/storage/test_state.py +++ b/tests/storage/test_state.py @@ -26,7 +26,7 @@ from synapse.types import JsonDict, RoomID, StateMap, UserID from synapse.types.state import StateFilter from synapse.util import Clock -from tests.unittest import HomeserverTestCase, TestCase +from tests.unittest import HomeserverTestCase logger = logging.getLogger(__name__) @@ -494,624 +494,3 @@ class StateStoreTestCase(HomeserverTestCase): self.assertEqual(is_all, True) self.assertDictEqual({(e5.type, e5.state_key): e5.event_id}, state_dict) - - -class StateFilterDifferenceTestCase(TestCase): - def assert_difference( - self, minuend: StateFilter, subtrahend: StateFilter, expected: StateFilter - ) -> None: - self.assertEqual( - minuend.approx_difference(subtrahend), - expected, - f"StateFilter difference not correct:\n\n\t{minuend!r}\nminus\n\t{subtrahend!r}\nwas\n\t{minuend.approx_difference(subtrahend)}\nexpected\n\t{expected}", - ) - - def test_state_filter_difference_no_include_other_minus_no_include_other( - self, - ) -> None: - """ - Tests the StateFilter.approx_difference method - where, in a.approx_difference(b), both a and b do not have the - include_others flag set. - """ - # (wildcard on state keys) - (wildcard on state keys): - self.assert_difference( - StateFilter.freeze( - {EventTypes.Member: None, EventTypes.Create: None}, - include_others=False, - ), - StateFilter.freeze( - {EventTypes.Member: None, EventTypes.CanonicalAlias: None}, - include_others=False, - ), - StateFilter.freeze({EventTypes.Create: None}, include_others=False), - ) - - # (wildcard on state keys) - (specific state keys) - # This one is an over-approximation because we can't represent - # 'all state keys except a few named examples' - self.assert_difference( - StateFilter.freeze({EventTypes.Member: None}, include_others=False), - StateFilter.freeze( - {EventTypes.Member: {"@wombat:spqr"}}, - include_others=False, - ), - StateFilter.freeze({EventTypes.Member: None}, include_others=False), - ) - - # (wildcard on state keys) - (no state keys) - self.assert_difference( - StateFilter.freeze( - {EventTypes.Member: None}, - include_others=False, - ), - StateFilter.freeze( - { - EventTypes.Member: set(), - }, - include_others=False, - ), - StateFilter.freeze( - {EventTypes.Member: None}, - include_others=False, - ), - ) - - # (specific state keys) - (wildcard on state keys): - self.assert_difference( - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, - EventTypes.CanonicalAlias: {""}, - }, - include_others=False, - ), - StateFilter.freeze( - {EventTypes.Member: None}, - include_others=False, - ), - StateFilter.freeze( - {EventTypes.CanonicalAlias: {""}}, - include_others=False, - ), - ) - - # (specific state keys) - (specific state keys) - self.assert_difference( - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, - EventTypes.CanonicalAlias: {""}, - }, - include_others=False, - ), - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:spqr"}, - }, - include_others=False, - ), - StateFilter.freeze( - { - EventTypes.Member: {"@spqr:spqr"}, - EventTypes.CanonicalAlias: {""}, - }, - include_others=False, - ), - ) - - # (specific state keys) - (no state keys) - self.assert_difference( - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, - EventTypes.CanonicalAlias: {""}, - }, - include_others=False, - ), - StateFilter.freeze( - { - EventTypes.Member: set(), - }, - include_others=False, - ), - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, - EventTypes.CanonicalAlias: {""}, - }, - include_others=False, - ), - ) - - def test_state_filter_difference_include_other_minus_no_include_other(self) -> None: - """ - Tests the StateFilter.approx_difference method - where, in a.approx_difference(b), only a has the include_others flag set. - """ - # (wildcard on state keys) - (wildcard on state keys): - self.assert_difference( - StateFilter.freeze( - {EventTypes.Member: None, EventTypes.Create: None}, - include_others=True, - ), - StateFilter.freeze( - {EventTypes.Member: None, EventTypes.CanonicalAlias: None}, - include_others=False, - ), - StateFilter.freeze( - { - EventTypes.Create: None, - EventTypes.Member: set(), - EventTypes.CanonicalAlias: set(), - }, - include_others=True, - ), - ) - - # (wildcard on state keys) - (specific state keys) - # This one is an over-approximation because we can't represent - # 'all state keys except a few named examples' - # This also shows that the resultant state filter is normalised. - self.assert_difference( - StateFilter.freeze({EventTypes.Member: None}, include_others=True), - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:spqr"}, - EventTypes.Create: {""}, - }, - include_others=False, - ), - StateFilter(types=frozendict(), include_others=True), - ) - - # (wildcard on state keys) - (no state keys) - self.assert_difference( - StateFilter.freeze( - {EventTypes.Member: None}, - include_others=True, - ), - StateFilter.freeze( - { - EventTypes.Member: set(), - }, - include_others=False, - ), - StateFilter( - types=frozendict(), - include_others=True, - ), - ) - - # (specific state keys) - (wildcard on state keys): - self.assert_difference( - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, - EventTypes.CanonicalAlias: {""}, - }, - include_others=True, - ), - StateFilter.freeze( - {EventTypes.Member: None}, - include_others=False, - ), - StateFilter.freeze( - { - EventTypes.CanonicalAlias: {""}, - EventTypes.Member: set(), - }, - include_others=True, - ), - ) - - # (specific state keys) - (specific state keys) - self.assert_difference( - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, - EventTypes.CanonicalAlias: {""}, - }, - include_others=True, - ), - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:spqr"}, - }, - include_others=False, - ), - StateFilter.freeze( - { - EventTypes.Member: {"@spqr:spqr"}, - EventTypes.CanonicalAlias: {""}, - }, - include_others=True, - ), - ) - - # (specific state keys) - (no state keys) - self.assert_difference( - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, - EventTypes.CanonicalAlias: {""}, - }, - include_others=True, - ), - StateFilter.freeze( - { - EventTypes.Member: set(), - }, - include_others=False, - ), - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, - EventTypes.CanonicalAlias: {""}, - }, - include_others=True, - ), - ) - - def test_state_filter_difference_include_other_minus_include_other(self) -> None: - """ - Tests the StateFilter.approx_difference method - where, in a.approx_difference(b), both a and b have the include_others - flag set. - """ - # (wildcard on state keys) - (wildcard on state keys): - self.assert_difference( - StateFilter.freeze( - {EventTypes.Member: None, EventTypes.Create: None}, - include_others=True, - ), - StateFilter.freeze( - {EventTypes.Member: None, EventTypes.CanonicalAlias: None}, - include_others=True, - ), - StateFilter(types=frozendict(), include_others=False), - ) - - # (wildcard on state keys) - (specific state keys) - # This one is an over-approximation because we can't represent - # 'all state keys except a few named examples' - self.assert_difference( - StateFilter.freeze({EventTypes.Member: None}, include_others=True), - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:spqr"}, - EventTypes.CanonicalAlias: {""}, - }, - include_others=True, - ), - StateFilter.freeze( - {EventTypes.Member: None, EventTypes.CanonicalAlias: None}, - include_others=False, - ), - ) - - # (wildcard on state keys) - (no state keys) - self.assert_difference( - StateFilter.freeze( - {EventTypes.Member: None}, - include_others=True, - ), - StateFilter.freeze( - { - EventTypes.Member: set(), - }, - include_others=True, - ), - StateFilter.freeze( - {EventTypes.Member: None}, - include_others=False, - ), - ) - - # (specific state keys) - (wildcard on state keys): - self.assert_difference( - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, - EventTypes.CanonicalAlias: {""}, - }, - include_others=True, - ), - StateFilter.freeze( - {EventTypes.Member: None}, - include_others=True, - ), - StateFilter( - types=frozendict(), - include_others=False, - ), - ) - - # (specific state keys) - (specific state keys) - # This one is an over-approximation because we can't represent - # 'all state keys except a few named examples' - self.assert_difference( - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, - EventTypes.CanonicalAlias: {""}, - EventTypes.Create: {""}, - }, - include_others=True, - ), - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:spqr"}, - EventTypes.Create: set(), - }, - include_others=True, - ), - StateFilter.freeze( - { - EventTypes.Member: {"@spqr:spqr"}, - EventTypes.Create: {""}, - }, - include_others=False, - ), - ) - - # (specific state keys) - (no state keys) - self.assert_difference( - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, - EventTypes.CanonicalAlias: {""}, - }, - include_others=True, - ), - StateFilter.freeze( - { - EventTypes.Member: set(), - }, - include_others=True, - ), - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, - }, - include_others=False, - ), - ) - - def test_state_filter_difference_no_include_other_minus_include_other(self) -> None: - """ - Tests the StateFilter.approx_difference method - where, in a.approx_difference(b), only b has the include_others flag set. - """ - # (wildcard on state keys) - (wildcard on state keys): - self.assert_difference( - StateFilter.freeze( - {EventTypes.Member: None, EventTypes.Create: None}, - include_others=False, - ), - StateFilter.freeze( - {EventTypes.Member: None, EventTypes.CanonicalAlias: None}, - include_others=True, - ), - StateFilter(types=frozendict(), include_others=False), - ) - - # (wildcard on state keys) - (specific state keys) - # This one is an over-approximation because we can't represent - # 'all state keys except a few named examples' - self.assert_difference( - StateFilter.freeze({EventTypes.Member: None}, include_others=False), - StateFilter.freeze( - {EventTypes.Member: {"@wombat:spqr"}}, - include_others=True, - ), - StateFilter.freeze({EventTypes.Member: None}, include_others=False), - ) - - # (wildcard on state keys) - (no state keys) - self.assert_difference( - StateFilter.freeze( - {EventTypes.Member: None}, - include_others=False, - ), - StateFilter.freeze( - { - EventTypes.Member: set(), - }, - include_others=True, - ), - StateFilter.freeze( - {EventTypes.Member: None}, - include_others=False, - ), - ) - - # (specific state keys) - (wildcard on state keys): - self.assert_difference( - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, - EventTypes.CanonicalAlias: {""}, - }, - include_others=False, - ), - StateFilter.freeze( - {EventTypes.Member: None}, - include_others=True, - ), - StateFilter( - types=frozendict(), - include_others=False, - ), - ) - - # (specific state keys) - (specific state keys) - # This one is an over-approximation because we can't represent - # 'all state keys except a few named examples' - self.assert_difference( - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, - EventTypes.CanonicalAlias: {""}, - }, - include_others=False, - ), - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:spqr"}, - }, - include_others=True, - ), - StateFilter.freeze( - { - EventTypes.Member: {"@spqr:spqr"}, - }, - include_others=False, - ), - ) - - # (specific state keys) - (no state keys) - self.assert_difference( - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, - EventTypes.CanonicalAlias: {""}, - }, - include_others=False, - ), - StateFilter.freeze( - { - EventTypes.Member: set(), - }, - include_others=True, - ), - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, - }, - include_others=False, - ), - ) - - def test_state_filter_difference_simple_cases(self) -> None: - """ - Tests some very simple cases of the StateFilter approx_difference, - that are not explicitly tested by the more in-depth tests. - """ - - self.assert_difference(StateFilter.all(), StateFilter.all(), StateFilter.none()) - - self.assert_difference( - StateFilter.all(), - StateFilter.none(), - StateFilter.all(), - ) - - -class StateFilterTestCase(TestCase): - def test_return_expanded(self) -> None: - """ - Tests the behaviour of the return_expanded() function that expands - StateFilters to include more state types (for the sake of cache hit rate). - """ - - self.assertEqual(StateFilter.all().return_expanded(), StateFilter.all()) - - self.assertEqual(StateFilter.none().return_expanded(), StateFilter.none()) - - # Concrete-only state filters stay the same - # (Case: mixed filter) - self.assertEqual( - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:test", "@alicia:test"}, - "some.other.state.type": {""}, - }, - include_others=False, - ).return_expanded(), - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:test", "@alicia:test"}, - "some.other.state.type": {""}, - }, - include_others=False, - ), - ) - - # Concrete-only state filters stay the same - # (Case: non-member-only filter) - self.assertEqual( - StateFilter.freeze( - {"some.other.state.type": {""}}, include_others=False - ).return_expanded(), - StateFilter.freeze({"some.other.state.type": {""}}, include_others=False), - ) - - # Concrete-only state filters stay the same - # (Case: member-only filter) - self.assertEqual( - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:test", "@alicia:test"}, - }, - include_others=False, - ).return_expanded(), - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:test", "@alicia:test"}, - }, - include_others=False, - ), - ) - - # Wildcard member-only state filters stay the same - self.assertEqual( - StateFilter.freeze( - {EventTypes.Member: None}, - include_others=False, - ).return_expanded(), - StateFilter.freeze( - {EventTypes.Member: None}, - include_others=False, - ), - ) - - # If there is a wildcard in the non-member portion of the filter, - # it's expanded to include ALL non-member events. - # (Case: mixed filter) - self.assertEqual( - StateFilter.freeze( - { - EventTypes.Member: {"@wombat:test", "@alicia:test"}, - "some.other.state.type": None, - }, - include_others=False, - ).return_expanded(), - StateFilter.freeze( - {EventTypes.Member: {"@wombat:test", "@alicia:test"}}, - include_others=True, - ), - ) - - # If there is a wildcard in the non-member portion of the filter, - # it's expanded to include ALL non-member events. - # (Case: non-member-only filter) - self.assertEqual( - StateFilter.freeze( - { - "some.other.state.type": None, - }, - include_others=False, - ).return_expanded(), - StateFilter.freeze({EventTypes.Member: set()}, include_others=True), - ) - self.assertEqual( - StateFilter.freeze( - { - "some.other.state.type": None, - "yet.another.state.type": {"wombat"}, - }, - include_others=False, - ).return_expanded(), - StateFilter.freeze({EventTypes.Member: set()}, include_others=True), - ) diff --git a/tests/types/__init__.py b/tests/types/__init__.py new file mode 100644 index 0000000000..e69de29bb2 diff --git a/tests/types/test_state.py b/tests/types/test_state.py new file mode 100644 index 0000000000..eb809f9fb7 --- /dev/null +++ b/tests/types/test_state.py @@ -0,0 +1,627 @@ +from frozendict import frozendict + +from synapse.api.constants import EventTypes +from synapse.types.state import StateFilter + +from tests.unittest import TestCase + + +class StateFilterDifferenceTestCase(TestCase): + def assert_difference( + self, minuend: StateFilter, subtrahend: StateFilter, expected: StateFilter + ) -> None: + self.assertEqual( + minuend.approx_difference(subtrahend), + expected, + f"StateFilter difference not correct:\n\n\t{minuend!r}\nminus\n\t{subtrahend!r}\nwas\n\t{minuend.approx_difference(subtrahend)}\nexpected\n\t{expected}", + ) + + def test_state_filter_difference_no_include_other_minus_no_include_other( + self, + ) -> None: + """ + Tests the StateFilter.approx_difference method + where, in a.approx_difference(b), both a and b do not have the + include_others flag set. + """ + # (wildcard on state keys) - (wildcard on state keys): + self.assert_difference( + StateFilter.freeze( + {EventTypes.Member: None, EventTypes.Create: None}, + include_others=False, + ), + StateFilter.freeze( + {EventTypes.Member: None, EventTypes.CanonicalAlias: None}, + include_others=False, + ), + StateFilter.freeze({EventTypes.Create: None}, include_others=False), + ) + + # (wildcard on state keys) - (specific state keys) + # This one is an over-approximation because we can't represent + # 'all state keys except a few named examples' + self.assert_difference( + StateFilter.freeze({EventTypes.Member: None}, include_others=False), + StateFilter.freeze( + {EventTypes.Member: {"@wombat:spqr"}}, + include_others=False, + ), + StateFilter.freeze({EventTypes.Member: None}, include_others=False), + ) + + # (wildcard on state keys) - (no state keys) + self.assert_difference( + StateFilter.freeze( + {EventTypes.Member: None}, + include_others=False, + ), + StateFilter.freeze( + { + EventTypes.Member: set(), + }, + include_others=False, + ), + StateFilter.freeze( + {EventTypes.Member: None}, + include_others=False, + ), + ) + + # (specific state keys) - (wildcard on state keys): + self.assert_difference( + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, + EventTypes.CanonicalAlias: {""}, + }, + include_others=False, + ), + StateFilter.freeze( + {EventTypes.Member: None}, + include_others=False, + ), + StateFilter.freeze( + {EventTypes.CanonicalAlias: {""}}, + include_others=False, + ), + ) + + # (specific state keys) - (specific state keys) + self.assert_difference( + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, + EventTypes.CanonicalAlias: {""}, + }, + include_others=False, + ), + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:spqr"}, + }, + include_others=False, + ), + StateFilter.freeze( + { + EventTypes.Member: {"@spqr:spqr"}, + EventTypes.CanonicalAlias: {""}, + }, + include_others=False, + ), + ) + + # (specific state keys) - (no state keys) + self.assert_difference( + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, + EventTypes.CanonicalAlias: {""}, + }, + include_others=False, + ), + StateFilter.freeze( + { + EventTypes.Member: set(), + }, + include_others=False, + ), + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, + EventTypes.CanonicalAlias: {""}, + }, + include_others=False, + ), + ) + + def test_state_filter_difference_include_other_minus_no_include_other(self) -> None: + """ + Tests the StateFilter.approx_difference method + where, in a.approx_difference(b), only a has the include_others flag set. + """ + # (wildcard on state keys) - (wildcard on state keys): + self.assert_difference( + StateFilter.freeze( + {EventTypes.Member: None, EventTypes.Create: None}, + include_others=True, + ), + StateFilter.freeze( + {EventTypes.Member: None, EventTypes.CanonicalAlias: None}, + include_others=False, + ), + StateFilter.freeze( + { + EventTypes.Create: None, + EventTypes.Member: set(), + EventTypes.CanonicalAlias: set(), + }, + include_others=True, + ), + ) + + # (wildcard on state keys) - (specific state keys) + # This one is an over-approximation because we can't represent + # 'all state keys except a few named examples' + # This also shows that the resultant state filter is normalised. + self.assert_difference( + StateFilter.freeze({EventTypes.Member: None}, include_others=True), + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:spqr"}, + EventTypes.Create: {""}, + }, + include_others=False, + ), + StateFilter(types=frozendict(), include_others=True), + ) + + # (wildcard on state keys) - (no state keys) + self.assert_difference( + StateFilter.freeze( + {EventTypes.Member: None}, + include_others=True, + ), + StateFilter.freeze( + { + EventTypes.Member: set(), + }, + include_others=False, + ), + StateFilter( + types=frozendict(), + include_others=True, + ), + ) + + # (specific state keys) - (wildcard on state keys): + self.assert_difference( + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, + EventTypes.CanonicalAlias: {""}, + }, + include_others=True, + ), + StateFilter.freeze( + {EventTypes.Member: None}, + include_others=False, + ), + StateFilter.freeze( + { + EventTypes.CanonicalAlias: {""}, + EventTypes.Member: set(), + }, + include_others=True, + ), + ) + + # (specific state keys) - (specific state keys) + self.assert_difference( + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, + EventTypes.CanonicalAlias: {""}, + }, + include_others=True, + ), + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:spqr"}, + }, + include_others=False, + ), + StateFilter.freeze( + { + EventTypes.Member: {"@spqr:spqr"}, + EventTypes.CanonicalAlias: {""}, + }, + include_others=True, + ), + ) + + # (specific state keys) - (no state keys) + self.assert_difference( + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, + EventTypes.CanonicalAlias: {""}, + }, + include_others=True, + ), + StateFilter.freeze( + { + EventTypes.Member: set(), + }, + include_others=False, + ), + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, + EventTypes.CanonicalAlias: {""}, + }, + include_others=True, + ), + ) + + def test_state_filter_difference_include_other_minus_include_other(self) -> None: + """ + Tests the StateFilter.approx_difference method + where, in a.approx_difference(b), both a and b have the include_others + flag set. + """ + # (wildcard on state keys) - (wildcard on state keys): + self.assert_difference( + StateFilter.freeze( + {EventTypes.Member: None, EventTypes.Create: None}, + include_others=True, + ), + StateFilter.freeze( + {EventTypes.Member: None, EventTypes.CanonicalAlias: None}, + include_others=True, + ), + StateFilter(types=frozendict(), include_others=False), + ) + + # (wildcard on state keys) - (specific state keys) + # This one is an over-approximation because we can't represent + # 'all state keys except a few named examples' + self.assert_difference( + StateFilter.freeze({EventTypes.Member: None}, include_others=True), + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:spqr"}, + EventTypes.CanonicalAlias: {""}, + }, + include_others=True, + ), + StateFilter.freeze( + {EventTypes.Member: None, EventTypes.CanonicalAlias: None}, + include_others=False, + ), + ) + + # (wildcard on state keys) - (no state keys) + self.assert_difference( + StateFilter.freeze( + {EventTypes.Member: None}, + include_others=True, + ), + StateFilter.freeze( + { + EventTypes.Member: set(), + }, + include_others=True, + ), + StateFilter.freeze( + {EventTypes.Member: None}, + include_others=False, + ), + ) + + # (specific state keys) - (wildcard on state keys): + self.assert_difference( + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, + EventTypes.CanonicalAlias: {""}, + }, + include_others=True, + ), + StateFilter.freeze( + {EventTypes.Member: None}, + include_others=True, + ), + StateFilter( + types=frozendict(), + include_others=False, + ), + ) + + # (specific state keys) - (specific state keys) + # This one is an over-approximation because we can't represent + # 'all state keys except a few named examples' + self.assert_difference( + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, + EventTypes.CanonicalAlias: {""}, + EventTypes.Create: {""}, + }, + include_others=True, + ), + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:spqr"}, + EventTypes.Create: set(), + }, + include_others=True, + ), + StateFilter.freeze( + { + EventTypes.Member: {"@spqr:spqr"}, + EventTypes.Create: {""}, + }, + include_others=False, + ), + ) + + # (specific state keys) - (no state keys) + self.assert_difference( + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, + EventTypes.CanonicalAlias: {""}, + }, + include_others=True, + ), + StateFilter.freeze( + { + EventTypes.Member: set(), + }, + include_others=True, + ), + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, + }, + include_others=False, + ), + ) + + def test_state_filter_difference_no_include_other_minus_include_other(self) -> None: + """ + Tests the StateFilter.approx_difference method + where, in a.approx_difference(b), only b has the include_others flag set. + """ + # (wildcard on state keys) - (wildcard on state keys): + self.assert_difference( + StateFilter.freeze( + {EventTypes.Member: None, EventTypes.Create: None}, + include_others=False, + ), + StateFilter.freeze( + {EventTypes.Member: None, EventTypes.CanonicalAlias: None}, + include_others=True, + ), + StateFilter(types=frozendict(), include_others=False), + ) + + # (wildcard on state keys) - (specific state keys) + # This one is an over-approximation because we can't represent + # 'all state keys except a few named examples' + self.assert_difference( + StateFilter.freeze({EventTypes.Member: None}, include_others=False), + StateFilter.freeze( + {EventTypes.Member: {"@wombat:spqr"}}, + include_others=True, + ), + StateFilter.freeze({EventTypes.Member: None}, include_others=False), + ) + + # (wildcard on state keys) - (no state keys) + self.assert_difference( + StateFilter.freeze( + {EventTypes.Member: None}, + include_others=False, + ), + StateFilter.freeze( + { + EventTypes.Member: set(), + }, + include_others=True, + ), + StateFilter.freeze( + {EventTypes.Member: None}, + include_others=False, + ), + ) + + # (specific state keys) - (wildcard on state keys): + self.assert_difference( + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, + EventTypes.CanonicalAlias: {""}, + }, + include_others=False, + ), + StateFilter.freeze( + {EventTypes.Member: None}, + include_others=True, + ), + StateFilter( + types=frozendict(), + include_others=False, + ), + ) + + # (specific state keys) - (specific state keys) + # This one is an over-approximation because we can't represent + # 'all state keys except a few named examples' + self.assert_difference( + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, + EventTypes.CanonicalAlias: {""}, + }, + include_others=False, + ), + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:spqr"}, + }, + include_others=True, + ), + StateFilter.freeze( + { + EventTypes.Member: {"@spqr:spqr"}, + }, + include_others=False, + ), + ) + + # (specific state keys) - (no state keys) + self.assert_difference( + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, + EventTypes.CanonicalAlias: {""}, + }, + include_others=False, + ), + StateFilter.freeze( + { + EventTypes.Member: set(), + }, + include_others=True, + ), + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:spqr", "@spqr:spqr"}, + }, + include_others=False, + ), + ) + + def test_state_filter_difference_simple_cases(self) -> None: + """ + Tests some very simple cases of the StateFilter approx_difference, + that are not explicitly tested by the more in-depth tests. + """ + + self.assert_difference(StateFilter.all(), StateFilter.all(), StateFilter.none()) + + self.assert_difference( + StateFilter.all(), + StateFilter.none(), + StateFilter.all(), + ) + + +class StateFilterTestCase(TestCase): + def test_return_expanded(self) -> None: + """ + Tests the behaviour of the return_expanded() function that expands + StateFilters to include more state types (for the sake of cache hit rate). + """ + + self.assertEqual(StateFilter.all().return_expanded(), StateFilter.all()) + + self.assertEqual(StateFilter.none().return_expanded(), StateFilter.none()) + + # Concrete-only state filters stay the same + # (Case: mixed filter) + self.assertEqual( + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:test", "@alicia:test"}, + "some.other.state.type": {""}, + }, + include_others=False, + ).return_expanded(), + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:test", "@alicia:test"}, + "some.other.state.type": {""}, + }, + include_others=False, + ), + ) + + # Concrete-only state filters stay the same + # (Case: non-member-only filter) + self.assertEqual( + StateFilter.freeze( + {"some.other.state.type": {""}}, include_others=False + ).return_expanded(), + StateFilter.freeze({"some.other.state.type": {""}}, include_others=False), + ) + + # Concrete-only state filters stay the same + # (Case: member-only filter) + self.assertEqual( + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:test", "@alicia:test"}, + }, + include_others=False, + ).return_expanded(), + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:test", "@alicia:test"}, + }, + include_others=False, + ), + ) + + # Wildcard member-only state filters stay the same + self.assertEqual( + StateFilter.freeze( + {EventTypes.Member: None}, + include_others=False, + ).return_expanded(), + StateFilter.freeze( + {EventTypes.Member: None}, + include_others=False, + ), + ) + + # If there is a wildcard in the non-member portion of the filter, + # it's expanded to include ALL non-member events. + # (Case: mixed filter) + self.assertEqual( + StateFilter.freeze( + { + EventTypes.Member: {"@wombat:test", "@alicia:test"}, + "some.other.state.type": None, + }, + include_others=False, + ).return_expanded(), + StateFilter.freeze( + {EventTypes.Member: {"@wombat:test", "@alicia:test"}}, + include_others=True, + ), + ) + + # If there is a wildcard in the non-member portion of the filter, + # it's expanded to include ALL non-member events. + # (Case: non-member-only filter) + self.assertEqual( + StateFilter.freeze( + { + "some.other.state.type": None, + }, + include_others=False, + ).return_expanded(), + StateFilter.freeze({EventTypes.Member: set()}, include_others=True), + ) + self.assertEqual( + StateFilter.freeze( + { + "some.other.state.type": None, + "yet.another.state.type": {"wombat"}, + }, + include_others=False, + ).return_expanded(), + StateFilter.freeze({EventTypes.Member: set()}, include_others=True), + ) -- cgit 1.5.1 From 62ed877433e23ba055cbc69a089c09d03c67681d Mon Sep 17 00:00:00 2001 From: reivilibre Date: Tue, 13 Dec 2022 13:19:19 +0000 Subject: Improve validation of field size limits in events. (#14664) --- changelog.d/14664.bugfix | 1 + stubs/synapse/synapse_rust/push.pyi | 2 +- synapse/api/constants.py | 1 + synapse/api/errors.py | 11 ++++- synapse/api/room_versions.py | 32 +++++++------- synapse/event_auth.py | 76 +++++++++++++++++++++++++++++--- synapse/handlers/federation_event.py | 20 +++++++++ synapse/push/bulk_push_rule_evaluator.py | 6 +-- 8 files changed, 119 insertions(+), 30 deletions(-) create mode 100644 changelog.d/14664.bugfix (limited to 'synapse/handlers') diff --git a/changelog.d/14664.bugfix b/changelog.d/14664.bugfix new file mode 100644 index 0000000000..a15df9a89d --- /dev/null +++ b/changelog.d/14664.bugfix @@ -0,0 +1 @@ +Improve validation of field size limits in events. \ No newline at end of file diff --git a/stubs/synapse/synapse_rust/push.pyi b/stubs/synapse/synapse_rust/push.pyi index a6a586a0b5..dab5d4aff7 100644 --- a/stubs/synapse/synapse_rust/push.pyi +++ b/stubs/synapse/synapse_rust/push.pyi @@ -45,7 +45,7 @@ class PushRuleEvaluator: notification_power_levels: Mapping[str, int], related_events_flattened: Mapping[str, Mapping[str, str]], related_event_match_enabled: bool, - room_version_feature_flags: list[str], + room_version_feature_flags: Tuple[str, ...], msc3931_enabled: bool, ): ... def run( diff --git a/synapse/api/constants.py b/synapse/api/constants.py index 89723d24fa..6a5e7171da 100644 --- a/synapse/api/constants.py +++ b/synapse/api/constants.py @@ -152,6 +152,7 @@ class EduTypes: class RejectedReason: AUTH_ERROR: Final = "auth_error" + OVERSIZED_EVENT: Final = "oversized_event" class RoomCreationPreset: diff --git a/synapse/api/errors.py b/synapse/api/errors.py index 76ef12ed3a..c2c177fd71 100644 --- a/synapse/api/errors.py +++ b/synapse/api/errors.py @@ -424,8 +424,17 @@ class ResourceLimitError(SynapseError): class EventSizeError(SynapseError): """An error raised when an event is too big.""" - def __init__(self, msg: str): + def __init__(self, msg: str, unpersistable: bool): + """ + unpersistable: + if True, the PDU must not be persisted, not even as a rejected PDU + when received over federation. + This is notably true when the entire PDU exceeds the size limit for a PDU, + (as opposed to an individual key's size limit being exceeded). + """ + super().__init__(413, msg, Codes.TOO_LARGE) + self.unpersistable = unpersistable class LoginError(SynapseError): diff --git a/synapse/api/room_versions.py b/synapse/api/room_versions.py index ac62011c9f..c397920fe5 100644 --- a/synapse/api/room_versions.py +++ b/synapse/api/room_versions.py @@ -12,7 +12,7 @@ # See the License for the specific language governing permissions and # limitations under the License. -from typing import Callable, Dict, List, Optional +from typing import Callable, Dict, Optional, Tuple import attr @@ -103,7 +103,7 @@ class RoomVersion: # is not enough to mark it "supported": the push rule evaluator also needs to # support the flag. Unknown flags are ignored by the evaluator, making conditions # fail if used. - msc3931_push_features: List[str] # values from PushRuleRoomFlag + msc3931_push_features: Tuple[str, ...] # values from PushRuleRoomFlag class RoomVersions: @@ -124,7 +124,7 @@ class RoomVersions: msc2716_redactions=False, msc3787_knock_restricted_join_rule=False, msc3667_int_only_power_levels=False, - msc3931_push_features=[], + msc3931_push_features=(), ) V2 = RoomVersion( "2", @@ -143,7 +143,7 @@ class RoomVersions: msc2716_redactions=False, msc3787_knock_restricted_join_rule=False, msc3667_int_only_power_levels=False, - msc3931_push_features=[], + msc3931_push_features=(), ) V3 = RoomVersion( "3", @@ -162,7 +162,7 @@ class RoomVersions: msc2716_redactions=False, msc3787_knock_restricted_join_rule=False, msc3667_int_only_power_levels=False, - msc3931_push_features=[], + msc3931_push_features=(), ) V4 = RoomVersion( "4", @@ -181,7 +181,7 @@ class RoomVersions: msc2716_redactions=False, msc3787_knock_restricted_join_rule=False, msc3667_int_only_power_levels=False, - msc3931_push_features=[], + msc3931_push_features=(), ) V5 = RoomVersion( "5", @@ -200,7 +200,7 @@ class RoomVersions: msc2716_redactions=False, msc3787_knock_restricted_join_rule=False, msc3667_int_only_power_levels=False, - msc3931_push_features=[], + msc3931_push_features=(), ) V6 = RoomVersion( "6", @@ -219,7 +219,7 @@ class RoomVersions: msc2716_redactions=False, msc3787_knock_restricted_join_rule=False, msc3667_int_only_power_levels=False, - msc3931_push_features=[], + msc3931_push_features=(), ) MSC2176 = RoomVersion( "org.matrix.msc2176", @@ -238,7 +238,7 @@ class RoomVersions: msc2716_redactions=False, msc3787_knock_restricted_join_rule=False, msc3667_int_only_power_levels=False, - msc3931_push_features=[], + msc3931_push_features=(), ) V7 = RoomVersion( "7", @@ -257,7 +257,7 @@ class RoomVersions: msc2716_redactions=False, msc3787_knock_restricted_join_rule=False, msc3667_int_only_power_levels=False, - msc3931_push_features=[], + msc3931_push_features=(), ) V8 = RoomVersion( "8", @@ -276,7 +276,7 @@ class RoomVersions: msc2716_redactions=False, msc3787_knock_restricted_join_rule=False, msc3667_int_only_power_levels=False, - msc3931_push_features=[], + msc3931_push_features=(), ) V9 = RoomVersion( "9", @@ -295,7 +295,7 @@ class RoomVersions: msc2716_redactions=False, msc3787_knock_restricted_join_rule=False, msc3667_int_only_power_levels=False, - msc3931_push_features=[], + msc3931_push_features=(), ) MSC3787 = RoomVersion( "org.matrix.msc3787", @@ -314,7 +314,7 @@ class RoomVersions: msc2716_redactions=False, msc3787_knock_restricted_join_rule=True, msc3667_int_only_power_levels=False, - msc3931_push_features=[], + msc3931_push_features=(), ) V10 = RoomVersion( "10", @@ -333,7 +333,7 @@ class RoomVersions: msc2716_redactions=False, msc3787_knock_restricted_join_rule=True, msc3667_int_only_power_levels=True, - msc3931_push_features=[], + msc3931_push_features=(), ) MSC2716v4 = RoomVersion( "org.matrix.msc2716v4", @@ -352,7 +352,7 @@ class RoomVersions: msc2716_redactions=True, msc3787_knock_restricted_join_rule=False, msc3667_int_only_power_levels=False, - msc3931_push_features=[], + msc3931_push_features=(), ) MSC1767v10 = RoomVersion( # MSC1767 (Extensible Events) based on room version "10" @@ -372,7 +372,7 @@ class RoomVersions: msc2716_redactions=False, msc3787_knock_restricted_join_rule=True, msc3667_int_only_power_levels=True, - msc3931_push_features=[PushRuleRoomFlag.EXTENSIBLE_EVENTS], + msc3931_push_features=(PushRuleRoomFlag.EXTENSIBLE_EVENTS,), ) diff --git a/synapse/event_auth.py b/synapse/event_auth.py index bab31e33c5..d437b7e5d1 100644 --- a/synapse/event_auth.py +++ b/synapse/event_auth.py @@ -52,6 +52,7 @@ from synapse.api.room_versions import ( KNOWN_ROOM_VERSIONS, EventFormatVersions, RoomVersion, + RoomVersions, ) from synapse.storage.databases.main.events_worker import EventRedactBehaviour from synapse.types import MutableStateMap, StateMap, UserID, get_domain_from_id @@ -341,19 +342,80 @@ def check_state_dependent_auth_rules( logger.debug("Allowing! %s", event) +# Set of room versions where Synapse did not apply event key size limits +# in bytes, but rather in codepoints. +# In these room versions, we are more lenient with event size validation. +LENIENT_EVENT_BYTE_LIMITS_ROOM_VERSIONS = { + RoomVersions.V1, + RoomVersions.V2, + RoomVersions.V3, + RoomVersions.V4, + RoomVersions.V5, + RoomVersions.V6, + RoomVersions.MSC2176, + RoomVersions.V7, + RoomVersions.V8, + RoomVersions.V9, + RoomVersions.MSC3787, + RoomVersions.V10, + RoomVersions.MSC2716v4, + RoomVersions.MSC1767v10, +} + + def _check_size_limits(event: "EventBase") -> None: + """ + Checks the size limits in a PDU. + + The entire size limit of the PDU is checked first. + Then the size of fields is checked, first in codepoints and then in bytes. + + The codepoint size limits are only for Synapse compatibility. + + Raises: + EventSizeError: + when a size limit has been violated. + + unpersistable=True if Synapse never would have accepted the event and + the PDU must NOT be persisted. + + unpersistable=False if a prior version of Synapse would have accepted the + event and so the PDU must be persisted as rejected to avoid + breaking the room. + """ + + # Whole PDU check + if len(encode_canonical_json(event.get_pdu_json())) > MAX_PDU_SIZE: + raise EventSizeError("event too large", unpersistable=True) + + # Codepoint size check: Synapse always enforced these limits, so apply + # them strictly. if len(event.user_id) > 255: - raise EventSizeError("'user_id' too large") + raise EventSizeError("'user_id' too large", unpersistable=True) if len(event.room_id) > 255: - raise EventSizeError("'room_id' too large") + raise EventSizeError("'room_id' too large", unpersistable=True) if event.is_state() and len(event.state_key) > 255: - raise EventSizeError("'state_key' too large") + raise EventSizeError("'state_key' too large", unpersistable=True) if len(event.type) > 255: - raise EventSizeError("'type' too large") + raise EventSizeError("'type' too large", unpersistable=True) if len(event.event_id) > 255: - raise EventSizeError("'event_id' too large") - if len(encode_canonical_json(event.get_pdu_json())) > MAX_PDU_SIZE: - raise EventSizeError("event too large") + raise EventSizeError("'event_id' too large", unpersistable=True) + + strict_byte_limits = ( + event.room_version not in LENIENT_EVENT_BYTE_LIMITS_ROOM_VERSIONS + ) + + # Byte size check: if these fail, then be lenient to avoid breaking rooms. + if len(event.user_id.encode("utf-8")) > 255: + raise EventSizeError("'user_id' too large", unpersistable=strict_byte_limits) + if len(event.room_id.encode("utf-8")) > 255: + raise EventSizeError("'room_id' too large", unpersistable=strict_byte_limits) + if event.is_state() and len(event.state_key.encode("utf-8")) > 255: + raise EventSizeError("'state_key' too large", unpersistable=strict_byte_limits) + if len(event.type.encode("utf-8")) > 255: + raise EventSizeError("'type' too large", unpersistable=strict_byte_limits) + if len(event.event_id.encode("utf-8")) > 255: + raise EventSizeError("'event_id' too large", unpersistable=strict_byte_limits) def _check_create(event: "EventBase") -> None: diff --git a/synapse/handlers/federation_event.py b/synapse/handlers/federation_event.py index d2facdab60..66aca2f864 100644 --- a/synapse/handlers/federation_event.py +++ b/synapse/handlers/federation_event.py @@ -43,6 +43,7 @@ from synapse.api.constants import ( from synapse.api.errors import ( AuthError, Codes, + EventSizeError, FederationError, FederationPullAttemptBackoffError, HttpResponseException, @@ -1736,6 +1737,15 @@ class FederationEventHandler: except AuthError as e: logger.warning("Rejecting %r because %s", event, e) context.rejected = RejectedReason.AUTH_ERROR + except EventSizeError as e: + if e.unpersistable: + # This event is completely unpersistable. + raise e + # Otherwise, we are somewhat lenient and just persist the event + # as rejected, for moderate compatibility with older Synapse + # versions. + logger.warning("While validating received event %r: %s", event, e) + context.rejected = RejectedReason.OVERSIZED_EVENT events_and_contexts_to_persist.append((event, context)) @@ -1781,6 +1791,16 @@ class FederationEventHandler: # TODO: use a different rejected reason here? context.rejected = RejectedReason.AUTH_ERROR return + except EventSizeError as e: + if e.unpersistable: + # This event is completely unpersistable. + raise e + # Otherwise, we are somewhat lenient and just persist the event + # as rejected, for moderate compatibility with older Synapse + # versions. + logger.warning("While validating received event %r: %s", event, e) + context.rejected = RejectedReason.OVERSIZED_EVENT + return # next, check that we have all of the event's auth events. # diff --git a/synapse/push/bulk_push_rule_evaluator.py b/synapse/push/bulk_push_rule_evaluator.py index 36e5b327ef..f27ba64d53 100644 --- a/synapse/push/bulk_push_rule_evaluator.py +++ b/synapse/push/bulk_push_rule_evaluator.py @@ -342,10 +342,6 @@ class BulkPushRuleEvaluator: for user_id, level in notification_levels.items(): notification_levels[user_id] = int(level) - room_version_features = event.room_version.msc3931_push_features - if not room_version_features: - room_version_features = [] - evaluator = PushRuleEvaluator( _flatten_dict(event, room_version=event.room_version), room_member_count, @@ -353,7 +349,7 @@ class BulkPushRuleEvaluator: notification_levels, related_events, self._related_event_match_enabled, - room_version_features, + event.room_version.msc3931_push_features, self.hs.config.experimental.msc1767_enabled, # MSC3931 flag ) -- cgit 1.5.1