diff options
| author | Patrick Cloke <patrickc@matrix.org> | 2023-11-09 09:48:24 -0500 |
|---|---|---|
| committer | Patrick Cloke <patrickc@matrix.org> | 2023-11-09 09:48:24 -0500 |
| commit | 396fa974a10a393be068ff2ab87134e89567b808 (patch) | |
| tree | fa9cc3cd57a77b71b68dd5f22afe952220ed936a /synapse | |
| parent | Don't use separate copy_read method. (diff) | |
| parent | Bump pyicu from 2.11 to 2.12 (#16603) (diff) | |
| download | synapse-396fa974a10a393be068ff2ab87134e89567b808.tar.xz | |
Merge remote-tracking branch 'origin/develop' into clokep/psycopg3
Diffstat (limited to 'synapse')
140 files changed, 4442 insertions, 2664 deletions
diff --git a/synapse/_scripts/register_new_matrix_user.py b/synapse/_scripts/register_new_matrix_user.py index 19ca399d44..9293808640 100644 --- a/synapse/_scripts/register_new_matrix_user.py +++ b/synapse/_scripts/register_new_matrix_user.py @@ -50,7 +50,7 @@ def request_registration( url = "%s/_synapse/admin/v1/register" % (server_location.rstrip("/"),) # Get the nonce - r = requests.get(url, verify=False) + r = requests.get(url) if r.status_code != 200: _print("ERROR! Received %d %s" % (r.status_code, r.reason)) @@ -88,7 +88,7 @@ def request_registration( } _print("Sending registration request...") - r = requests.post(url, json=data, verify=False) + r = requests.post(url, json=data) if r.status_code != 200: _print("ERROR! Received %d %s" % (r.status_code, r.reason)) diff --git a/synapse/_scripts/synapse_port_db.py b/synapse/_scripts/synapse_port_db.py index ab2b29cf1b..ef8590db65 100755 --- a/synapse/_scripts/synapse_port_db.py +++ b/synapse/_scripts/synapse_port_db.py @@ -191,7 +191,7 @@ IGNORED_TABLES = { "user_directory_search_stat", "user_directory_search_pos", "users_who_share_private_rooms", - "users_in_public_room", + "users_in_public_rooms", # UI auth sessions have foreign keys so additional care needs to be taken, # the sessions are transient anyway, so ignore them. "ui_auth_sessions", diff --git a/synapse/api/auth/internal.py b/synapse/api/auth/internal.py index a75f6f2cc4..36ee9c8b8f 100644 --- a/synapse/api/auth/internal.py +++ b/synapse/api/auth/internal.py @@ -115,7 +115,7 @@ class InternalAuth(BaseAuth): Once get_user_by_req has set up the opentracing span, this does the actual work. """ try: - ip_addr = request.getClientAddress().host + ip_addr = request.get_client_ip_if_available() user_agent = get_request_user_agent(request) access_token = self.get_access_token_from_request(request) diff --git a/synapse/api/presence.py b/synapse/api/presence.py index b78f419994..afef6712e1 100644 --- a/synapse/api/presence.py +++ b/synapse/api/presence.py @@ -80,10 +80,6 @@ class UserPresenceState: def as_dict(self) -> JsonDict: return attr.asdict(self) - @staticmethod - def from_dict(d: JsonDict) -> "UserPresenceState": - return UserPresenceState(**d) - def copy_and_replace(self, **kwargs: Any) -> "UserPresenceState": return attr.evolve(self, **kwargs) diff --git a/synapse/config/server.py b/synapse/config/server.py index 72d30da300..f9e18d2053 100644 --- a/synapse/config/server.py +++ b/synapse/config/server.py @@ -368,9 +368,14 @@ class ServerConfig(Config): # Whether to enable user presence. presence_config = config.get("presence") or {} - self.use_presence = presence_config.get("enabled") - if self.use_presence is None: - self.use_presence = config.get("use_presence", True) + presence_enabled = presence_config.get("enabled") + if presence_enabled is None: + presence_enabled = config.get("use_presence", True) + + # Whether presence is enabled *at all*. + self.presence_enabled = bool(presence_enabled) + # Whether to internally track presence, requires that presence is enabled, + self.track_presence = self.presence_enabled and presence_enabled != "untracked" # Custom presence router module # This is the legacy way of configuring it (the config should now be put in the modules section) diff --git a/synapse/config/workers.py b/synapse/config/workers.py index f1766088fc..6d67a8cd5c 100644 --- a/synapse/config/workers.py +++ b/synapse/config/workers.py @@ -358,9 +358,9 @@ class WorkerConfig(Config): "Must only specify one instance to handle `account_data` messages." ) - if len(self.writers.receipts) != 1: + if len(self.writers.receipts) == 0: raise ConfigError( - "Must only specify one instance to handle `receipts` messages." + "Must specify at least one instance to handle `receipts` messages." ) if len(self.writers.events) == 0: diff --git a/synapse/events/utils.py b/synapse/events/utils.py index 53af423a5a..ac2cf83d9f 100644 --- a/synapse/events/utils.py +++ b/synapse/events/utils.py @@ -17,6 +17,7 @@ import re from typing import ( TYPE_CHECKING, Any, + Awaitable, Callable, Dict, Iterable, @@ -45,6 +46,7 @@ from . import EventBase if TYPE_CHECKING: from synapse.handlers.relations import BundledAggregations + from synapse.server import HomeServer # Split strings on "." but not "\." (or "\\\."). @@ -56,6 +58,13 @@ CANONICALJSON_MAX_INT = (2**53) - 1 CANONICALJSON_MIN_INT = -CANONICALJSON_MAX_INT +# Module API callback that allows adding fields to the unsigned section of +# events that are sent to clients. +ADD_EXTRA_FIELDS_TO_UNSIGNED_CLIENT_EVENT_CALLBACK = Callable[ + [EventBase], Awaitable[JsonDict] +] + + def prune_event(event: EventBase) -> EventBase: """Returns a pruned version of the given event, which removes all keys we don't know about or think could potentially be dodgy. @@ -509,7 +518,13 @@ class EventClientSerializer: clients. """ - def serialize_event( + def __init__(self, hs: "HomeServer") -> None: + self._store = hs.get_datastores().main + self._add_extra_fields_to_unsigned_client_event_callbacks: List[ + ADD_EXTRA_FIELDS_TO_UNSIGNED_CLIENT_EVENT_CALLBACK + ] = [] + + async def serialize_event( self, event: Union[JsonDict, EventBase], time_now: int, @@ -535,10 +550,21 @@ class EventClientSerializer: serialized_event = serialize_event(event, time_now, config=config) + new_unsigned = {} + for callback in self._add_extra_fields_to_unsigned_client_event_callbacks: + u = await callback(event) + new_unsigned.update(u) + + if new_unsigned: + # We do the `update` this way round so that modules can't clobber + # existing fields. + new_unsigned.update(serialized_event["unsigned"]) + serialized_event["unsigned"] = new_unsigned + # Check if there are any bundled aggregations to include with the event. if bundle_aggregations: if event.event_id in bundle_aggregations: - self._inject_bundled_aggregations( + await self._inject_bundled_aggregations( event, time_now, config, @@ -548,7 +574,7 @@ class EventClientSerializer: return serialized_event - def _inject_bundled_aggregations( + async def _inject_bundled_aggregations( self, event: EventBase, time_now: int, @@ -590,7 +616,7 @@ class EventClientSerializer: # said that we should only include the `event_id`, `origin_server_ts` and # `sender` of the edit; however MSC3925 proposes extending it to the whole # of the edit, which is what we do here. - serialized_aggregations[RelationTypes.REPLACE] = self.serialize_event( + serialized_aggregations[RelationTypes.REPLACE] = await self.serialize_event( event_aggregations.replace, time_now, config=config, @@ -600,7 +626,7 @@ class EventClientSerializer: if event_aggregations.thread: thread = event_aggregations.thread - serialized_latest_event = self.serialize_event( + serialized_latest_event = await self.serialize_event( thread.latest_event, time_now, config=config, @@ -623,7 +649,7 @@ class EventClientSerializer: "m.relations", {} ).update(serialized_aggregations) - def serialize_events( + async def serialize_events( self, events: Iterable[Union[JsonDict, EventBase]], time_now: int, @@ -645,7 +671,7 @@ class EventClientSerializer: The list of serialized events """ return [ - self.serialize_event( + await self.serialize_event( event, time_now, config=config, @@ -654,6 +680,14 @@ class EventClientSerializer: for event in events ] + def register_add_extra_fields_to_unsigned_client_event_callback( + self, callback: ADD_EXTRA_FIELDS_TO_UNSIGNED_CLIENT_EVENT_CALLBACK + ) -> None: + """Register a callback that returns additions to the unsigned section of + serialized events. + """ + self._add_extra_fields_to_unsigned_client_event_callbacks.append(callback) + _PowerLevel = Union[str, int] PowerLevelsContent = Mapping[str, Union[_PowerLevel, Mapping[str, _PowerLevel]]] diff --git a/synapse/federation/federation_client.py b/synapse/federation/federation_client.py index c8bc46415d..1a7fa175ec 100644 --- a/synapse/federation/federation_client.py +++ b/synapse/federation/federation_client.py @@ -1402,7 +1402,7 @@ class FederationClient(FederationBase): The remote homeserver return some state from the room. The response dictionary is in the form: - {"knock_state_events": [<state event dict>, ...]} + {"knock_room_state": [<state event dict>, ...]} The list of state events may be empty. @@ -1429,7 +1429,7 @@ class FederationClient(FederationBase): The remote homeserver can optionally return some state from the room. The response dictionary is in the form: - {"knock_state_events": [<state event dict>, ...]} + {"knock_room_state": [<state event dict>, ...]} The list of state events may be empty. """ diff --git a/synapse/federation/federation_server.py b/synapse/federation/federation_server.py index ec8e770430..2bb2c64ebe 100644 --- a/synapse/federation/federation_server.py +++ b/synapse/federation/federation_server.py @@ -84,7 +84,7 @@ from synapse.replication.http.federation import ( 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.types import JsonDict, StateMap, UserID, get_domain_from_id from synapse.util import unwrapFirstError from synapse.util.async_helpers import Linearizer, concurrently_execute, gather_results from synapse.util.caches.response_cache import ResponseCache @@ -850,14 +850,7 @@ class FederationServer(FederationBase): context, self._room_prejoin_state_types ) ) - 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, - } + return {"knock_room_state": stripped_room_state} async def _on_send_membership_event( self, origin: str, content: JsonDict, membership_type: str, room_id: str @@ -1006,6 +999,12 @@ class FederationServer(FederationBase): async def on_claim_client_keys( self, query: List[Tuple[str, str, str, int]], always_include_fallback_keys: bool ) -> Dict[str, Any]: + if any( + not self.hs.is_mine(UserID.from_string(user_id)) + for user_id, _, _, _ in query + ): + raise SynapseError(400, "User is not hosted on this homeserver") + log_kv({"message": "Claiming one time keys.", "user, device pairs": query}) results = await self._e2e_keys_handler.claim_local_one_time_keys( query, always_include_fallback_keys=always_include_fallback_keys @@ -1402,7 +1401,7 @@ class FederationHandlerRegistry: self._edu_type_to_instance[edu_type] = instance_names async def on_edu(self, edu_type: str, origin: str, content: dict) -> None: - if not self.config.server.use_presence and edu_type == EduTypes.PRESENCE: + if not self.config.server.track_presence and edu_type == EduTypes.PRESENCE: return # Check if we have a handler on this instance diff --git a/synapse/federation/send_queue.py b/synapse/federation/send_queue.py index 6520795635..525968bcba 100644 --- a/synapse/federation/send_queue.py +++ b/synapse/federation/send_queue.py @@ -395,7 +395,7 @@ class PresenceDestinationsRow(BaseFederationRow): @staticmethod def from_data(data: JsonDict) -> "PresenceDestinationsRow": return PresenceDestinationsRow( - state=UserPresenceState.from_dict(data["state"]), destinations=data["dests"] + state=UserPresenceState(**data["state"]), destinations=data["dests"] ) def to_data(self) -> JsonDict: diff --git a/synapse/federation/sender/__init__.py b/synapse/federation/sender/__init__.py index fb20fd8a10..7980d1a322 100644 --- a/synapse/federation/sender/__init__.py +++ b/synapse/federation/sender/__init__.py @@ -67,7 +67,7 @@ The loop continues so long as there is anything to send. At each iteration of th When the `PerDestinationQueue` has the catch-up flag set, the *Catch-Up Transmission Loop* (`_catch_up_transmission_loop`) is used in lieu of the regular `_transaction_transmission_loop`. -(Only once the catch-up mode has been exited can the regular tranaction transmission behaviour +(Only once the catch-up mode has been exited can the regular transaction transmission behaviour be resumed.) *Catch-Up Mode*, entered upon Synapse startup or once a homeserver has fallen behind due to @@ -844,7 +844,7 @@ class FederationSender(AbstractFederationSender): destinations (list[str]) """ - if not states or not self.hs.config.server.use_presence: + if not states or not self.hs.config.server.track_presence: # No-op if presence is disabled. return diff --git a/synapse/federation/transport/client.py b/synapse/federation/transport/client.py index b5e4b2680e..fab4800717 100644 --- a/synapse/federation/transport/client.py +++ b/synapse/federation/transport/client.py @@ -431,7 +431,7 @@ class TransportLayerClient: The remote homeserver can optionally return some state from the room. The response dictionary is in the form: - {"knock_state_events": [<state event dict>, ...]} + {"knock_room_state": [<state event dict>, ...]} The list of state events may be empty. """ diff --git a/synapse/handlers/account_validity.py b/synapse/handlers/account_validity.py index f1a7a05df6..6c2a49a3b9 100644 --- a/synapse/handlers/account_validity.py +++ b/synapse/handlers/account_validity.py @@ -212,8 +212,8 @@ class AccountValidityHandler: addresses = [] for threepid in threepids: - if threepid["medium"] == "email": - addresses.append(threepid["address"]) + if threepid.medium == "email": + addresses.append(threepid.address) return addresses diff --git a/synapse/handlers/admin.py b/synapse/handlers/admin.py index ba9704a065..d06f8e3296 100644 --- a/synapse/handlers/admin.py +++ b/synapse/handlers/admin.py @@ -16,6 +16,8 @@ import abc import logging from typing import TYPE_CHECKING, Any, Dict, List, Mapping, Optional, Sequence, Set +import attr + from synapse.api.constants import Direction, Membership from synapse.events import EventBase from synapse.types import JsonMapping, RoomStreamToken, StateMap, UserID, UserInfo @@ -93,7 +95,7 @@ class AdminHandler: ] user_info_dict["displayname"] = profile.display_name user_info_dict["avatar_url"] = profile.avatar_url - user_info_dict["threepids"] = threepids + user_info_dict["threepids"] = [attr.asdict(t) for t in threepids] user_info_dict["external_ids"] = external_ids user_info_dict["erased"] = await self._store.is_user_erased(user.to_string()) @@ -171,8 +173,8 @@ class AdminHandler: else: stream_ordering = room.stream_ordering - from_key = RoomStreamToken(0, 0) - to_key = RoomStreamToken(None, stream_ordering) + from_key = RoomStreamToken(topological=0, stream=0) + to_key = RoomStreamToken(stream=stream_ordering) # Events that we've processed in this room written_events: Set[str] = set() @@ -281,7 +283,7 @@ class AdminHandler: start, limit, user_id ) for media in media_ids: - writer.write_media_id(media["media_id"], media) + writer.write_media_id(media.media_id, attr.asdict(media)) logger.info( "[%s] Written %d media_ids of %s", diff --git a/synapse/handlers/appservice.py b/synapse/handlers/appservice.py index 7de7bd3289..873dadc3bd 100644 --- a/synapse/handlers/appservice.py +++ b/synapse/handlers/appservice.py @@ -47,6 +47,7 @@ from synapse.types import ( DeviceListUpdates, JsonDict, JsonMapping, + MultiWriterStreamToken, RoomAlias, RoomStreamToken, StreamKeyType, @@ -216,8 +217,8 @@ class ApplicationServicesHandler: def notify_interested_services_ephemeral( self, - stream_key: str, - new_token: Union[int, RoomStreamToken], + stream_key: StreamKeyType, + new_token: Union[int, RoomStreamToken, MultiWriterStreamToken], users: Collection[Union[str, UserID]], ) -> None: """ @@ -259,19 +260,6 @@ class ApplicationServicesHandler: ): return - # Assert that new_token is an integer (and not a RoomStreamToken). - # All of the supported streams that this function handles use an - # integer to track progress (rather than a RoomStreamToken - a - # vector clock implementation) as they don't support multiple - # stream writers. - # - # As a result, we simply assert that new_token is an integer. - # If we do end up needing to pass a RoomStreamToken down here - # in the future, using RoomStreamToken.stream (the minimum stream - # position) to convert to an ascending integer value should work. - # Additional context: https://github.com/matrix-org/synapse/pull/11137 - assert isinstance(new_token, int) - # Ignore to-device messages if the feature flag is not enabled if ( stream_key == StreamKeyType.TO_DEVICE @@ -286,6 +274,9 @@ class ApplicationServicesHandler: ): return + # We know we're not a `RoomStreamToken` at this point. + assert not isinstance(new_token, RoomStreamToken) + # Check whether there are any appservices which have registered to receive # ephemeral events. # @@ -326,8 +317,8 @@ class ApplicationServicesHandler: async def _notify_interested_services_ephemeral( self, services: List[ApplicationService], - stream_key: str, - new_token: int, + stream_key: StreamKeyType, + new_token: Union[int, MultiWriterStreamToken], users: Collection[Union[str, UserID]], ) -> None: logger.debug("Checking interested services for %s", stream_key) @@ -340,6 +331,7 @@ class ApplicationServicesHandler: # # Instead we simply grab the latest typing updates in _handle_typing # and, if they apply to this application service, send it off. + assert isinstance(new_token, int) events = await self._handle_typing(service, new_token) if events: self.scheduler.enqueue_for_appservice(service, ephemeral=events) @@ -350,15 +342,23 @@ class ApplicationServicesHandler: (service.id, stream_key) ): if stream_key == StreamKeyType.RECEIPT: + assert isinstance(new_token, MultiWriterStreamToken) + + # We store appservice tokens as integers, so we ignore + # the `instance_map` components and instead simply + # follow the base stream position. + new_token = MultiWriterStreamToken(stream=new_token.stream) + events = await self._handle_receipts(service, new_token) self.scheduler.enqueue_for_appservice(service, ephemeral=events) # Persist the latest handled stream token for this appservice await self.store.set_appservice_stream_type_pos( - service, "read_receipt", new_token + service, "read_receipt", new_token.stream ) elif stream_key == StreamKeyType.PRESENCE: + assert isinstance(new_token, int) events = await self._handle_presence(service, users, new_token) self.scheduler.enqueue_for_appservice(service, ephemeral=events) @@ -368,6 +368,7 @@ class ApplicationServicesHandler: ) elif stream_key == StreamKeyType.TO_DEVICE: + assert isinstance(new_token, int) # Retrieve a list of to-device message events, as well as the # maximum stream token of the messages we were able to retrieve. to_device_messages = await self._get_to_device_messages( @@ -383,6 +384,7 @@ class ApplicationServicesHandler: ) elif stream_key == StreamKeyType.DEVICE_LIST: + assert isinstance(new_token, int) device_list_summary = await self._get_device_list_summary( service, new_token ) @@ -432,7 +434,7 @@ class ApplicationServicesHandler: return typing async def _handle_receipts( - self, service: ApplicationService, new_token: int + self, service: ApplicationService, new_token: MultiWriterStreamToken ) -> List[JsonMapping]: """ Return the latest read receipts that the given application service should receive. @@ -455,15 +457,17 @@ class ApplicationServicesHandler: from_key = await self.store.get_type_stream_id_for_appservice( service, "read_receipt" ) - if new_token is not None and new_token <= from_key: + if new_token is not None and new_token.stream <= from_key: logger.debug( "Rejecting token lower than or equal to stored: %s" % (new_token,) ) return [] + from_token = MultiWriterStreamToken(stream=from_key) + receipts_source = self.event_sources.sources.receipt receipts, _ = await receipts_source.get_new_events_as( - service=service, from_key=from_key, to_key=new_token + service=service, from_key=from_token, to_key=new_token ) return receipts diff --git a/synapse/handlers/deactivate_account.py b/synapse/handlers/deactivate_account.py index 67adeae6a7..370f4041fb 100644 --- a/synapse/handlers/deactivate_account.py +++ b/synapse/handlers/deactivate_account.py @@ -103,10 +103,10 @@ class DeactivateAccountHandler: # Attempt to unbind any known bound threepids to this account from identity # server(s). bound_threepids = await self.store.user_get_bound_threepids(user_id) - for threepid in bound_threepids: + for medium, address in bound_threepids: try: result = await self._identity_handler.try_unbind_threepid( - user_id, threepid["medium"], threepid["address"], id_server + user_id, medium, address, id_server ) except Exception: # Do we want this to be a fatal error or should we carry on? @@ -117,9 +117,9 @@ class DeactivateAccountHandler: # Remove any local threepid associations for this account. local_threepids = await self.store.user_get_threepids(user_id) - for threepid in local_threepids: + for local_threepid in local_threepids: await self._auth_handler.delete_local_threepid( - user_id, threepid["medium"], threepid["address"] + user_id, local_threepid.medium, local_threepid.address ) # delete any devices belonging to the user, which will also diff --git a/synapse/handlers/device.py b/synapse/handlers/device.py index 86ad96d030..93472d0117 100644 --- a/synapse/handlers/device.py +++ b/synapse/handlers/device.py @@ -14,17 +14,7 @@ # See the License for the specific language governing permissions and # limitations under the License. import logging -from typing import ( - TYPE_CHECKING, - Any, - Dict, - Iterable, - List, - Mapping, - Optional, - Set, - Tuple, -) +from typing import TYPE_CHECKING, Dict, Iterable, List, Mapping, Optional, Set, Tuple from synapse.api import errors from synapse.api.constants import EduTypes, EventTypes @@ -41,6 +31,7 @@ from synapse.metrics.background_process_metrics import ( run_as_background_process, wrap_as_background_process, ) +from synapse.storage.databases.main.client_ips import DeviceLastConnectionInfo from synapse.types import ( JsonDict, JsonMapping, @@ -337,6 +328,9 @@ class DeviceWorkerHandler: return result async def on_federation_query_user_devices(self, user_id: str) -> JsonDict: + if not self.hs.is_mine(UserID.from_string(user_id)): + raise SynapseError(400, "User is not hosted on this homeserver") + stream_id, devices = await self.store.get_e2e_device_keys_for_federation_query( user_id ) @@ -601,6 +595,8 @@ class DeviceHandler(DeviceWorkerHandler): ) # Delete device messages asynchronously and in batches using the task scheduler + # We specify an upper stream id to avoid deleting non delivered messages + # if an user re-uses a device ID. await self._task_scheduler.schedule_task( DELETE_DEVICE_MSGS_TASK_NAME, resource_id=device_id, @@ -845,7 +841,6 @@ class DeviceHandler(DeviceWorkerHandler): else: assert max_stream_id == stream_id # Avoid moving `room_id` backwards. - pass if self._handle_new_device_update_new_data: continue @@ -1009,14 +1004,14 @@ class DeviceHandler(DeviceWorkerHandler): def _update_device_from_client_ips( - device: JsonDict, client_ips: Mapping[Tuple[str, str], Mapping[str, Any]] + device: JsonDict, client_ips: Mapping[Tuple[str, str], DeviceLastConnectionInfo] ) -> None: - ip = client_ips.get((device["user_id"], device["device_id"]), {}) + ip = client_ips.get((device["user_id"], device["device_id"])) device.update( { - "last_seen_user_agent": ip.get("user_agent"), - "last_seen_ts": ip.get("last_seen"), - "last_seen_ip": ip.get("ip"), + "last_seen_user_agent": ip.user_agent if ip else None, + "last_seen_ts": ip.last_seen if ip else None, + "last_seen_ip": ip.ip if ip else None, } ) diff --git a/synapse/handlers/e2e_keys.py b/synapse/handlers/e2e_keys.py index 8c6432035d..d06524495f 100644 --- a/synapse/handlers/e2e_keys.py +++ b/synapse/handlers/e2e_keys.py @@ -542,6 +542,12 @@ class E2eKeysHandler: device_keys_query: Dict[str, Optional[List[str]]] = query_body.get( "device_keys", {} ) + if any( + not self.is_mine(UserID.from_string(user_id)) + for user_id in device_keys_query + ): + raise SynapseError(400, "User is not hosted on this homeserver") + res = await self.query_local_devices( device_keys_query, include_displaynames=( @@ -659,6 +665,20 @@ class E2eKeysHandler: timeout: Optional[int], always_include_fallback_keys: bool, ) -> JsonDict: + """ + Args: + query: A chain of maps from (user_id, device_id, algorithm) to the requested + number of keys to claim. + user: The user who is claiming these keys. + timeout: How long to wait for any federation key claim requests before + giving up. + always_include_fallback_keys: always include a fallback key for local users' + devices, even if we managed to claim a one-time-key. + + Returns: a heterogeneous dict with two keys: + one_time_keys: chain of maps user ID -> device ID -> key ID -> key. + failures: map from remote destination to a JsonDict describing the error. + """ local_query: List[Tuple[str, str, str, int]] = [] remote_queries: Dict[str, Dict[str, Dict[str, Dict[str, int]]]] = {} @@ -739,6 +759,16 @@ class E2eKeysHandler: async def upload_keys_for_user( self, user_id: str, device_id: str, keys: JsonDict ) -> JsonDict: + """ + Args: + user_id: user whose keys are being uploaded. + device_id: device whose keys are being uploaded. + keys: the body of a /keys/upload request. + + Returns a dictionary with one field: + "one_time_keys": A mapping from algorithm to number of keys for that + algorithm, including those previously persisted. + """ # This can only be called from the main process. assert isinstance(self.device_handler, DeviceHandler) diff --git a/synapse/handlers/events.py b/synapse/handlers/events.py index d12803bf0f..756825061c 100644 --- a/synapse/handlers/events.py +++ b/synapse/handlers/events.py @@ -120,7 +120,7 @@ class EventStreamHandler: events.extend(to_add) - chunks = self._event_serializer.serialize_events( + chunks = await self._event_serializer.serialize_events( events, time_now, config=SerializeEventConfig( diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index 29cd45550a..9d72794e8b 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -868,19 +868,10 @@ class FederationHandler: # 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. - 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") - ) + stripped_room_state = knock_response.get("knock_room_state") if stripped_room_state is None: - raise KeyError( - "Missing 'knock_room_state' (or legacy 'knock_state_events') field in " - "send_knock response" - ) + raise KeyError("Missing 'knock_room_state' field in send_knock response") event.unsigned["knock_room_state"] = stripped_room_state @@ -1506,7 +1497,6 @@ class FederationHandler: # in the meantime and context needs to be recomputed, so let's do so. if i == max_retries - 1: raise e - pass else: destinations = {x.split(":", 1)[-1] for x in (sender_user_id, room_id)} @@ -1582,7 +1572,6 @@ class FederationHandler: # in the meantime and context needs to be recomputed, so let's do so. if i == max_retries - 1: raise e - pass async def add_display_name_to_third_party_invite( self, diff --git a/synapse/handlers/identity.py b/synapse/handlers/identity.py index 472879c964..c041b67993 100644 --- a/synapse/handlers/identity.py +++ b/synapse/handlers/identity.py @@ -19,6 +19,8 @@ import logging import urllib.parse from typing import TYPE_CHECKING, Awaitable, Callable, Dict, List, Optional, Tuple +import attr + from synapse.api.errors import ( CodeMessageException, Codes, @@ -357,9 +359,9 @@ class IdentityHandler: # Check to see if a session already exists and that it is not yet # marked as validated - if session and session.get("validated_at") is None: - session_id = session["session_id"] - last_send_attempt = session["last_send_attempt"] + if session and session.validated_at is None: + session_id = session.session_id + last_send_attempt = session.last_send_attempt # Check that the send_attempt is higher than previous attempts if send_attempt <= last_send_attempt: @@ -480,7 +482,6 @@ class IdentityHandler: # We don't actually know which medium this 3PID is. Thus we first assume it's email, # and if validation fails we try msisdn - validation_session = None # Try to validate as email if self.hs.config.email.can_verify_email: @@ -488,19 +489,18 @@ class IdentityHandler: validation_session = await self.store.get_threepid_validation_session( "email", client_secret, sid=sid, validated=True ) - - if validation_session: - return validation_session + if validation_session: + return attr.asdict(validation_session) # Try to validate as msisdn if self.hs.config.registration.account_threepid_delegate_msisdn: # Ask our delegated msisdn identity server - validation_session = await self.threepid_from_creds( + return await self.threepid_from_creds( self.hs.config.registration.account_threepid_delegate_msisdn, threepid_creds, ) - return validation_session + return None async def proxy_msisdn_submit_token( self, id_server: str, client_secret: str, sid: str, token: str diff --git a/synapse/handlers/initial_sync.py b/synapse/handlers/initial_sync.py index 5737f8014d..c4bec955fe 100644 --- a/synapse/handlers/initial_sync.py +++ b/synapse/handlers/initial_sync.py @@ -145,7 +145,7 @@ class InitialSyncHandler: joined_rooms = [r.room_id for r in room_list if r.membership == Membership.JOIN] receipt = await self.store.get_linearized_receipts_for_rooms( joined_rooms, - to_key=int(now_token.receipt_key), + to_key=now_token.receipt_key, ) receipt = ReceiptEventSource.filter_out_private_receipts(receipt, user_id) @@ -173,7 +173,7 @@ class InitialSyncHandler: d["inviter"] = event.sender invite_event = await self.store.get_event(event.event_id) - d["invite"] = self._event_serializer.serialize_event( + d["invite"] = await self._event_serializer.serialize_event( invite_event, time_now, config=serializer_options, @@ -192,8 +192,7 @@ class InitialSyncHandler: ) elif event.membership == Membership.LEAVE: room_end_token = RoomStreamToken( - None, - event.stream_ordering, + stream=event.stream_ordering, ) deferred_room_state = run_in_background( self._state_storage_controller.get_state_for_events, @@ -226,7 +225,7 @@ class InitialSyncHandler: d["messages"] = { "chunk": ( - self._event_serializer.serialize_events( + await self._event_serializer.serialize_events( messages, time_now=time_now, config=serializer_options, @@ -236,7 +235,7 @@ class InitialSyncHandler: "end": await end_token.to_string(self.store), } - d["state"] = self._event_serializer.serialize_events( + d["state"] = await self._event_serializer.serialize_events( current_state.values(), time_now=time_now, config=serializer_options, @@ -388,7 +387,7 @@ class InitialSyncHandler: "messages": { "chunk": ( # Don't bundle aggregations as this is a deprecated API. - self._event_serializer.serialize_events( + await self._event_serializer.serialize_events( messages, time_now, config=serialize_options ) ), @@ -397,7 +396,7 @@ class InitialSyncHandler: }, "state": ( # Don't bundle aggregations as this is a deprecated API. - self._event_serializer.serialize_events( + await self._event_serializer.serialize_events( room_state.values(), time_now, config=serialize_options ) ), @@ -421,7 +420,7 @@ class InitialSyncHandler: time_now = self.clock.time_msec() serialize_options = SerializeEventConfig(requester=requester) # Don't bundle aggregations as this is a deprecated API. - state = self._event_serializer.serialize_events( + state = await self._event_serializer.serialize_events( current_state.values(), time_now, config=serialize_options, @@ -440,7 +439,7 @@ class InitialSyncHandler: async def get_presence() -> List[JsonDict]: # If presence is disabled, return an empty list - if not self.hs.config.server.use_presence: + if not self.hs.config.server.presence_enabled: return [] states = await presence_handler.get_states( @@ -498,7 +497,7 @@ class InitialSyncHandler: "messages": { "chunk": ( # Don't bundle aggregations as this is a deprecated API. - self._event_serializer.serialize_events( + await self._event_serializer.serialize_events( messages, time_now, config=serialize_options ) ), diff --git a/synapse/handlers/message.py b/synapse/handlers/message.py index 44dbbf81dd..811a41f161 100644 --- a/synapse/handlers/message.py +++ b/synapse/handlers/message.py @@ -244,7 +244,7 @@ class MessageHandler: ) room_state = room_state_events[membership_event_id] - events = self._event_serializer.serialize_events( + events = await self._event_serializer.serialize_events( room_state.values(), self.clock.time_msec(), config=SerializeEventConfig(requester=requester), @@ -999,7 +999,26 @@ class EventCreationHandler: raise ShadowBanError() if ratelimit: - await self.request_ratelimiter.ratelimit(requester, update=False) + room_id = event_dict["room_id"] + try: + room_version = await self.store.get_room_version(room_id) + except NotFoundError: + # The room doesn't exist. + raise AuthError(403, f"User {requester.user} not in room {room_id}") + + if room_version.updated_redaction_rules: + redacts = event_dict["content"].get("redacts") + else: + redacts = event_dict.get("redacts") + + is_admin_redaction = await self.is_admin_redaction( + event_type=event_dict["type"], + sender=event_dict["sender"], + redacts=redacts, + ) + await self.request_ratelimiter.ratelimit( + requester, is_admin_redaction=is_admin_redaction, update=False + ) # We limit the number of concurrent event sends in a room so that we # don't fork the DAG too much. If we don't limit then we can end up in @@ -1133,7 +1152,6 @@ class EventCreationHandler: # in the meantime and context needs to be recomputed, so let's do so. if i == max_retries - 1: raise e - pass # we know it was persisted, so must have a stream ordering assert ev.internal_metadata.stream_ordering @@ -1509,6 +1527,18 @@ class EventCreationHandler: first_event.room_id ) if writer_instance != self._instance_name: + # Ratelimit before sending to the other event persister, to + # ensure that we correctly have ratelimits on both the event + # creators and event persisters. + if ratelimit: + for event, _ in events_and_context: + is_admin_redaction = await self.is_admin_redaction( + event.type, event.sender, event.redacts + ) + await self.request_ratelimiter.ratelimit( + requester, is_admin_redaction=is_admin_redaction + ) + try: result = await self.send_events( instance_name=writer_instance, @@ -1539,6 +1569,7 @@ class EventCreationHandler: # stream_ordering entry manually (as it was persisted on # another worker). event.internal_metadata.stream_ordering = stream_id + return event event = await self.persist_and_notify_client_events( @@ -1697,21 +1728,9 @@ class EventCreationHandler: # can apply different ratelimiting. We do this by simply checking # it's not a self-redaction (to avoid having to look up whether the # user is actually admin or not). - is_admin_redaction = False - if event.type == EventTypes.Redaction: - assert event.redacts is not None - - original_event = await self.store.get_event( - event.redacts, - redact_behaviour=EventRedactBehaviour.as_is, - get_prev_content=False, - allow_rejected=False, - allow_none=True, - ) - - is_admin_redaction = bool( - original_event and event.sender != original_event.sender - ) + is_admin_redaction = await self.is_admin_redaction( + event.type, event.sender, event.redacts + ) await self.request_ratelimiter.ratelimit( requester, is_admin_redaction=is_admin_redaction @@ -1931,6 +1950,27 @@ class EventCreationHandler: return persisted_events[-1] + async def is_admin_redaction( + self, event_type: str, sender: str, redacts: Optional[str] + ) -> bool: + """Return whether the event is a redaction made by an admin, and thus + should use a different ratelimiter. + """ + if event_type != EventTypes.Redaction: + return False + + assert redacts is not None + + original_event = await self.store.get_event( + redacts, + redact_behaviour=EventRedactBehaviour.as_is, + get_prev_content=False, + allow_rejected=False, + allow_none=True, + ) + + return bool(original_event and sender != original_event.sender) + async def _maybe_kick_guest_users( self, event: EventBase, context: EventContext ) -> None: @@ -2038,7 +2078,6 @@ class EventCreationHandler: # in the meantime and context needs to be recomputed, so let's do so. if i == max_retries - 1: raise e - pass return True except AuthError: logger.info( diff --git a/synapse/handlers/pagination.py b/synapse/handlers/pagination.py index 878f267a4e..87e51bca48 100644 --- a/synapse/handlers/pagination.py +++ b/synapse/handlers/pagination.py @@ -657,7 +657,7 @@ class PaginationHandler: chunk = { "chunk": ( - self._event_serializer.serialize_events( + await self._event_serializer.serialize_events( events, time_now, config=serialize_options, @@ -669,7 +669,7 @@ class PaginationHandler: } if state: - chunk["state"] = self._event_serializer.serialize_events( + chunk["state"] = await self._event_serializer.serialize_events( state, time_now, config=serialize_options ) diff --git a/synapse/handlers/presence.py b/synapse/handlers/presence.py index 7c7cda3e95..202beee738 100644 --- a/synapse/handlers/presence.py +++ b/synapse/handlers/presence.py @@ -110,6 +110,7 @@ from synapse.replication.http.streams import ReplicationGetStreamUpdates from synapse.replication.tcp.commands import ClearUserSyncsCommand from synapse.replication.tcp.streams import PresenceFederationStream, PresenceStream from synapse.storage.databases.main import DataStore +from synapse.storage.databases.main.state_deltas import StateDelta from synapse.streams import EventSource from synapse.types import ( JsonDict, @@ -191,7 +192,8 @@ class BasePresenceHandler(abc.ABC): self.state = hs.get_state_handler() self.is_mine_id = hs.is_mine_id - self._presence_enabled = hs.config.server.use_presence + self._presence_enabled = hs.config.server.presence_enabled + self._track_presence = hs.config.server.track_presence self._federation = None if hs.should_send_federation(): @@ -511,7 +513,7 @@ class WorkerPresenceHandler(BasePresenceHandler): ) async def _on_shutdown(self) -> None: - if self._presence_enabled: + if self._track_presence: self.hs.get_replication_command_handler().send_command( ClearUserSyncsCommand(self.instance_id) ) @@ -523,7 +525,7 @@ class WorkerPresenceHandler(BasePresenceHandler): is_syncing: bool, last_sync_ms: int, ) -> None: - if self._presence_enabled: + if self._track_presence: self.hs.get_replication_command_handler().send_user_sync( self.instance_id, user_id, device_id, is_syncing, last_sync_ms ) @@ -570,7 +572,7 @@ class WorkerPresenceHandler(BasePresenceHandler): Called by the sync and events servlets to record that a user has connected to this worker and is waiting for some events. """ - if not affect_presence or not self._presence_enabled: + if not affect_presence or not self._track_presence: return _NullContextManager() # Note that this causes last_active_ts to be incremented which is not @@ -701,8 +703,8 @@ class WorkerPresenceHandler(BasePresenceHandler): user_id = target_user.to_string() - # If presence is disabled, no-op - if not self._presence_enabled: + # If tracking of presence is disabled, no-op + if not self._track_presence: return # Proxy request to instance that writes presence @@ -722,7 +724,7 @@ class WorkerPresenceHandler(BasePresenceHandler): with the app. """ # If presence is disabled, no-op - if not self._presence_enabled: + if not self._track_presence: return # Proxy request to instance that writes presence @@ -759,7 +761,7 @@ class PresenceHandler(BasePresenceHandler): ] = {} now = self.clock.time_msec() - if self._presence_enabled: + if self._track_presence: for state in self.user_to_current_state.values(): # Create a psuedo-device to properly handle time outs. This will # be overridden by any "real" devices within SYNC_ONLINE_TIMEOUT. @@ -830,7 +832,7 @@ class PresenceHandler(BasePresenceHandler): self.external_sync_linearizer = Linearizer(name="external_sync_linearizer") - if self._presence_enabled: + if self._track_presence: # Start a LoopingCall in 30s that fires every 5s. # The initial delay is to allow disconnected clients a chance to # reconnect before we treat them as offline. @@ -838,6 +840,9 @@ class PresenceHandler(BasePresenceHandler): 30, self.clock.looping_call, self._handle_timeouts, 5000 ) + # Presence information is persisted, whether or not it is being tracked + # internally. + if self._presence_enabled: self.clock.call_later( 60, self.clock.looping_call, @@ -853,7 +858,7 @@ class PresenceHandler(BasePresenceHandler): ) # Used to handle sending of presence to newly joined users/servers - if self._presence_enabled: + if self._track_presence: self.notifier.add_replication_callback(self.notify_new_event) # Presence is best effort and quickly heals itself, so lets just always @@ -904,7 +909,9 @@ class PresenceHandler(BasePresenceHandler): ) async def _update_states( - self, new_states: Iterable[UserPresenceState], force_notify: bool = False + self, + new_states: Iterable[UserPresenceState], + force_notify: bool = False, ) -> None: """Updates presence of users. Sets the appropriate timeouts. Pokes the notifier and federation if and only if the changed presence state @@ -942,7 +949,7 @@ class PresenceHandler(BasePresenceHandler): for new_state in new_states: user_id = new_state.user_id - # Its fine to not hit the database here, as the only thing not in + # It's fine to not hit the database here, as the only thing not in # the current state cache are OFFLINE states, where the only field # of interest is last_active which is safe enough to assume is 0 # here. @@ -956,6 +963,9 @@ class PresenceHandler(BasePresenceHandler): is_mine=self.is_mine_id(user_id), wheel_timer=self.wheel_timer, now=now, + # When overriding disabled presence, don't kick off all the + # wheel timers. + persist=not self._track_presence, ) if force_notify: @@ -1071,7 +1081,7 @@ class PresenceHandler(BasePresenceHandler): with the app. """ # If presence is disabled, no-op - if not self._presence_enabled: + if not self._track_presence: return user_id = user.to_string() @@ -1123,7 +1133,7 @@ class PresenceHandler(BasePresenceHandler): client that is being used by a user. presence_state: The presence state indicated in the sync request """ - if not affect_presence or not self._presence_enabled: + if not affect_presence or not self._track_presence: return _NullContextManager() curr_sync = self._user_device_to_num_current_syncs.get((user_id, device_id), 0) @@ -1283,7 +1293,7 @@ class PresenceHandler(BasePresenceHandler): async def incoming_presence(self, origin: str, content: JsonDict) -> None: """Called when we receive a `m.presence` EDU from a remote server.""" - if not self._presence_enabled: + if not self._track_presence: return now = self.clock.time_msec() @@ -1358,7 +1368,7 @@ class PresenceHandler(BasePresenceHandler): raise SynapseError(400, "Invalid presence state") # If presence is disabled, no-op - if not self._presence_enabled: + if not self._track_presence: return user_id = target_user.to_string() @@ -1499,9 +1509,9 @@ class PresenceHandler(BasePresenceHandler): # We may get multiple deltas for different rooms, but we want to # handle them on a room by room basis, so we batch them up by # room. - deltas_by_room: Dict[str, List[JsonDict]] = {} + deltas_by_room: Dict[str, List[StateDelta]] = {} for delta in deltas: - deltas_by_room.setdefault(delta["room_id"], []).append(delta) + deltas_by_room.setdefault(delta.room_id, []).append(delta) for room_id, deltas_for_room in deltas_by_room.items(): await self._handle_state_delta(room_id, deltas_for_room) @@ -1513,7 +1523,7 @@ class PresenceHandler(BasePresenceHandler): max_pos ) - async def _handle_state_delta(self, room_id: str, deltas: List[JsonDict]) -> None: + async def _handle_state_delta(self, room_id: str, deltas: List[StateDelta]) -> None: """Process current state deltas for the room to find new joins that need to be handled. """ @@ -1524,31 +1534,30 @@ class PresenceHandler(BasePresenceHandler): newly_joined_users = set() for delta in deltas: - assert room_id == delta["room_id"] + assert room_id == delta.room_id - typ = delta["type"] - state_key = delta["state_key"] - event_id = delta["event_id"] - prev_event_id = delta["prev_event_id"] - - logger.debug("Handling: %r %r, %s", typ, state_key, event_id) + logger.debug( + "Handling: %r %r, %s", delta.event_type, delta.state_key, delta.event_id + ) # Drop any event that isn't a membership join - if typ != EventTypes.Member: + if delta.event_type != EventTypes.Member: continue - if event_id is None: + if delta.event_id is None: # state has been deleted, so this is not a join. We only care about # joins. continue - event = await self.store.get_event(event_id, allow_none=True) + event = await self.store.get_event(delta.event_id, allow_none=True) if not event or event.content.get("membership") != Membership.JOIN: # We only care about joins continue - if prev_event_id: - prev_event = await self.store.get_event(prev_event_id, allow_none=True) + if delta.prev_event_id: + prev_event = await self.store.get_event( + delta.prev_event_id, allow_none=True + ) if ( prev_event and prev_event.content.get("membership") == Membership.JOIN @@ -1556,7 +1565,7 @@ class PresenceHandler(BasePresenceHandler): # Ignore changes to join events. continue - newly_joined_users.add(state_key) + newly_joined_users.add(delta.state_key) if not newly_joined_users: # If nobody has joined then there's nothing to do. @@ -2118,6 +2127,7 @@ def handle_update( is_mine: bool, wheel_timer: WheelTimer, now: int, + persist: bool, ) -> Tuple[UserPresenceState, bool, bool]: """Given a presence update: 1. Add any appropriate timers. @@ -2129,6 +2139,8 @@ def handle_update( is_mine: Whether the user is ours wheel_timer now: Time now in ms + persist: True if this state should persist until another update occurs. + Skips insertion into wheel timers. Returns: 3-tuple: `(new_state, persist_and_notify, federation_ping)` where: @@ -2146,14 +2158,15 @@ def handle_update( if is_mine: if new_state.state == PresenceState.ONLINE: # Idle timer - wheel_timer.insert( - now=now, obj=user_id, then=new_state.last_active_ts + IDLE_TIMER - ) + if not persist: + wheel_timer.insert( + now=now, obj=user_id, then=new_state.last_active_ts + IDLE_TIMER + ) active = now - new_state.last_active_ts < LAST_ACTIVE_GRANULARITY new_state = new_state.copy_and_replace(currently_active=active) - if active: + if active and not persist: wheel_timer.insert( now=now, obj=user_id, @@ -2162,11 +2175,12 @@ def handle_update( if new_state.state != PresenceState.OFFLINE: # User has stopped syncing - wheel_timer.insert( - now=now, - obj=user_id, - then=new_state.last_user_sync_ts + SYNC_ONLINE_TIMEOUT, - ) + if not persist: + wheel_timer.insert( + now=now, + obj=user_id, + then=new_state.last_user_sync_ts + SYNC_ONLINE_TIMEOUT, + ) last_federate = new_state.last_federation_update_ts if now - last_federate > FEDERATION_PING_INTERVAL: @@ -2174,7 +2188,7 @@ def handle_update( new_state = new_state.copy_and_replace(last_federation_update_ts=now) federation_ping = True - if new_state.state == PresenceState.BUSY: + if new_state.state == PresenceState.BUSY and not persist: wheel_timer.insert( now=now, obj=user_id, @@ -2182,11 +2196,13 @@ def handle_update( ) else: - wheel_timer.insert( - now=now, - obj=user_id, - then=new_state.last_federation_update_ts + FEDERATION_TIMEOUT, - ) + # An update for a remote user was received. + if not persist: + wheel_timer.insert( + now=now, + obj=user_id, + then=new_state.last_federation_update_ts + FEDERATION_TIMEOUT, + ) # Check whether the change was something worth notifying about if should_notify(prev_state, new_state, is_mine): diff --git a/synapse/handlers/push_rules.py b/synapse/handlers/push_rules.py index 7ed88a3611..87b428ab1c 100644 --- a/synapse/handlers/push_rules.py +++ b/synapse/handlers/push_rules.py @@ -19,7 +19,7 @@ from synapse.api.errors import SynapseError, UnrecognizedRequestError from synapse.push.clientformat import format_push_rules_for_user from synapse.storage.push_rule import RuleNotFoundException from synapse.synapse_rust.push import get_base_rule_ids -from synapse.types import JsonDict, UserID +from synapse.types import JsonDict, StreamKeyType, UserID if TYPE_CHECKING: from synapse.server import HomeServer @@ -114,7 +114,9 @@ class PushRulesHandler: user_id: the user ID the change is for. """ stream_id = self._main_store.get_max_push_rules_stream_id() - self._notifier.on_new_event("push_rules_key", stream_id, users=[user_id]) + self._notifier.on_new_event( + StreamKeyType.PUSH_RULES, stream_id, users=[user_id] + ) async def push_rules_for_user( self, user: UserID diff --git a/synapse/handlers/receipts.py b/synapse/handlers/receipts.py index a7a29b758b..b5f7a8b47e 100644 --- a/synapse/handlers/receipts.py +++ b/synapse/handlers/receipts.py @@ -20,6 +20,7 @@ from synapse.streams import EventSource from synapse.types import ( JsonDict, JsonMapping, + MultiWriterStreamToken, ReadReceipt, StreamKeyType, UserID, @@ -130,11 +131,10 @@ class ReceiptsHandler: async def _handle_new_receipts(self, receipts: List[ReadReceipt]) -> bool: """Takes a list of receipts, stores them and informs the notifier.""" - min_batch_id: Optional[int] = None - max_batch_id: Optional[int] = None + receipts_persisted: List[ReadReceipt] = [] for receipt in receipts: - res = await self.store.insert_receipt( + stream_id = await self.store.insert_receipt( receipt.room_id, receipt.receipt_type, receipt.user_id, @@ -143,30 +143,26 @@ class ReceiptsHandler: receipt.data, ) - if not res: - # res will be None if this receipt is 'old' + if stream_id is None: + # stream_id will be None if this receipt is 'old' continue - stream_id, max_persisted_id = res + receipts_persisted.append(receipt) - if min_batch_id is None or stream_id < min_batch_id: - min_batch_id = stream_id - if max_batch_id is None or max_persisted_id > max_batch_id: - max_batch_id = max_persisted_id - - # Either both of these should be None or neither. - if min_batch_id is None or max_batch_id is None: + if not receipts_persisted: # no new receipts return False - affected_room_ids = list({r.room_id for r in receipts}) + max_batch_id = self.store.get_max_receipt_stream_id() + + affected_room_ids = list({r.room_id for r in receipts_persisted}) self.notifier.on_new_event( StreamKeyType.RECEIPT, max_batch_id, rooms=affected_room_ids ) # Note that the min here shouldn't be relied upon to be accurate. await self.hs.get_pusherpool().on_new_receipts( - min_batch_id, max_batch_id, affected_room_ids + {r.user_id for r in receipts_persisted} ) return True @@ -205,7 +201,7 @@ class ReceiptsHandler: await self.federation_sender.send_read_receipt(receipt) -class ReceiptEventSource(EventSource[int, JsonMapping]): +class ReceiptEventSource(EventSource[MultiWriterStreamToken, JsonMapping]): def __init__(self, hs: "HomeServer"): self.store = hs.get_datastores().main self.config = hs.config @@ -278,13 +274,12 @@ class ReceiptEventSource(EventSource[int, JsonMapping]): async def get_new_events( self, user: UserID, - from_key: int, + from_key: MultiWriterStreamToken, limit: int, room_ids: Iterable[str], is_guest: bool, explicit_room_id: Optional[str] = None, - ) -> Tuple[List[JsonMapping], int]: - from_key = int(from_key) + ) -> Tuple[List[JsonMapping], MultiWriterStreamToken]: to_key = self.get_current_key() if from_key == to_key: @@ -301,8 +296,11 @@ class ReceiptEventSource(EventSource[int, JsonMapping]): return events, to_key async def get_new_events_as( - self, from_key: int, to_key: int, service: ApplicationService - ) -> Tuple[List[JsonMapping], int]: + self, + from_key: MultiWriterStreamToken, + to_key: MultiWriterStreamToken, + service: ApplicationService, + ) -> Tuple[List[JsonMapping], MultiWriterStreamToken]: """Returns a set of new read receipt events that an appservice may be interested in. @@ -317,8 +315,6 @@ class ReceiptEventSource(EventSource[int, JsonMapping]): appservice may be interested in. * The current read receipt stream token. """ - from_key = int(from_key) - if from_key == to_key: return [], to_key @@ -338,5 +334,5 @@ class ReceiptEventSource(EventSource[int, JsonMapping]): return events, to_key - def get_current_key(self) -> int: + def get_current_key(self) -> MultiWriterStreamToken: return self.store.get_max_receipt_stream_id() diff --git a/synapse/handlers/relations.py b/synapse/handlers/relations.py index 9b13448cdd..a15983afae 100644 --- a/synapse/handlers/relations.py +++ b/synapse/handlers/relations.py @@ -167,7 +167,7 @@ class RelationsHandler: now = self._clock.time_msec() serialize_options = SerializeEventConfig(requester=requester) return_value: JsonDict = { - "chunk": self._event_serializer.serialize_events( + "chunk": await self._event_serializer.serialize_events( events, now, bundle_aggregations=aggregations, @@ -177,7 +177,9 @@ class RelationsHandler: if include_original_event: # Do not bundle aggregations when retrieving the original event because # we want the content before relations are applied to it. - return_value["original_event"] = self._event_serializer.serialize_event( + return_value[ + "original_event" + ] = await self._event_serializer.serialize_event( event, now, bundle_aggregations=None, @@ -602,7 +604,7 @@ class RelationsHandler: ) now = self._clock.time_msec() - serialized_events = self._event_serializer.serialize_events( + serialized_events = await self._event_serializer.serialize_events( events, now, bundle_aggregations=aggregations ) diff --git a/synapse/handlers/room.py b/synapse/handlers/room.py index a0c3b16819..6d680b0795 100644 --- a/synapse/handlers/room.py +++ b/synapse/handlers/room.py @@ -261,7 +261,6 @@ class RoomCreationHandler: # in the meantime and context needs to be recomputed, so let's do so. if i == max_retries - 1: raise e - pass # This is to satisfy mypy and should never happen raise PartialStateConflictError() @@ -1708,7 +1707,7 @@ class RoomEventSource(EventSource[RoomStreamToken, EventBase]): if from_key.topological: logger.warning("Stream has topological part!!!! %r", from_key) - from_key = RoomStreamToken(None, from_key.stream) + from_key = RoomStreamToken(stream=from_key.stream) app_service = self.store.get_app_service_by_user_id(user.to_string()) if app_service: @@ -1940,9 +1939,10 @@ class RoomShutdownHandler: else: logger.info("Shutting down room %r", room_id) - users = await self.store.get_users_in_room(room_id) - for user_id in users: - if not self.hs.is_mine_id(user_id): + users = await self.store.get_local_users_related_to_room(room_id) + for user_id, membership in users: + # If the user is not in the room (or is banned), nothing to do. + if membership not in (Membership.JOIN, Membership.INVITE, Membership.KNOCK): continue logger.info("Kicking %r from %r...", user_id, room_id) diff --git a/synapse/handlers/room_list.py b/synapse/handlers/room_list.py index bb0bdb8e6f..2947e154be 100644 --- a/synapse/handlers/room_list.py +++ b/synapse/handlers/room_list.py @@ -33,7 +33,8 @@ from synapse.api.errors import ( RequestSendFailed, SynapseError, ) -from synapse.types import JsonDict, ThirdPartyInstanceID +from synapse.storage.databases.main.room import LargestRoomStats +from synapse.types import JsonDict, JsonMapping, ThirdPartyInstanceID from synapse.util.caches.descriptors import _CacheContext, cached from synapse.util.caches.response_cache import ResponseCache @@ -170,26 +171,24 @@ class RoomListHandler: ignore_non_federatable=from_federation, ) - def build_room_entry(room: JsonDict) -> JsonDict: + def build_room_entry(room: LargestRoomStats) -> JsonDict: entry = { - "room_id": room["room_id"], - "name": room["name"], - "topic": room["topic"], - "canonical_alias": room["canonical_alias"], - "num_joined_members": room["joined_members"], - "avatar_url": room["avatar"], - "world_readable": room["history_visibility"] + "room_id": room.room_id, + "name": room.name, + "topic": room.topic, + "canonical_alias": room.canonical_alias, + "num_joined_members": room.joined_members, + "avatar_url": room.avatar, + "world_readable": room.history_visibility == HistoryVisibility.WORLD_READABLE, - "guest_can_join": room["guest_access"] == "can_join", - "join_rule": room["join_rules"], - "room_type": room["room_type"], + "guest_can_join": room.guest_access == "can_join", + "join_rule": room.join_rules, + "room_type": room.room_type, } # Filter out Nones – rather omit the field altogether return {k: v for k, v in entry.items() if v is not None} - results = [build_room_entry(r) for r in results] - response: JsonDict = {} num_results = len(results) if limit is not None: @@ -212,33 +211,33 @@ class RoomListHandler: # If there was a token given then we assume that there # must be previous results. response["prev_batch"] = RoomListNextBatch( - last_joined_members=initial_entry["num_joined_members"], - last_room_id=initial_entry["room_id"], + last_joined_members=initial_entry.joined_members, + last_room_id=initial_entry.room_id, direction_is_forward=False, ).to_token() if more_to_come: response["next_batch"] = RoomListNextBatch( - last_joined_members=final_entry["num_joined_members"], - last_room_id=final_entry["room_id"], + last_joined_members=final_entry.joined_members, + last_room_id=final_entry.room_id, direction_is_forward=True, ).to_token() else: if has_batch_token: response["next_batch"] = RoomListNextBatch( - last_joined_members=final_entry["num_joined_members"], - last_room_id=final_entry["room_id"], + last_joined_members=final_entry.joined_members, + last_room_id=final_entry.room_id, direction_is_forward=True, ).to_token() if more_to_come: response["prev_batch"] = RoomListNextBatch( - last_joined_members=initial_entry["num_joined_members"], - last_room_id=initial_entry["room_id"], + last_joined_members=initial_entry.joined_members, + last_room_id=initial_entry.room_id, direction_is_forward=False, ).to_token() - response["chunk"] = results + response["chunk"] = [build_room_entry(r) for r in results] response["total_room_count_estimate"] = await self.store.count_public_rooms( network_tuple, @@ -256,7 +255,7 @@ class RoomListHandler: cache_context: _CacheContext, with_alias: bool = True, allow_private: bool = False, - ) -> Optional[JsonDict]: + ) -> Optional[JsonMapping]: """Returns the entry for a room Args: diff --git a/synapse/handlers/room_member.py b/synapse/handlers/room_member.py index 90343c2306..918eb203e2 100644 --- a/synapse/handlers/room_member.py +++ b/synapse/handlers/room_member.py @@ -16,7 +16,7 @@ import abc import logging import random from http import HTTPStatus -from typing import TYPE_CHECKING, Any, Dict, Iterable, List, Optional, Set, Tuple +from typing import TYPE_CHECKING, Iterable, List, Optional, Set, Tuple from synapse import types from synapse.api.constants import ( @@ -44,6 +44,7 @@ from synapse.handlers.worker_lock import NEW_EVENT_DURING_PURGE_LOCK_NAME from synapse.logging import opentracing from synapse.metrics import event_processing_positions from synapse.metrics.background_process_metrics import run_as_background_process +from synapse.storage.databases.main.state_deltas import StateDelta from synapse.types import ( JsonDict, Requester, @@ -382,8 +383,10 @@ class RoomMemberHandler(metaclass=abc.ABCMeta): and persist a new event for the new membership change. Args: - requester: - target: + requester: User requesting the membership change, i.e. the sender of the + desired membership event. + target: Use whose membership should change, i.e. the state_key of the + desired membership event. room_id: membership: @@ -415,7 +418,6 @@ class RoomMemberHandler(metaclass=abc.ABCMeta): Returns: Tuple of event ID and stream ordering position """ - user_id = target.to_string() if content is None: @@ -475,21 +477,6 @@ class RoomMemberHandler(metaclass=abc.ABCMeta): (EventTypes.Member, user_id), None ) - if event.membership == Membership.JOIN: - newly_joined = True - if prev_member_event_id: - prev_member_event = await self.store.get_event( - prev_member_event_id - ) - newly_joined = prev_member_event.membership != Membership.JOIN - - # Only rate-limit if the user actually joined the room, otherwise we'll end - # up blocking profile updates. - if newly_joined and ratelimit: - await self._join_rate_limiter_local.ratelimit(requester) - await self._join_rate_per_room_limiter.ratelimit( - requester, key=room_id, update=False - ) with opentracing.start_active_span("handle_new_client_event"): result_event = ( await self.event_creation_handler.handle_new_client_event( @@ -514,7 +501,6 @@ class RoomMemberHandler(metaclass=abc.ABCMeta): # in the meantime and context needs to be recomputed, so let's do so. if i == max_retries - 1: raise e - pass # we know it was persisted, so should have a stream ordering assert result_event.internal_metadata.stream_ordering @@ -618,6 +604,25 @@ class RoomMemberHandler(metaclass=abc.ABCMeta): Raises: ShadowBanError if a shadow-banned requester attempts to send an invite. """ + if ratelimit: + if action == Membership.JOIN: + # Only rate-limit if the user isn't already joined to the room, otherwise + # we'll end up blocking profile updates. + ( + current_membership, + _, + ) = await self.store.get_local_current_membership_for_user_in_room( + requester.user.to_string(), + room_id, + ) + if current_membership != Membership.JOIN: + await self._join_rate_limiter_local.ratelimit(requester) + await self._join_rate_per_room_limiter.ratelimit( + requester, key=room_id, update=False + ) + elif action == Membership.INVITE: + await self.ratelimit_invite(requester, room_id, target.to_string()) + if action == Membership.INVITE and requester.shadow_banned: # We randomly sleep a bit just to annoy the requester. await self.clock.sleep(random.randint(1, 10)) @@ -794,8 +799,6 @@ class RoomMemberHandler(metaclass=abc.ABCMeta): if effective_membership_state == Membership.INVITE: target_id = target.to_string() - if ratelimit: - await self.ratelimit_invite(requester, room_id, target_id) # block any attempts to invite the server notices mxid if target_id == self._server_notices_mxid: @@ -2002,7 +2005,6 @@ class RoomMemberMasterHandler(RoomMemberHandler): # in the meantime and context needs to be recomputed, so let's do so. if i == max_retries - 1: raise e - pass # we know it was persisted, so must have a stream ordering assert result_event.internal_metadata.stream_ordering @@ -2145,24 +2147,18 @@ class RoomForgetterHandler(StateDeltasHandler): await self._store.update_room_forgetter_stream_pos(max_pos) - async def _handle_deltas(self, deltas: List[Dict[str, Any]]) -> None: + async def _handle_deltas(self, deltas: List[StateDelta]) -> None: """Called with the state deltas to process""" for delta in deltas: - typ = delta["type"] - state_key = delta["state_key"] - room_id = delta["room_id"] - event_id = delta["event_id"] - prev_event_id = delta["prev_event_id"] - - if typ != EventTypes.Member: + if delta.event_type != EventTypes.Member: continue - if not self._hs.is_mine_id(state_key): + if not self._hs.is_mine_id(delta.state_key): continue change = await self._get_key_change( - prev_event_id, - event_id, + delta.prev_event_id, + delta.event_id, key_name="membership", public_value=Membership.JOIN, ) @@ -2171,7 +2167,7 @@ class RoomForgetterHandler(StateDeltasHandler): if is_leave: try: await self._room_member_handler.forget( - UserID.from_string(state_key), room_id + UserID.from_string(delta.state_key), delta.room_id ) except SynapseError as e: if e.code == 400: diff --git a/synapse/handlers/room_summary.py b/synapse/handlers/room_summary.py index dd559b4c45..1dfb12e065 100644 --- a/synapse/handlers/room_summary.py +++ b/synapse/handlers/room_summary.py @@ -703,24 +703,24 @@ class RoomSummaryHandler: # there should always be an entry assert stats is not None, "unable to retrieve stats for %s" % (room_id,) - entry = { - "room_id": stats["room_id"], - "name": stats["name"], - "topic": stats["topic"], - "canonical_alias": stats["canonical_alias"], - "num_joined_members": stats["joined_members"], - "avatar_url": stats["avatar"], - "join_rule": stats["join_rules"], + entry: JsonDict = { + "room_id": stats.room_id, + "name": stats.name, + "topic": stats.topic, + "canonical_alias": stats.canonical_alias, + "num_joined_members": stats.joined_members, + "avatar_url": stats.avatar, + "join_rule": stats.join_rules, "world_readable": ( - stats["history_visibility"] == HistoryVisibility.WORLD_READABLE + stats.history_visibility == HistoryVisibility.WORLD_READABLE ), - "guest_can_join": stats["guest_access"] == "can_join", - "room_type": stats["room_type"], + "guest_can_join": stats.guest_access == "can_join", + "room_type": stats.room_type, } if self._msc3266_enabled: - entry["im.nheko.summary.version"] = stats["version"] - entry["im.nheko.summary.encryption"] = stats["encryption"] + entry["im.nheko.summary.version"] = stats.version + entry["im.nheko.summary.encryption"] = stats.encryption # Federation requests need to provide additional information so the # requested server is able to filter the response appropriately. diff --git a/synapse/handlers/search.py b/synapse/handlers/search.py index aad4706f14..f51ed9d5bb 100644 --- a/synapse/handlers/search.py +++ b/synapse/handlers/search.py @@ -374,13 +374,13 @@ class SearchHandler: serialize_options = SerializeEventConfig(requester=requester) for context in contexts.values(): - context["events_before"] = self._event_serializer.serialize_events( + context["events_before"] = await self._event_serializer.serialize_events( context["events_before"], time_now, bundle_aggregations=aggregations, config=serialize_options, ) - context["events_after"] = self._event_serializer.serialize_events( + context["events_after"] = await self._event_serializer.serialize_events( context["events_after"], time_now, bundle_aggregations=aggregations, @@ -390,7 +390,7 @@ class SearchHandler: results = [ { "rank": search_result.rank_map[e.event_id], - "result": self._event_serializer.serialize_event( + "result": await self._event_serializer.serialize_event( e, time_now, bundle_aggregations=aggregations, @@ -409,7 +409,7 @@ class SearchHandler: if state_results: rooms_cat_res["state"] = { - room_id: self._event_serializer.serialize_events( + room_id: await self._event_serializer.serialize_events( state_events, time_now, config=serialize_options ) for room_id, state_events in state_results.items() diff --git a/synapse/handlers/sso.py b/synapse/handlers/sso.py index e9a544e754..62f2454f5d 100644 --- a/synapse/handlers/sso.py +++ b/synapse/handlers/sso.py @@ -1206,10 +1206,7 @@ class SsoHandler: # 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"] - + for user_id, device_id in devices: # 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: diff --git a/synapse/handlers/stats.py b/synapse/handlers/stats.py index 3dde19fc81..817b41aa37 100644 --- a/synapse/handlers/stats.py +++ b/synapse/handlers/stats.py @@ -27,6 +27,7 @@ from typing import ( from synapse.api.constants import EventContentFields, EventTypes, Membership from synapse.metrics import event_processing_positions from synapse.metrics.background_process_metrics import run_as_background_process +from synapse.storage.databases.main.state_deltas import StateDelta from synapse.types import JsonDict if TYPE_CHECKING: @@ -142,7 +143,7 @@ class StatsHandler: self.pos = max_pos async def _handle_deltas( - self, deltas: Iterable[JsonDict] + self, deltas: Iterable[StateDelta] ) -> Tuple[Dict[str, CounterType[str]], Dict[str, CounterType[str]]]: """Called with the state deltas to process @@ -157,51 +158,50 @@ class StatsHandler: room_to_state_updates: Dict[str, Dict[str, Any]] = {} for delta in deltas: - typ = delta["type"] - state_key = delta["state_key"] - room_id = delta["room_id"] - event_id = delta["event_id"] - stream_id = delta["stream_id"] - prev_event_id = delta["prev_event_id"] - - logger.debug("Handling: %r, %r %r, %s", room_id, typ, state_key, event_id) + logger.debug( + "Handling: %r, %r %r, %s", + delta.room_id, + delta.event_type, + delta.state_key, + delta.event_id, + ) - token = await self.store.get_earliest_token_for_stats("room", room_id) + token = await self.store.get_earliest_token_for_stats("room", delta.room_id) # If the earliest token to begin from is larger than our current # stream ID, skip processing this delta. - if token is not None and token >= stream_id: + if token is not None and token >= delta.stream_id: logger.debug( "Ignoring: %s as earlier than this room's initial ingestion event", - event_id, + delta.event_id, ) continue - if event_id is None and prev_event_id is None: + if delta.event_id is None and delta.prev_event_id is None: logger.error( "event ID is None and so is the previous event ID. stream_id: %s", - stream_id, + delta.stream_id, ) continue event_content: JsonDict = {} - if event_id is not None: - event = await self.store.get_event(event_id, allow_none=True) + if delta.event_id is not None: + event = await self.store.get_event(delta.event_id, allow_none=True) if event: event_content = event.content or {} # All the values in this dict are deltas (RELATIVE changes) - room_stats_delta = room_to_stats_deltas.setdefault(room_id, Counter()) + room_stats_delta = room_to_stats_deltas.setdefault(delta.room_id, Counter()) - room_state = room_to_state_updates.setdefault(room_id, {}) + room_state = room_to_state_updates.setdefault(delta.room_id, {}) - if prev_event_id is None: + if delta.prev_event_id is None: # this state event doesn't overwrite another, # so it is a new effective/current state event room_stats_delta["current_state_events"] += 1 - if typ == EventTypes.Member: + if delta.event_type == EventTypes.Member: # we could use StateDeltasHandler._get_key_change here but it's # a bit inefficient given we're not testing for a specific # result; might as well just grab the prev_membership and @@ -210,9 +210,9 @@ class StatsHandler: # in the absence of a previous event because we do not want to # reduce the leave count when a new-to-the-room user joins. prev_membership = None - if prev_event_id is not None: + if delta.prev_event_id is not None: prev_event = await self.store.get_event( - prev_event_id, allow_none=True + delta.prev_event_id, allow_none=True ) if prev_event: prev_event_content = prev_event.content @@ -256,7 +256,7 @@ class StatsHandler: else: raise ValueError("%r is not a valid membership" % (membership,)) - user_id = state_key + user_id = delta.state_key if self.is_mine_id(user_id): # this accounts for transitions like leave → ban and so on. has_changed_joinedness = (prev_membership == Membership.JOIN) != ( @@ -272,30 +272,30 @@ class StatsHandler: room_stats_delta["local_users_in_room"] += membership_delta - elif typ == EventTypes.Create: + elif delta.event_type == EventTypes.Create: room_state["is_federatable"] = ( event_content.get(EventContentFields.FEDERATE, True) is True ) room_type = event_content.get(EventContentFields.ROOM_TYPE) if isinstance(room_type, str): room_state["room_type"] = room_type - elif typ == EventTypes.JoinRules: + elif delta.event_type == EventTypes.JoinRules: room_state["join_rules"] = event_content.get("join_rule") - elif typ == EventTypes.RoomHistoryVisibility: + elif delta.event_type == EventTypes.RoomHistoryVisibility: room_state["history_visibility"] = event_content.get( "history_visibility" ) - elif typ == EventTypes.RoomEncryption: + elif delta.event_type == EventTypes.RoomEncryption: room_state["encryption"] = event_content.get("algorithm") - elif typ == EventTypes.Name: + elif delta.event_type == EventTypes.Name: room_state["name"] = event_content.get("name") - elif typ == EventTypes.Topic: + elif delta.event_type == EventTypes.Topic: room_state["topic"] = event_content.get("topic") - elif typ == EventTypes.RoomAvatar: + elif delta.event_type == EventTypes.RoomAvatar: room_state["avatar"] = event_content.get("url") - elif typ == EventTypes.CanonicalAlias: + elif delta.event_type == EventTypes.CanonicalAlias: room_state["canonical_alias"] = event_content.get("alias") - elif typ == EventTypes.GuestAccess: + elif delta.event_type == EventTypes.GuestAccess: room_state["guest_access"] = event_content.get( EventContentFields.GUEST_ACCESS ) diff --git a/synapse/handlers/sync.py b/synapse/handlers/sync.py index 7bd42f635f..2f1bc5a015 100644 --- a/synapse/handlers/sync.py +++ b/synapse/handlers/sync.py @@ -40,7 +40,6 @@ 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.device import DELETE_DEVICE_MSGS_TASK_NAME from synapse.handlers.relations import BundledAggregations from synapse.logging import issue9533_logger from synapse.logging.context import current_context @@ -58,6 +57,7 @@ from synapse.types import ( DeviceListUpdates, JsonDict, JsonMapping, + MultiWriterStreamToken, MutableStateMap, Requester, RoomStreamToken, @@ -363,36 +363,15 @@ class SyncHandler: # (since we now know that the device has received them) if since_token is not None: since_stream_id = since_token.to_device_key - # Fast path: delete a limited number of to-device messages up front. - # We do this to avoid the overhead of scheduling a task for every - # sync. - device_deletion_limit = 100 deleted = await self.store.delete_messages_for_device( sync_config.user.to_string(), sync_config.device_id, since_stream_id, - limit=device_deletion_limit, ) logger.debug( "Deleted %d to-device messages up to %d", deleted, since_stream_id ) - # If we hit the limit, schedule a background task to delete the rest. - if deleted >= device_deletion_limit: - await self._task_scheduler.schedule_task( - DELETE_DEVICE_MSGS_TASK_NAME, - resource_id=sync_config.device_id, - params={ - "user_id": sync_config.user.to_string(), - "device_id": sync_config.device_id, - "up_to_stream_id": since_stream_id, - }, - ) - logger.debug( - "Deletion of to-device messages up to %d scheduled", - since_stream_id, - ) - if timeout == 0 or since_token is None or full_state: # we are going to return immediately, so don't bother calling # notifier.wait_for_events. @@ -499,7 +478,11 @@ class SyncHandler: event_copy = {k: v for (k, v) in event.items() if k != "room_id"} ephemeral_by_room.setdefault(room_id, []).append(event_copy) - receipt_key = since_token.receipt_key if since_token else 0 + receipt_key = ( + since_token.receipt_key + if since_token + else MultiWriterStreamToken(stream=0) + ) receipt_source = self.event_sources.sources.receipt receipts, receipt_key = await receipt_source.get_new_events( @@ -522,12 +505,27 @@ class SyncHandler: async def _load_filtered_recents( self, room_id: str, + sync_result_builder: "SyncResultBuilder", sync_config: SyncConfig, - now_token: StreamToken, + upto_token: StreamToken, since_token: Optional[StreamToken] = None, potential_recents: Optional[List[EventBase]] = None, newly_joined_room: bool = False, ) -> TimelineBatch: + """Create a timeline batch for the room + + Args: + room_id + sync_result_builder + sync_config + upto_token: The token up to which we should fetch (more) events. + If `potential_results` is non-empty then this is *start* of + the the list. + since_token + potential_recents: If non-empty, the events between the since token + and current token to send down to clients. + newly_joined_room + """ with Measure(self.clock, "load_filtered_recents"): timeline_limit = sync_config.filter_collection.timeline_limit() block_all_timeline = ( @@ -543,6 +541,20 @@ class SyncHandler: else: limited = False + # Check if there is a gap, if so we need to mark this as limited and + # recalculate which events to send down. + gap_token = await self.store.get_timeline_gaps( + room_id, + since_token.room_key if since_token else None, + sync_result_builder.now_token.room_key, + ) + if gap_token: + # There's a gap, so we need to ignore the passed in + # `potential_recents`, and reset `upto_token` to match. + potential_recents = None + upto_token = sync_result_builder.now_token + limited = True + log_kv({"limited": limited}) if potential_recents: @@ -581,10 +593,10 @@ class SyncHandler: recents = [] if not limited or block_all_timeline: - prev_batch_token = now_token + prev_batch_token = upto_token if recents: room_key = recents[0].internal_metadata.before - prev_batch_token = now_token.copy_and_replace( + prev_batch_token = upto_token.copy_and_replace( StreamKeyType.ROOM, room_key ) @@ -595,11 +607,15 @@ class SyncHandler: filtering_factor = 2 load_limit = max(timeline_limit * filtering_factor, 10) max_repeat = 5 # Only try a few times per room, otherwise - room_key = now_token.room_key + room_key = upto_token.room_key end_key = room_key since_key = None - if since_token and not newly_joined_room: + if since_token and gap_token: + # If there is a gap then we need to only include events after + # it. + since_key = gap_token + elif since_token and not newly_joined_room: since_key = since_token.room_key while limited and len(recents) < timeline_limit and max_repeat: @@ -669,7 +685,7 @@ class SyncHandler: recents = recents[-timeline_limit:] room_key = recents[0].internal_metadata.before - prev_batch_token = now_token.copy_and_replace(StreamKeyType.ROOM, room_key) + prev_batch_token = upto_token.copy_and_replace(StreamKeyType.ROOM, room_key) # Don't bother to bundle aggregations if the timeline is unlimited, # as clients will have all the necessary information. @@ -684,7 +700,9 @@ class SyncHandler: return TimelineBatch( events=recents, prev_batch=prev_batch_token, - limited=limited or newly_joined_room, + # Also mark as limited if this is a new room or there has been a gap + # (to force client to paginate the gap). + limited=limited or newly_joined_room or gap_token is not None, bundled_aggregations=bundled_aggregations, ) @@ -1499,7 +1517,7 @@ class SyncHandler: # Presence data is included if the server has it enabled and not filtered out. include_presence_data = bool( - self.hs_config.server.use_presence + self.hs_config.server.presence_enabled and not sync_config.filter_collection.blocks_all_presence() ) # Device list updates are sent if a since token is provided. @@ -2333,7 +2351,7 @@ class SyncHandler: continue leave_token = now_token.copy_and_replace( - StreamKeyType.ROOM, RoomStreamToken(None, event.stream_ordering) + StreamKeyType.ROOM, RoomStreamToken(stream=event.stream_ordering) ) room_entries.append( RoomSyncResultBuilder( @@ -2419,8 +2437,9 @@ class SyncHandler: batch = await self._load_filtered_recents( room_id, + sync_result_builder, sync_config, - now_token=upto_token, + upto_token=upto_token, since_token=since_token, potential_recents=events, newly_joined_room=newly_joined, diff --git a/synapse/handlers/ui_auth/checkers.py b/synapse/handlers/ui_auth/checkers.py index 78a75bfed6..ab8f7610e9 100644 --- a/synapse/handlers/ui_auth/checkers.py +++ b/synapse/handlers/ui_auth/checkers.py @@ -187,9 +187,9 @@ class _BaseThreepidAuthChecker: if row: threepid = { - "medium": row["medium"], - "address": row["address"], - "validated_at": row["validated_at"], + "medium": row.medium, + "address": row.address, + "validated_at": row.validated_at, } # Valid threepid returned, delete from the db diff --git a/synapse/handlers/user_directory.py b/synapse/handlers/user_directory.py index a0f5568000..75717ba4f9 100644 --- a/synapse/handlers/user_directory.py +++ b/synapse/handlers/user_directory.py @@ -14,7 +14,7 @@ import logging from http import HTTPStatus -from typing import TYPE_CHECKING, Any, Dict, List, Optional, Set, Tuple +from typing import TYPE_CHECKING, List, Optional, Set, Tuple from twisted.internet.interfaces import IDelayedCall @@ -23,6 +23,7 @@ from synapse.api.constants import EventTypes, HistoryVisibility, JoinRules, Memb from synapse.api.errors import Codes, SynapseError from synapse.handlers.state_deltas import MatchChange, StateDeltasHandler from synapse.metrics.background_process_metrics import run_as_background_process +from synapse.storage.databases.main.state_deltas import StateDelta from synapse.storage.databases.main.user_directory import SearchResult from synapse.storage.roommember import ProfileInfo from synapse.types import UserID @@ -247,32 +248,31 @@ class UserDirectoryHandler(StateDeltasHandler): await self.store.update_user_directory_stream_pos(max_pos) - async def _handle_deltas(self, deltas: List[Dict[str, Any]]) -> None: + async def _handle_deltas(self, deltas: List[StateDelta]) -> None: """Called with the state deltas to process""" for delta in deltas: - typ = delta["type"] - state_key = delta["state_key"] - room_id = delta["room_id"] - event_id: Optional[str] = delta["event_id"] - prev_event_id: Optional[str] = delta["prev_event_id"] - - logger.debug("Handling: %r %r, %s", typ, state_key, event_id) + logger.debug( + "Handling: %r %r, %s", delta.event_type, delta.state_key, delta.event_id + ) # For join rule and visibility changes we need to check if the room # may have become public or not and add/remove the users in said room - if typ in (EventTypes.RoomHistoryVisibility, EventTypes.JoinRules): + if delta.event_type in ( + EventTypes.RoomHistoryVisibility, + EventTypes.JoinRules, + ): await self._handle_room_publicity_change( - room_id, prev_event_id, event_id, typ + delta.room_id, delta.prev_event_id, delta.event_id, delta.event_type ) - elif typ == EventTypes.Member: + elif delta.event_type == EventTypes.Member: await self._handle_room_membership_event( - room_id, - prev_event_id, - event_id, - state_key, + delta.room_id, + delta.prev_event_id, + delta.event_id, + delta.state_key, ) else: - logger.debug("Ignoring irrelevant type: %r", typ) + logger.debug("Ignoring irrelevant type: %r", delta.event_type) async def _handle_room_publicity_change( self, diff --git a/synapse/http/connectproxyclient.py b/synapse/http/connectproxyclient.py index 636efc33e8..59b914b87e 100644 --- a/synapse/http/connectproxyclient.py +++ b/synapse/http/connectproxyclient.py @@ -59,7 +59,7 @@ class BasicProxyCredentials(ProxyCredentials): a Proxy-Authorization header. """ # Encode as base64 and prepend the authorization type - return b"Basic " + base64.encodebytes(self.username_password) + return b"Basic " + base64.b64encode(self.username_password) @attr.s(auto_attribs=True) diff --git a/synapse/http/server.py b/synapse/http/server.py index 3bbf91298e..1e4e56f36b 100644 --- a/synapse/http/server.py +++ b/synapse/http/server.py @@ -266,7 +266,7 @@ class HttpServer(Protocol): def register_paths( self, method: str, - path_patterns: Iterable[Pattern], + path_patterns: Iterable[Pattern[str]], callback: ServletCallback, servlet_classname: str, ) -> None: diff --git a/synapse/logging/opentracing.py b/synapse/logging/opentracing.py index 4454fe29a5..e297fa9c8b 100644 --- a/synapse/logging/opentracing.py +++ b/synapse/logging/opentracing.py @@ -1019,11 +1019,14 @@ def tag_args(func: Callable[P, R]) -> Callable[P, R]: if not opentracing: return func + # getfullargspec is somewhat expensive, so ensure it is only called a single + # time (the function signature shouldn't change anyway). + argspec = inspect.getfullargspec(func) + @contextlib.contextmanager def _wrapping_logic( - func: Callable[P, R], *args: P.args, **kwargs: P.kwargs + _func: Callable[P, R], *args: P.args, **kwargs: P.kwargs ) -> Generator[None, None, None]: - argspec = inspect.getfullargspec(func) # We use `[1:]` to skip the `self` object reference and `start=1` to # make the index line up with `argspec.args`. # diff --git a/synapse/media/_base.py b/synapse/media/_base.py index 20cb8b9010..860e5ddca2 100644 --- a/synapse/media/_base.py +++ b/synapse/media/_base.py @@ -26,11 +26,11 @@ from twisted.internet.interfaces import IConsumer from twisted.protocols.basic import FileSender from twisted.web.server import Request -from synapse.api.errors import Codes, SynapseError, cs_error +from synapse.api.errors import Codes, cs_error from synapse.http.server import finish_request, respond_with_json from synapse.http.site import SynapseRequest from synapse.logging.context import make_deferred_yieldable -from synapse.util.stringutils import is_ascii, parse_and_validate_server_name +from synapse.util.stringutils import is_ascii logger = logging.getLogger(__name__) @@ -50,53 +50,46 @@ TEXT_CONTENT_TYPES = [ "text/xml", ] - -def parse_media_id(request: Request) -> Tuple[str, str, Optional[str]]: - """Parses the server name, media ID and optional file name from the request URI - - Also performs some rough validation on the server name. - - Args: - request: The `Request`. - - Returns: - A tuple containing the parsed server name, media ID and optional file name. - - Raises: - SynapseError(404): if parsing or validation fail for any reason - """ - try: - # The type on postpath seems incorrect in Twisted 21.2.0. - postpath: List[bytes] = request.postpath # type: ignore - assert postpath - - # This allows users to append e.g. /test.png to the URL. Useful for - # clients that parse the URL to see content type. - server_name_bytes, media_id_bytes = postpath[:2] - server_name = server_name_bytes.decode("utf-8") - media_id = media_id_bytes.decode("utf8") - - # Validate the server name, raising if invalid - parse_and_validate_server_name(server_name) - - file_name = None - if len(postpath) > 2: - try: - file_name = urllib.parse.unquote(postpath[-1].decode("utf-8")) - except UnicodeDecodeError: - pass - return server_name, media_id, file_name - except Exception: - raise SynapseError( - 404, "Invalid media id token %r" % (request.postpath,), Codes.UNKNOWN - ) +# A list of all content types that are "safe" to be rendered inline in a browser. +INLINE_CONTENT_TYPES = [ + "text/css", + "text/plain", + "text/csv", + "application/json", + "application/ld+json", + # We allow some media files deemed as safe, which comes from the matrix-react-sdk. + # https://github.com/matrix-org/matrix-react-sdk/blob/a70fcfd0bcf7f8c85986da18001ea11597989a7c/src/utils/blobs.ts#L51 + # SVGs are *intentionally* omitted. + "image/jpeg", + "image/gif", + "image/png", + "image/apng", + "image/webp", + "image/avif", + "video/mp4", + "video/webm", + "video/ogg", + "video/quicktime", + "audio/mp4", + "audio/webm", + "audio/aac", + "audio/mpeg", + "audio/ogg", + "audio/wave", + "audio/wav", + "audio/x-wav", + "audio/x-pn-wav", + "audio/flac", + "audio/x-flac", +] def respond_404(request: SynapseRequest) -> None: + assert request.path is not None respond_with_json( request, 404, - cs_error("Not found %r" % (request.postpath,), code=Codes.NOT_FOUND), + cs_error("Not found '%s'" % (request.path.decode(),), code=Codes.NOT_FOUND), send_cors=True, ) @@ -153,8 +146,15 @@ def add_file_headers( request.setHeader(b"Content-Type", content_type.encode("UTF-8")) - # Use a Content-Disposition of attachment to force download of media. - disposition = "attachment" + # A strict subset of content types is allowed to be inlined so that they may + # be viewed directly in a browser. Other file types are forced to be downloads. + # + # Only the type & subtype are important, parameters can be ignored. + if media_type.lower().split(";", 1)[0] in INLINE_CONTENT_TYPES: + disposition = "inline" + else: + disposition = "attachment" + if upload_name: # RFC6266 section 4.1 [1] defines both `filename` and `filename*`. # @@ -334,7 +334,7 @@ class ThumbnailInfo: # Content type of thumbnail, e.g. image/png type: str # The size of the media file, in bytes. - length: Optional[int] = None + length: int @attr.s(slots=True, frozen=True, auto_attribs=True) diff --git a/synapse/media/media_repository.py b/synapse/media/media_repository.py index 1b7b014f9a..72b0f1c5de 100644 --- a/synapse/media/media_repository.py +++ b/synapse/media/media_repository.py @@ -48,6 +48,7 @@ from synapse.media.filepath import MediaFilePaths from synapse.media.media_storage import MediaStorage from synapse.media.storage_provider import StorageProviderWrapper from synapse.media.thumbnailer import Thumbnailer, ThumbnailError +from synapse.media.url_previewer import UrlPreviewer from synapse.metrics.background_process_metrics import run_as_background_process from synapse.types import UserID from synapse.util.async_helpers import Linearizer @@ -114,7 +115,7 @@ class MediaRepository: ) storage_providers.append(provider) - self.media_storage = MediaStorage( + self.media_storage: MediaStorage = MediaStorage( self.hs, self.primary_base_path, self.filepaths, storage_providers ) @@ -142,6 +143,13 @@ class MediaRepository: MEDIA_RETENTION_CHECK_PERIOD_MS, ) + if hs.config.media.url_preview_enabled: + self.url_previewer: Optional[UrlPreviewer] = UrlPreviewer( + hs, self, self.media_storage + ) + else: + self.url_previewer = None + def _start_update_recently_accessed(self) -> Deferred: return run_as_background_process( "update_recently_accessed_media", self._update_recently_accessed @@ -616,6 +624,7 @@ class MediaRepository: height=t_height, method=t_method, type=t_type, + length=t_byte_source.tell(), ), ) @@ -686,6 +695,7 @@ class MediaRepository: height=t_height, method=t_method, type=t_type, + length=t_byte_source.tell(), ), ) @@ -831,6 +841,7 @@ class MediaRepository: height=t_height, method=t_method, type=t_type, + length=t_byte_source.tell(), ), ) @@ -938,10 +949,7 @@ class MediaRepository: deleted = 0 - for media in old_media: - origin = media["media_origin"] - media_id = media["media_id"] - file_id = media["filesystem_id"] + for origin, media_id, file_id in old_media: key = (origin, media_id) logger.info("Deleting: %r", key) diff --git a/synapse/metrics/_reactor_metrics.py b/synapse/metrics/_reactor_metrics.py index a2c6e6842d..dd486dd3e2 100644 --- a/synapse/metrics/_reactor_metrics.py +++ b/synapse/metrics/_reactor_metrics.py @@ -12,17 +12,45 @@ # See the License for the specific language governing permissions and # limitations under the License. -import select +import logging import time -from typing import Any, Iterable, List, Tuple +from selectors import SelectSelector, _PollLikeSelector # type: ignore[attr-defined] +from typing import Any, Callable, Iterable from prometheus_client import Histogram, Metric from prometheus_client.core import REGISTRY, GaugeMetricFamily -from twisted.internet import reactor +from twisted.internet import reactor, selectreactor +from twisted.internet.asyncioreactor import AsyncioSelectorReactor from synapse.metrics._types import Collector +try: + from selectors import KqueueSelector +except ImportError: + + class KqueueSelector: # type: ignore[no-redef] + pass + + +try: + from twisted.internet.epollreactor import EPollReactor +except ImportError: + + class EPollReactor: # type: ignore[no-redef] + pass + + +try: + from twisted.internet.pollreactor import PollReactor +except ImportError: + + class PollReactor: # type: ignore[no-redef] + pass + + +logger = logging.getLogger(__name__) + # # Twisted reactor metrics # @@ -34,52 +62,100 @@ tick_time = Histogram( ) -class EpollWrapper: - """a wrapper for an epoll object which records the time between polls""" +class CallWrapper: + """A wrapper for a callable which records the time between calls""" - def __init__(self, poller: "select.epoll"): # type: ignore[name-defined] + def __init__(self, wrapped: Callable[..., Any]): self.last_polled = time.time() - self._poller = poller + self._wrapped = wrapped - def poll(self, *args, **kwargs) -> List[Tuple[int, int]]: # type: ignore[no-untyped-def] - # record the time since poll() was last called. This gives a good proxy for + def __call__(self, *args, **kwargs) -> Any: # type: ignore[no-untyped-def] + # record the time since this was last called. This gives a good proxy for # how long it takes to run everything in the reactor - ie, how long anything # waiting for the next tick will have to wait. tick_time.observe(time.time() - self.last_polled) - ret = self._poller.poll(*args, **kwargs) + ret = self._wrapped(*args, **kwargs) self.last_polled = time.time() return ret + +class ObjWrapper: + """A wrapper for an object which wraps a specified method in CallWrapper. + + Other methods/attributes are passed to the original object. + + This is necessary when the wrapped object does not allow the attribute to be + overwritten. + """ + + def __init__(self, wrapped: Any, method_name: str): + self._wrapped = wrapped + self._method_name = method_name + self._wrapped_method = CallWrapper(getattr(wrapped, method_name)) + def __getattr__(self, item: str) -> Any: - return getattr(self._poller, item) + if item == self._method_name: + return self._wrapped_method + + return getattr(self._wrapped, item) class ReactorLastSeenMetric(Collector): - def __init__(self, epoll_wrapper: EpollWrapper): - self._epoll_wrapper = epoll_wrapper + def __init__(self, call_wrapper: CallWrapper): + self._call_wrapper = call_wrapper def collect(self) -> Iterable[Metric]: cm = GaugeMetricFamily( "python_twisted_reactor_last_seen", "Seconds since the Twisted reactor was last seen", ) - cm.add_metric([], time.time() - self._epoll_wrapper.last_polled) + cm.add_metric([], time.time() - self._call_wrapper.last_polled) yield cm +# Twisted has already select a reasonable reactor for us, so assumptions can be +# made about the shape. +wrapper = None try: - # if the reactor has a `_poller` attribute, which is an `epoll` object - # (ie, it's an EPollReactor), we wrap the `epoll` with a thing that will - # measure the time between ticks - from select import epoll # type: ignore[attr-defined] - - poller = reactor._poller # type: ignore[attr-defined] -except (AttributeError, ImportError): - pass -else: - if isinstance(poller, epoll): - poller = EpollWrapper(poller) - reactor._poller = poller # type: ignore[attr-defined] - REGISTRY.register(ReactorLastSeenMetric(poller)) + if isinstance(reactor, (PollReactor, EPollReactor)): + reactor._poller = ObjWrapper(reactor._poller, "poll") # type: ignore[attr-defined] + wrapper = reactor._poller._wrapped_method # type: ignore[attr-defined] + + elif isinstance(reactor, selectreactor.SelectReactor): + # Twisted uses a module-level _select function. + wrapper = selectreactor._select = CallWrapper(selectreactor._select) + + elif isinstance(reactor, AsyncioSelectorReactor): + # For asyncio look at the underlying asyncio event loop. + asyncio_loop = reactor._asyncioEventloop # A sub-class of BaseEventLoop, + + # A sub-class of BaseSelector. + selector = asyncio_loop._selector # type: ignore[attr-defined] + + if isinstance(selector, SelectSelector): + wrapper = selector._select = CallWrapper(selector._select) # type: ignore[attr-defined] + + # poll, epoll, and /dev/poll. + elif isinstance(selector, _PollLikeSelector): + selector._selector = ObjWrapper(selector._selector, "poll") # type: ignore[attr-defined] + wrapper = selector._selector._wrapped_method # type: ignore[attr-defined] + + elif isinstance(selector, KqueueSelector): + selector._selector = ObjWrapper(selector._selector, "control") # type: ignore[attr-defined] + wrapper = selector._selector._wrapped_method # type: ignore[attr-defined] + + else: + # E.g. this does not support the (Windows-only) ProactorEventLoop. + logger.warning( + "Skipping configuring ReactorLastSeenMetric: unexpected asyncio loop selector: %r via %r", + selector, + asyncio_loop, + ) +except Exception as e: + logger.warning("Configuring ReactorLastSeenMetric failed: %r", e) + + +if wrapper: + REGISTRY.register(ReactorLastSeenMetric(wrapper)) diff --git a/synapse/module_api/__init__.py b/synapse/module_api/__init__.py index 65e2aca456..755c59274c 100644 --- a/synapse/module_api/__init__.py +++ b/synapse/module_api/__init__.py @@ -23,6 +23,7 @@ from typing import ( Generator, Iterable, List, + Mapping, Optional, Tuple, TypeVar, @@ -39,6 +40,7 @@ from twisted.web.resource import Resource from synapse.api import errors from synapse.api.errors import SynapseError +from synapse.api.presence import UserPresenceState from synapse.config import ConfigError from synapse.events import EventBase from synapse.events.presence_router import ( @@ -46,6 +48,7 @@ from synapse.events.presence_router import ( GET_USERS_FOR_STATES_CALLBACK, PresenceRouter, ) +from synapse.events.utils import ADD_EXTRA_FIELDS_TO_UNSIGNED_CLIENT_EVENT_CALLBACK from synapse.handlers.account_data import ON_ACCOUNT_DATA_UPDATED_CALLBACK from synapse.handlers.auth import ( CHECK_3PID_AUTH_CALLBACK, @@ -257,6 +260,7 @@ class ModuleApi: self.custom_template_dir = hs.config.server.custom_template_directory self._callbacks = hs.get_module_api_callbacks() self.msc3861_oauth_delegation_enabled = hs.config.experimental.msc3861.enabled + self._event_serializer = hs.get_event_client_serializer() try: app_name = self._hs.config.email.email_app_name @@ -488,6 +492,25 @@ class ModuleApi: """ self._hs.register_module_web_resource(path, resource) + def register_add_extra_fields_to_unsigned_client_event_callbacks( + self, + *, + add_field_to_unsigned_callback: Optional[ + ADD_EXTRA_FIELDS_TO_UNSIGNED_CLIENT_EVENT_CALLBACK + ] = None, + ) -> None: + """Registers a callback that can be used to add fields to the unsigned + section of events. + + The callback is called every time an event is sent down to a client. + + Added in Synapse 1.96.0 + """ + if add_field_to_unsigned_callback is not None: + self._event_serializer.register_add_extra_fields_to_unsigned_client_event_callback( + add_field_to_unsigned_callback + ) + ######################################################################### # The following methods can be called by the module at any point in time. @@ -678,7 +701,7 @@ class ModuleApi: "msisdn" for phone numbers, and an "address" key which value is the threepid's address. """ - return await self._store.user_get_threepids(user_id) + return [attr.asdict(t) for t in await self._store.user_get_threepids(user_id)] def check_user_exists(self, user_id: str) -> "defer.Deferred[Optional[str]]": """Check if user exists. @@ -1184,6 +1207,37 @@ class ModuleApi: presence_events, [destination] ) + async def set_presence_for_users( + self, users: Mapping[str, Tuple[str, Optional[str]]] + ) -> None: + """ + Update the internal presence state of users. + + This can be used for either local or remote users. + + Note that this method can only be run on the process that is configured to write to the + presence stream. By default, this is the main process. + + Added in Synapse v1.96.0. + """ + + # We pull out the presence handler here to break a cyclic + # dependency between the presence router and module API. + presence_handler = self._hs.get_presence_handler() + + from synapse.handlers.presence import PresenceHandler + + assert isinstance(presence_handler, PresenceHandler) + + states = await presence_handler.current_state_for_users(users.keys()) + for user_id, (state, status_msg) in users.items(): + prev_state = states.setdefault(user_id, UserPresenceState.default(user_id)) + states[user_id] = prev_state.copy_and_replace( + state=state, status_msg=status_msg + ) + + await presence_handler._update_states(states.values(), force_notify=True) + def looping_background_call( self, f: Callable, diff --git a/synapse/notifier.py b/synapse/notifier.py index fc39e5c963..ee0bd84f1e 100644 --- a/synapse/notifier.py +++ b/synapse/notifier.py @@ -21,11 +21,13 @@ from typing import ( Dict, Iterable, List, + Literal, Optional, Set, Tuple, TypeVar, Union, + overload, ) import attr @@ -44,6 +46,7 @@ from synapse.metrics import LaterGauge from synapse.streams.config import PaginationConfig from synapse.types import ( JsonDict, + MultiWriterStreamToken, PersistedEventPosition, RoomStreamToken, StrCollection, @@ -126,8 +129,8 @@ class _NotifierUserStream: def notify( self, - stream_key: str, - stream_id: Union[int, RoomStreamToken], + stream_key: StreamKeyType, + stream_id: Union[int, RoomStreamToken, MultiWriterStreamToken], time_now_ms: int, ) -> None: """Notify any listeners for this user of a new event from an @@ -452,10 +455,48 @@ class Notifier: except Exception: logger.exception("Error pusher pool of event") + @overload def on_new_event( self, - stream_key: str, - new_token: Union[int, RoomStreamToken], + stream_key: Literal[StreamKeyType.ROOM], + new_token: RoomStreamToken, + users: Optional[Collection[Union[str, UserID]]] = None, + rooms: Optional[StrCollection] = None, + ) -> None: + ... + + @overload + def on_new_event( + self, + stream_key: Literal[StreamKeyType.RECEIPT], + new_token: MultiWriterStreamToken, + users: Optional[Collection[Union[str, UserID]]] = None, + rooms: Optional[StrCollection] = None, + ) -> None: + ... + + @overload + def on_new_event( + self, + stream_key: Literal[ + StreamKeyType.ACCOUNT_DATA, + StreamKeyType.DEVICE_LIST, + StreamKeyType.PRESENCE, + StreamKeyType.PUSH_RULES, + StreamKeyType.TO_DEVICE, + StreamKeyType.TYPING, + StreamKeyType.UN_PARTIAL_STATED_ROOMS, + ], + new_token: int, + users: Optional[Collection[Union[str, UserID]]] = None, + rooms: Optional[StrCollection] = None, + ) -> None: + ... + + def on_new_event( + self, + stream_key: StreamKeyType, + new_token: Union[int, RoomStreamToken, MultiWriterStreamToken], users: Optional[Collection[Union[str, UserID]]] = None, rooms: Optional[StrCollection] = None, ) -> None: @@ -655,30 +696,29 @@ class Notifier: events: List[Union[JsonDict, EventBase]] = [] end_token = from_token - for name, source in self.event_sources.sources.get_sources(): - keyname = "%s_key" % name - before_id = getattr(before_token, keyname) - after_id = getattr(after_token, keyname) + for keyname, source in self.event_sources.sources.get_sources(): + before_id = before_token.get_field(keyname) + after_id = after_token.get_field(keyname) if before_id == after_id: continue new_events, new_key = await source.get_new_events( user=user, - from_key=getattr(from_token, keyname), + from_key=from_token.get_field(keyname), limit=limit, is_guest=is_peeking, room_ids=room_ids, explicit_room_id=explicit_room_id, ) - if name == "room": + if keyname == StreamKeyType.ROOM: new_events = await filter_events_for_client( self._storage_controllers, user.to_string(), new_events, is_peeking=is_peeking, ) - elif name == "presence": + elif keyname == StreamKeyType.PRESENCE: now = self.clock.time_msec() new_events[:] = [ { diff --git a/synapse/push/__init__.py b/synapse/push/__init__.py index 9e3a98741a..4d405f2a0c 100644 --- a/synapse/push/__init__.py +++ b/synapse/push/__init__.py @@ -101,7 +101,7 @@ if TYPE_CHECKING: class PusherConfig: """Parameters necessary to configure a pusher.""" - id: Optional[str] + id: Optional[int] user_name: str profile_tag: str @@ -182,7 +182,7 @@ class Pusher(metaclass=abc.ABCMeta): raise NotImplementedError() @abc.abstractmethod - def on_new_receipts(self, min_stream_id: int, max_stream_id: int) -> None: + def on_new_receipts(self) -> None: raise NotImplementedError() @abc.abstractmethod diff --git a/synapse/push/bulk_push_rule_evaluator.py b/synapse/push/bulk_push_rule_evaluator.py index 14784312dc..5934b1ef34 100644 --- a/synapse/push/bulk_push_rule_evaluator.py +++ b/synapse/push/bulk_push_rule_evaluator.py @@ -25,10 +25,13 @@ from typing import ( Sequence, Tuple, Union, + cast, ) from prometheus_client import Counter +from twisted.internet.defer import Deferred + from synapse.api.constants import ( MAIN_TIMELINE, EventContentFields, @@ -40,11 +43,15 @@ from synapse.api.room_versions import PushRuleRoomFlag from synapse.event_auth import auth_types_for_event, get_user_power_level from synapse.events import EventBase, relation_from_event from synapse.events.snapshot import EventContext +from synapse.logging.context import make_deferred_yieldable, run_in_background from synapse.state import POWER_KEY from synapse.storage.databases.main.roommember import EventIdMembership +from synapse.storage.roommember import ProfileInfo from synapse.synapse_rust.push import FilteredPushRules, PushRuleEvaluator from synapse.types import JsonValue from synapse.types.state import StateFilter +from synapse.util import unwrapFirstError +from synapse.util.async_helpers import gather_results from synapse.util.caches import register_cache from synapse.util.metrics import measure_func from synapse.visibility import filter_event_for_clients_with_state @@ -342,15 +349,41 @@ class BulkPushRuleEvaluator: rules_by_user = await self._get_rules_for_event(event) actions_by_user: Dict[str, Collection[Union[Mapping, str]]] = {} - room_member_count = await self.store.get_number_joined_users_in_room( - event.room_id - ) - + # Gather a bunch of info in parallel. + # + # This has a lot of ignored types and casting due to the use of @cached + # decorated functions passed into run_in_background. + # + # See https://github.com/matrix-org/synapse/issues/16606 ( - power_levels, - sender_power_level, - ) = await self._get_power_levels_and_sender_level( - event, context, event_id_to_event + room_member_count, + (power_levels, sender_power_level), + related_events, + profiles, + ) = await make_deferred_yieldable( + cast( + "Deferred[Tuple[int, Tuple[dict, Optional[int]], Dict[str, Dict[str, JsonValue]], Mapping[str, ProfileInfo]]]", + gather_results( + ( + run_in_background( # type: ignore[call-arg] + self.store.get_number_joined_users_in_room, event.room_id # type: ignore[arg-type] + ), + run_in_background( + self._get_power_levels_and_sender_level, + event, + context, + event_id_to_event, + ), + run_in_background(self._related_events, event), + run_in_background( # type: ignore[call-arg] + self.store.get_subset_users_in_room_with_profiles, + event.room_id, # type: ignore[arg-type] + rules_by_user.keys(), # type: ignore[arg-type] + ), + ), + consumeErrors=True, + ).addErrback(unwrapFirstError), + ) ) # Find the event's thread ID. @@ -366,8 +399,6 @@ class BulkPushRuleEvaluator: # the parent is part of a thread. thread_id = await self.store.get_thread_id(relation.parent_id) - related_events = await self._related_events(event) - # It's possible that old room versions have non-integer power levels (floats or # strings; even the occasional `null`). For old rooms, we interpret these as if # they were integers. Do this here for the `@room` power level threshold. @@ -400,11 +431,6 @@ class BulkPushRuleEvaluator: self.hs.config.experimental.msc1767_enabled, # MSC3931 flag ) - users = rules_by_user.keys() - profiles = await self.store.get_subset_users_in_room_with_profiles( - event.room_id, users - ) - for uid, rules in rules_by_user.items(): if event.sender == uid: continue diff --git a/synapse/push/emailpusher.py b/synapse/push/emailpusher.py index 1710dd51b9..cf45fd09a8 100644 --- a/synapse/push/emailpusher.py +++ b/synapse/push/emailpusher.py @@ -99,7 +99,7 @@ class EmailPusher(Pusher): pass self.timed_call = None - def on_new_receipts(self, min_stream_id: int, max_stream_id: int) -> None: + def on_new_receipts(self) -> None: # We could wake up and cancel the timer but there tend to be quite a # lot of read receipts so it's probably less work to just let the # timer fire diff --git a/synapse/push/httppusher.py b/synapse/push/httppusher.py index 50027680cb..725910a659 100644 --- a/synapse/push/httppusher.py +++ b/synapse/push/httppusher.py @@ -160,7 +160,7 @@ class HttpPusher(Pusher): if should_check_for_notifs: self._start_processing() - def on_new_receipts(self, min_stream_id: int, max_stream_id: int) -> None: + def on_new_receipts(self) -> None: # Note that the min here shouldn't be relied upon to be accurate. # We could check the receipts are actually m.read receipts here, diff --git a/synapse/push/pusherpool.py b/synapse/push/pusherpool.py index 6517e3566f..15a2cc932f 100644 --- a/synapse/push/pusherpool.py +++ b/synapse/push/pusherpool.py @@ -292,20 +292,12 @@ class PusherPool: except Exception: logger.exception("Exception in pusher on_new_notifications") - async def on_new_receipts( - self, min_stream_id: int, max_stream_id: int, affected_room_ids: Iterable[str] - ) -> None: + async def on_new_receipts(self, users_affected: StrCollection) -> None: if not self.pushers: # nothing to do here. return try: - # Need to subtract 1 from the minimum because the lower bound here - # is not inclusive - users_affected = await self.store.get_users_sent_receipts_between( - min_stream_id - 1, max_stream_id - ) - for u in users_affected: # Don't push if the user account has expired expired = await self._account_validity_handler.is_user_expired(u) @@ -314,7 +306,7 @@ class PusherPool: if u in self.pushers: for p in self.pushers[u].values(): - p.on_new_receipts(min_stream_id, max_stream_id) + p.on_new_receipts() except Exception: logger.exception("Exception in pusher on_new_receipts") diff --git a/synapse/replication/http/_base.py b/synapse/replication/http/_base.py index 63cf24a14d..38701aea72 100644 --- a/synapse/replication/http/_base.py +++ b/synapse/replication/http/_base.py @@ -238,7 +238,7 @@ class ReplicationEndpoint(metaclass=abc.ABCMeta): data[_STREAM_POSITION_KEY] = { "streams": { - stream.NAME: stream.current_token(local_instance_name) + stream.NAME: stream.minimal_local_current_token() for stream in streams }, "instance_name": local_instance_name, @@ -433,7 +433,7 @@ class ReplicationEndpoint(metaclass=abc.ABCMeta): if self.WAIT_FOR_STREAMS: response[_STREAM_POSITION_KEY] = { - stream.NAME: stream.current_token(self._instance_name) + stream.NAME: stream.minimal_local_current_token() for stream in self._streams } diff --git a/synapse/replication/http/federation.py b/synapse/replication/http/federation.py index 53ad327030..e728297dce 100644 --- a/synapse/replication/http/federation.py +++ b/synapse/replication/http/federation.py @@ -138,7 +138,11 @@ class ReplicationFederationSendEventsRestServlet(ReplicationEndpoint): event_and_contexts.append((event, context)) - logger.info("Got %d events from federation", len(event_and_contexts)) + logger.info( + "Got batch of %i events to persist to room %s", + len(event_and_contexts), + room_id, + ) max_stream_id = await self.federation_event_handler.persist_events_and_notify( room_id, event_and_contexts, backfilled diff --git a/synapse/replication/http/send_events.py b/synapse/replication/http/send_events.py index 4f82c9f96d..8eea256063 100644 --- a/synapse/replication/http/send_events.py +++ b/synapse/replication/http/send_events.py @@ -118,6 +118,7 @@ class ReplicationSendEventsRestServlet(ReplicationEndpoint): with Measure(self.clock, "repl_send_events_parse"): events_and_context = [] events = payload["events"] + rooms = set() for event_payload in events: event_dict = event_payload["event"] @@ -144,11 +145,13 @@ class ReplicationSendEventsRestServlet(ReplicationEndpoint): UserID.from_string(u) for u in event_payload["extra_users"] ] - logger.info( - "Got batch of events to send, last ID of batch is: %s, sending into room: %s", - event.event_id, - event.room_id, - ) + # all the rooms *should* be the same, but we'll log separately to be + # sure. + rooms.add(event.room_id) + + logger.info( + "Got batch of %i events to persist to rooms %s", len(events), rooms + ) last_event = ( await self.event_creation_handler.persist_and_notify_client_events( diff --git a/synapse/replication/tcp/client.py b/synapse/replication/tcp/client.py index 1c7946522a..1312b6f21e 100644 --- a/synapse/replication/tcp/client.py +++ b/synapse/replication/tcp/client.py @@ -126,12 +126,11 @@ class ReplicationDataHandler: StreamKeyType.ACCOUNT_DATA, token, users=[row.user_id for row in rows] ) elif stream_name == ReceiptsStream.NAME: + new_token = self.store.get_max_receipt_stream_id() self.notifier.on_new_event( - StreamKeyType.RECEIPT, token, rooms=[row.room_id for row in rows] - ) - await self._pusher_pool.on_new_receipts( - token, token, {row.room_id for row in rows} + StreamKeyType.RECEIPT, new_token, rooms=[row.room_id for row in rows] ) + await self._pusher_pool.on_new_receipts({row.user_id for row in rows}) elif stream_name == ToDeviceStream.NAME: entities = [row.entity for row in rows if row.entity.startswith("@")] if entities: @@ -281,14 +280,6 @@ class ReplicationDataHandler: # may be streaming. self.notifier.notify_replication() - def on_remote_server_up(self, server: str) -> None: - """Called when get a new REMOTE_SERVER_UP command.""" - - # Let's wake up the transaction queue for the server in case we have - # pending stuff to send to it. - if self.send_handler: - self.send_handler.wake_destination(server) - async def wait_for_stream_position( self, instance_name: str, @@ -339,7 +330,7 @@ class ReplicationDataHandler: try: await make_deferred_yieldable(deferred) except defer.TimeoutError: - logger.error( + logger.warning( "Timed out waiting for repl stream %r to reach %s (%s)" "; currently at: %s", stream_name, @@ -407,9 +398,6 @@ class FederationSenderHandler: self._fed_position_linearizer = Linearizer(name="_fed_position_linearizer") - def wake_destination(self, server: str) -> None: - self.federation_sender.wake_destination(server) - async def process_replication_rows( self, stream_name: str, token: int, rows: list ) -> None: diff --git a/synapse/replication/tcp/commands.py b/synapse/replication/tcp/commands.py index e616b5e1c8..0f0f851b79 100644 --- a/synapse/replication/tcp/commands.py +++ b/synapse/replication/tcp/commands.py @@ -18,7 +18,7 @@ allowed to be sent by which side. """ import abc import logging -from typing import Optional, Tuple, Type, TypeVar +from typing import List, Optional, Tuple, Type, TypeVar from synapse.replication.tcp.streams._base import StreamRow from synapse.util import json_decoder, json_encoder @@ -74,6 +74,8 @@ SC = TypeVar("SC", bound="_SimpleCommand") class _SimpleCommand(Command): """An implementation of Command whose argument is just a 'data' string.""" + __slots__ = ["data"] + def __init__(self, data: str): self.data = data @@ -122,6 +124,8 @@ class RdataCommand(Command): RDATA presence master 59 ["@baz:example.com", "online", ...] """ + __slots__ = ["stream_name", "instance_name", "token", "row"] + NAME = "RDATA" def __init__( @@ -179,6 +183,8 @@ class PositionCommand(Command): of the stream. """ + __slots__ = ["stream_name", "instance_name", "prev_token", "new_token"] + NAME = "POSITION" def __init__( @@ -235,6 +241,8 @@ class ReplicateCommand(Command): REPLICATE """ + __slots__: List[str] = [] + NAME = "REPLICATE" def __init__(self) -> None: @@ -264,6 +272,8 @@ class UserSyncCommand(Command): Where <state> is either "start" or "end" """ + __slots__ = ["instance_id", "user_id", "device_id", "is_syncing", "last_sync_ms"] + NAME = "USER_SYNC" def __init__( @@ -316,6 +326,8 @@ class ClearUserSyncsCommand(Command): CLEAR_USER_SYNC <instance_id> """ + __slots__ = ["instance_id"] + NAME = "CLEAR_USER_SYNC" def __init__(self, instance_id: str): @@ -343,6 +355,8 @@ class FederationAckCommand(Command): FEDERATION_ACK <instance_name> <token> """ + __slots__ = ["instance_name", "token"] + NAME = "FEDERATION_ACK" def __init__(self, instance_name: str, token: int): @@ -368,6 +382,15 @@ class UserIpCommand(Command): USER_IP <user_id>, <access_token>, <ip>, <device_id>, <last_seen>, <user_agent> """ + __slots__ = [ + "user_id", + "access_token", + "ip", + "user_agent", + "device_id", + "last_seen", + ] + NAME = "USER_IP" def __init__( @@ -423,8 +446,6 @@ class RemoteServerUpCommand(_SimpleCommand): """Sent when a worker has detected that a remote server is no longer "down" and retry timings should be reset. - If sent from a client the server will relay to all other workers. - Format:: REMOTE_SERVER_UP <server> @@ -441,6 +462,8 @@ class LockReleasedCommand(Command): LOCK_RELEASED ["<instance_name>", "<lock_name>", "<lock_key>"] """ + __slots__ = ["instance_name", "lock_name", "lock_key"] + NAME = "LOCK_RELEASED" def __init__( diff --git a/synapse/replication/tcp/handler.py b/synapse/replication/tcp/handler.py index b668bb5da1..afd03137f0 100644 --- a/synapse/replication/tcp/handler.py +++ b/synapse/replication/tcp/handler.py @@ -611,10 +611,14 @@ class ReplicationCommandHandler: # Find where we previously streamed up to. current_token = stream.current_token(cmd.instance_name) - # If the position token matches our current token then we're up to - # date and there's nothing to do. Otherwise, fetch all updates - # between then and now. - missing_updates = cmd.prev_token != current_token + # If the incoming previous position is less than our current position + # then we're up to date and there's nothing to do. Otherwise, fetch + # all updates between then and now. + # + # Note: We also have to check that `current_token` is at most the + # new position, to handle the case where the stream gets "reset" + # (e.g. for `caches` and `typing` after the writer's restart). + missing_updates = not (cmd.prev_token <= current_token <= cmd.new_token) while missing_updates: # Note: There may very well not be any new updates, but we check to # make sure. This can particularly happen for the event stream where @@ -644,7 +648,7 @@ class ReplicationCommandHandler: [stream.parse_row(row) for row in rows], ) - logger.info("Caught up with stream '%s' to %i", stream_name, cmd.new_token) + logger.info("Caught up with stream '%s' to %i", stream_name, current_token) # We've now caught up to position sent to us, notify handler. await self._replication_data_handler.on_position( @@ -657,8 +661,6 @@ class ReplicationCommandHandler: self, conn: IReplicationConnection, cmd: RemoteServerUpCommand ) -> None: """Called when get a new REMOTE_SERVER_UP command.""" - self._replication_data_handler.on_remote_server_up(cmd.data) - self._notifier.notify_remote_server_up(cmd.data) def on_LOCK_RELEASED( diff --git a/synapse/replication/tcp/resource.py b/synapse/replication/tcp/resource.py index 1d9a29d22e..38abb5df54 100644 --- a/synapse/replication/tcp/resource.py +++ b/synapse/replication/tcp/resource.py @@ -27,7 +27,7 @@ from synapse.metrics.background_process_metrics import run_as_background_process from synapse.replication.tcp.commands import PositionCommand from synapse.replication.tcp.protocol import ServerReplicationStreamProtocol from synapse.replication.tcp.streams import EventsStream -from synapse.replication.tcp.streams._base import StreamRow, Token +from synapse.replication.tcp.streams._base import CachesStream, StreamRow, Token from synapse.util.metrics import Measure if TYPE_CHECKING: @@ -204,6 +204,23 @@ class ReplicationStreamer: # The token has advanced but there is no data to # send, so we send a `POSITION` to inform other # workers of the updated position. + # + # There are two reasons for this: 1) this instance + # requested a stream ID but didn't use it, or 2) + # this instance advanced its own stream position due + # to receiving notifications about other instances + # advancing their stream position. + + # We skip sending `POSITION` for the `caches` stream + # for the second case as a) it generates a lot of + # traffic as every worker would echo each write, and + # b) nothing cares if a given worker's caches stream + # position lags. + if stream.NAME == CachesStream.NAME: + # If there haven't been any writes since the + # `last_token` then we're in the second case. + if stream.minimal_local_current_token() <= last_token: + continue # Note: `last_token` may not *actually* be the # last token we sent out in a RDATA or POSITION. diff --git a/synapse/replication/tcp/streams/_base.py b/synapse/replication/tcp/streams/_base.py index c6088a0f99..58a44029aa 100644 --- a/synapse/replication/tcp/streams/_base.py +++ b/synapse/replication/tcp/streams/_base.py @@ -33,6 +33,7 @@ from synapse.replication.http.streams import ReplicationGetStreamUpdates if TYPE_CHECKING: from synapse.server import HomeServer + from synapse.storage.util.id_generators import AbstractStreamIdGenerator logger = logging.getLogger(__name__) @@ -107,22 +108,10 @@ class Stream: def __init__( self, local_instance_name: str, - current_token_function: Callable[[str], Token], update_function: UpdateFunction, ): """Instantiate a Stream - `current_token_function` and `update_function` are callbacks which - should be implemented by subclasses. - - `current_token_function` takes an instance name, which is a writer to - the stream, and returns the position in the stream of the writer (as - viewed from the current process). On the writer process this is where - the writer has successfully written up to, whereas on other processes - this is the position which we have received updates up to over - replication. (Note that most streams have a single writer and so their - implementations ignore the instance name passed in). - `update_function` is called to get updates for this stream between a pair of stream tokens. See the `UpdateFunction` type definition for more info. @@ -133,12 +122,28 @@ class Stream: update_function: callback go get stream updates, as above """ self.local_instance_name = local_instance_name - self.current_token = current_token_function self.update_function = update_function # The token from which we last asked for updates self.last_token = self.current_token(self.local_instance_name) + def current_token(self, instance_name: str) -> Token: + """This takes an instance name, which is a writer to + the stream, and returns the position in the stream of the writer (as + viewed from the current process). + """ + # We can't make this an abstract class as it makes mypy unhappy. + raise NotImplementedError() + + def minimal_local_current_token(self) -> Token: + """Tries to return a minimal current token for the local instance, + i.e. for writers this would be the last successful write. + + If local instance is not a writer (or has written yet) then falls back + to returning the normal "current token". + """ + raise NotImplementedError() + def discard_updates_and_advance(self) -> None: """Called when the stream should advance but the updates would be discarded, e.g. when there are no currently connected workers. @@ -156,6 +161,14 @@ class Stream: and `limited` is whether there are more updates to fetch. """ current_token = self.current_token(self.local_instance_name) + + # If the minimum current token for the local instance is less than or + # equal to the last thing we published, we know that there are no + # updates. + if self.last_token >= self.minimal_local_current_token(): + self.last_token = current_token + return [], current_token, False + updates, current_token, limited = await self.get_updates_since( self.local_instance_name, self.last_token, current_token ) @@ -190,6 +203,25 @@ class Stream: return updates, upto_token, limited +class _StreamFromIdGen(Stream): + """Helper class for simple streams that use a stream ID generator""" + + def __init__( + self, + local_instance_name: str, + update_function: UpdateFunction, + stream_id_gen: "AbstractStreamIdGenerator", + ): + self._stream_id_gen = stream_id_gen + super().__init__(local_instance_name, update_function) + + def current_token(self, instance_name: str) -> Token: + return self._stream_id_gen.get_current_token_for_writer(instance_name) + + def minimal_local_current_token(self) -> Token: + return self._stream_id_gen.get_minimal_local_current_token() + + def current_token_without_instance( current_token: Callable[[], int] ) -> Callable[[str], int]: @@ -242,17 +274,21 @@ class BackfillStream(Stream): self.store = hs.get_datastores().main super().__init__( hs.get_instance_name(), - self._current_token, self.store.get_all_new_backfill_event_rows, ) - def _current_token(self, instance_name: str) -> int: + def current_token(self, instance_name: str) -> Token: # The backfill stream over replication operates on *positive* numbers, # which means we need to negate it. return -self.store._backfill_id_gen.get_current_token_for_writer(instance_name) + def minimal_local_current_token(self) -> Token: + # The backfill stream over replication operates on *positive* numbers, + # which means we need to negate it. + return -self.store._backfill_id_gen.get_minimal_local_current_token() + -class PresenceStream(Stream): +class PresenceStream(_StreamFromIdGen): @attr.s(slots=True, frozen=True, auto_attribs=True) class PresenceStreamRow: user_id: str @@ -283,9 +319,7 @@ class PresenceStream(Stream): update_function = make_http_update_function(hs, self.NAME) super().__init__( - hs.get_instance_name(), - current_token_without_instance(store.get_current_presence_token), - update_function, + hs.get_instance_name(), update_function, store._presence_id_gen ) @@ -305,13 +339,18 @@ class PresenceFederationStream(Stream): ROW_TYPE = PresenceFederationStreamRow def __init__(self, hs: "HomeServer"): - federation_queue = hs.get_presence_handler().get_federation_queue() + self._federation_queue = hs.get_presence_handler().get_federation_queue() super().__init__( hs.get_instance_name(), - federation_queue.get_current_token, - federation_queue.get_replication_rows, + self._federation_queue.get_replication_rows, ) + def current_token(self, instance_name: str) -> Token: + return self._federation_queue.get_current_token(instance_name) + + def minimal_local_current_token(self) -> Token: + return self._federation_queue.get_current_token(self.local_instance_name) + class TypingStream(Stream): @attr.s(slots=True, frozen=True, auto_attribs=True) @@ -341,20 +380,25 @@ class TypingStream(Stream): update_function: Callable[ [str, int, int, int], Awaitable[Tuple[List[Tuple[int, Any]], int, bool]] ] = typing_writer_handler.get_all_typing_updates - current_token_function = typing_writer_handler.get_current_token + self.current_token_function = typing_writer_handler.get_current_token else: # Query the typing writer process update_function = make_http_update_function(hs, self.NAME) - current_token_function = hs.get_typing_handler().get_current_token + self.current_token_function = hs.get_typing_handler().get_current_token super().__init__( hs.get_instance_name(), - current_token_without_instance(current_token_function), update_function, ) + def current_token(self, instance_name: str) -> Token: + return self.current_token_function() -class ReceiptsStream(Stream): + def minimal_local_current_token(self) -> Token: + return self.current_token_function() + + +class ReceiptsStream(_StreamFromIdGen): @attr.s(slots=True, frozen=True, auto_attribs=True) class ReceiptsStreamRow: room_id: str @@ -371,12 +415,12 @@ class ReceiptsStream(Stream): store = hs.get_datastores().main super().__init__( hs.get_instance_name(), - current_token_without_instance(store.get_max_receipt_stream_id), store.get_all_updated_receipts, + store._receipts_id_gen, ) -class PushRulesStream(Stream): +class PushRulesStream(_StreamFromIdGen): """A user has changed their push rules""" @attr.s(slots=True, frozen=True, auto_attribs=True) @@ -387,20 +431,16 @@ class PushRulesStream(Stream): ROW_TYPE = PushRulesStreamRow def __init__(self, hs: "HomeServer"): - self.store = hs.get_datastores().main + store = hs.get_datastores().main super().__init__( hs.get_instance_name(), - self._current_token, - self.store.get_all_push_rule_updates, + store.get_all_push_rule_updates, + store._push_rules_stream_id_gen, ) - def _current_token(self, instance_name: str) -> int: - push_rules_token = self.store.get_max_push_rules_stream_id() - return push_rules_token - -class PushersStream(Stream): +class PushersStream(_StreamFromIdGen): """A user has added/changed/removed a pusher""" @attr.s(slots=True, frozen=True, auto_attribs=True) @@ -418,8 +458,8 @@ class PushersStream(Stream): super().__init__( hs.get_instance_name(), - current_token_without_instance(store.get_pushers_stream_token), store.get_all_updated_pushers_rows, + store._pushers_id_gen, ) @@ -447,15 +487,22 @@ class CachesStream(Stream): ROW_TYPE = CachesStreamRow def __init__(self, hs: "HomeServer"): - store = hs.get_datastores().main + self.store = hs.get_datastores().main super().__init__( hs.get_instance_name(), - store.get_cache_stream_token_for_writer, - store.get_all_updated_caches, + self.store.get_all_updated_caches, ) + def current_token(self, instance_name: str) -> Token: + return self.store.get_cache_stream_token_for_writer(instance_name) + + def minimal_local_current_token(self) -> Token: + if self.store._cache_id_gen: + return self.store._cache_id_gen.get_minimal_local_current_token() + return self.current_token(self.local_instance_name) + -class DeviceListsStream(Stream): +class DeviceListsStream(_StreamFromIdGen): """Either a user has updated their devices or a remote server needs to be told about a device update. """ @@ -473,8 +520,8 @@ class DeviceListsStream(Stream): self.store = hs.get_datastores().main super().__init__( hs.get_instance_name(), - current_token_without_instance(self.store.get_device_stream_token), self._update_function, + self.store._device_list_id_gen, ) async def _update_function( @@ -525,7 +572,7 @@ class DeviceListsStream(Stream): return updates, upper_limit_token, devices_limited or signatures_limited -class ToDeviceStream(Stream): +class ToDeviceStream(_StreamFromIdGen): """New to_device messages for a client""" @attr.s(slots=True, frozen=True, auto_attribs=True) @@ -539,12 +586,12 @@ class ToDeviceStream(Stream): store = hs.get_datastores().main super().__init__( hs.get_instance_name(), - current_token_without_instance(store.get_to_device_stream_token), store.get_all_new_device_messages, + store._device_inbox_id_gen, ) -class AccountDataStream(Stream): +class AccountDataStream(_StreamFromIdGen): """Global or per room account data was changed""" @attr.s(slots=True, frozen=True, auto_attribs=True) @@ -560,8 +607,8 @@ class AccountDataStream(Stream): self.store = hs.get_datastores().main super().__init__( hs.get_instance_name(), - current_token_without_instance(self.store.get_max_account_data_stream_id), self._update_function, + self.store._account_data_id_gen, ) async def _update_function( diff --git a/synapse/replication/tcp/streams/events.py b/synapse/replication/tcp/streams/events.py index ad9b760713..57138fea80 100644 --- a/synapse/replication/tcp/streams/events.py +++ b/synapse/replication/tcp/streams/events.py @@ -13,15 +13,16 @@ # See the License for the specific language governing permissions and # limitations under the License. import heapq +from collections import defaultdict from typing import TYPE_CHECKING, Iterable, Optional, Tuple, Type, TypeVar, cast import attr from synapse.replication.tcp.streams._base import ( - Stream, StreamRow, StreamUpdateResult, Token, + _StreamFromIdGen, ) if TYPE_CHECKING: @@ -51,8 +52,19 @@ data part are: * The state_key of the state which has changed * The event id of the new state +A "state-all" row is sent whenever the "current state" in a room changes, but there are +too many state updates for a particular room in the same update. This replaces any +"state" rows on a per-room basis. The fields in the data part are: + +* The room id for the state changes + """ +# Any room with more than _MAX_STATE_UPDATES_PER_ROOM will send a EventsStreamAllStateRow +# instead of individual EventsStreamEventRow. This is predominantly useful when +# purging large rooms. +_MAX_STATE_UPDATES_PER_ROOM = 150 + @attr.s(slots=True, frozen=True, auto_attribs=True) class EventsStreamRow: @@ -111,15 +123,23 @@ class EventsStreamCurrentStateRow(BaseEventsStreamRow): event_id: Optional[str] +@attr.s(slots=True, frozen=True, auto_attribs=True) +class EventsStreamAllStateRow(BaseEventsStreamRow): + TypeId = "state-all" + + room_id: str + + _EventRows: Tuple[Type[BaseEventsStreamRow], ...] = ( EventsStreamEventRow, EventsStreamCurrentStateRow, + EventsStreamAllStateRow, ) TypeToRow = {Row.TypeId: Row for Row in _EventRows} -class EventsStream(Stream): +class EventsStream(_StreamFromIdGen): """We received a new event, or an event went from being an outlier to not""" NAME = "events" @@ -127,9 +147,7 @@ class EventsStream(Stream): def __init__(self, hs: "HomeServer"): self._store = hs.get_datastores().main super().__init__( - hs.get_instance_name(), - self._store._stream_id_gen.get_current_token_for_writer, - self._update_function, + hs.get_instance_name(), self._update_function, self._store._stream_id_gen ) async def _update_function( @@ -139,6 +157,12 @@ class EventsStream(Stream): current_token: Token, target_row_count: int, ) -> StreamUpdateResult: + # The events stream cannot be "reset", so its safe to return early if + # the from token is larger than the current token (the DB query will + # trivially return 0 rows anyway). + if from_token >= current_token: + return [], current_token, False + # the events stream merges together three separate sources: # * new events # * current_state changes @@ -213,9 +237,28 @@ class EventsStream(Stream): if stream_id <= upper_limit ) + # Separate out rooms that have many state updates, listeners should clear + # all state for those rooms. + state_updates_by_room = defaultdict(list) + for stream_id, room_id, _type, _state_key, _event_id in state_rows: + state_updates_by_room[room_id].append(stream_id) + + state_all_rows = [ + (stream_ids[-1], room_id) + for room_id, stream_ids in state_updates_by_room.items() + if len(stream_ids) >= _MAX_STATE_UPDATES_PER_ROOM + ] + state_all_updates: Iterable[Tuple[int, Tuple]] = ( + (max_stream_id, (EventsStreamAllStateRow.TypeId, (room_id,))) + for (max_stream_id, room_id) in state_all_rows + ) + + # Any remaining state updates are sent individually. + state_all_rooms = {room_id for _, room_id in state_all_rows} state_updates: Iterable[Tuple[int, Tuple]] = ( (stream_id, (EventsStreamCurrentStateRow.TypeId, rest)) for (stream_id, *rest) in state_rows + if rest[0] not in state_all_rooms ) ex_outliers_updates: Iterable[Tuple[int, Tuple]] = ( @@ -224,7 +267,11 @@ class EventsStream(Stream): ) # we need to return a sorted list, so merge them together. - updates = list(heapq.merge(event_updates, state_updates, ex_outliers_updates)) + updates = list( + heapq.merge( + event_updates, state_all_updates, state_updates, ex_outliers_updates + ) + ) return updates, upper_limit, limited @classmethod diff --git a/synapse/replication/tcp/streams/federation.py b/synapse/replication/tcp/streams/federation.py index 4046bdec69..7f5af5852c 100644 --- a/synapse/replication/tcp/streams/federation.py +++ b/synapse/replication/tcp/streams/federation.py @@ -18,6 +18,7 @@ import attr from synapse.replication.tcp.streams._base import ( Stream, + Token, current_token_without_instance, make_http_update_function, ) @@ -47,7 +48,7 @@ class FederationStream(Stream): # will be a real FederationSender, which has stubs for current_token and # get_replication_rows.) federation_sender = hs.get_federation_sender() - current_token = current_token_without_instance( + self.current_token_func = current_token_without_instance( federation_sender.get_current_token ) update_function: Callable[ @@ -57,15 +58,21 @@ class FederationStream(Stream): elif hs.should_send_federation(): # federation sender: Query master process update_function = make_http_update_function(hs, self.NAME) - current_token = self._stub_current_token + self.current_token_func = self._stub_current_token else: # other worker: stub out the update function (we're not interested in # any updates so when we get a POSITION we do nothing) update_function = self._stub_update_function - current_token = self._stub_current_token + self.current_token_func = self._stub_current_token - super().__init__(hs.get_instance_name(), current_token, update_function) + super().__init__(hs.get_instance_name(), update_function) + + def current_token(self, instance_name: str) -> Token: + return self.current_token_func(instance_name) + + def minimal_local_current_token(self) -> Token: + return self.current_token(self.local_instance_name) @staticmethod def _stub_current_token(instance_name: str) -> int: diff --git a/synapse/replication/tcp/streams/partial_state.py b/synapse/replication/tcp/streams/partial_state.py index a8ce5ffd72..ad181d7e93 100644 --- a/synapse/replication/tcp/streams/partial_state.py +++ b/synapse/replication/tcp/streams/partial_state.py @@ -15,7 +15,7 @@ from typing import TYPE_CHECKING import attr -from synapse.replication.tcp.streams import Stream +from synapse.replication.tcp.streams._base import _StreamFromIdGen if TYPE_CHECKING: from synapse.server import HomeServer @@ -27,7 +27,7 @@ class UnPartialStatedRoomStreamRow: room_id: str -class UnPartialStatedRoomStream(Stream): +class UnPartialStatedRoomStream(_StreamFromIdGen): """ Stream to notify about rooms becoming un-partial-stated; that is, when the background sync finishes such that we now have full state for @@ -41,8 +41,8 @@ class UnPartialStatedRoomStream(Stream): store = hs.get_datastores().main super().__init__( hs.get_instance_name(), - store.get_un_partial_stated_rooms_token, store.get_un_partial_stated_rooms_from_stream, + store._un_partial_stated_rooms_stream_id_gen, ) @@ -56,7 +56,7 @@ class UnPartialStatedEventStreamRow: rejection_status_changed: bool -class UnPartialStatedEventStream(Stream): +class UnPartialStatedEventStream(_StreamFromIdGen): """ Stream to notify about events becoming un-partial-stated. """ @@ -68,6 +68,6 @@ class UnPartialStatedEventStream(Stream): store = hs.get_datastores().main super().__init__( hs.get_instance_name(), - store.get_un_partial_stated_events_token, store.get_un_partial_stated_events_from_stream, + store._un_partial_stated_events_stream_id_gen, ) diff --git a/synapse/rest/admin/__init__.py b/synapse/rest/admin/__init__.py index 7d0b4b55a0..9bd0d764f8 100644 --- a/synapse/rest/admin/__init__.py +++ b/synapse/rest/admin/__init__.py @@ -16,7 +16,6 @@ # limitations under the License. import logging -import platform from http import HTTPStatus from typing import TYPE_CHECKING, Optional, Tuple @@ -107,10 +106,7 @@ class VersionServlet(RestServlet): PATTERNS = admin_patterns("/server_version$") def __init__(self, hs: "HomeServer"): - self.res = { - "server_version": SYNAPSE_VERSION, - "python_version": platform.python_version(), - } + self.res = {"server_version": SYNAPSE_VERSION} def on_GET(self, request: SynapseRequest) -> Tuple[int, JsonDict]: return HTTPStatus.OK, self.res @@ -150,7 +146,7 @@ class PurgeHistoryRestServlet(RestServlet): # RoomStreamToken expects [int] not Optional[int] assert event.internal_metadata.stream_ordering is not None room_token = RoomStreamToken( - event.depth, event.internal_metadata.stream_ordering + topological=event.depth, stream=event.internal_metadata.stream_ordering ) token = await room_token.to_string(self.store) diff --git a/synapse/rest/admin/federation.py b/synapse/rest/admin/federation.py index e0ee55bd0e..a6ce787da1 100644 --- a/synapse/rest/admin/federation.py +++ b/synapse/rest/admin/federation.py @@ -85,7 +85,19 @@ class ListDestinationsRestServlet(RestServlet): destinations, total = await self._store.get_destinations_paginate( start, limit, destination, order_by, direction ) - response = {"destinations": destinations, "total": total} + response = { + "destinations": [ + { + "destination": r[0], + "retry_last_ts": r[1], + "retry_interval": r[2], + "failure_ts": r[3], + "last_successful_stream_ordering": r[4], + } + for r in destinations + ], + "total": total, + } if (start + limit) < total: response["next_token"] = str(start + len(destinations)) @@ -198,7 +210,13 @@ class DestinationMembershipRestServlet(RestServlet): rooms, total = await self._store.get_destination_rooms_paginate( destination, start, limit, direction ) - response = {"rooms": rooms, "total": total} + response = { + "rooms": [ + {"room_id": room_id, "stream_ordering": stream_ordering} + for room_id, stream_ordering in rooms + ], + "total": total, + } if (start + limit) < total: response["next_token"] = str(start + len(rooms)) diff --git a/synapse/rest/admin/media.py b/synapse/rest/admin/media.py index b7637dff0b..8cf5268854 100644 --- a/synapse/rest/admin/media.py +++ b/synapse/rest/admin/media.py @@ -17,6 +17,8 @@ import logging from http import HTTPStatus from typing import TYPE_CHECKING, Optional, Tuple +import attr + from synapse.api.constants import Direction from synapse.api.errors import Codes, NotFoundError, SynapseError from synapse.http.server import HttpServer @@ -418,7 +420,7 @@ class UserMediaRestServlet(RestServlet): start, limit, user_id, order_by, direction ) - ret = {"media": media, "total": total} + ret = {"media": [attr.asdict(m) for m in media], "total": total} if (start + limit) < total: ret["next_token"] = start + len(media) @@ -477,7 +479,7 @@ class UserMediaRestServlet(RestServlet): ) deleted_media, total = await self.media_repository.delete_local_media_ids( - [row["media_id"] for row in media] + [m.media_id for m in media] ) return HTTPStatus.OK, {"deleted_media": deleted_media, "total": total} diff --git a/synapse/rest/admin/registration_tokens.py b/synapse/rest/admin/registration_tokens.py index ffce92d45e..f3e06d3da3 100644 --- a/synapse/rest/admin/registration_tokens.py +++ b/synapse/rest/admin/registration_tokens.py @@ -77,7 +77,18 @@ class ListRegistrationTokensRestServlet(RestServlet): await assert_requester_is_admin(self.auth, request) valid = parse_boolean(request, "valid") token_list = await self.store.get_registration_tokens(valid) - return HTTPStatus.OK, {"registration_tokens": token_list} + return HTTPStatus.OK, { + "registration_tokens": [ + { + "token": t[0], + "uses_allowed": t[1], + "pending": t[2], + "completed": t[3], + "expiry_time": t[4], + } + for t in token_list + ] + } class NewRegistrationTokenRestServlet(RestServlet): diff --git a/synapse/rest/admin/rooms.py b/synapse/rest/admin/rooms.py index 436718c8b2..23a034522c 100644 --- a/synapse/rest/admin/rooms.py +++ b/synapse/rest/admin/rooms.py @@ -16,6 +16,8 @@ from http import HTTPStatus from typing import TYPE_CHECKING, List, Optional, Tuple, cast from urllib import parse as urlparse +import attr + from synapse.api.constants import Direction, EventTypes, JoinRules, Membership from synapse.api.errors import AuthError, Codes, NotFoundError, SynapseError from synapse.api.filtering import Filter @@ -306,10 +308,13 @@ class RoomRestServlet(RestServlet): raise NotFoundError("Room not found") members = await self.store.get_users_in_room(room_id) - ret["joined_local_devices"] = await self.store.count_devices_by_users(members) - ret["forgotten"] = await self.store.is_locally_forgotten_room(room_id) + result = attr.asdict(ret) + result["joined_local_devices"] = await self.store.count_devices_by_users( + members + ) + result["forgotten"] = await self.store.is_locally_forgotten_room(room_id) - return HTTPStatus.OK, ret + return HTTPStatus.OK, result async def on_DELETE( self, request: SynapseRequest, room_id: str @@ -444,7 +449,7 @@ class RoomStateRestServlet(RestServlet): event_ids = await self._storage_controllers.state.get_current_state_ids(room_id) events = await self.store.get_events(event_ids.values()) now = self.clock.time_msec() - room_state = self._event_serializer.serialize_events(events.values(), now) + room_state = await self._event_serializer.serialize_events(events.values(), now) ret = {"state": room_state} return HTTPStatus.OK, ret @@ -724,7 +729,17 @@ class ForwardExtremitiesRestServlet(ResolveRoomIdMixin, RestServlet): room_id, _ = await self.resolve_room_id(room_identifier) extremities = await self.store.get_forward_extremities_for_room(room_id) - return HTTPStatus.OK, {"count": len(extremities), "results": extremities} + result = [ + { + "event_id": ex[0], + "state_group": ex[1], + "depth": ex[2], + "received_ts": ex[3], + } + for ex in extremities + ] + + return HTTPStatus.OK, {"count": len(extremities), "results": result} class RoomEventContextServlet(RestServlet): @@ -779,22 +794,22 @@ class RoomEventContextServlet(RestServlet): time_now = self.clock.time_msec() results = { - "events_before": self._event_serializer.serialize_events( + "events_before": await self._event_serializer.serialize_events( event_context.events_before, time_now, bundle_aggregations=event_context.aggregations, ), - "event": self._event_serializer.serialize_event( + "event": await self._event_serializer.serialize_event( event_context.event, time_now, bundle_aggregations=event_context.aggregations, ), - "events_after": self._event_serializer.serialize_events( + "events_after": await self._event_serializer.serialize_events( event_context.events_after, time_now, bundle_aggregations=event_context.aggregations, ), - "state": self._event_serializer.serialize_events( + "state": await self._event_serializer.serialize_events( event_context.state, time_now ), "start": event_context.start, diff --git a/synapse/rest/admin/statistics.py b/synapse/rest/admin/statistics.py index 19780e4b4c..75d8a37ccf 100644 --- a/synapse/rest/admin/statistics.py +++ b/synapse/rest/admin/statistics.py @@ -108,7 +108,18 @@ class UserMediaStatisticsRestServlet(RestServlet): users_media, total = await self.store.get_users_media_usage_paginate( start, limit, from_ts, until_ts, order_by, direction, search_term ) - ret = {"users": users_media, "total": total} + ret = { + "users": [ + { + "user_id": r[0], + "displayname": r[1], + "media_count": r[2], + "media_length": r[3], + } + for r in users_media + ], + "total": total, + } if (start + limit) < total: ret["next_token"] = start + len(users_media) diff --git a/synapse/rest/admin/users.py b/synapse/rest/admin/users.py index 5b743a1d03..73878dd99d 100644 --- a/synapse/rest/admin/users.py +++ b/synapse/rest/admin/users.py @@ -18,6 +18,8 @@ import secrets from http import HTTPStatus from typing import TYPE_CHECKING, Dict, List, Optional, Tuple +import attr + from synapse.api.constants import Direction, UserTypes from synapse.api.errors import Codes, NotFoundError, SynapseError from synapse.http.servlet import ( @@ -161,11 +163,13 @@ class UsersRestServletV2(RestServlet): ) # If support for MSC3866 is not enabled, don't show the approval flag. + filter = None if not self._msc3866_enabled: - for user in users: - del user["approved"] - ret = {"users": users, "total": total} + def _filter(a: attr.Attribute) -> bool: + return a.name != "approved" + + ret = {"users": [attr.asdict(u, filter=filter) for u in users], "total": total} if (start + limit) < total: ret["next_token"] = str(start + len(users)) @@ -329,9 +333,8 @@ class UserRestServletV2(RestServlet): if threepids is not None: # get changed threepids (added and removed) - # convert List[Dict[str, Any]] into Set[Tuple[str, str]] cur_threepids = { - (threepid["medium"], threepid["address"]) + (threepid.medium, threepid.address) for threepid in await self.store.user_get_threepids(user_id) } add_threepids = new_threepids - cur_threepids @@ -842,7 +845,18 @@ class SearchUsersRestServlet(RestServlet): logger.info("term: %s ", term) ret = await self.store.search_users(term) - return HTTPStatus.OK, ret + results = [ + { + "name": name, + "password_hash": password_hash, + "is_guest": bool(is_guest), + "admin": bool(admin), + "user_type": user_type, + } + for name, password_hash, is_guest, admin, user_type in ret + ] + + return HTTPStatus.OK, results class UserAdminServlet(RestServlet): diff --git a/synapse/rest/client/account.py b/synapse/rest/client/account.py index e74a87af4d..641390cb30 100644 --- a/synapse/rest/client/account.py +++ b/synapse/rest/client/account.py @@ -24,6 +24,8 @@ if TYPE_CHECKING or HAS_PYDANTIC_V2: from pydantic.v1 import StrictBool, StrictStr, constr else: from pydantic import StrictBool, StrictStr, constr + +import attr from typing_extensions import Literal from twisted.web.server import Request @@ -595,7 +597,7 @@ class ThreepidRestServlet(RestServlet): threepids = await self.datastore.user_get_threepids(requester.user.to_string()) - return 200, {"threepids": threepids} + return 200, {"threepids": [attr.asdict(t) for t in threepids]} # NOTE(dmr): I have chosen not to use Pydantic to parse this request's body, because # the endpoint is deprecated. (If you really want to, you could do this by reusing diff --git a/synapse/rest/client/events.py b/synapse/rest/client/events.py index 3eca4fe21f..5705f812a5 100644 --- a/synapse/rest/client/events.py +++ b/synapse/rest/client/events.py @@ -93,7 +93,7 @@ class EventRestServlet(RestServlet): event = await self.event_handler.get_event(requester.user, None, event_id) if event: - result = self._event_serializer.serialize_event( + result = await self._event_serializer.serialize_event( event, self.clock.time_msec(), config=SerializeEventConfig(requester=requester), diff --git a/synapse/rest/client/notifications.py b/synapse/rest/client/notifications.py index e7fe1332e7..5688d8593d 100644 --- a/synapse/rest/client/notifications.py +++ b/synapse/rest/client/notifications.py @@ -87,7 +87,7 @@ class NotificationsServlet(RestServlet): "actions": pa.actions, "ts": pa.received_ts, "event": ( - self._event_serializer.serialize_event( + await self._event_serializer.serialize_event( notif_events[pa.event_id], now, config=serialize_options, diff --git a/synapse/rest/client/presence.py b/synapse/rest/client/presence.py index d578faa969..054a391f26 100644 --- a/synapse/rest/client/presence.py +++ b/synapse/rest/client/presence.py @@ -42,15 +42,13 @@ class PresenceStatusRestServlet(RestServlet): self.clock = hs.get_clock() self.auth = hs.get_auth() - self._use_presence = hs.config.server.use_presence - async def on_GET( self, request: SynapseRequest, user_id: str ) -> Tuple[int, JsonDict]: requester = await self.auth.get_user_by_req(request) user = UserID.from_string(user_id) - if not self._use_presence: + if not self.hs.config.server.presence_enabled: return 200, {"presence": "offline"} if requester.user != user: @@ -96,7 +94,7 @@ class PresenceStatusRestServlet(RestServlet): except Exception: raise SynapseError(400, "Unable to parse state") - if self._use_presence: + if self.hs.config.server.track_presence: await self.presence_handler.set_state(user, requester.device_id, state) return 200, {} diff --git a/synapse/rest/client/room.py b/synapse/rest/client/room.py index 553938ce9d..96f5726911 100644 --- a/synapse/rest/client/room.py +++ b/synapse/rest/client/room.py @@ -859,7 +859,7 @@ class RoomEventServlet(RestServlet): # per MSC2676, /rooms/{roomId}/event/{eventId}, should return the # *original* event, rather than the edited version - event_dict = self._event_serializer.serialize_event( + event_dict = await self._event_serializer.serialize_event( event, self.clock.time_msec(), bundle_aggregations=aggregations, @@ -911,25 +911,25 @@ class RoomEventContextServlet(RestServlet): time_now = self.clock.time_msec() serializer_options = SerializeEventConfig(requester=requester) results = { - "events_before": self._event_serializer.serialize_events( + "events_before": await self._event_serializer.serialize_events( event_context.events_before, time_now, bundle_aggregations=event_context.aggregations, config=serializer_options, ), - "event": self._event_serializer.serialize_event( + "event": await self._event_serializer.serialize_event( event_context.event, time_now, bundle_aggregations=event_context.aggregations, config=serializer_options, ), - "events_after": self._event_serializer.serialize_events( + "events_after": await self._event_serializer.serialize_events( event_context.events_after, time_now, bundle_aggregations=event_context.aggregations, config=serializer_options, ), - "state": self._event_serializer.serialize_events( + "state": await self._event_serializer.serialize_events( event_context.state, time_now, config=serializer_options, diff --git a/synapse/rest/client/sync.py b/synapse/rest/client/sync.py index 42bdd3bb10..33fde6c6f8 100644 --- a/synapse/rest/client/sync.py +++ b/synapse/rest/client/sync.py @@ -384,7 +384,7 @@ class SyncRestServlet(RestServlet): """ invited = {} for room in rooms: - invite = self._event_serializer.serialize_event( + invite = await self._event_serializer.serialize_event( room.invite, time_now, config=serialize_options ) unsigned = dict(invite.get("unsigned", {})) @@ -415,7 +415,7 @@ class SyncRestServlet(RestServlet): """ knocked = {} for room in rooms: - knock = self._event_serializer.serialize_event( + knock = await self._event_serializer.serialize_event( room.knock, time_now, config=serialize_options ) @@ -506,10 +506,10 @@ class SyncRestServlet(RestServlet): event.room_id, ) - serialized_state = self._event_serializer.serialize_events( + serialized_state = await self._event_serializer.serialize_events( state_events, time_now, config=serialize_options ) - serialized_timeline = self._event_serializer.serialize_events( + serialized_timeline = await self._event_serializer.serialize_events( timeline_events, time_now, config=serialize_options, diff --git a/synapse/rest/media/config_resource.py b/synapse/rest/media/config_resource.py index a95804d327..dbf5133c72 100644 --- a/synapse/rest/media/config_resource.py +++ b/synapse/rest/media/config_resource.py @@ -14,17 +14,19 @@ # limitations under the License. # +import re from typing import TYPE_CHECKING -from synapse.http.server import DirectServeJsonResource, respond_with_json +from synapse.http.server import respond_with_json +from synapse.http.servlet import RestServlet from synapse.http.site import SynapseRequest if TYPE_CHECKING: from synapse.server import HomeServer -class MediaConfigResource(DirectServeJsonResource): - isLeaf = True +class MediaConfigResource(RestServlet): + PATTERNS = [re.compile("/_matrix/media/(r0|v3|v1)/config$")] def __init__(self, hs: "HomeServer"): super().__init__() @@ -33,9 +35,6 @@ class MediaConfigResource(DirectServeJsonResource): self.auth = hs.get_auth() self.limits_dict = {"m.upload.size": config.media.max_upload_size} - async def _async_render_GET(self, request: SynapseRequest) -> None: + async def on_GET(self, request: SynapseRequest) -> None: await self.auth.get_user_by_req(request) respond_with_json(request, 200, self.limits_dict, send_cors=True) - - async def _async_render_OPTIONS(self, request: SynapseRequest) -> None: - respond_with_json(request, 200, {}, send_cors=True) diff --git a/synapse/rest/media/download_resource.py b/synapse/rest/media/download_resource.py index 3c618ef60a..65b9ff52fa 100644 --- a/synapse/rest/media/download_resource.py +++ b/synapse/rest/media/download_resource.py @@ -13,16 +13,14 @@ # See the License for the specific language governing permissions and # limitations under the License. import logging -from typing import TYPE_CHECKING +import re +from typing import TYPE_CHECKING, Optional -from synapse.http.server import ( - DirectServeJsonResource, - set_corp_headers, - set_cors_headers, -) -from synapse.http.servlet import parse_boolean +from synapse.http.server import set_corp_headers, set_cors_headers +from synapse.http.servlet import RestServlet, parse_boolean from synapse.http.site import SynapseRequest -from synapse.media._base import parse_media_id, respond_404 +from synapse.media._base import respond_404 +from synapse.util.stringutils import parse_and_validate_server_name if TYPE_CHECKING: from synapse.media.media_repository import MediaRepository @@ -31,15 +29,28 @@ if TYPE_CHECKING: logger = logging.getLogger(__name__) -class DownloadResource(DirectServeJsonResource): - isLeaf = True +class DownloadResource(RestServlet): + PATTERNS = [ + re.compile( + "/_matrix/media/(r0|v3|v1)/download/(?P<server_name>[^/]*)/(?P<media_id>[^/]*)(/(?P<file_name>[^/]*))?$" + ) + ] def __init__(self, hs: "HomeServer", media_repo: "MediaRepository"): super().__init__() self.media_repo = media_repo self._is_mine_server_name = hs.is_mine_server_name - async def _async_render_GET(self, request: SynapseRequest) -> None: + async def on_GET( + self, + request: SynapseRequest, + server_name: str, + media_id: str, + file_name: Optional[str] = None, + ) -> None: + # Validate the server name, raising if invalid + parse_and_validate_server_name(server_name) + set_cors_headers(request) set_corp_headers(request) request.setHeader( @@ -58,9 +69,8 @@ class DownloadResource(DirectServeJsonResource): b"Referrer-Policy", b"no-referrer", ) - server_name, media_id, name = parse_media_id(request) if self._is_mine_server_name(server_name): - await self.media_repo.get_local_media(request, media_id, name) + await self.media_repo.get_local_media(request, media_id, file_name) else: allow_remote = parse_boolean(request, "allow_remote", default=True) if not allow_remote: @@ -72,4 +82,6 @@ class DownloadResource(DirectServeJsonResource): respond_404(request) return - await self.media_repo.get_remote_media(request, server_name, media_id, name) + await self.media_repo.get_remote_media( + request, server_name, media_id, file_name + ) diff --git a/synapse/rest/media/media_repository_resource.py b/synapse/rest/media/media_repository_resource.py index 5ebaa3b032..2089bb1029 100644 --- a/synapse/rest/media/media_repository_resource.py +++ b/synapse/rest/media/media_repository_resource.py @@ -15,7 +15,7 @@ from typing import TYPE_CHECKING from synapse.config._base import ConfigError -from synapse.http.server import UnrecognizedRequestResource +from synapse.http.server import HttpServer, JsonResource from .config_resource import MediaConfigResource from .download_resource import DownloadResource @@ -27,7 +27,7 @@ if TYPE_CHECKING: from synapse.server import HomeServer -class MediaRepositoryResource(UnrecognizedRequestResource): +class MediaRepositoryResource(JsonResource): """File uploading and downloading. Uploads are POSTed to a resource which returns a token which is used to GET @@ -70,6 +70,11 @@ class MediaRepositoryResource(UnrecognizedRequestResource): width and height are close to the requested size and the aspect matches the requested size. The client should scale the image if it needs to fit within a given rectangle. + + This gets mounted at various points under /_matrix/media, including: + * /_matrix/media/r0 + * /_matrix/media/v1 + * /_matrix/media/v3 """ def __init__(self, hs: "HomeServer"): @@ -77,17 +82,23 @@ class MediaRepositoryResource(UnrecognizedRequestResource): if not hs.config.media.can_load_media_repo: raise ConfigError("Synapse is not configured to use a media repo.") - super().__init__() + JsonResource.__init__(self, hs, canonical_json=False) + self.register_servlets(self, hs) + + @staticmethod + def register_servlets(http_server: HttpServer, hs: "HomeServer") -> None: media_repo = hs.get_media_repository() - self.putChild(b"upload", UploadResource(hs, media_repo)) - self.putChild(b"download", DownloadResource(hs, media_repo)) - self.putChild( - b"thumbnail", ThumbnailResource(hs, media_repo, media_repo.media_storage) + # Note that many of these should not exist as v1 endpoints, but empirically + # a lot of traffic still goes to them. + + UploadResource(hs, media_repo).register(http_server) + DownloadResource(hs, media_repo).register(http_server) + ThumbnailResource(hs, media_repo, media_repo.media_storage).register( + http_server ) if hs.config.media.url_preview_enabled: - self.putChild( - b"preview_url", - PreviewUrlResource(hs, media_repo, media_repo.media_storage), + PreviewUrlResource(hs, media_repo, media_repo.media_storage).register( + http_server ) - self.putChild(b"config", MediaConfigResource(hs)) + MediaConfigResource(hs).register(http_server) diff --git a/synapse/rest/media/preview_url_resource.py b/synapse/rest/media/preview_url_resource.py index 58513c4be4..c8acb65dca 100644 --- a/synapse/rest/media/preview_url_resource.py +++ b/synapse/rest/media/preview_url_resource.py @@ -13,24 +13,20 @@ # See the License for the specific language governing permissions and # limitations under the License. +import re from typing import TYPE_CHECKING -from synapse.http.server import ( - DirectServeJsonResource, - respond_with_json, - respond_with_json_bytes, -) -from synapse.http.servlet import parse_integer, parse_string +from synapse.http.server import respond_with_json_bytes +from synapse.http.servlet import RestServlet, parse_integer, parse_string from synapse.http.site import SynapseRequest from synapse.media.media_storage import MediaStorage -from synapse.media.url_previewer import UrlPreviewer if TYPE_CHECKING: from synapse.media.media_repository import MediaRepository from synapse.server import HomeServer -class PreviewUrlResource(DirectServeJsonResource): +class PreviewUrlResource(RestServlet): """ The `GET /_matrix/media/r0/preview_url` endpoint provides a generic preview API for URLs which outputs Open Graph (https://ogp.me/) responses (with some Matrix @@ -48,7 +44,7 @@ class PreviewUrlResource(DirectServeJsonResource): * Matrix cannot be used to distribute the metadata between homeservers. """ - isLeaf = True + PATTERNS = [re.compile("/_matrix/media/(r0|v3|v1)/preview_url$")] def __init__( self, @@ -62,14 +58,10 @@ class PreviewUrlResource(DirectServeJsonResource): self.clock = hs.get_clock() self.media_repo = media_repo self.media_storage = media_storage + assert self.media_repo.url_previewer is not None + self.url_previewer = self.media_repo.url_previewer - self._url_previewer = UrlPreviewer(hs, media_repo, media_storage) - - async def _async_render_OPTIONS(self, request: SynapseRequest) -> None: - request.setHeader(b"Allow", b"OPTIONS, GET") - respond_with_json(request, 200, {}, send_cors=True) - - async def _async_render_GET(self, request: SynapseRequest) -> None: + async def on_GET(self, request: SynapseRequest) -> None: # XXX: if get_user_by_req fails, what should we do in an async render? requester = await self.auth.get_user_by_req(request) url = parse_string(request, "url", required=True) @@ -77,5 +69,5 @@ class PreviewUrlResource(DirectServeJsonResource): if ts is None: ts = self.clock.time_msec() - og = await self._url_previewer.preview(url, requester.user, ts) + og = await self.url_previewer.preview(url, requester.user, ts) respond_with_json_bytes(request, 200, og, send_cors=True) diff --git a/synapse/rest/media/thumbnail_resource.py b/synapse/rest/media/thumbnail_resource.py index 661e604b85..85b6bdbe72 100644 --- a/synapse/rest/media/thumbnail_resource.py +++ b/synapse/rest/media/thumbnail_resource.py @@ -13,29 +13,24 @@ # See the License for the specific language governing permissions and # limitations under the License. - import logging -from typing import TYPE_CHECKING, Any, Dict, List, Optional, Tuple +import re +from typing import TYPE_CHECKING, List, Optional, Tuple from synapse.api.errors import Codes, SynapseError, cs_error from synapse.config.repository import THUMBNAIL_SUPPORTED_MEDIA_FORMAT_MAP -from synapse.http.server import ( - DirectServeJsonResource, - respond_with_json, - set_corp_headers, - set_cors_headers, -) -from synapse.http.servlet import parse_integer, parse_string +from synapse.http.server import respond_with_json, set_corp_headers, set_cors_headers +from synapse.http.servlet import RestServlet, parse_integer, parse_string from synapse.http.site import SynapseRequest from synapse.media._base import ( FileInfo, ThumbnailInfo, - parse_media_id, respond_404, respond_with_file, respond_with_responder, ) from synapse.media.media_storage import MediaStorage +from synapse.util.stringutils import parse_and_validate_server_name if TYPE_CHECKING: from synapse.media.media_repository import MediaRepository @@ -44,8 +39,12 @@ if TYPE_CHECKING: logger = logging.getLogger(__name__) -class ThumbnailResource(DirectServeJsonResource): - isLeaf = True +class ThumbnailResource(RestServlet): + PATTERNS = [ + re.compile( + "/_matrix/media/(r0|v3|v1)/thumbnail/(?P<server_name>[^/]*)/(?P<media_id>[^/]*)$" + ) + ] def __init__( self, @@ -60,12 +59,17 @@ class ThumbnailResource(DirectServeJsonResource): self.media_storage = media_storage self.dynamic_thumbnails = hs.config.media.dynamic_thumbnails self._is_mine_server_name = hs.is_mine_server_name + self._server_name = hs.hostname self.prevent_media_downloads_from = hs.config.media.prevent_media_downloads_from - async def _async_render_GET(self, request: SynapseRequest) -> None: + async def on_GET( + self, request: SynapseRequest, server_name: str, media_id: str + ) -> None: + # Validate the server name, raising if invalid + parse_and_validate_server_name(server_name) + set_cors_headers(request) set_corp_headers(request) - server_name, media_id, _ = parse_media_id(request) width = parse_integer(request, "width", required=True) height = parse_integer(request, "height", required=True) method = parse_string(request, "method", "scale") @@ -155,30 +159,24 @@ class ThumbnailResource(DirectServeJsonResource): thumbnail_infos = await self.store.get_local_media_thumbnails(media_id) for info in thumbnail_infos: - t_w = info["thumbnail_width"] == desired_width - t_h = info["thumbnail_height"] == desired_height - t_method = info["thumbnail_method"] == desired_method - t_type = info["thumbnail_type"] == desired_type + t_w = info.width == desired_width + t_h = info.height == desired_height + t_method = info.method == desired_method + t_type = info.type == desired_type if t_w and t_h and t_method and t_type: file_info = FileInfo( server_name=None, file_id=media_id, url_cache=media_info["url_cache"], - thumbnail=ThumbnailInfo( - width=info["thumbnail_width"], - height=info["thumbnail_height"], - type=info["thumbnail_type"], - method=info["thumbnail_method"], - ), + thumbnail=info, ) - t_type = file_info.thumbnail_type - t_length = info["thumbnail_length"] - responder = await self.media_storage.fetch_media(file_info) if responder: - await respond_with_responder(request, responder, t_type, t_length) + await respond_with_responder( + request, responder, info.type, info.length + ) return logger.debug("We don't have a thumbnail of that size. Generating") @@ -218,29 +216,23 @@ class ThumbnailResource(DirectServeJsonResource): file_id = media_info["filesystem_id"] for info in thumbnail_infos: - t_w = info["thumbnail_width"] == desired_width - t_h = info["thumbnail_height"] == desired_height - t_method = info["thumbnail_method"] == desired_method - t_type = info["thumbnail_type"] == desired_type + t_w = info.width == desired_width + t_h = info.height == desired_height + t_method = info.method == desired_method + t_type = info.type == desired_type if t_w and t_h and t_method and t_type: file_info = FileInfo( server_name=server_name, file_id=media_info["filesystem_id"], - thumbnail=ThumbnailInfo( - width=info["thumbnail_width"], - height=info["thumbnail_height"], - type=info["thumbnail_type"], - method=info["thumbnail_method"], - ), + thumbnail=info, ) - t_type = file_info.thumbnail_type - t_length = info["thumbnail_length"] - responder = await self.media_storage.fetch_media(file_info) if responder: - await respond_with_responder(request, responder, t_type, t_length) + await respond_with_responder( + request, responder, info.type, info.length + ) return logger.debug("We don't have a thumbnail of that size. Generating") @@ -300,7 +292,7 @@ class ThumbnailResource(DirectServeJsonResource): desired_height: int, desired_method: str, desired_type: str, - thumbnail_infos: List[Dict[str, Any]], + thumbnail_infos: List[ThumbnailInfo], media_id: str, file_id: str, url_cache: bool, @@ -315,7 +307,7 @@ class ThumbnailResource(DirectServeJsonResource): desired_height: The desired height, the returned thumbnail may be larger than this. desired_method: The desired method used to generate the thumbnail. desired_type: The desired content-type of the thumbnail. - thumbnail_infos: A list of dictionaries of candidate thumbnails. + thumbnail_infos: A list of thumbnail info of candidate thumbnails. file_id: The ID of the media that a thumbnail is being requested for. url_cache: True if this is from a URL cache. server_name: The server name, if this is a remote thumbnail. @@ -418,13 +410,14 @@ class ThumbnailResource(DirectServeJsonResource): # `dynamic_thumbnails` is disabled. logger.info("Failed to find any generated thumbnails") + assert request.path is not None respond_with_json( request, 400, cs_error( - "Cannot find any thumbnails for the requested media (%r). This might mean the media is not a supported_media_format=(%s) or that thumbnailing failed for some other reason. (Dynamic thumbnails are disabled on this server.)" + "Cannot find any thumbnails for the requested media ('%s'). This might mean the media is not a supported_media_format=(%s) or that thumbnailing failed for some other reason. (Dynamic thumbnails are disabled on this server.)" % ( - request.postpath, + request.path.decode(), ", ".join(THUMBNAIL_SUPPORTED_MEDIA_FORMAT_MAP.keys()), ), code=Codes.UNKNOWN, @@ -438,7 +431,7 @@ class ThumbnailResource(DirectServeJsonResource): desired_height: int, desired_method: str, desired_type: str, - thumbnail_infos: List[Dict[str, Any]], + thumbnail_infos: List[ThumbnailInfo], file_id: str, url_cache: bool, server_name: Optional[str], @@ -451,7 +444,7 @@ class ThumbnailResource(DirectServeJsonResource): desired_height: The desired height, the returned thumbnail may be larger than this. desired_method: The desired method used to generate the thumbnail. desired_type: The desired content-type of the thumbnail. - thumbnail_infos: A list of dictionaries of candidate thumbnails. + thumbnail_infos: A list of thumbnail infos of candidate thumbnails. file_id: The ID of the media that a thumbnail is being requested for. url_cache: True if this is from a URL cache. server_name: The server name, if this is a remote thumbnail. @@ -469,21 +462,25 @@ class ThumbnailResource(DirectServeJsonResource): if desired_method == "crop": # Thumbnails that match equal or larger sizes of desired width/height. - crop_info_list: List[Tuple[int, int, int, bool, int, Dict[str, Any]]] = [] + crop_info_list: List[ + Tuple[int, int, int, bool, Optional[int], ThumbnailInfo] + ] = [] # Other thumbnails. - crop_info_list2: List[Tuple[int, int, int, bool, int, Dict[str, Any]]] = [] + crop_info_list2: List[ + Tuple[int, int, int, bool, Optional[int], ThumbnailInfo] + ] = [] for info in thumbnail_infos: # Skip thumbnails generated with different methods. - if info["thumbnail_method"] != "crop": + if info.method != "crop": continue - t_w = info["thumbnail_width"] - t_h = info["thumbnail_height"] + t_w = info.width + t_h = info.height aspect_quality = abs(d_w * t_h - d_h * t_w) min_quality = 0 if d_w <= t_w and d_h <= t_h else 1 size_quality = abs((d_w - t_w) * (d_h - t_h)) - type_quality = desired_type != info["thumbnail_type"] - length_quality = info["thumbnail_length"] + type_quality = desired_type != info.type + length_quality = info.length if t_w >= d_w or t_h >= d_h: crop_info_list.append( ( @@ -508,7 +505,7 @@ class ThumbnailResource(DirectServeJsonResource): ) # Pick the most appropriate thumbnail. Some values of `desired_width` and # `desired_height` may result in a tie, in which case we avoid comparing on - # the thumbnail info dictionary and pick the thumbnail that appears earlier + # the thumbnail info and pick the thumbnail that appears earlier # in the list of candidates. if crop_info_list: thumbnail_info = min(crop_info_list, key=lambda t: t[:-1])[-1] @@ -516,20 +513,20 @@ class ThumbnailResource(DirectServeJsonResource): thumbnail_info = min(crop_info_list2, key=lambda t: t[:-1])[-1] elif desired_method == "scale": # Thumbnails that match equal or larger sizes of desired width/height. - info_list: List[Tuple[int, bool, int, Dict[str, Any]]] = [] + info_list: List[Tuple[int, bool, int, ThumbnailInfo]] = [] # Other thumbnails. - info_list2: List[Tuple[int, bool, int, Dict[str, Any]]] = [] + info_list2: List[Tuple[int, bool, int, ThumbnailInfo]] = [] for info in thumbnail_infos: # Skip thumbnails generated with different methods. - if info["thumbnail_method"] != "scale": + if info.method != "scale": continue - t_w = info["thumbnail_width"] - t_h = info["thumbnail_height"] + t_w = info.width + t_h = info.height size_quality = abs((d_w - t_w) * (d_h - t_h)) - type_quality = desired_type != info["thumbnail_type"] - length_quality = info["thumbnail_length"] + type_quality = desired_type != info.type + length_quality = info.length if t_w >= d_w or t_h >= d_h: info_list.append((size_quality, type_quality, length_quality, info)) else: @@ -538,7 +535,7 @@ class ThumbnailResource(DirectServeJsonResource): ) # Pick the most appropriate thumbnail. Some values of `desired_width` and # `desired_height` may result in a tie, in which case we avoid comparing on - # the thumbnail info dictionary and pick the thumbnail that appears earlier + # the thumbnail info and pick the thumbnail that appears earlier # in the list of candidates. if info_list: thumbnail_info = min(info_list, key=lambda t: t[:-1])[-1] @@ -550,13 +547,7 @@ class ThumbnailResource(DirectServeJsonResource): file_id=file_id, url_cache=url_cache, server_name=server_name, - thumbnail=ThumbnailInfo( - width=thumbnail_info["thumbnail_width"], - height=thumbnail_info["thumbnail_height"], - type=thumbnail_info["thumbnail_type"], - method=thumbnail_info["thumbnail_method"], - length=thumbnail_info["thumbnail_length"], - ), + thumbnail=thumbnail_info, ) # No matching thumbnail was found. diff --git a/synapse/rest/media/upload_resource.py b/synapse/rest/media/upload_resource.py index 043e8d6077..949326d85d 100644 --- a/synapse/rest/media/upload_resource.py +++ b/synapse/rest/media/upload_resource.py @@ -14,11 +14,12 @@ # limitations under the License. import logging +import re from typing import IO, TYPE_CHECKING, Dict, List, Optional from synapse.api.errors import Codes, SynapseError -from synapse.http.server import DirectServeJsonResource, respond_with_json -from synapse.http.servlet import parse_bytes_from_args +from synapse.http.server import respond_with_json +from synapse.http.servlet import RestServlet, parse_bytes_from_args from synapse.http.site import SynapseRequest from synapse.media.media_storage import SpamMediaException @@ -29,8 +30,8 @@ if TYPE_CHECKING: logger = logging.getLogger(__name__) -class UploadResource(DirectServeJsonResource): - isLeaf = True +class UploadResource(RestServlet): + PATTERNS = [re.compile("/_matrix/media/(r0|v3|v1)/upload")] def __init__(self, hs: "HomeServer", media_repo: "MediaRepository"): super().__init__() @@ -43,10 +44,7 @@ class UploadResource(DirectServeJsonResource): self.max_upload_size = hs.config.media.max_upload_size self.clock = hs.get_clock() - async def _async_render_OPTIONS(self, request: SynapseRequest) -> None: - respond_with_json(request, 200, {}, send_cors=True) - - async def _async_render_POST(self, request: SynapseRequest) -> None: + async def on_POST(self, request: SynapseRequest) -> None: requester = await self.auth.get_user_by_req(request) raw_content_length = request.getHeader("Content-Length") if raw_content_length is None: diff --git a/synapse/server.py b/synapse/server.py index 71ead524d6..5bfb4ba4eb 100644 --- a/synapse/server.py +++ b/synapse/server.py @@ -786,7 +786,7 @@ class HomeServer(metaclass=abc.ABCMeta): @cache_in_self def get_event_client_serializer(self) -> EventClientSerializer: - return EventClientSerializer() + return EventClientSerializer(self) @cache_in_self def get_password_policy_handler(self) -> PasswordPolicyHandler: diff --git a/synapse/storage/background_updates.py b/synapse/storage/background_updates.py index b66347a79f..1b1e46a64c 100644 --- a/synapse/storage/background_updates.py +++ b/synapse/storage/background_updates.py @@ -28,6 +28,7 @@ from typing import ( Sequence, Tuple, Type, + cast, ) import attr @@ -492,14 +493,14 @@ class BackgroundUpdater: True if we have finished running all the background updates, otherwise False """ - def get_background_updates_txn(txn: Cursor) -> List[Dict[str, Any]]: + def get_background_updates_txn(txn: Cursor) -> List[Tuple[str, Optional[str]]]: txn.execute( """ SELECT update_name, depends_on FROM background_updates ORDER BY ordering, update_name """ ) - return self.db_pool.cursor_to_dict(txn) + return cast(List[Tuple[str, Optional[str]]], txn.fetchall()) if not self._current_background_update: all_pending_updates = await self.db_pool.runInteraction( @@ -511,14 +512,13 @@ class BackgroundUpdater: return True # find the first update which isn't dependent on another one in the queue. - pending = {update["update_name"] for update in all_pending_updates} - for upd in all_pending_updates: - depends_on = upd["depends_on"] + pending = {update_name for update_name, depends_on in all_pending_updates} + for update_name, depends_on in all_pending_updates: if not depends_on or depends_on not in pending: break logger.info( "Not starting on bg update %s until %s is done", - upd["update_name"], + update_name, depends_on, ) else: @@ -528,7 +528,7 @@ class BackgroundUpdater: "another: dependency cycle?" ) - self._current_background_update = upd["update_name"] + self._current_background_update = update_name # We have a background update to run, otherwise we would have returned # early. diff --git a/synapse/storage/controllers/persist_events.py b/synapse/storage/controllers/persist_events.py index f39ae2d635..1529c86cc5 100644 --- a/synapse/storage/controllers/persist_events.py +++ b/synapse/storage/controllers/persist_events.py @@ -542,13 +542,15 @@ class EventsPersistenceStorageController: return await res.get_state(self._state_controller, StateFilter.all()) async def _persist_event_batch( - self, _room_id: str, task: _PersistEventsTask + self, room_id: str, task: _PersistEventsTask ) -> Dict[str, str]: """Callback for the _event_persist_queue Calculates the change to current state and forward extremities, and persists the given events and with those updates. + Assumes that we are only persisting events for one room at a time. + Returns: A dictionary of event ID to event ID we didn't persist as we already had another event persisted with the same TXN ID. @@ -594,140 +596,23 @@ class EventsPersistenceStorageController: # We can't easily parallelize these since different chunks # might contain the same event. :( - # NB: Assumes that we are only persisting events for one room - # at a time. - - # map room_id->set[event_ids] giving the new forward - # extremities in each room - new_forward_extremities: Dict[str, Set[str]] = {} - - # map room_id->(to_delete, to_insert) where to_delete is a list - # of type/state keys to remove from current state, and to_insert - # is a map (type,key)->event_id giving the state delta in each - # room - state_delta_for_room: Dict[str, DeltaState] = {} + new_forward_extremities = None + state_delta_for_room = None if not backfilled: with Measure(self._clock, "_calculate_state_and_extrem"): - # Work out the new "current state" for each room. + # Work out the new "current state" for the room. # We do this by working out what the new extremities are and then # calculating the state from that. - events_by_room: Dict[str, List[Tuple[EventBase, EventContext]]] = {} - for event, context in chunk: - events_by_room.setdefault(event.room_id, []).append( - (event, context) - ) - - for room_id, ev_ctx_rm in events_by_room.items(): - latest_event_ids = ( - await self.main_store.get_latest_event_ids_in_room(room_id) - ) - new_latest_event_ids = await self._calculate_new_extremities( - room_id, ev_ctx_rm, latest_event_ids - ) - - if new_latest_event_ids == latest_event_ids: - # No change in extremities, so no change in state - continue - - # there should always be at least one forward extremity. - # (except during the initial persistence of the send_join - # results, in which case there will be no existing - # extremities, so we'll `continue` above and skip this bit.) - assert new_latest_event_ids, "No forward extremities left!" - - new_forward_extremities[room_id] = new_latest_event_ids - - len_1 = ( - len(latest_event_ids) == 1 - and len(new_latest_event_ids) == 1 - ) - if len_1: - all_single_prev_not_state = all( - len(event.prev_event_ids()) == 1 - and not event.is_state() - for event, ctx in ev_ctx_rm - ) - # Don't bother calculating state if they're just - # a long chain of single ancestor non-state events. - if all_single_prev_not_state: - continue - - state_delta_counter.inc() - if len(new_latest_event_ids) == 1: - state_delta_single_event_counter.inc() - - # This is a fairly handwavey check to see if we could - # have guessed what the delta would have been when - # processing one of these events. - # What we're interested in is if the latest extremities - # were the same when we created the event as they are - # now. When this server creates a new event (as opposed - # to receiving it over federation) it will use the - # forward extremities as the prev_events, so we can - # guess this by looking at the prev_events and checking - # if they match the current forward extremities. - for ev, _ in ev_ctx_rm: - prev_event_ids = set(ev.prev_event_ids()) - if latest_event_ids == prev_event_ids: - state_delta_reuse_delta_counter.inc() - break - - logger.debug("Calculating state delta for room %s", room_id) - with Measure( - self._clock, "persist_events.get_new_state_after_events" - ): - res = await self._get_new_state_after_events( - room_id, - ev_ctx_rm, - latest_event_ids, - new_latest_event_ids, - ) - current_state, delta_ids, new_latest_event_ids = res - - # there should always be at least one forward extremity. - # (except during the initial persistence of the send_join - # results, in which case there will be no existing - # extremities, so we'll `continue` above and skip this bit.) - assert new_latest_event_ids, "No forward extremities left!" - - new_forward_extremities[room_id] = new_latest_event_ids - - # If either are not None then there has been a change, - # and we need to work out the delta (or use that - # given) - delta = None - if delta_ids is not None: - # If there is a delta we know that we've - # only added or replaced state, never - # removed keys entirely. - delta = DeltaState([], delta_ids) - elif current_state is not None: - with Measure( - self._clock, "persist_events.calculate_state_delta" - ): - delta = await self._calculate_state_delta( - room_id, current_state - ) - - if delta: - # If we have a change of state then lets check - # whether we're actually still a member of the room, - # or if our last user left. If we're no longer in - # the room then we delete the current state and - # extremities. - is_still_joined = await self._is_server_still_joined( - room_id, - ev_ctx_rm, - delta, - ) - if not is_still_joined: - logger.info("Server no longer in room %s", room_id) - delta.no_longer_in_room = True - - state_delta_for_room[room_id] = delta + ( + new_forward_extremities, + state_delta_for_room, + ) = await self._calculate_new_forward_extremities_and_state_delta( + room_id, chunk + ) await self.persist_events_store._persist_events_and_state_updates( + room_id, chunk, state_delta_for_room=state_delta_for_room, new_forward_extremities=new_forward_extremities, @@ -737,6 +622,117 @@ class EventsPersistenceStorageController: return replaced_events + async def _calculate_new_forward_extremities_and_state_delta( + self, room_id: str, ev_ctx_rm: List[Tuple[EventBase, EventContext]] + ) -> Tuple[Optional[Set[str]], Optional[DeltaState]]: + """Calculates the new forward extremities and state delta for a room + given events to persist. + + Assumes that we are only persisting events for one room at a time. + + Returns: + A tuple of: + A set of str giving the new forward extremities the room + + The state delta for the room. + """ + + latest_event_ids = await self.main_store.get_latest_event_ids_in_room(room_id) + new_latest_event_ids = await self._calculate_new_extremities( + room_id, ev_ctx_rm, latest_event_ids + ) + + if new_latest_event_ids == latest_event_ids: + # No change in extremities, so no change in state + return (None, None) + + # there should always be at least one forward extremity. + # (except during the initial persistence of the send_join + # results, in which case there will be no existing + # extremities, so we'll `continue` above and skip this bit.) + assert new_latest_event_ids, "No forward extremities left!" + + new_forward_extremities = new_latest_event_ids + + len_1 = len(latest_event_ids) == 1 and len(new_latest_event_ids) == 1 + if len_1: + all_single_prev_not_state = all( + len(event.prev_event_ids()) == 1 and not event.is_state() + for event, ctx in ev_ctx_rm + ) + # Don't bother calculating state if they're just + # a long chain of single ancestor non-state events. + if all_single_prev_not_state: + return (new_forward_extremities, None) + + state_delta_counter.inc() + if len(new_latest_event_ids) == 1: + state_delta_single_event_counter.inc() + + # This is a fairly handwavey check to see if we could + # have guessed what the delta would have been when + # processing one of these events. + # What we're interested in is if the latest extremities + # were the same when we created the event as they are + # now. When this server creates a new event (as opposed + # to receiving it over federation) it will use the + # forward extremities as the prev_events, so we can + # guess this by looking at the prev_events and checking + # if they match the current forward extremities. + for ev, _ in ev_ctx_rm: + prev_event_ids = set(ev.prev_event_ids()) + if latest_event_ids == prev_event_ids: + state_delta_reuse_delta_counter.inc() + break + + logger.debug("Calculating state delta for room %s", room_id) + with Measure(self._clock, "persist_events.get_new_state_after_events"): + res = await self._get_new_state_after_events( + room_id, + ev_ctx_rm, + latest_event_ids, + new_latest_event_ids, + ) + current_state, delta_ids, new_latest_event_ids = res + + # there should always be at least one forward extremity. + # (except during the initial persistence of the send_join + # results, in which case there will be no existing + # extremities, so we'll `continue` above and skip this bit.) + assert new_latest_event_ids, "No forward extremities left!" + + new_forward_extremities = new_latest_event_ids + + # If either are not None then there has been a change, + # and we need to work out the delta (or use that + # given) + delta = None + if delta_ids is not None: + # If there is a delta we know that we've + # only added or replaced state, never + # removed keys entirely. + delta = DeltaState([], delta_ids) + elif current_state is not None: + with Measure(self._clock, "persist_events.calculate_state_delta"): + delta = await self._calculate_state_delta(room_id, current_state) + + if delta: + # If we have a change of state then lets check + # whether we're actually still a member of the room, + # or if our last user left. If we're no longer in + # the room then we delete the current state and + # extremities. + is_still_joined = await self._is_server_still_joined( + room_id, + ev_ctx_rm, + delta, + ) + if not is_still_joined: + logger.info("Server no longer in room %s", room_id) + delta.no_longer_in_room = True + + return (new_forward_extremities, delta) + async def _calculate_new_extremities( self, room_id: str, diff --git a/synapse/storage/controllers/state.py b/synapse/storage/controllers/state.py index 46957723a1..9f7959c45d 100644 --- a/synapse/storage/controllers/state.py +++ b/synapse/storage/controllers/state.py @@ -16,7 +16,6 @@ from itertools import chain from typing import ( TYPE_CHECKING, AbstractSet, - Any, Callable, Collection, Dict, @@ -32,6 +31,7 @@ from typing import ( from synapse.api.constants import EventTypes, Membership from synapse.events import EventBase from synapse.logging.opentracing import tag_args, trace +from synapse.storage.databases.main.state_deltas import StateDelta from synapse.storage.roommember import ProfileInfo from synapse.storage.util.partial_state_events_tracker import ( PartialCurrentStateTracker, @@ -531,19 +531,9 @@ class StateStorageController: @tag_args async def get_current_state_deltas( self, prev_stream_id: int, max_stream_id: int - ) -> Tuple[int, List[Dict[str, Any]]]: + ) -> Tuple[int, List[StateDelta]]: """Fetch a list of room state changes since the given stream id - Each entry in the result contains the following fields: - - stream_id (int) - - room_id (str) - - type (str): event type - - state_key (str): - - event_id (str|None): new event_id for this state key. None if the - state has been deleted. - - prev_event_id (str|None): previous event_id for this state key. None - if it's new state. - Args: prev_stream_id: point to get changes since (exclusive) max_stream_id: the point that we know has been correctly persisted diff --git a/synapse/storage/database.py b/synapse/storage/database.py index 695229bc91..05775425b7 100644 --- a/synapse/storage/database.py +++ b/synapse/storage/database.py @@ -19,7 +19,6 @@ import logging import time import types from collections import defaultdict -from sys import intern from time import monotonic as monotonic_time from typing import ( TYPE_CHECKING, @@ -36,7 +35,6 @@ from typing import ( Tuple, Type, TypeVar, - Union, cast, overload, ) @@ -63,7 +61,8 @@ from synapse.storage.engines import ( BaseDatabaseEngine, Psycopg2Engine, PsycopgEngine, - Sqlite3Engine, PostgresEngine, + Sqlite3Engine, + PostgresEngine, ) from synapse.storage.types import Connection, Cursor, SQLQueryParameters from synapse.util.async_helpers import delay_cancellation @@ -416,7 +415,6 @@ class LoggingTransaction: assert isinstance(self.database_engine, PostgresEngine) if isinstance(self.database_engine, Psycopg2Engine): - from psycopg2.extras import execute_values return self._do_execute( @@ -470,6 +468,16 @@ class LoggingTransaction: self._do_execute(self.txn.execute, sql, parameters) def executemany(self, sql: str, *args: Any) -> None: + """Repeatedly execute the same piece of SQL with different parameters. + + See https://peps.python.org/pep-0249/#executemany. Note in particular that + + > Use of this method for an operation which produces one or more result sets + > constitutes undefined behavior + + so you can't use this for e.g. a SELECT, an UPDATE ... RETURNING, or a + DELETE FROM... RETURNING. + """ # TODO: we should add a type for *args here. Looking at Cursor.executemany # and DBAPI2 it ought to be Sequence[_Parameter], but we pass in # Iterable[Iterable[Any]] in execute_batch and execute_values above, which mypy @@ -661,13 +669,16 @@ class DatabasePool: If the background updates have not completed, wait 15 sec and check again. """ - updates = await self.simple_select_list( - "background_updates", - keyvalues=None, - retcols=["update_name"], - desc="check_background_updates", + updates = cast( + List[Tuple[str]], + await self.simple_select_list( + "background_updates", + keyvalues=None, + retcols=["update_name"], + desc="check_background_updates", + ), ) - background_update_names = [x["update_name"] for x in updates] + background_update_names = [x[0] for x in updates] for table, update_name in UNIQUE_INDEX_BACKGROUND_UPDATES.items(): if update_name not in background_update_names: @@ -1085,57 +1096,20 @@ class DatabasePool: self._db_pool.runWithConnection(inner_func, *args, **kwargs) ) - @staticmethod - def cursor_to_dict(cursor: Cursor) -> List[Dict[str, Any]]: - """Converts a SQL cursor into an list of dicts. - - Args: - cursor: The DBAPI cursor which has executed a query. - Returns: - A list of dicts where the key is the column header. - """ - assert cursor.description is not None, "cursor.description was None" - col_headers = [intern(str(column[0])) for column in cursor.description] - results = [dict(zip(col_headers, row)) for row in cursor] - return results - - @overload - async def execute( - self, desc: str, decoder: Literal[None], query: str, *args: Any - ) -> List[Tuple[Any, ...]]: - ... - - @overload - async def execute( - self, desc: str, decoder: Callable[[Cursor], R], query: str, *args: Any - ) -> R: - ... - - async def execute( - self, - desc: str, - decoder: Optional[Callable[[Cursor], R]], - query: str, - *args: Any, - ) -> Union[List[Tuple[Any, ...]], R]: + async def execute(self, desc: str, query: str, *args: Any) -> List[Tuple[Any, ...]]: """Runs a single query for a result set. Args: desc: description of the transaction, for logging and metrics - decoder - The function which can resolve the cursor results to - something meaningful. query - The query string to execute *args - Query args. Returns: The result of decoder(results) """ - def interaction(txn: LoggingTransaction) -> Union[List[Tuple[Any, ...]], R]: + def interaction(txn: LoggingTransaction) -> List[Tuple[Any, ...]]: txn.execute(query, args) - if decoder: - return decoder(txn) - else: - return txn.fetchall() + return txn.fetchall() return await self.runInteraction(desc, interaction) @@ -1211,6 +1185,9 @@ class DatabasePool: keys: list of column names values: for each row, a list of values in the same order as `keys` """ + # If there's nothing to insert, then skip executing the query. + if not values: + return if isinstance(txn.database_engine, Psycopg2Engine): # We use `execute_values` as it can be a lot faster than `execute_batch`, @@ -1490,12 +1467,12 @@ class DatabasePool: allvalues.update(values) latter = "UPDATE SET " + ", ".join(k + "=EXCLUDED." + k for k in values) - sql = "INSERT INTO %s (%s) VALUES (%s) ON CONFLICT (%s) %s DO %s" % ( + sql = "INSERT INTO %s (%s) VALUES (%s) ON CONFLICT (%s) %sDO %s" % ( table, ", ".join(k for k in allvalues), ", ".join("?" for _ in allvalues), ", ".join(k for k in keyvalues), - f"WHERE {where_clause}" if where_clause else "", + f"WHERE {where_clause} " if where_clause else "", latter, ) txn.execute(sql, list(allvalues.values())) @@ -1544,7 +1521,7 @@ class DatabasePool: key_names: Collection[str], key_values: Collection[Iterable[Any]], value_names: Collection[str], - value_values: Iterable[Iterable[Any]], + value_values: Collection[Iterable[Any]], ) -> None: """ Upsert, many times. @@ -1557,6 +1534,19 @@ class DatabasePool: value_values: A list of each row's value column values. Ignored if value_names is empty. """ + # If there's nothing to upsert, then skip executing the query. + if not key_values: + return + + # No value columns, therefore make a blank list so that the following + # zip() works correctly. + if not value_names: + value_values = [() for x in range(len(key_values))] + elif len(value_values) != len(key_values): + raise ValueError( + f"{len(key_values)} key rows and {len(value_values)} value rows: should be the same number." + ) + if table not in self._unsafe_to_upsert_tables: return self.simple_upsert_many_txn_native_upsert( txn, table, key_names, key_values, value_names, value_values @@ -1591,10 +1581,6 @@ class DatabasePool: value_values: A list of each row's value column values. Ignored if value_names is empty. """ - # No value columns, therefore make a blank list so that the following - # zip() works correctly. - if not value_names: - value_values = [() for x in range(len(key_values))] # Lock the table just once, to prevent it being done once per row. # Note that, according to Postgres' documentation, once obtained, @@ -1632,10 +1618,7 @@ class DatabasePool: allnames.extend(value_names) if not value_names: - # No value columns, therefore make a blank list so that the - # following zip() works correctly. latter = "NOTHING" - value_values = [() for x in range(len(key_values))] else: latter = "UPDATE SET " + ", ".join( k + "=EXCLUDED." + k for k in value_names @@ -1867,9 +1850,9 @@ class DatabasePool: keyvalues: Optional[Dict[str, Any]], retcols: Collection[str], desc: str = "simple_select_list", - ) -> List[Dict[str, Any]]: + ) -> List[Tuple[Any, ...]]: """Executes a SELECT query on the named table, which may return zero or - more rows, returning the result as a list of dicts. + more rows, returning the result as a list of tuples. Args: table: the table name @@ -1880,8 +1863,7 @@ class DatabasePool: desc: description of the transaction, for logging and metrics Returns: - A list of dictionaries, one per result row, each a mapping between the - column names from `retcols` and that column's value for the row. + A list of tuples, one per result row, each the retcolumn's value for the row. """ return await self.runInteraction( desc, @@ -1899,9 +1881,9 @@ class DatabasePool: table: str, keyvalues: Optional[Dict[str, Any]], retcols: Iterable[str], - ) -> List[Dict[str, Any]]: + ) -> List[Tuple[Any, ...]]: """Executes a SELECT query on the named table, which may return zero or - more rows, returning the result as a list of dicts. + more rows, returning the result as a list of tuples. Args: txn: Transaction object @@ -1912,8 +1894,7 @@ class DatabasePool: retcols: the names of the columns to return Returns: - A list of dictionaries, one per result row, each a mapping between the - column names from `retcols` and that column's value for the row. + A list of tuples, one per result row, each the retcolumn's value for the row. """ if keyvalues: sql = "SELECT %s FROM %s WHERE %s" % ( @@ -1926,7 +1907,7 @@ class DatabasePool: sql = "SELECT %s FROM %s" % (", ".join(retcols), table) txn.execute(sql) - return cls.cursor_to_dict(txn) + return txn.fetchall() async def simple_select_many_batch( self, @@ -1937,9 +1918,9 @@ class DatabasePool: keyvalues: Optional[Dict[str, Any]] = None, desc: str = "simple_select_many_batch", batch_size: int = 100, - ) -> List[Dict[str, Any]]: + ) -> List[Tuple[Any, ...]]: """Executes a SELECT query on the named table, which may return zero or - more rows, returning the result as a list of dicts. + more rows. Filters rows by whether the value of `column` is in `iterable`. @@ -1951,10 +1932,13 @@ class DatabasePool: keyvalues: dict of column names and values to select the rows with desc: description of the transaction, for logging and metrics batch_size: the number of rows for each select query + + Returns: + The results as a list of tuples. """ keyvalues = keyvalues or {} - results: List[Dict[str, Any]] = [] + results: List[Tuple[Any, ...]] = [] for chunk in batch_iter(iterable, batch_size): rows = await self.runInteraction( @@ -1981,9 +1965,9 @@ class DatabasePool: iterable: Collection[Any], keyvalues: Dict[str, Any], retcols: Iterable[str], - ) -> List[Dict[str, Any]]: + ) -> List[Tuple[Any, ...]]: """Executes a SELECT query on the named table, which may return zero or - more rows, returning the result as a list of dicts. + more rows. Filters rows by whether the value of `column` is in `iterable`. @@ -1994,7 +1978,11 @@ class DatabasePool: iterable: list keyvalues: dict of column names and values to select the rows with retcols: list of strings giving the names of the columns to return + + Returns: + The results as a list of tuples. """ + # If there's nothing to select, then skip executing the query. if not iterable: return [] @@ -2012,7 +2000,7 @@ class DatabasePool: ) txn.execute(sql, values) - return cls.cursor_to_dict(txn) + return txn.fetchall() async def simple_update( self, @@ -2129,13 +2117,13 @@ class DatabasePool: raise ValueError( f"{len(key_values)} key rows and {len(value_values)} value rows: should be the same number." ) + # If there is nothing to update, then skip executing the query. + if not key_values: + return # List of tuples of (value values, then key values) # (This matches the order needed for the query) - args = [tuple(x) + tuple(y) for x, y in zip(value_values, key_values)] - - for ks, vs in zip(key_values, value_values): - args.append(tuple(vs) + tuple(ks)) + args = [tuple(vv) + tuple(kv) for vv, kv in zip(value_values, key_values)] # 'col1 = ?, col2 = ?, ...' set_clause = ", ".join(f"{n} = ?" for n in value_names) @@ -2147,9 +2135,7 @@ class DatabasePool: where_clause = "" # UPDATE mytable SET col1 = ?, col2 = ? WHERE col3 = ? AND col4 = ? - sql = f""" - UPDATE {table} SET {set_clause} {where_clause} - """ + sql = f"UPDATE {table} SET {set_clause} {where_clause}" txn.execute_batch(sql, args) @@ -2365,11 +2351,10 @@ class DatabasePool: Returns: Number rows deleted """ + # If there's nothing to delete, then skip executing the query. if not values: return 0 - sql = "DELETE FROM %s" % table - clause, values = make_in_list_sql_clause(txn.database_engine, column, values) clauses = [clause] @@ -2377,8 +2362,7 @@ class DatabasePool: clauses.append("%s = ?" % (key,)) values.append(value) - if clauses: - sql = "%s WHERE %s" % (sql, " AND ".join(clauses)) + sql = "DELETE FROM %s WHERE %s" % (table, " AND ".join(clauses)) txn.execute(sql, values) return txn.rowcount @@ -2481,7 +2465,7 @@ class DatabasePool: keyvalues: Optional[Dict[str, Any]] = None, exclude_keyvalues: Optional[Dict[str, Any]] = None, order_direction: str = "ASC", - ) -> List[Dict[str, Any]]: + ) -> List[Tuple[Any, ...]]: """ Executes a SELECT query on the named table with start and limit, of row numbers, which may return zero or number of rows from start to limit, @@ -2510,7 +2494,7 @@ class DatabasePool: order_direction: Whether the results should be ordered "ASC" or "DESC". Returns: - The result as a list of dictionaries. + The result as a list of tuples. """ if order_direction not in ["ASC", "DESC"]: raise ValueError("order_direction must be one of 'ASC' or 'DESC'.") @@ -2537,69 +2521,7 @@ class DatabasePool: ) txn.execute(sql, arg_list + [limit, start]) - return cls.cursor_to_dict(txn) - - async def simple_search_list( - self, - table: str, - term: Optional[str], - col: str, - retcols: Collection[str], - desc: str = "simple_search_list", - ) -> Optional[List[Dict[str, Any]]]: - """Executes a SELECT query on the named table, which may return zero or - more rows, returning the result as a list of dicts. - - Args: - table: the table name - term: term for searching the table matched to a column. - col: column to query term should be matched to - retcols: the names of the columns to return - - Returns: - A list of dictionaries or None. - """ - - return await self.runInteraction( - desc, - self.simple_search_list_txn, - table, - term, - col, - retcols, - db_autocommit=True, - ) - - @classmethod - def simple_search_list_txn( - cls, - txn: LoggingTransaction, - table: str, - term: Optional[str], - col: str, - retcols: Iterable[str], - ) -> Optional[List[Dict[str, Any]]]: - """Executes a SELECT query on the named table, which may return zero or - more rows, returning the result as a list of dicts. - - Args: - txn: Transaction object - table: the table name - term: term for searching the table matched to a column. - col: column to query term should be matched to - retcols: the names of the columns to return - - Returns: - None if no term is given, otherwise a list of dictionaries. - """ - if term: - sql = "SELECT %s FROM %s WHERE %s LIKE ?" % (", ".join(retcols), table, col) - termvalues = ["%%" + term + "%%"] - txn.execute(sql, termvalues) - else: - return None - - return cls.cursor_to_dict(txn) + return txn.fetchall() def make_in_list_sql_clause( diff --git a/synapse/storage/databases/main/__init__.py b/synapse/storage/databases/main/__init__.py index 101403578c..89f4077351 100644 --- a/synapse/storage/databases/main/__init__.py +++ b/synapse/storage/databases/main/__init__.py @@ -15,7 +15,9 @@ # limitations under the License. import logging -from typing import TYPE_CHECKING, List, Optional, Tuple, cast +from typing import TYPE_CHECKING, List, Optional, Tuple, Union, cast + +import attr from synapse.api.constants import Direction from synapse.config.homeserver import HomeServerConfig @@ -28,7 +30,7 @@ from synapse.storage.database import ( from synapse.storage.databases.main.stats import UserSortOrder from synapse.storage.engines import BaseDatabaseEngine from synapse.storage.types import Cursor -from synapse.types import JsonDict, get_domain_from_id +from synapse.types import get_domain_from_id from .account_data import AccountDataStore from .appservice import ApplicationServiceStore, ApplicationServiceTransactionStore @@ -82,6 +84,25 @@ if TYPE_CHECKING: logger = logging.getLogger(__name__) +@attr.s(slots=True, frozen=True, auto_attribs=True) +class UserPaginateResponse: + """This is very similar to UserInfo, but not quite the same.""" + + name: str + user_type: Optional[str] + is_guest: bool + admin: bool + deactivated: bool + shadow_banned: bool + displayname: Optional[str] + avatar_url: Optional[str] + creation_ts: Optional[int] + approved: bool + erased: bool + last_seen_ts: int + locked: bool + + class DataStore( EventsBackgroundUpdatesStore, ExperimentalFeaturesStore, @@ -142,26 +163,6 @@ class DataStore( super().__init__(database, db_conn, hs) - async def get_users(self) -> List[JsonDict]: - """Function to retrieve a list of users in users table. - - Returns: - A list of dictionaries representing users. - """ - return await self.db_pool.simple_select_list( - table="users", - keyvalues={}, - retcols=[ - "name", - "password_hash", - "is_guest", - "admin", - "user_type", - "deactivated", - ], - desc="get_users", - ) - async def get_users_paginate( self, start: int, @@ -176,7 +177,7 @@ class DataStore( approved: bool = True, not_user_types: Optional[List[str]] = None, locked: bool = False, - ) -> Tuple[List[JsonDict], int]: + ) -> Tuple[List[UserPaginateResponse], int]: """Function to retrieve a paginated list of users from users list. This will return a json list of users and the total number of users matching the filter criteria. @@ -202,7 +203,7 @@ class DataStore( def get_users_paginate_txn( txn: LoggingTransaction, - ) -> Tuple[List[JsonDict], int]: + ) -> Tuple[List[UserPaginateResponse], int]: filters = [] args: list = [] @@ -302,13 +303,24 @@ 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]) + users = [ + UserPaginateResponse( + name=row[0], + user_type=row[1], + is_guest=bool(row[2]), + admin=bool(row[3]), + deactivated=bool(row[4]), + shadow_banned=bool(row[5]), + displayname=row[6], + avatar_url=row[7], + creation_ts=row[8], + approved=bool(row[9]), + erased=bool(row[10]), + last_seen_ts=row[11], + locked=bool(row[12]), + ) + for row in txn + ] return users, count @@ -316,7 +328,11 @@ class DataStore( "get_users_paginate_txn", get_users_paginate_txn ) - async def search_users(self, term: str) -> Optional[List[JsonDict]]: + async def search_users( + self, term: str + ) -> List[ + Tuple[str, Optional[str], Union[int, bool], Union[int, bool], Optional[str]] + ]: """Function to search users list for one or more users with the matched term. @@ -324,15 +340,37 @@ class DataStore( term: search term Returns: - A list of dictionaries or None. + A list of tuples of name, password_hash, is_guest, admin, user_type or None. """ - return await self.db_pool.simple_search_list( - table="users", - term=term, - col="name", - retcols=["name", "password_hash", "is_guest", "admin", "user_type"], - desc="search_users", - ) + + def search_users( + txn: LoggingTransaction, + ) -> List[ + Tuple[str, Optional[str], Union[int, bool], Union[int, bool], Optional[str]] + ]: + search_term = "%%" + term + "%%" + + sql = """ + SELECT name, password_hash, is_guest, admin, user_type + FROM users + WHERE name LIKE ? + """ + txn.execute(sql, (search_term,)) + + return cast( + List[ + Tuple[ + str, + Optional[str], + Union[int, bool], + Union[int, bool], + Optional[str], + ] + ], + txn.fetchall(), + ) + + return await self.db_pool.runInteraction("search_users", search_users) def check_database_before_upgrade( diff --git a/synapse/storage/databases/main/account_data.py b/synapse/storage/databases/main/account_data.py index 80f146dd53..d7482a1f4e 100644 --- a/synapse/storage/databases/main/account_data.py +++ b/synapse/storage/databases/main/account_data.py @@ -94,7 +94,10 @@ class AccountDataWorkerStore(PushRulesWorkerStore, CacheInvalidationWorkerStore) hs.get_replication_notifier(), "room_account_data", "stream_id", - extra_tables=[("room_tags_revisions", "stream_id")], + extra_tables=[ + ("account_data", "stream_id"), + ("room_tags_revisions", "stream_id"), + ], is_writer=self._instance_name in hs.config.worker.writers.account_data, ) @@ -103,6 +106,13 @@ class AccountDataWorkerStore(PushRulesWorkerStore, CacheInvalidationWorkerStore) "AccountDataAndTagsChangeCache", account_max ) + self.db_pool.updates.register_background_index_update( + update_name="room_account_data_index_room_id", + index_name="room_account_data_room_id", + table="room_account_data", + columns=("room_id",), + ) + self.db_pool.updates.register_background_update_handler( "delete_account_data_for_deactivated_users", self._delete_account_data_for_deactivated_users, @@ -151,10 +161,10 @@ class AccountDataWorkerStore(PushRulesWorkerStore, CacheInvalidationWorkerStore) sql += " AND content != '{}'" txn.execute(sql, (user_id,)) - rows = self.db_pool.cursor_to_dict(txn) return { - row["account_data_type"]: db_to_json(row["content"]) for row in rows + account_data_type: db_to_json(content) + for account_data_type, content in txn } return await self.db_pool.runInteraction( @@ -196,13 +206,12 @@ class AccountDataWorkerStore(PushRulesWorkerStore, CacheInvalidationWorkerStore) sql += " AND content != '{}'" txn.execute(sql, (user_id,)) - rows = self.db_pool.cursor_to_dict(txn) by_room: Dict[str, Dict[str, JsonDict]] = {} - for row in rows: - room_data = by_room.setdefault(row["room_id"], {}) + for room_id, account_data_type, content in txn: + room_data = by_room.setdefault(room_id, {}) - room_data[row["account_data_type"]] = db_to_json(row["content"]) + room_data[account_data_type] = db_to_json(content) return by_room @@ -277,16 +286,20 @@ class AccountDataWorkerStore(PushRulesWorkerStore, CacheInvalidationWorkerStore) def get_account_data_for_room_txn( txn: LoggingTransaction, - ) -> Dict[str, JsonDict]: - rows = self.db_pool.simple_select_list_txn( - txn, - "room_account_data", - {"user_id": user_id, "room_id": room_id}, - ["account_data_type", "content"], + ) -> Dict[str, JsonMapping]: + rows = cast( + List[Tuple[str, str]], + self.db_pool.simple_select_list_txn( + txn, + table="room_account_data", + keyvalues={"user_id": user_id, "room_id": room_id}, + retcols=["account_data_type", "content"], + ), ) return { - row["account_data_type"]: db_to_json(row["content"]) for row in rows + account_data_type: db_to_json(content) + for account_data_type, content in rows } return await self.db_pool.runInteraction( diff --git a/synapse/storage/databases/main/appservice.py b/synapse/storage/databases/main/appservice.py index 0553a0621a..fa7d1c469a 100644 --- a/synapse/storage/databases/main/appservice.py +++ b/synapse/storage/databases/main/appservice.py @@ -14,17 +14,7 @@ # limitations under the License. import logging import re -from typing import ( - TYPE_CHECKING, - Any, - Dict, - List, - Optional, - Pattern, - Sequence, - Tuple, - cast, -) +from typing import TYPE_CHECKING, List, Optional, Pattern, Sequence, Tuple, cast from synapse.appservice import ( ApplicationService, @@ -207,16 +197,21 @@ class ApplicationServiceTransactionWorkerStore( Returns: A list of ApplicationServices, which may be empty. """ - results = await self.db_pool.simple_select_list( - "application_services_state", {"state": state.value}, ["as_id"] + results = cast( + List[Tuple[str]], + await self.db_pool.simple_select_list( + table="application_services_state", + keyvalues={"state": state.value}, + retcols=("as_id",), + ), ) # NB: This assumes this class is linked with ApplicationServiceStore as_list = self.get_app_services() services = [] - for res in results: + for (as_id,) in results: for service in as_list: - if service.id == res["as_id"]: + if service.id == as_id: services.append(service) return services @@ -353,21 +348,15 @@ class ApplicationServiceTransactionWorkerStore( def _get_oldest_unsent_txn( txn: LoggingTransaction, - ) -> Optional[Dict[str, Any]]: + ) -> Optional[Tuple[int, str]]: # Monotonically increasing txn ids, so just select the smallest # one in the txns table (we delete them when they are sent) txn.execute( - "SELECT * FROM application_services_txns WHERE as_id=?" + "SELECT txn_id, event_ids FROM application_services_txns WHERE as_id=?" " ORDER BY txn_id ASC LIMIT 1", (service.id,), ) - rows = self.db_pool.cursor_to_dict(txn) - if not rows: - return None - - entry = rows[0] - - return entry + return cast(Optional[Tuple[int, str]], txn.fetchone()) entry = await self.db_pool.runInteraction( "get_oldest_unsent_appservice_txn", _get_oldest_unsent_txn @@ -376,8 +365,9 @@ class ApplicationServiceTransactionWorkerStore( if not entry: return None - event_ids = db_to_json(entry["event_ids"]) + txn_id, event_ids_str = entry + event_ids = db_to_json(event_ids_str) events = await self.get_events_as_list(event_ids) # TODO: to-device messages, one-time key counts, device list summaries and unused @@ -385,7 +375,7 @@ class ApplicationServiceTransactionWorkerStore( # We likely want to populate those for reliability. return AppServiceTransaction( service=service, - id=entry["txn_id"], + id=txn_id, events=events, ephemeral=[], to_device_messages=[], diff --git a/synapse/storage/databases/main/cache.py b/synapse/storage/databases/main/cache.py index 2fbd389c71..4d0470ffd9 100644 --- a/synapse/storage/databases/main/cache.py +++ b/synapse/storage/databases/main/cache.py @@ -23,6 +23,7 @@ from synapse.metrics.background_process_metrics import wrap_as_background_proces from synapse.replication.tcp.streams import BackfillStream, CachesStream from synapse.replication.tcp.streams.events import ( EventsStream, + EventsStreamAllStateRow, EventsStreamCurrentStateRow, EventsStreamEventRow, EventsStreamRow, @@ -264,6 +265,13 @@ class CacheInvalidationWorkerStore(SQLBaseStore): (data.state_key,) ) self.get_rooms_for_user.invalidate((data.state_key,)) # type: ignore[attr-defined] + elif row.type == EventsStreamAllStateRow.TypeId: + assert isinstance(data, EventsStreamAllStateRow) + # Similar to the above, but the entire caches are invalidated. This is + # unfortunate for the membership caches, but should recover quickly. + self._curr_state_delta_stream_cache.entity_has_changed(data.room_id, token) # type: ignore[attr-defined] + self.get_rooms_for_user_with_stream_ordering.invalidate_all() # type: ignore[attr-defined] + self.get_rooms_for_user.invalidate_all() # type: ignore[attr-defined] else: raise Exception("Unknown events stream row type %s" % (row.type,)) diff --git a/synapse/storage/databases/main/censor_events.py b/synapse/storage/databases/main/censor_events.py index 58177ecec1..711fdddd4e 100644 --- a/synapse/storage/databases/main/censor_events.py +++ b/synapse/storage/databases/main/censor_events.py @@ -93,7 +93,7 @@ class CensorEventsStore(EventsWorkerStore, CacheInvalidationWorkerStore, SQLBase """ rows = await self.db_pool.execute( - "_censor_redactions_fetch", None, sql, before_ts, 100 + "_censor_redactions_fetch", sql, before_ts, 100 ) updates = [] diff --git a/synapse/storage/databases/main/client_ips.py b/synapse/storage/databases/main/client_ips.py index 7da47c3dd7..c006129625 100644 --- a/synapse/storage/databases/main/client_ips.py +++ b/synapse/storage/databases/main/client_ips.py @@ -15,6 +15,7 @@ import logging from typing import TYPE_CHECKING, Dict, List, Mapping, Optional, Tuple, Union, cast +import attr from typing_extensions import TypedDict from synapse.metrics.background_process_metrics import wrap_as_background_process @@ -42,7 +43,8 @@ logger = logging.getLogger(__name__) LAST_SEEN_GRANULARITY = 120 * 1000 -class DeviceLastConnectionInfo(TypedDict): +@attr.s(slots=True, frozen=True, auto_attribs=True) +class DeviceLastConnectionInfo: """Metadata for the last connection seen for a user and device combination""" # These types must match the columns in the `devices` table @@ -499,8 +501,7 @@ class ClientIpWorkerStore(ClientIpBackgroundUpdateStore, MonthlyActiveUsersWorke device_id: If None fetches all devices for the user Returns: - A dictionary mapping a tuple of (user_id, device_id) to dicts, with - keys giving the column names from the devices table. + A dictionary mapping a tuple of (user_id, device_id) to DeviceLastConnectionInfo. """ keyvalues = {"user_id": user_id} @@ -508,7 +509,7 @@ class ClientIpWorkerStore(ClientIpBackgroundUpdateStore, MonthlyActiveUsersWorke keyvalues["device_id"] = device_id res = cast( - List[DeviceLastConnectionInfo], + List[Tuple[str, Optional[str], Optional[str], str, Optional[int]]], await self.db_pool.simple_select_list( table="devices", keyvalues=keyvalues, @@ -516,7 +517,16 @@ class ClientIpWorkerStore(ClientIpBackgroundUpdateStore, MonthlyActiveUsersWorke ), ) - return {(d["user_id"], d["device_id"]): d for d in res} + return { + (user_id, device_id): DeviceLastConnectionInfo( + user_id=user_id, + device_id=device_id, + ip=ip, + user_agent=user_agent, + last_seen=last_seen, + ) + for user_id, ip, user_agent, device_id, last_seen in res + } async def _get_user_ip_and_agents_from_database( self, user: UserID, since_ts: int = 0 @@ -683,8 +693,7 @@ class ClientIpWorkerStore(ClientIpBackgroundUpdateStore, MonthlyActiveUsersWorke device_id: If None fetches all devices for the user Returns: - A dictionary mapping a tuple of (user_id, device_id) to dicts, with - keys giving the column names from the devices table. + A dictionary mapping a tuple of (user_id, device_id) to DeviceLastConnectionInfo. """ ret = await self._get_last_client_ip_by_device_from_database(user_id, device_id) @@ -705,13 +714,13 @@ class ClientIpWorkerStore(ClientIpBackgroundUpdateStore, MonthlyActiveUsersWorke continue if not device_id or did == device_id: - ret[(user_id, did)] = { - "user_id": user_id, - "ip": ip, - "user_agent": user_agent, - "device_id": did, - "last_seen": last_seen, - } + ret[(user_id, did)] = DeviceLastConnectionInfo( + user_id=user_id, + ip=ip, + user_agent=user_agent, + device_id=did, + last_seen=last_seen, + ) return ret async def get_user_ip_and_agents( diff --git a/synapse/storage/databases/main/deviceinbox.py b/synapse/storage/databases/main/deviceinbox.py index 744e98c6d0..3e7425d4a6 100644 --- a/synapse/storage/databases/main/deviceinbox.py +++ b/synapse/storage/databases/main/deviceinbox.py @@ -344,18 +344,19 @@ class DeviceInboxWorkerStore(SQLBaseStore): # Note that this is more efficient than just dropping `device_id` from the query, # since device_inbox has an index on `(user_id, device_id, stream_id)` if not device_ids_to_query: - user_device_dicts = self.db_pool.simple_select_many_txn( - txn, - table="devices", - column="user_id", - iterable=user_ids_to_query, - keyvalues={"hidden": False}, - retcols=("device_id",), + user_device_dicts = cast( + List[Tuple[str]], + self.db_pool.simple_select_many_txn( + txn, + table="devices", + column="user_id", + iterable=user_ids_to_query, + keyvalues={"hidden": False}, + retcols=("device_id",), + ), ) - device_ids_to_query.update( - {row["device_id"] for row in user_device_dicts} - ) + device_ids_to_query.update({row[0] for row in user_device_dicts}) if not device_ids_to_query: # We've ended up with no devices to query. @@ -449,7 +450,7 @@ class DeviceInboxWorkerStore(SQLBaseStore): user_id: str, device_id: Optional[str], up_to_stream_id: int, - limit: int, + limit: Optional[int] = None, ) -> int: """ Args: @@ -477,17 +478,19 @@ class DeviceInboxWorkerStore(SQLBaseStore): log_kv({"message": "No changes in cache since last check"}) return 0 - ROW_ID_NAME = self.database_engine.row_id_name - def delete_messages_for_device_txn(txn: LoggingTransaction) -> int: + limit_statement = "" if limit is None else f"LIMIT {limit}" sql = f""" - DELETE FROM device_inbox WHERE {ROW_ID_NAME} IN ( - SELECT {ROW_ID_NAME} FROM device_inbox - WHERE user_id = ? AND device_id = ? AND stream_id <= ? - LIMIT {limit} + DELETE FROM device_inbox WHERE user_id = ? AND device_id = ? AND stream_id <= ( + SELECT MAX(stream_id) FROM ( + SELECT stream_id FROM device_inbox + WHERE user_id = ? AND device_id = ? AND stream_id <= ? + ORDER BY stream_id + {limit_statement} + ) AS q1 ) """ - txn.execute(sql, (user_id, device_id, up_to_stream_id)) + txn.execute(sql, (user_id, device_id, user_id, device_id, up_to_stream_id)) return txn.rowcount count = await self.db_pool.runInteraction( @@ -845,20 +848,21 @@ class DeviceInboxWorkerStore(SQLBaseStore): # We exclude hidden devices (such as cross-signing keys) here as they are # not expected to receive to-device messages. - rows = self.db_pool.simple_select_many_txn( - txn, - table="devices", - keyvalues={"user_id": user_id, "hidden": False}, - column="device_id", - iterable=devices, - retcols=("device_id",), + rows = cast( + List[Tuple[str]], + self.db_pool.simple_select_many_txn( + txn, + table="devices", + keyvalues={"user_id": user_id, "hidden": False}, + column="device_id", + iterable=devices, + retcols=("device_id",), + ), ) - for row in rows: + for (device_id,) in rows: # Only insert into the local inbox if the device exists on # this server - device_id = row["device_id"] - with start_active_span("serialise_to_device_message"): msg = messages_by_device[device_id] set_tag(SynapseTags.TO_DEVICE_TYPE, msg["type"]) diff --git a/synapse/storage/databases/main/devices.py b/synapse/storage/databases/main/devices.py index df596f35f9..04d12a876c 100644 --- a/synapse/storage/databases/main/devices.py +++ b/synapse/storage/databases/main/devices.py @@ -283,7 +283,9 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): allow_none=True, ) - async def get_devices_by_user(self, user_id: str) -> Dict[str, Dict[str, str]]: + async def get_devices_by_user( + self, user_id: str + ) -> Dict[str, Dict[str, Optional[str]]]: """Retrieve all of a user's registered devices. Only returns devices that are not marked as hidden. @@ -291,20 +293,26 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): user_id: Returns: A mapping from device_id to a dict containing "device_id", "user_id" - and "display_name" for each device. + and "display_name" for each device. Display name may be null. """ - devices = await self.db_pool.simple_select_list( - table="devices", - keyvalues={"user_id": user_id, "hidden": False}, - retcols=("user_id", "device_id", "display_name"), - desc="get_devices_by_user", + devices = cast( + List[Tuple[str, str, Optional[str]]], + await self.db_pool.simple_select_list( + table="devices", + keyvalues={"user_id": user_id, "hidden": False}, + retcols=("user_id", "device_id", "display_name"), + desc="get_devices_by_user", + ), ) - return {d["device_id"]: d for d in devices} + return { + d[1]: {"user_id": d[0], "device_id": d[1], "display_name": d[2]} + for d in devices + } async def get_devices_by_auth_provider_session_id( self, auth_provider_id: str, auth_provider_session_id: str - ) -> List[Dict[str, Any]]: + ) -> List[Tuple[str, str]]: """Retrieve the list of devices associated with a SSO IdP session ID. Args: @@ -313,14 +321,17 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): Returns: A list of dicts containing the device_id and the user_id of each device """ - return await self.db_pool.simple_select_list( - table="device_auth_providers", - keyvalues={ - "auth_provider_id": auth_provider_id, - "auth_provider_session_id": auth_provider_session_id, - }, - retcols=("user_id", "device_id"), - desc="get_devices_by_auth_provider_session_id", + return cast( + List[Tuple[str, str]], + await self.db_pool.simple_select_list( + table="device_auth_providers", + keyvalues={ + "auth_provider_id": auth_provider_id, + "auth_provider_session_id": auth_provider_session_id, + }, + retcols=("user_id", "device_id"), + desc="get_devices_by_auth_provider_session_id", + ), ) @trace @@ -692,7 +703,7 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): key_names=("destination", "user_id"), key_values=[(destination, user_id) for user_id, _ in rows], value_names=("stream_id",), - value_values=((stream_id,) for _, stream_id in rows), + value_values=[(stream_id,) for _, stream_id in rows], ) # Delete all sent outbound pokes @@ -821,15 +832,16 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): async def get_cached_devices_for_user( self, user_id: str ) -> Mapping[str, JsonMapping]: - devices = await self.db_pool.simple_select_list( - table="device_lists_remote_cache", - keyvalues={"user_id": user_id}, - retcols=("device_id", "content"), - desc="get_cached_devices_for_user", + devices = cast( + List[Tuple[str, str]], + await self.db_pool.simple_select_list( + table="device_lists_remote_cache", + keyvalues={"user_id": user_id}, + retcols=("device_id", "content"), + desc="get_cached_devices_for_user", + ), ) - return { - device["device_id"]: db_to_json(device["content"]) for device in devices - } + return {device[0]: db_to_json(device[1]) for device in devices} def get_cached_device_list_changes( self, @@ -882,7 +894,6 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): rows = await self.db_pool.execute( "get_all_devices_changed", - None, sql, from_key, to_key, @@ -966,7 +977,7 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): WHERE from_user_id = ? AND stream_id > ? """ rows = await self.db_pool.execute( - "get_users_whose_signatures_changed", None, sql, user_id, from_key + "get_users_whose_signatures_changed", sql, user_id, from_key ) return {user for row in rows for user in db_to_json(row[0])} else: @@ -1052,16 +1063,19 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): async def get_device_list_last_stream_id_for_remotes( self, user_ids: Iterable[str] ) -> Mapping[str, Optional[str]]: - rows = await self.db_pool.simple_select_many_batch( - table="device_lists_remote_extremeties", - column="user_id", - iterable=user_ids, - retcols=("user_id", "stream_id"), - desc="get_device_list_last_stream_id_for_remotes", + rows = cast( + List[Tuple[str, str]], + await self.db_pool.simple_select_many_batch( + table="device_lists_remote_extremeties", + column="user_id", + iterable=user_ids, + retcols=("user_id", "stream_id"), + desc="get_device_list_last_stream_id_for_remotes", + ), ) results: Dict[str, Optional[str]] = {user_id: None for user_id in user_ids} - results.update({row["user_id"]: row["stream_id"] for row in rows}) + results.update(rows) return results @@ -1077,22 +1091,28 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): The IDs of users whose device lists need resync. """ if user_ids: - rows = await self.db_pool.simple_select_many_batch( - table="device_lists_remote_resync", - column="user_id", - iterable=user_ids, - retcols=("user_id",), - desc="get_user_ids_requiring_device_list_resync_with_iterable", + rows = cast( + List[Tuple[str]], + await self.db_pool.simple_select_many_batch( + table="device_lists_remote_resync", + column="user_id", + iterable=user_ids, + retcols=("user_id",), + desc="get_user_ids_requiring_device_list_resync_with_iterable", + ), ) else: - rows = await self.db_pool.simple_select_list( - table="device_lists_remote_resync", - keyvalues=None, - retcols=("user_id",), - desc="get_user_ids_requiring_device_list_resync", + rows = cast( + List[Tuple[str]], + await self.db_pool.simple_select_list( + table="device_lists_remote_resync", + keyvalues=None, + retcols=("user_id",), + desc="get_user_ids_requiring_device_list_resync", + ), ) - return {row["user_id"] for row in rows} + return {row[0] for row in rows} async def mark_remote_users_device_caches_as_stale( self, user_ids: StrCollection @@ -1413,13 +1433,13 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): def get_devices_not_accessed_since_txn( txn: LoggingTransaction, - ) -> List[Dict[str, str]]: + ) -> List[Tuple[str, str]]: sql = """ SELECT user_id, device_id FROM devices WHERE last_seen < ? AND hidden = FALSE """ txn.execute(sql, (since_ms,)) - return self.db_pool.cursor_to_dict(txn) + return cast(List[Tuple[str, str]], txn.fetchall()) rows = await self.db_pool.runInteraction( "get_devices_not_accessed_since", @@ -1427,11 +1447,11 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): ) devices: Dict[str, List[str]] = {} - for row in rows: + for user_id, device_id in rows: # Remote devices are never stale from our point of view. - if self.hs.is_mine_id(row["user_id"]): - user_devices = devices.setdefault(row["user_id"], []) - user_devices.append(row["device_id"]) + if self.hs.is_mine_id(user_id): + user_devices = devices.setdefault(user_id, []) + user_devices.append(device_id) return devices @@ -1600,7 +1620,6 @@ class DeviceBackgroundUpdateStore(SQLBaseStore): # # For each duplicate, we delete all the existing rows and put one back. - KEY_COLS = ["stream_id", "destination", "user_id", "device_id"] last_row = progress.get( "last_row", {"stream_id": 0, "destination": "", "user_id": "", "device_id": ""}, @@ -1608,44 +1627,62 @@ class DeviceBackgroundUpdateStore(SQLBaseStore): def _txn(txn: LoggingTransaction) -> int: clause, args = make_tuple_comparison_clause( - [(x, last_row[x]) for x in KEY_COLS] + [ + ("stream_id", last_row["stream_id"]), + ("destination", last_row["destination"]), + ("user_id", last_row["user_id"]), + ("device_id", last_row["device_id"]), + ] ) - sql = """ + sql = f""" SELECT stream_id, destination, user_id, device_id, MAX(ts) AS ts FROM device_lists_outbound_pokes - WHERE %s - GROUP BY %s + WHERE {clause} + GROUP BY stream_id, destination, user_id, device_id HAVING count(*) > 1 - ORDER BY %s + ORDER BY stream_id, destination, user_id, device_id LIMIT ? - """ % ( - clause, # WHERE - ",".join(KEY_COLS), # GROUP BY - ",".join(KEY_COLS), # ORDER BY - ) + """ txn.execute(sql, args + [batch_size]) - rows = self.db_pool.cursor_to_dict(txn) + rows = txn.fetchall() - row = None - for row in rows: + stream_id, destination, user_id, device_id = None, None, None, None + for stream_id, destination, user_id, device_id, _ in rows: self.db_pool.simple_delete_txn( txn, "device_lists_outbound_pokes", - {x: row[x] for x in KEY_COLS}, + { + "stream_id": stream_id, + "destination": destination, + "user_id": user_id, + "device_id": device_id, + }, ) - row["sent"] = False self.db_pool.simple_insert_txn( txn, "device_lists_outbound_pokes", - row, + { + "stream_id": stream_id, + "destination": destination, + "user_id": user_id, + "device_id": device_id, + "sent": False, + }, ) - if row: + if rows: self.db_pool.updates._background_update_progress_txn( txn, BG_UPDATE_REMOVE_DUP_OUTBOUND_POKES, - {"last_row": row}, + { + "last_row": { + "stream_id": stream_id, + "destination": destination, + "user_id": user_id, + "device_id": device_id, + } + }, ) return len(rows) diff --git a/synapse/storage/databases/main/e2e_room_keys.py b/synapse/storage/databases/main/e2e_room_keys.py index d01f28cc80..ad904a26a6 100644 --- a/synapse/storage/databases/main/e2e_room_keys.py +++ b/synapse/storage/databases/main/e2e_room_keys.py @@ -13,7 +13,7 @@ # See the License for the specific language governing permissions and # limitations under the License. -from typing import TYPE_CHECKING, Dict, Iterable, Mapping, Optional, Tuple, cast +from typing import TYPE_CHECKING, Dict, Iterable, List, Mapping, Optional, Tuple, cast from typing_extensions import Literal, TypedDict @@ -53,6 +53,13 @@ class EndToEndRoomKeyBackgroundStore(SQLBaseStore): ): super().__init__(database, db_conn, hs) + self.db_pool.updates.register_background_index_update( + update_name="e2e_room_keys_index_room_id", + index_name="e2e_room_keys_room_id", + table="e2e_room_keys", + columns=("room_id",), + ) + self.db_pool.updates.register_background_update_handler( "delete_e2e_backup_keys_for_deactivated_users", self._delete_e2e_backup_keys_for_deactivated_users, @@ -208,7 +215,7 @@ class EndToEndRoomKeyStore(EndToEndRoomKeyBackgroundStore): "message": "Set room key", "room_id": room_id, "session_id": session_id, - StreamKeyType.ROOM: room_key, + StreamKeyType.ROOM.value: room_key, } ) @@ -267,32 +274,41 @@ class EndToEndRoomKeyStore(EndToEndRoomKeyBackgroundStore): if session_id: keyvalues["session_id"] = session_id - rows = await self.db_pool.simple_select_list( - table="e2e_room_keys", - keyvalues=keyvalues, - retcols=( - "user_id", - "room_id", - "session_id", - "first_message_index", - "forwarded_count", - "is_verified", - "session_data", + rows = cast( + List[Tuple[str, str, int, int, int, str]], + await self.db_pool.simple_select_list( + table="e2e_room_keys", + keyvalues=keyvalues, + retcols=( + "room_id", + "session_id", + "first_message_index", + "forwarded_count", + "is_verified", + "session_data", + ), + desc="get_e2e_room_keys", ), - desc="get_e2e_room_keys", ) sessions: Dict[ Literal["rooms"], Dict[str, Dict[Literal["sessions"], Dict[str, RoomKey]]] ] = {"rooms": {}} - for row in rows: - room_entry = sessions["rooms"].setdefault(row["room_id"], {"sessions": {}}) - room_entry["sessions"][row["session_id"]] = { - "first_message_index": row["first_message_index"], - "forwarded_count": row["forwarded_count"], + for ( + room_id, + session_id, + first_message_index, + forwarded_count, + is_verified, + session_data, + ) in rows: + room_entry = sessions["rooms"].setdefault(room_id, {"sessions": {}}) + room_entry["sessions"][session_id] = { + "first_message_index": first_message_index, + "forwarded_count": forwarded_count, # is_verified must be returned to the client as a boolean - "is_verified": bool(row["is_verified"]), - "session_data": db_to_json(row["session_data"]), + "is_verified": bool(is_verified), + "session_data": db_to_json(session_data), } return sessions diff --git a/synapse/storage/databases/main/end_to_end_keys.py b/synapse/storage/databases/main/end_to_end_keys.py index 89fac23f93..4f96ac25c7 100644 --- a/synapse/storage/databases/main/end_to_end_keys.py +++ b/synapse/storage/databases/main/end_to_end_keys.py @@ -24,6 +24,7 @@ from typing import ( Mapping, Optional, Sequence, + Set, Tuple, Union, cast, @@ -155,7 +156,6 @@ class EndToEndKeyWorkerStore(EndToEndKeyBackgroundStore, CacheInvalidationWorker """ rows = await self.db_pool.execute( "get_e2e_device_keys_for_federation_query_check", - None, sql, now_stream_id, user_id, @@ -493,15 +493,18 @@ class EndToEndKeyWorkerStore(EndToEndKeyBackgroundStore, CacheInvalidationWorker A map from (algorithm, key_id) to json string for key """ - rows = await self.db_pool.simple_select_many_batch( - table="e2e_one_time_keys_json", - column="key_id", - iterable=key_ids, - retcols=("algorithm", "key_id", "key_json"), - keyvalues={"user_id": user_id, "device_id": device_id}, - desc="add_e2e_one_time_keys_check", + rows = cast( + List[Tuple[str, str, str]], + await self.db_pool.simple_select_many_batch( + table="e2e_one_time_keys_json", + column="key_id", + iterable=key_ids, + retcols=("algorithm", "key_id", "key_json"), + keyvalues={"user_id": user_id, "device_id": device_id}, + desc="add_e2e_one_time_keys_check", + ), ) - result = {(row["algorithm"], row["key_id"]): row["key_json"] for row in rows} + result = {(algorithm, key_id): key_json for algorithm, key_id, key_json in rows} log_kv({"message": "Fetched one time keys for user", "one_time_keys": result}) return result @@ -921,14 +924,10 @@ class EndToEndKeyWorkerStore(EndToEndKeyBackgroundStore, CacheInvalidationWorker } txn.execute(sql, params) - rows = self.db_pool.cursor_to_dict(txn) - for row in rows: - user_id = row["user_id"] - key_type = row["keytype"] - key = db_to_json(row["keydata"]) + for user_id, key_type, key_data, _ in txn: user_keys = result.setdefault(user_id, {}) - user_keys[key_type] = key + user_keys[key_type] = db_to_json(key_data) return result @@ -988,13 +987,9 @@ class EndToEndKeyWorkerStore(EndToEndKeyBackgroundStore, CacheInvalidationWorker query_params.extend(item) txn.execute(sql, query_params) - rows = self.db_pool.cursor_to_dict(txn) # and add the signatures to the appropriate keys - for row in rows: - key_id: str = row["key_id"] - target_user_id: str = row["target_user_id"] - target_device_id: str = row["target_device_id"] + for target_user_id, target_device_id, key_id, signature in txn: key_type = devices[(target_user_id, target_device_id)] # We need to copy everything, because the result may have come # from the cache. dict.copy only does a shallow copy, so we @@ -1012,13 +1007,11 @@ class EndToEndKeyWorkerStore(EndToEndKeyBackgroundStore, CacheInvalidationWorker ].copy() if from_user_id in signatures: user_sigs = signatures[from_user_id] = signatures[from_user_id] - user_sigs[key_id] = row["signature"] + user_sigs[key_id] = signature else: - signatures[from_user_id] = {key_id: row["signature"]} + signatures[from_user_id] = {key_id: signature} else: - target_user_key["signatures"] = { - from_user_id: {key_id: row["signature"]} - } + target_user_key["signatures"] = {from_user_id: {key_id: signature}} return keys @@ -1118,7 +1111,7 @@ class EndToEndKeyWorkerStore(EndToEndKeyBackgroundStore, CacheInvalidationWorker ... async def claim_e2e_one_time_keys( - self, query_list: Iterable[Tuple[str, str, str, int]] + self, query_list: Collection[Tuple[str, str, str, int]] ) -> Tuple[ Dict[str, Dict[str, Dict[str, JsonDict]]], List[Tuple[str, str, str, int]] ]: @@ -1128,131 +1121,63 @@ class EndToEndKeyWorkerStore(EndToEndKeyBackgroundStore, CacheInvalidationWorker query_list: An iterable of tuples of (user ID, device ID, algorithm). Returns: - A tuple pf: + A tuple (results, missing) of: A map of user ID -> a map device ID -> a map of key ID -> JSON. - A copy of the input which has not been fulfilled. + A copy of the input which has not been fulfilled. The returned counts + may be less than the input counts. In this case, the returned counts + are the number of claims that were not fulfilled. """ - - @trace - def _claim_e2e_one_time_key_simple( - txn: LoggingTransaction, - user_id: str, - device_id: str, - algorithm: str, - count: int, - ) -> List[Tuple[str, str]]: - """Claim OTK for device for DBs that don't support RETURNING. - - Returns: - A tuple of key name (algorithm + key ID) and key JSON, if an - OTK was found. - """ - - sql = """ - SELECT key_id, key_json FROM e2e_one_time_keys_json - WHERE user_id = ? AND device_id = ? AND algorithm = ? - LIMIT ? - """ - - txn.execute(sql, (user_id, device_id, algorithm, count)) - otk_rows = list(txn) - if not otk_rows: - return [] - - self.db_pool.simple_delete_many_txn( - txn, - table="e2e_one_time_keys_json", - column="key_id", - values=[otk_row[0] for otk_row in otk_rows], - keyvalues={ - "user_id": user_id, - "device_id": device_id, - "algorithm": algorithm, - }, - ) - self._invalidate_cache_and_stream( - txn, self.count_e2e_one_time_keys, (user_id, device_id) - ) - - return [ - (f"{algorithm}:{key_id}", key_json) for key_id, key_json in otk_rows - ] - - @trace - def _claim_e2e_one_time_key_returning( - txn: LoggingTransaction, - user_id: str, - device_id: str, - algorithm: str, - count: int, - ) -> List[Tuple[str, str]]: - """Claim OTK for device for DBs that support RETURNING. - - Returns: - A tuple of key name (algorithm + key ID) and key JSON, if an - OTK was found. - """ - - # We can use RETURNING to do the fetch and DELETE in once step. - sql = """ - DELETE FROM e2e_one_time_keys_json - WHERE user_id = ? AND device_id = ? AND algorithm = ? - AND key_id IN ( - SELECT key_id FROM e2e_one_time_keys_json - WHERE user_id = ? AND device_id = ? AND algorithm = ? - LIMIT ? - ) - RETURNING key_id, key_json - """ - - txn.execute( - sql, - (user_id, device_id, algorithm, user_id, device_id, algorithm, count), - ) - otk_rows = list(txn) - if not otk_rows: - return [] - - self._invalidate_cache_and_stream( - txn, self.count_e2e_one_time_keys, (user_id, device_id) - ) - - return [ - (f"{algorithm}:{key_id}", key_json) for key_id, key_json in otk_rows - ] - results: Dict[str, Dict[str, Dict[str, JsonDict]]] = {} missing: List[Tuple[str, str, str, int]] = [] - for user_id, device_id, algorithm, count in query_list: - if self.database_engine.supports_returning: - # If we support RETURNING clause we can use a single query that - # allows us to use autocommit mode. - _claim_e2e_one_time_key = _claim_e2e_one_time_key_returning - db_autocommit = True - else: - _claim_e2e_one_time_key = _claim_e2e_one_time_key_simple - db_autocommit = False - - claim_rows = await self.db_pool.runInteraction( + if isinstance(self.database_engine, PostgresEngine): + # If we can use execute_values we can use a single batch query + # in autocommit mode. + unfulfilled_claim_counts: Dict[Tuple[str, str, str], int] = {} + for user_id, device_id, algorithm, count in query_list: + unfulfilled_claim_counts[user_id, device_id, algorithm] = count + + bulk_claims = await self.db_pool.runInteraction( "claim_e2e_one_time_keys", - _claim_e2e_one_time_key, - user_id, - device_id, - algorithm, - count, - db_autocommit=db_autocommit, + self._claim_e2e_one_time_keys_bulk, + query_list, + db_autocommit=True, ) - if claim_rows: + + for user_id, device_id, algorithm, key_id, key_json in bulk_claims: device_results = results.setdefault(user_id, {}).setdefault( device_id, {} ) - for claim_row in claim_rows: - device_results[claim_row[0]] = json_decoder.decode(claim_row[1]) + device_results[f"{algorithm}:{key_id}"] = json_decoder.decode(key_json) + unfulfilled_claim_counts[(user_id, device_id, algorithm)] -= 1 + # Did we get enough OTKs? - count -= len(claim_rows) - if count: - missing.append((user_id, device_id, algorithm, count)) + missing = [ + (user, device, alg, count) + for (user, device, alg), count in unfulfilled_claim_counts.items() + if count > 0 + ] + else: + for user_id, device_id, algorithm, count in query_list: + claim_rows = await self.db_pool.runInteraction( + "claim_e2e_one_time_keys", + self._claim_e2e_one_time_key_simple, + user_id, + device_id, + algorithm, + count, + db_autocommit=False, + ) + if claim_rows: + device_results = results.setdefault(user_id, {}).setdefault( + device_id, {} + ) + for claim_row in claim_rows: + device_results[claim_row[0]] = json_decoder.decode(claim_row[1]) + # Did we get enough OTKs? + count -= len(claim_rows) + if count: + missing.append((user_id, device_id, algorithm, count)) return results, missing @@ -1268,6 +1193,65 @@ class EndToEndKeyWorkerStore(EndToEndKeyBackgroundStore, CacheInvalidationWorker Returns: A map of user ID -> a map device ID -> a map of key ID -> JSON. """ + if isinstance(self.database_engine, PostgresEngine): + return await self.db_pool.runInteraction( + "_claim_e2e_fallback_keys_bulk", + self._claim_e2e_fallback_keys_bulk_txn, + query_list, + db_autocommit=True, + ) + # Use an UPDATE FROM... RETURNING combined with a VALUES block to do + # everything in one query. Note: this is also supported in SQLite 3.33.0, + # (see https://www.sqlite.org/lang_update.html#update_from), but we do not + # have an equivalent of psycopg2's execute_values to do this in one query. + else: + return await self._claim_e2e_fallback_keys_simple(query_list) + + def _claim_e2e_fallback_keys_bulk_txn( + self, + txn: LoggingTransaction, + query_list: Iterable[Tuple[str, str, str, bool]], + ) -> Dict[str, Dict[str, Dict[str, JsonDict]]]: + """Efficient implementation of claim_e2e_fallback_keys for Postgres. + + Safe to autocommit: this is a single query. + """ + results: Dict[str, Dict[str, Dict[str, JsonDict]]] = {} + + sql = """ + WITH claims(user_id, device_id, algorithm, mark_as_used) AS ( + VALUES ? + ) + UPDATE e2e_fallback_keys_json k + SET used = used OR mark_as_used + FROM claims + WHERE (k.user_id, k.device_id, k.algorithm) = (claims.user_id, claims.device_id, claims.algorithm) + RETURNING k.user_id, k.device_id, k.algorithm, k.key_id, k.key_json; + """ + claimed_keys = cast( + List[Tuple[str, str, str, str, str]], + txn.execute_values(sql, query_list), + ) + + seen_user_device: Set[Tuple[str, str]] = set() + for user_id, device_id, algorithm, key_id, key_json in claimed_keys: + device_results = results.setdefault(user_id, {}).setdefault(device_id, {}) + device_results[f"{algorithm}:{key_id}"] = json_decoder.decode(key_json) + + if (user_id, device_id) in seen_user_device: + continue + seen_user_device.add((user_id, device_id)) + self._invalidate_cache_and_stream( + txn, self.get_e2e_unused_fallback_key_types, (user_id, device_id) + ) + + return results + + async def _claim_e2e_fallback_keys_simple( + self, + query_list: Iterable[Tuple[str, str, str, bool]], + ) -> Dict[str, Dict[str, Dict[str, JsonDict]]]: + """Naive, inefficient implementation of claim_e2e_fallback_keys for SQLite.""" results: Dict[str, Dict[str, Dict[str, JsonDict]]] = {} for user_id, device_id, algorithm, mark_as_used in query_list: row = await self.db_pool.simple_select_one( @@ -1310,6 +1294,99 @@ class EndToEndKeyWorkerStore(EndToEndKeyBackgroundStore, CacheInvalidationWorker return results + @trace + def _claim_e2e_one_time_key_simple( + self, + txn: LoggingTransaction, + user_id: str, + device_id: str, + algorithm: str, + count: int, + ) -> List[Tuple[str, str]]: + """Claim OTK for device for DBs that don't support RETURNING. + + Returns: + A tuple of key name (algorithm + key ID) and key JSON, if an + OTK was found. + """ + + sql = """ + SELECT key_id, key_json FROM e2e_one_time_keys_json + WHERE user_id = ? AND device_id = ? AND algorithm = ? + LIMIT ? + """ + + txn.execute(sql, (user_id, device_id, algorithm, count)) + otk_rows = list(txn) + if not otk_rows: + return [] + + self.db_pool.simple_delete_many_txn( + txn, + table="e2e_one_time_keys_json", + column="key_id", + values=[otk_row[0] for otk_row in otk_rows], + keyvalues={ + "user_id": user_id, + "device_id": device_id, + "algorithm": algorithm, + }, + ) + self._invalidate_cache_and_stream( + txn, self.count_e2e_one_time_keys, (user_id, device_id) + ) + + return [(f"{algorithm}:{key_id}", key_json) for key_id, key_json in otk_rows] + + @trace + def _claim_e2e_one_time_keys_bulk( + self, + txn: LoggingTransaction, + query_list: Iterable[Tuple[str, str, str, int]], + ) -> List[Tuple[str, str, str, str, str]]: + """Bulk claim OTKs, for DBs that support DELETE FROM... RETURNING. + + Args: + query_list: Collection of tuples (user_id, device_id, algorithm, count) + as passed to claim_e2e_one_time_keys. + + Returns: + A list of tuples (user_id, device_id, algorithm, key_id, key_json) + for each OTK claimed. + """ + sql = """ + WITH claims(user_id, device_id, algorithm, claim_count) AS ( + VALUES ? + ), ranked_keys AS ( + SELECT + user_id, device_id, algorithm, key_id, claim_count, + ROW_NUMBER() OVER (PARTITION BY (user_id, device_id, algorithm)) AS r + FROM e2e_one_time_keys_json + JOIN claims USING (user_id, device_id, algorithm) + ) + DELETE FROM e2e_one_time_keys_json k + WHERE (user_id, device_id, algorithm, key_id) IN ( + SELECT user_id, device_id, algorithm, key_id + FROM ranked_keys + WHERE r <= claim_count + ) + RETURNING user_id, device_id, algorithm, key_id, key_json; + """ + otk_rows = cast( + List[Tuple[str, str, str, str, str]], txn.execute_values(sql, query_list) + ) + + seen_user_device: Set[Tuple[str, str]] = set() + for user_id, device_id, _, _, _ in otk_rows: + if (user_id, device_id) in seen_user_device: + continue + seen_user_device.add((user_id, device_id)) + self._invalidate_cache_and_stream( + txn, self.count_e2e_one_time_keys, (user_id, device_id) + ) + + return otk_rows + class EndToEndKeyStore(EndToEndKeyWorkerStore, SQLBaseStore): def __init__( diff --git a/synapse/storage/databases/main/event_federation.py b/synapse/storage/databases/main/event_federation.py index d4251be7e7..b8bbd1eccd 100644 --- a/synapse/storage/databases/main/event_federation.py +++ b/synapse/storage/databases/main/event_federation.py @@ -1048,15 +1048,18 @@ class EventFederationWorkerStore(SignatureWorkerStore, EventsWorkerStore, SQLBas Args: event_ids: The event IDs to calculate the max depth of. """ - rows = await self.db_pool.simple_select_many_batch( - table="events", - column="event_id", - iterable=event_ids, - retcols=( - "event_id", - "depth", + rows = cast( + List[Tuple[str, int]], + await self.db_pool.simple_select_many_batch( + table="events", + column="event_id", + iterable=event_ids, + retcols=( + "event_id", + "depth", + ), + desc="get_max_depth_of", ), - desc="get_max_depth_of", ) if not rows: @@ -1064,10 +1067,10 @@ class EventFederationWorkerStore(SignatureWorkerStore, EventsWorkerStore, SQLBas else: max_depth_event_id = "" current_max_depth = 0 - for row in rows: - if row["depth"] > current_max_depth: - max_depth_event_id = row["event_id"] - current_max_depth = row["depth"] + for event_id, depth in rows: + if depth > current_max_depth: + max_depth_event_id = event_id + current_max_depth = depth return max_depth_event_id, current_max_depth @@ -1077,15 +1080,18 @@ class EventFederationWorkerStore(SignatureWorkerStore, EventsWorkerStore, SQLBas Args: event_ids: The event IDs to calculate the max depth of. """ - rows = await self.db_pool.simple_select_many_batch( - table="events", - column="event_id", - iterable=event_ids, - retcols=( - "event_id", - "depth", + rows = cast( + List[Tuple[str, int]], + await self.db_pool.simple_select_many_batch( + table="events", + column="event_id", + iterable=event_ids, + retcols=( + "event_id", + "depth", + ), + desc="get_min_depth_of", ), - desc="get_min_depth_of", ) if not rows: @@ -1093,10 +1099,10 @@ class EventFederationWorkerStore(SignatureWorkerStore, EventsWorkerStore, SQLBas else: min_depth_event_id = "" current_min_depth = MAX_DEPTH - for row in rows: - if row["depth"] < current_min_depth: - min_depth_event_id = row["event_id"] - current_min_depth = row["depth"] + for event_id, depth in rows: + if depth < current_min_depth: + min_depth_event_id = event_id + current_min_depth = depth return min_depth_event_id, current_min_depth @@ -1552,19 +1558,18 @@ class EventFederationWorkerStore(SignatureWorkerStore, EventsWorkerStore, SQLBas A filtered down list of `event_ids` that have previous failed pull attempts. """ - rows = await self.db_pool.simple_select_many_batch( - table="event_failed_pull_attempts", - column="event_id", - iterable=event_ids, - keyvalues={}, - retcols=("event_id",), - desc="get_event_ids_with_failed_pull_attempts", + rows = cast( + List[Tuple[str]], + await self.db_pool.simple_select_many_batch( + table="event_failed_pull_attempts", + column="event_id", + iterable=event_ids, + keyvalues={}, + retcols=("event_id",), + desc="get_event_ids_with_failed_pull_attempts", + ), ) - event_ids_with_failed_pull_attempts: Set[str] = { - row["event_id"] for row in rows - } - - return event_ids_with_failed_pull_attempts + return {row[0] for row in rows} @trace async def get_event_ids_to_not_pull_from_backoff( @@ -1584,32 +1589,34 @@ class EventFederationWorkerStore(SignatureWorkerStore, EventsWorkerStore, SQLBas A dictionary of event_ids that should not be attempted to be pulled and the next timestamp at which we may try pulling them again. """ - 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", + event_failed_pull_attempts = cast( + List[Tuple[str, int, int]], + 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", ), - desc="get_event_ids_to_not_pull_from_backoff", ) current_time = self._clock.time_msec() event_ids_with_backoff = {} - for event_failed_pull_attempt in event_failed_pull_attempts: - event_id = event_failed_pull_attempt["event_id"] + for event_id, last_attempt_ts, num_attempts 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. backoff_end_time = ( - event_failed_pull_attempt["last_attempt_ts"] + last_attempt_ts + ( 2 ** min( - event_failed_pull_attempt["num_attempts"], + num_attempts, BACKFILL_EVENT_EXPONENTIAL_BACKOFF_MAXIMUM_DOUBLING_STEPS, ) ) @@ -1890,21 +1897,23 @@ class EventFederationWorkerStore(SignatureWorkerStore, EventsWorkerStore, SQLBas # keeping only the forward extremities (i.e. the events not referenced # by other events in the queue). We do this so that we can always # backpaginate in all the events we have dropped. - rows = await self.db_pool.simple_select_list( - table="federation_inbound_events_staging", - keyvalues={"room_id": room_id}, - retcols=("event_id", "event_json"), - desc="prune_staged_events_in_room_fetch", + rows = cast( + List[Tuple[str, str]], + await self.db_pool.simple_select_list( + table="federation_inbound_events_staging", + keyvalues={"room_id": room_id}, + retcols=("event_id", "event_json"), + desc="prune_staged_events_in_room_fetch", + ), ) # Find the set of events referenced by those in the queue, as well as # collecting all the event IDs in the queue. referenced_events: Set[str] = set() seen_events: Set[str] = set() - for row in rows: - event_id = row["event_id"] + for event_id, event_json in rows: seen_events.add(event_id) - event_d = db_to_json(row["event_json"]) + event_d = db_to_json(event_json) # We don't bother parsing the dicts into full blown event objects, # as that is needlessly expensive. diff --git a/synapse/storage/databases/main/event_push_actions.py b/synapse/storage/databases/main/event_push_actions.py index ed29d1fa5d..e4dc68c0d8 100644 --- a/synapse/storage/databases/main/event_push_actions.py +++ b/synapse/storage/databases/main/event_push_actions.py @@ -182,6 +182,7 @@ class UserPushAction(EmailPushAction): profile_tag: str +# TODO This is used as a cached value and is mutable. @attr.s(slots=True, auto_attribs=True) class NotifCounts: """ @@ -193,7 +194,7 @@ class NotifCounts: highlight_count: int = 0 -@attr.s(slots=True, auto_attribs=True) +@attr.s(slots=True, frozen=True, auto_attribs=True) class RoomNotifCounts: """ The per-user, per-room count of notifications. Used by sync and push. @@ -201,7 +202,7 @@ class RoomNotifCounts: main_timeline: NotifCounts # Map of thread ID to the notification counts. - threads: Dict[str, NotifCounts] + threads: Mapping[str, NotifCounts] @staticmethod def empty() -> "RoomNotifCounts": @@ -483,7 +484,7 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas return room_to_count - @cached(tree=True, max_entries=5000, iterable=True) + @cached(tree=True, max_entries=5000, iterable=True) # type: ignore[synapse-@cached-mutable] async def get_unread_event_push_actions_by_room_for_user( self, room_id: str, diff --git a/synapse/storage/databases/main/events.py b/synapse/storage/databases/main/events.py index 790d058c43..7c34bde3e5 100644 --- a/synapse/storage/databases/main/events.py +++ b/synapse/storage/databases/main/events.py @@ -27,6 +27,7 @@ from typing import ( Optional, Set, Tuple, + Union, cast, ) @@ -78,7 +79,7 @@ class DeltaState: Attributes: to_delete: List of type/state_keys to delete from current state to_insert: Map of state to upsert into current state - no_longer_in_room: The server is not longer in the room, so the room + no_longer_in_room: The server is no longer in the room, so the room should e.g. be removed from `current_state_events` table. """ @@ -130,22 +131,25 @@ class PersistEventsStore: @trace async def _persist_events_and_state_updates( self, + room_id: str, events_and_contexts: List[Tuple[EventBase, EventContext]], *, - state_delta_for_room: Dict[str, DeltaState], - new_forward_extremities: Dict[str, Set[str]], + state_delta_for_room: Optional[DeltaState], + new_forward_extremities: Optional[Set[str]], use_negative_stream_ordering: bool = False, inhibit_local_membership_updates: bool = False, ) -> None: """Persist a set of events alongside updates to the current state and - forward extremities tables. + forward extremities tables. + + Assumes that we are only persisting events for one room at a time. Args: + room_id: events_and_contexts: - state_delta_for_room: Map from room_id to the delta to apply to - room state - new_forward_extremities: Map from room_id to set of event IDs - that are the new forward extremities of the room. + state_delta_for_room: The delta to apply to the room state + new_forward_extremities: A set of event IDs that are the new forward + extremities of the room. use_negative_stream_ordering: Whether to start stream_ordering on the negative side and decrement. This should be set as True for backfilled events because backfilled events get a negative @@ -195,6 +199,7 @@ class PersistEventsStore: await self.db_pool.runInteraction( "persist_events", self._persist_events_txn, + room_id=room_id, events_and_contexts=events_and_contexts, inhibit_local_membership_updates=inhibit_local_membership_updates, state_delta_for_room=state_delta_for_room, @@ -220,9 +225,9 @@ class PersistEventsStore: event_counter.labels(event.type, origin_type, origin_entity).inc() - for room_id, latest_event_ids in new_forward_extremities.items(): + if new_forward_extremities: self.store.get_latest_event_ids_in_room.prefill( - (room_id,), frozenset(latest_event_ids) + (room_id,), frozenset(new_forward_extremities) ) async def _get_events_which_are_prevs(self, event_ids: Iterable[str]) -> List[str]: @@ -335,10 +340,11 @@ class PersistEventsStore: self, txn: LoggingTransaction, *, + room_id: str, events_and_contexts: List[Tuple[EventBase, EventContext]], inhibit_local_membership_updates: bool, - state_delta_for_room: Dict[str, DeltaState], - new_forward_extremities: Dict[str, Set[str]], + state_delta_for_room: Optional[DeltaState], + new_forward_extremities: Optional[Set[str]], ) -> None: """Insert some number of room events into the necessary database tables. @@ -346,8 +352,11 @@ class PersistEventsStore: and the rejections table. Things reading from those table will need to check whether the event was rejected. + Assumes that we are only persisting events for one room at a time. + Args: txn + room_id: The room the events are from events_and_contexts: events to persist inhibit_local_membership_updates: Stop the local_current_membership from being updated by these events. This should be set to True @@ -356,10 +365,9 @@ class PersistEventsStore: delete_existing True to purge existing table rows for the events from the database. This is useful when retrying due to IntegrityError. - state_delta_for_room: The current-state delta for each room. - new_forward_extremities: The new forward extremities for each room. - For each room, a list of the event ids which are the forward - extremities. + state_delta_for_room: The current-state delta for the room. + new_forward_extremities: The new forward extremities for the room: + a set of the event ids which are the forward extremities. Raises: PartialStateConflictError: if attempting to persist a partial state event in @@ -375,14 +383,13 @@ class PersistEventsStore: # # Annoyingly SQLite doesn't support row level locking. if isinstance(self.database_engine, PostgresEngine): - for room_id in {e.room_id for e, _ in events_and_contexts}: - txn.execute( - "SELECT room_version FROM rooms WHERE room_id = ? FOR SHARE", - (room_id,), - ) - row = txn.fetchone() - if row is None: - raise Exception(f"Room does not exist {room_id}") + txn.execute( + "SELECT room_version FROM rooms WHERE room_id = ? FOR SHARE", + (room_id,), + ) + row = txn.fetchone() + if row is None: + raise Exception(f"Room does not exist {room_id}") # stream orderings should have been assigned by now assert min_stream_order @@ -418,7 +425,9 @@ class PersistEventsStore: events_and_contexts ) - self._update_room_depths_txn(txn, events_and_contexts=events_and_contexts) + self._update_room_depths_txn( + txn, room_id, events_and_contexts=events_and_contexts + ) # _update_outliers_txn filters out any events which have already been # persisted, and returns the filtered list. @@ -431,11 +440,13 @@ class PersistEventsStore: self._store_event_txn(txn, events_and_contexts=events_and_contexts) - self._update_forward_extremities_txn( - txn, - new_forward_extremities=new_forward_extremities, - max_stream_order=max_stream_order, - ) + if new_forward_extremities: + self._update_forward_extremities_txn( + txn, + room_id, + new_forward_extremities=new_forward_extremities, + max_stream_order=max_stream_order, + ) self._persist_transaction_ids_txn(txn, events_and_contexts) @@ -463,7 +474,10 @@ class PersistEventsStore: # We call this last as it assumes we've inserted the events into # room_memberships, where applicable. # NB: This function invalidates all state related caches - self._update_current_state_txn(txn, state_delta_for_room, min_stream_order) + if state_delta_for_room: + self._update_current_state_txn( + txn, room_id, state_delta_for_room, min_stream_order + ) def _persist_event_auth_chain_txn( self, @@ -501,16 +515,19 @@ class PersistEventsStore: # We ignore legacy rooms that we aren't filling the chain cover index # for. - rows = self.db_pool.simple_select_many_txn( - txn, - table="rooms", - column="room_id", - iterable={event.room_id for event in events if event.is_state()}, - keyvalues={}, - retcols=("room_id", "has_auth_chain_index"), + rows = cast( + List[Tuple[str, Optional[Union[int, bool]]]], + self.db_pool.simple_select_many_txn( + txn, + table="rooms", + column="room_id", + iterable={event.room_id for event in events if event.is_state()}, + keyvalues={}, + retcols=("room_id", "has_auth_chain_index"), + ), ) rooms_using_chain_index = { - row["room_id"] for row in rows if row["has_auth_chain_index"] + room_id for room_id, has_auth_chain_index in rows if has_auth_chain_index } state_events = { @@ -571,19 +588,18 @@ class PersistEventsStore: # We check if there are any events that need to be handled in the rooms # we're looking at. These should just be out of band memberships, where # we didn't have the auth chain when we first persisted. - rows = db_pool.simple_select_many_txn( - txn, - table="event_auth_chain_to_calculate", - keyvalues={}, - column="room_id", - iterable=set(event_to_room_id.values()), - retcols=("event_id", "type", "state_key"), + auth_chain_to_calc_rows = cast( + List[Tuple[str, str, str]], + db_pool.simple_select_many_txn( + txn, + table="event_auth_chain_to_calculate", + keyvalues={}, + column="room_id", + iterable=set(event_to_room_id.values()), + retcols=("event_id", "type", "state_key"), + ), ) - for row in rows: - event_id = row["event_id"] - event_type = row["type"] - state_key = row["state_key"] - + for event_id, event_type, state_key in auth_chain_to_calc_rows: # (We could pull out the auth events for all rows at once using # simple_select_many, but this case happens rarely and almost always # with a single row.) @@ -753,23 +769,31 @@ class PersistEventsStore: # Step 1, fetch all existing links from all the chains we've seen # referenced. chain_links = _LinkMap() - rows = db_pool.simple_select_many_txn( - txn, - table="event_auth_chain_links", - column="origin_chain_id", - iterable={chain_id for chain_id, _ in chain_map.values()}, - keyvalues={}, - retcols=( - "origin_chain_id", - "origin_sequence_number", - "target_chain_id", - "target_sequence_number", + auth_chain_rows = cast( + List[Tuple[int, int, int, int]], + db_pool.simple_select_many_txn( + txn, + table="event_auth_chain_links", + column="origin_chain_id", + iterable={chain_id for chain_id, _ in chain_map.values()}, + keyvalues={}, + retcols=( + "origin_chain_id", + "origin_sequence_number", + "target_chain_id", + "target_sequence_number", + ), ), ) - for row in rows: + for ( + origin_chain_id, + origin_sequence_number, + target_chain_id, + target_sequence_number, + ) in auth_chain_rows: chain_links.add_link( - (row["origin_chain_id"], row["origin_sequence_number"]), - (row["target_chain_id"], row["target_sequence_number"]), + (origin_chain_id, origin_sequence_number), + (target_chain_id, target_sequence_number), new=False, ) @@ -1015,74 +1039,75 @@ class PersistEventsStore: await self.db_pool.runInteraction( "update_current_state", self._update_current_state_txn, - state_delta_by_room={room_id: state_delta}, + room_id, + delta_state=state_delta, stream_id=stream_ordering, ) def _update_current_state_txn( self, txn: LoggingTransaction, - state_delta_by_room: Dict[str, DeltaState], + room_id: str, + delta_state: DeltaState, stream_id: int, ) -> None: - for room_id, delta_state in state_delta_by_room.items(): - to_delete = delta_state.to_delete - to_insert = delta_state.to_insert - - # Figure out the changes of membership to invalidate the - # `get_rooms_for_user` cache. - # We find out which membership events we may have deleted - # and which we have added, then we invalidate the caches for all - # those users. - members_changed = { - state_key - for ev_type, state_key in itertools.chain(to_delete, to_insert) - if ev_type == EventTypes.Member - } + to_delete = delta_state.to_delete + to_insert = delta_state.to_insert + + # Figure out the changes of membership to invalidate the + # `get_rooms_for_user` cache. + # We find out which membership events we may have deleted + # and which we have added, then we invalidate the caches for all + # those users. + members_changed = { + state_key + for ev_type, state_key in itertools.chain(to_delete, to_insert) + if ev_type == EventTypes.Member + } - if delta_state.no_longer_in_room: - # Server is no longer in the room so we delete the room from - # current_state_events, being careful we've already updated the - # rooms.room_version column (which gets populated in a - # background task). - self._upsert_room_version_txn(txn, room_id) + if delta_state.no_longer_in_room: + # Server is no longer in the room so we delete the room from + # current_state_events, being careful we've already updated the + # rooms.room_version column (which gets populated in a + # background task). + self._upsert_room_version_txn(txn, room_id) - # Before deleting we populate the current_state_delta_stream - # so that async background tasks get told what happened. - sql = """ + # Before deleting we populate the current_state_delta_stream + # so that async background tasks get told what happened. + sql = """ INSERT INTO current_state_delta_stream (stream_id, instance_name, room_id, type, state_key, event_id, prev_event_id) SELECT ?, ?, room_id, type, state_key, null, event_id FROM current_state_events WHERE room_id = ? """ - txn.execute(sql, (stream_id, self._instance_name, room_id)) + txn.execute(sql, (stream_id, self._instance_name, room_id)) - # We also want to invalidate the membership caches for users - # that were in the room. - users_in_room = self.store.get_users_in_room_txn(txn, room_id) - members_changed.update(users_in_room) + # We also want to invalidate the membership caches for users + # that were in the room. + users_in_room = self.store.get_users_in_room_txn(txn, room_id) + members_changed.update(users_in_room) - self.db_pool.simple_delete_txn( - txn, - table="current_state_events", - keyvalues={"room_id": room_id}, - ) - else: - # We're still in the room, so we update the current state as normal. + self.db_pool.simple_delete_txn( + txn, + table="current_state_events", + keyvalues={"room_id": room_id}, + ) + else: + # We're still in the room, so we update the current state as normal. - # First we add entries to the current_state_delta_stream. We - # do this before updating the current_state_events table so - # that we can use it to calculate the `prev_event_id`. (This - # allows us to not have to pull out the existing state - # unnecessarily). - # - # The stream_id for the update is chosen to be the minimum of the stream_ids - # for the batch of the events that we are persisting; that means we do not - # end up in a situation where workers see events before the - # current_state_delta updates. - # - sql = """ + # First we add entries to the current_state_delta_stream. We + # do this before updating the current_state_events table so + # that we can use it to calculate the `prev_event_id`. (This + # allows us to not have to pull out the existing state + # unnecessarily). + # + # The stream_id for the update is chosen to be the minimum of the stream_ids + # for the batch of the events that we are persisting; that means we do not + # end up in a situation where workers see events before the + # current_state_delta updates. + # + sql = """ INSERT INTO current_state_delta_stream (stream_id, instance_name, room_id, type, state_key, event_id, prev_event_id) SELECT ?, ?, ?, ?, ?, ?, ( @@ -1090,39 +1115,39 @@ class PersistEventsStore: WHERE room_id = ? AND type = ? AND state_key = ? ) """ - txn.execute_batch( - sql, + txn.execute_batch( + sql, + ( ( - ( - stream_id, - self._instance_name, - room_id, - etype, - state_key, - to_insert.get((etype, state_key)), - room_id, - etype, - state_key, - ) - for etype, state_key in itertools.chain(to_delete, to_insert) - ), - ) - # Now we actually update the current_state_events table + stream_id, + self._instance_name, + room_id, + etype, + state_key, + to_insert.get((etype, state_key)), + room_id, + etype, + state_key, + ) + for etype, state_key in itertools.chain(to_delete, to_insert) + ), + ) + # Now we actually update the current_state_events table - txn.execute_batch( - "DELETE FROM current_state_events" - " WHERE room_id = ? AND type = ? AND state_key = ?", - ( - (room_id, etype, state_key) - for etype, state_key in itertools.chain(to_delete, to_insert) - ), - ) + txn.execute_batch( + "DELETE FROM current_state_events" + " WHERE room_id = ? AND type = ? AND state_key = ?", + ( + (room_id, etype, state_key) + for etype, state_key in itertools.chain(to_delete, to_insert) + ), + ) - # We include the membership in the current state table, hence we do - # a lookup when we insert. This assumes that all events have already - # been inserted into room_memberships. - txn.execute_batch( - """INSERT INTO current_state_events + # We include the membership in the current state table, hence we do + # a lookup when we insert. This assumes that all events have already + # been inserted into room_memberships. + txn.execute_batch( + """INSERT INTO current_state_events (room_id, type, state_key, event_id, membership, event_stream_ordering) VALUES ( ?, ?, ?, ?, @@ -1130,34 +1155,34 @@ class PersistEventsStore: (SELECT stream_ordering FROM events WHERE event_id = ?) ) """, - [ - (room_id, key[0], key[1], ev_id, ev_id, ev_id) - for key, ev_id in to_insert.items() - ], - ) + [ + (room_id, key[0], key[1], ev_id, ev_id, ev_id) + for key, ev_id in to_insert.items() + ], + ) - # We now update `local_current_membership`. We do this regardless - # of whether we're still in the room or not to handle the case where - # e.g. we just got banned (where we need to record that fact here). - - # Note: Do we really want to delete rows here (that we do not - # subsequently reinsert below)? While technically correct it means - # we have no record of the fact the user *was* a member of the - # room but got, say, state reset out of it. - if to_delete or to_insert: - txn.execute_batch( - "DELETE FROM local_current_membership" - " WHERE room_id = ? AND user_id = ?", - ( - (room_id, state_key) - for etype, state_key in itertools.chain(to_delete, to_insert) - if etype == EventTypes.Member and self.is_mine_id(state_key) - ), - ) + # We now update `local_current_membership`. We do this regardless + # of whether we're still in the room or not to handle the case where + # e.g. we just got banned (where we need to record that fact here). - if to_insert: - txn.execute_batch( - """INSERT INTO local_current_membership + # Note: Do we really want to delete rows here (that we do not + # subsequently reinsert below)? While technically correct it means + # we have no record of the fact the user *was* a member of the + # room but got, say, state reset out of it. + if to_delete or to_insert: + txn.execute_batch( + "DELETE FROM local_current_membership" + " WHERE room_id = ? AND user_id = ?", + ( + (room_id, state_key) + for etype, state_key in itertools.chain(to_delete, to_insert) + if etype == EventTypes.Member and self.is_mine_id(state_key) + ), + ) + + if to_insert: + txn.execute_batch( + """INSERT INTO local_current_membership (room_id, user_id, event_id, membership, event_stream_ordering) VALUES ( ?, ?, ?, @@ -1165,29 +1190,27 @@ class PersistEventsStore: (SELECT stream_ordering FROM events WHERE event_id = ?) ) """, - [ - (room_id, key[1], ev_id, ev_id, ev_id) - for key, ev_id in to_insert.items() - if key[0] == EventTypes.Member and self.is_mine_id(key[1]) - ], - ) - - txn.call_after( - self.store._curr_state_delta_stream_cache.entity_has_changed, - room_id, - stream_id, + [ + (room_id, key[1], ev_id, ev_id, ev_id) + for key, ev_id in to_insert.items() + if key[0] == EventTypes.Member and self.is_mine_id(key[1]) + ], ) - # Invalidate the various caches - self.store._invalidate_state_caches_and_stream( - txn, room_id, members_changed - ) + txn.call_after( + self.store._curr_state_delta_stream_cache.entity_has_changed, + room_id, + stream_id, + ) - # Check if any of the remote membership changes requires us to - # unsubscribe from their device lists. - self.store.handle_potentially_left_users_txn( - txn, {m for m in members_changed if not self.hs.is_mine_id(m)} - ) + # Invalidate the various caches + self.store._invalidate_state_caches_and_stream(txn, room_id, members_changed) + + # Check if any of the remote membership changes requires us to + # unsubscribe from their device lists. + self.store.handle_potentially_left_users_txn( + txn, {m for m in members_changed if not self.hs.is_mine_id(m)} + ) def _upsert_room_version_txn(self, txn: LoggingTransaction, room_id: str) -> None: """Update the room version in the database based off current state @@ -1221,23 +1244,19 @@ class PersistEventsStore: def _update_forward_extremities_txn( self, txn: LoggingTransaction, - new_forward_extremities: Dict[str, Set[str]], + room_id: str, + new_forward_extremities: Set[str], max_stream_order: int, ) -> None: - for room_id in new_forward_extremities.keys(): - self.db_pool.simple_delete_txn( - txn, table="event_forward_extremities", keyvalues={"room_id": room_id} - ) + self.db_pool.simple_delete_txn( + txn, table="event_forward_extremities", keyvalues={"room_id": room_id} + ) self.db_pool.simple_insert_many_txn( txn, table="event_forward_extremities", keys=("event_id", "room_id"), - values=[ - (ev_id, room_id) - for room_id, new_extrem in new_forward_extremities.items() - for ev_id in new_extrem - ], + values=[(ev_id, room_id) for ev_id in new_forward_extremities], ) # We now insert into stream_ordering_to_exterm a mapping from room_id, # new stream_ordering to new forward extremeties in the room. @@ -1249,8 +1268,7 @@ class PersistEventsStore: keys=("room_id", "event_id", "stream_ordering"), values=[ (room_id, event_id, max_stream_order) - for room_id, new_extrem in new_forward_extremities.items() - for event_id in new_extrem + for event_id in new_forward_extremities ], ) @@ -1287,36 +1305,45 @@ class PersistEventsStore: def _update_room_depths_txn( self, txn: LoggingTransaction, + room_id: str, events_and_contexts: List[Tuple[EventBase, EventContext]], ) -> None: """Update min_depth for each room Args: txn: db connection + room_id: The room ID events_and_contexts: events we are persisting """ - depth_updates: Dict[str, int] = {} + stream_ordering: Optional[int] = None + depth_update = 0 for event, context in events_and_contexts: - # Then update the `stream_ordering` position to mark the latest - # event as the front of the room. This should not be done for - # backfilled events because backfilled events have negative - # stream_ordering and happened in the past so we know that we don't - # need to update the stream_ordering tip/front for the room. + # Don't update the stream ordering for backfilled events because + # backfilled events have negative stream_ordering and happened in the + # past, so we know that we don't need to update the stream_ordering + # tip/front for the room. assert event.internal_metadata.stream_ordering is not None if event.internal_metadata.stream_ordering >= 0: - txn.call_after( - self.store._events_stream_cache.entity_has_changed, - event.room_id, - event.internal_metadata.stream_ordering, - ) + if stream_ordering is None: + stream_ordering = event.internal_metadata.stream_ordering + else: + stream_ordering = max( + stream_ordering, event.internal_metadata.stream_ordering + ) if not event.internal_metadata.is_outlier() and not context.rejected: - depth_updates[event.room_id] = max( - event.depth, depth_updates.get(event.room_id, event.depth) - ) + depth_update = max(event.depth, depth_update) - for room_id, depth in depth_updates.items(): - self._update_min_depth_for_room_txn(txn, room_id, depth) + # Then update the `stream_ordering` position to mark the latest event as + # the front of the room. + if stream_ordering is not None: + txn.call_after( + self.store._events_stream_cache.entity_has_changed, + room_id, + stream_ordering, + ) + + self._update_min_depth_for_room_txn(txn, room_id, depth_update) def _update_outliers_txn( self, @@ -1339,13 +1366,19 @@ class PersistEventsStore: PartialStateConflictError: if attempting to persist a partial state event in a room that has been un-partial stated. """ - txn.execute( - "SELECT event_id, outlier FROM events WHERE event_id in (%s)" - % (",".join(["?"] * len(events_and_contexts)),), - [event.event_id for event, _ in events_and_contexts], + rows = cast( + List[Tuple[str, bool]], + self.db_pool.simple_select_many_txn( + txn, + "events", + "event_id", + [event.event_id for event, _ in events_and_contexts], + keyvalues={}, + retcols=("event_id", "outlier"), + ), ) - have_persisted = dict(cast(Iterable[Tuple[str, bool]], txn)) + have_persisted = dict(rows) logger.debug( "_update_outliers_txn: events=%s have_persisted=%s", @@ -1443,7 +1476,7 @@ class PersistEventsStore: txn, table="event_json", keys=("event_id", "room_id", "internal_metadata", "json", "format_version"), - values=( + values=[ ( event.event_id, event.room_id, @@ -1452,7 +1485,7 @@ class PersistEventsStore: event.format_version, ) for event, _ in events_and_contexts - ), + ], ) self.db_pool.simple_insert_many_txn( @@ -1475,7 +1508,7 @@ class PersistEventsStore: "state_key", "rejection_reason", ), - values=( + values=[ ( self._instance_name, event.internal_metadata.stream_ordering, @@ -1494,7 +1527,7 @@ class PersistEventsStore: context.rejected, ) for event, context in events_and_contexts - ), + ], ) # If we're persisting an unredacted event we go and ensure @@ -1517,11 +1550,11 @@ class PersistEventsStore: txn, table="state_events", keys=("event_id", "room_id", "type", "state_key"), - values=( + values=[ (event.event_id, event.room_id, event.type, event.state_key) for event, _ in events_and_contexts if event.is_state() - ), + ], ) def _store_rejected_events_txn( @@ -1654,8 +1687,6 @@ class PersistEventsStore: ) -> None: to_prefill = [] - rows = [] - ev_map = {e.event_id: e for e, _ in events_and_contexts} if not ev_map: return @@ -1676,10 +1707,9 @@ class PersistEventsStore: ) txn.execute(sql + clause, args) - rows = self.db_pool.cursor_to_dict(txn) - for row in rows: - event = ev_map[row["event_id"]] - if not row["rejects"] and not row["redacts"]: + for event_id, redacts, rejects in txn: + event = ev_map[event_id] + if not rejects and not redacts: to_prefill.append(EventCacheEntry(event=event, redacted_event=None)) async def external_prefill() -> None: @@ -2259,35 +2289,59 @@ class PersistEventsStore: Forward extremities are handled when we first start persisting the events. """ - # From the events passed in, add all of the prev events as backwards extremities. - # Ignore any events that are already backwards extrems or outliers. - query = ( - "INSERT INTO event_backward_extremities (event_id, room_id)" - " SELECT ?, ? WHERE NOT EXISTS (" - " SELECT 1 FROM event_backward_extremities" - " WHERE event_id = ? AND room_id = ?" - " )" - # 1. Don't add an event as a extremity again if we already persisted it - # as a non-outlier. - # 2. Don't add an outlier as an extremity if it has no prev_events - " AND NOT EXISTS (" - " SELECT 1 FROM events" - " LEFT JOIN event_edges edge" - " ON edge.event_id = events.event_id" - " WHERE events.event_id = ? AND events.room_id = ? AND (events.outlier = FALSE OR edge.event_id IS NULL)" - " )" + + room_id = events[0].room_id + + potential_backwards_extremities = { + e_id + for ev in events + for e_id in ev.prev_event_ids() + if not ev.internal_metadata.is_outlier() + } + + if not potential_backwards_extremities: + return + + existing_events_outliers = self.db_pool.simple_select_many_txn( + txn, + table="events", + column="event_id", + iterable=potential_backwards_extremities, + keyvalues={"outlier": False}, + retcols=("event_id",), ) - txn.execute_batch( - query, - [ - (e_id, ev.room_id, e_id, ev.room_id, e_id, ev.room_id) - for ev in events - for e_id in ev.prev_event_ids() - if not ev.internal_metadata.is_outlier() - ], + potential_backwards_extremities.difference_update( + e for e, in existing_events_outliers ) + if potential_backwards_extremities: + self.db_pool.simple_upsert_many_txn( + txn, + table="event_backward_extremities", + key_names=("room_id", "event_id"), + key_values=[(room_id, ev) for ev in potential_backwards_extremities], + value_names=(), + value_values=(), + ) + + # Record the stream orderings where we have new gaps. + gap_events = [ + (room_id, self._instance_name, ev.internal_metadata.stream_ordering) + for ev in events + if any( + e_id in potential_backwards_extremities + for e_id in ev.prev_event_ids() + ) + ] + + self.db_pool.simple_insert_many_txn( + txn, + table="timeline_gaps", + keys=("room_id", "instance_name", "stream_ordering"), + values=gap_events, + ) + # Delete all these events that we've already fetched and now know that their # prev events are the new backwards extremeties. query = ( diff --git a/synapse/storage/databases/main/events_bg_updates.py b/synapse/storage/databases/main/events_bg_updates.py index daef3685b0..0061805150 100644 --- a/synapse/storage/databases/main/events_bg_updates.py +++ b/synapse/storage/databases/main/events_bg_updates.py @@ -369,18 +369,20 @@ class EventsBackgroundUpdatesStore(SQLBaseStore): chunks = [event_ids[i : i + 100] for i in range(0, len(event_ids), 100)] for chunk in chunks: - ev_rows = self.db_pool.simple_select_many_txn( - txn, - table="event_json", - column="event_id", - iterable=chunk, - retcols=["event_id", "json"], - keyvalues={}, + ev_rows = cast( + List[Tuple[str, str]], + self.db_pool.simple_select_many_txn( + txn, + table="event_json", + column="event_id", + iterable=chunk, + retcols=["event_id", "json"], + keyvalues={}, + ), ) - for row in ev_rows: - event_id = row["event_id"] - event_json = db_to_json(row["json"]) + for event_id, json in ev_rows: + event_json = db_to_json(json) try: origin_server_ts = event_json["origin_server_ts"] except (KeyError, AttributeError): @@ -563,15 +565,18 @@ class EventsBackgroundUpdatesStore(SQLBaseStore): if deleted: # We now need to invalidate the caches of these rooms - rows = self.db_pool.simple_select_many_txn( - txn, - table="events", - column="event_id", - iterable=to_delete, - keyvalues={}, - retcols=("room_id",), + rows = cast( + List[Tuple[str]], + self.db_pool.simple_select_many_txn( + txn, + table="events", + column="event_id", + iterable=to_delete, + keyvalues={}, + retcols=("room_id",), + ), ) - room_ids = {row["room_id"] for row in rows} + room_ids = {row[0] for row in rows} for room_id in room_ids: txn.call_after( self.get_latest_event_ids_in_room.invalidate, (room_id,) # type: ignore[attr-defined] @@ -1038,18 +1043,21 @@ class EventsBackgroundUpdatesStore(SQLBaseStore): count = len(rows) # We also need to fetch the auth events for them. - auth_events = self.db_pool.simple_select_many_txn( - txn, - table="event_auth", - column="event_id", - iterable=event_to_room_id, - keyvalues={}, - retcols=("event_id", "auth_id"), + auth_events = cast( + List[Tuple[str, str]], + self.db_pool.simple_select_many_txn( + txn, + table="event_auth", + column="event_id", + iterable=event_to_room_id, + keyvalues={}, + retcols=("event_id", "auth_id"), + ), ) event_to_auth_chain: Dict[str, List[str]] = {} - for row in auth_events: - event_to_auth_chain.setdefault(row["event_id"], []).append(row["auth_id"]) + for event_id, auth_id in auth_events: + event_to_auth_chain.setdefault(event_id, []).append(auth_id) # Calculate and persist the chain cover index for this set of events. # @@ -1302,12 +1310,9 @@ class EventsBackgroundUpdatesStore(SQLBaseStore): # ANALYZE the new column to build stats on it, to encourage PostgreSQL to use the # indexes on it. - # We need to pass execute a dummy function to handle the txn's result otherwise - # it tries to call fetchall() on it and fails because there's no result to fetch. - await self.db_pool.execute( + await self.db_pool.runInteraction( "background_analyze_new_stream_ordering_column", - lambda txn: None, - "ANALYZE events(stream_ordering2)", + lambda txn: txn.execute("ANALYZE events(stream_ordering2)"), ) await self.db_pool.runInteraction( diff --git a/synapse/storage/databases/main/events_forward_extremities.py b/synapse/storage/databases/main/events_forward_extremities.py index f851bff604..0ba84b1469 100644 --- a/synapse/storage/databases/main/events_forward_extremities.py +++ b/synapse/storage/databases/main/events_forward_extremities.py @@ -13,7 +13,7 @@ # limitations under the License. import logging -from typing import Any, Dict, List +from typing import List, Optional, Tuple, cast from synapse.api.errors import SynapseError from synapse.storage.database import LoggingTransaction @@ -91,12 +91,17 @@ class EventForwardExtremitiesStore( async def get_forward_extremities_for_room( self, room_id: str - ) -> List[Dict[str, Any]]: - """Get list of forward extremities for a room.""" + ) -> List[Tuple[str, int, int, Optional[int]]]: + """ + Get list of forward extremities for a room. + + Returns: + A list of tuples of event_id, state_group, depth, and received_ts. + """ def get_forward_extremities_for_room_txn( txn: LoggingTransaction, - ) -> List[Dict[str, Any]]: + ) -> List[Tuple[str, int, int, Optional[int]]]: sql = """ SELECT event_id, state_group, depth, received_ts FROM event_forward_extremities @@ -106,7 +111,7 @@ class EventForwardExtremitiesStore( """ txn.execute(sql, (room_id,)) - return self.db_pool.cursor_to_dict(txn) + return cast(List[Tuple[str, int, int, Optional[int]]], txn.fetchall()) return await self.db_pool.runInteraction( "get_forward_extremities_for_room", diff --git a/synapse/storage/databases/main/events_worker.py b/synapse/storage/databases/main/events_worker.py index b788d70fc5..5bf864c1fb 100644 --- a/synapse/storage/databases/main/events_worker.py +++ b/synapse/storage/databases/main/events_worker.py @@ -1584,16 +1584,19 @@ class EventsWorkerStore(SQLBaseStore): """Given a list of event ids, check if we have already processed and stored them as non outliers. """ - rows = await self.db_pool.simple_select_many_batch( - table="events", - retcols=("event_id",), - column="event_id", - iterable=list(event_ids), - keyvalues={"outlier": False}, - desc="have_events_in_timeline", + rows = cast( + List[Tuple[str]], + await self.db_pool.simple_select_many_batch( + table="events", + retcols=("event_id",), + column="event_id", + iterable=list(event_ids), + keyvalues={"outlier": False}, + desc="have_events_in_timeline", + ), ) - return {r["event_id"] for r in rows} + return {r[0] for r in rows} @trace @tag_args @@ -2093,12 +2096,6 @@ class EventsWorkerStore(SQLBaseStore): def _cleanup_old_transaction_ids_txn(txn: LoggingTransaction) -> None: one_day_ago = self._clock.time_msec() - 24 * 60 * 60 * 1000 sql = """ - DELETE FROM event_txn_id - WHERE inserted_ts < ? - """ - txn.execute(sql, (one_day_ago,)) - - sql = """ DELETE FROM event_txn_id_device_id WHERE inserted_ts < ? """ @@ -2336,15 +2333,18 @@ class EventsWorkerStore(SQLBaseStore): a dict mapping from event id to partial-stateness. We return True for any of the events which are unknown (or are outliers). """ - result = await self.db_pool.simple_select_many_batch( - table="partial_state_events", - column="event_id", - iterable=event_ids, - retcols=["event_id"], - desc="get_partial_state_events", + result = cast( + List[Tuple[str]], + await self.db_pool.simple_select_many_batch( + table="partial_state_events", + column="event_id", + iterable=event_ids, + retcols=["event_id"], + desc="get_partial_state_events", + ), ) # convert the result to a dict, to make @cachedList work - partial = {r["event_id"] for r in result} + partial = {r[0] for r in result} return {e_id: e_id in partial for e_id in event_ids} @cached() diff --git a/synapse/storage/databases/main/experimental_features.py b/synapse/storage/databases/main/experimental_features.py index 654f924019..60621edeef 100644 --- a/synapse/storage/databases/main/experimental_features.py +++ b/synapse/storage/databases/main/experimental_features.py @@ -12,7 +12,7 @@ # See the License for the specific language governing permissions and # limitations under the License. -from typing import TYPE_CHECKING, Dict, FrozenSet +from typing import TYPE_CHECKING, Dict, FrozenSet, List, Tuple, cast from synapse.storage.database import DatabasePool, LoggingDatabaseConnection from synapse.storage.databases.main import CacheInvalidationWorkerStore @@ -42,13 +42,16 @@ class ExperimentalFeaturesStore(CacheInvalidationWorkerStore): Returns: the features currently enabled for the user """ - enabled = await self.db_pool.simple_select_list( - "per_user_experimental_features", - {"user_id": user_id, "enabled": True}, - ["feature"], + enabled = cast( + List[Tuple[str]], + await self.db_pool.simple_select_list( + table="per_user_experimental_features", + keyvalues={"user_id": user_id, "enabled": True}, + retcols=("feature",), + ), ) - return frozenset(feature["feature"] for feature in enabled) + return frozenset(feature[0] for feature in enabled) async def set_features_for_user( self, diff --git a/synapse/storage/databases/main/keys.py b/synapse/storage/databases/main/keys.py index 889c578b9c..ce88772f9e 100644 --- a/synapse/storage/databases/main/keys.py +++ b/synapse/storage/databases/main/keys.py @@ -16,7 +16,7 @@ import itertools import json import logging -from typing import Dict, Iterable, Mapping, Optional, Tuple +from typing import Dict, Iterable, List, Mapping, Optional, Tuple, Union, cast from canonicaljson import encode_canonical_json from signedjson.key import decode_verify_key_bytes @@ -205,35 +205,39 @@ class KeyStore(CacheInvalidationWorkerStore): If we have multiple entries for a given key ID, returns the most recent. """ - rows = await self.db_pool.simple_select_many_batch( - table="server_keys_json", - column="key_id", - iterable=key_ids, - keyvalues={"server_name": server_name}, - retcols=( - "key_id", - "from_server", - "ts_added_ms", - "ts_valid_until_ms", - "key_json", + rows = cast( + List[Tuple[str, str, int, int, Union[bytes, memoryview]]], + await self.db_pool.simple_select_many_batch( + table="server_keys_json", + column="key_id", + iterable=key_ids, + keyvalues={"server_name": server_name}, + retcols=( + "key_id", + "from_server", + "ts_added_ms", + "ts_valid_until_ms", + "key_json", + ), + desc="get_server_keys_json_for_remote", ), - desc="get_server_keys_json_for_remote", ) if not rows: return {} - # We sort the rows so that the most recently added entry is picked up. - rows.sort(key=lambda r: r["ts_added_ms"]) + # We sort the rows by ts_added_ms so that the most recently added entry + # will stomp over older entries in the dictionary. + rows.sort(key=lambda r: r[2]) return { - row["key_id"]: FetchKeyResultForRemote( + key_id: FetchKeyResultForRemote( # Cast to bytes since postgresql returns a memoryview. - key_json=bytes(row["key_json"]), - valid_until_ts=row["ts_valid_until_ms"], - added_ts=row["ts_added_ms"], + key_json=bytes(key_json), + valid_until_ts=ts_valid_until_ms, + added_ts=ts_added_ms, ) - for row in rows + for key_id, from_server, ts_added_ms, ts_valid_until_ms, key_json in rows } async def get_all_server_keys_json_for_remote( @@ -244,30 +248,35 @@ class KeyStore(CacheInvalidationWorkerStore): If we have multiple entries for a given key ID, returns the most recent. """ - rows = await self.db_pool.simple_select_list( - table="server_keys_json", - keyvalues={"server_name": server_name}, - retcols=( - "key_id", - "from_server", - "ts_added_ms", - "ts_valid_until_ms", - "key_json", + rows = cast( + List[Tuple[str, str, int, int, Union[bytes, memoryview]]], + await self.db_pool.simple_select_list( + table="server_keys_json", + keyvalues={"server_name": server_name}, + retcols=( + "key_id", + "from_server", + "ts_added_ms", + "ts_valid_until_ms", + "key_json", + ), + desc="get_server_keys_json_for_remote", ), - desc="get_server_keys_json_for_remote", ) if not rows: return {} - rows.sort(key=lambda r: r["ts_added_ms"]) + # We sort the rows by ts_added_ms so that the most recently added entry + # will stomp over older entries in the dictionary. + rows.sort(key=lambda r: r[2]) return { - row["key_id"]: FetchKeyResultForRemote( + key_id: FetchKeyResultForRemote( # Cast to bytes since postgresql returns a memoryview. - key_json=bytes(row["key_json"]), - valid_until_ts=row["ts_valid_until_ms"], - added_ts=row["ts_added_ms"], + key_json=bytes(key_json), + valid_until_ts=ts_valid_until_ms, + added_ts=ts_added_ms, ) - for row in rows + for key_id, from_server, ts_added_ms, ts_valid_until_ms, key_json in rows } diff --git a/synapse/storage/databases/main/media_repository.py b/synapse/storage/databases/main/media_repository.py index 8cebeb5189..c8d7c9fd32 100644 --- a/synapse/storage/databases/main/media_repository.py +++ b/synapse/storage/databases/main/media_repository.py @@ -26,8 +26,11 @@ from typing import ( cast, ) +import attr + from synapse.api.constants import Direction from synapse.logging.opentracing import trace +from synapse.media._base import ThumbnailInfo from synapse.storage._base import SQLBaseStore from synapse.storage.database import ( DatabasePool, @@ -44,6 +47,18 @@ BG_UPDATE_REMOVE_MEDIA_REPO_INDEX_WITHOUT_METHOD_2 = ( ) +@attr.s(slots=True, frozen=True, auto_attribs=True) +class LocalMedia: + media_id: str + media_type: str + media_length: int + upload_name: str + created_ts: int + last_access_ts: int + quarantined_by: Optional[str] + safe_from_quarantine: bool + + class MediaSortOrder(Enum): """ Enum to define the sorting method used when returning media with @@ -179,7 +194,7 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore): user_id: str, order_by: str = MediaSortOrder.CREATED_TS.value, direction: Direction = Direction.FORWARDS, - ) -> Tuple[List[Dict[str, Any]], int]: + ) -> Tuple[List[LocalMedia], int]: """Get a paginated list of metadata for a local piece of media which an user_id has uploaded @@ -196,7 +211,7 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore): def get_local_media_by_user_paginate_txn( txn: LoggingTransaction, - ) -> Tuple[List[Dict[str, Any]], int]: + ) -> Tuple[List[LocalMedia], int]: # Set ordering order_by_column = MediaSortOrder(order_by).value @@ -216,14 +231,14 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore): sql = """ SELECT - "media_id", - "media_type", - "media_length", - "upload_name", - "created_ts", - "last_access_ts", - "quarantined_by", - "safe_from_quarantine" + media_id, + media_type, + media_length, + upload_name, + created_ts, + last_access_ts, + quarantined_by, + safe_from_quarantine FROM local_media_repository WHERE user_id = ? ORDER BY {order_by_column} {order}, media_id ASC @@ -235,7 +250,19 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore): args += [limit, start] txn.execute(sql, args) - media = self.db_pool.cursor_to_dict(txn) + media = [ + LocalMedia( + media_id=row[0], + media_type=row[1], + media_length=row[2], + upload_name=row[3], + created_ts=row[4], + last_access_ts=row[5], + quarantined_by=row[6], + safe_from_quarantine=bool(row[7]), + ) + for row in txn + ] return media, count return await self.db_pool.runInteraction( @@ -435,19 +462,28 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore): desc="store_url_cache", ) - async def get_local_media_thumbnails(self, media_id: str) -> List[Dict[str, Any]]: - return await self.db_pool.simple_select_list( - "local_media_repository_thumbnails", - {"media_id": media_id}, - ( - "thumbnail_width", - "thumbnail_height", - "thumbnail_method", - "thumbnail_type", - "thumbnail_length", + async def get_local_media_thumbnails(self, media_id: str) -> List[ThumbnailInfo]: + rows = cast( + List[Tuple[int, int, str, str, int]], + await self.db_pool.simple_select_list( + "local_media_repository_thumbnails", + {"media_id": media_id}, + ( + "thumbnail_width", + "thumbnail_height", + "thumbnail_method", + "thumbnail_type", + "thumbnail_length", + ), + desc="get_local_media_thumbnails", ), - desc="get_local_media_thumbnails", ) + return [ + ThumbnailInfo( + width=row[0], height=row[1], method=row[2], type=row[3], length=row[4] + ) + for row in rows + ] @trace async def store_local_thumbnail( @@ -556,20 +592,28 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore): async def get_remote_media_thumbnails( self, origin: str, media_id: str - ) -> List[Dict[str, Any]]: - return await self.db_pool.simple_select_list( - "remote_media_cache_thumbnails", - {"media_origin": origin, "media_id": media_id}, - ( - "thumbnail_width", - "thumbnail_height", - "thumbnail_method", - "thumbnail_type", - "thumbnail_length", - "filesystem_id", + ) -> List[ThumbnailInfo]: + rows = cast( + List[Tuple[int, int, str, str, int]], + await self.db_pool.simple_select_list( + "remote_media_cache_thumbnails", + {"media_origin": origin, "media_id": media_id}, + ( + "thumbnail_width", + "thumbnail_height", + "thumbnail_method", + "thumbnail_type", + "thumbnail_length", + ), + desc="get_remote_media_thumbnails", ), - desc="get_remote_media_thumbnails", ) + return [ + ThumbnailInfo( + width=row[0], height=row[1], method=row[2], type=row[3], length=row[4] + ) + for row in rows + ] @trace async def get_remote_media_thumbnail( @@ -632,7 +676,7 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore): async def get_remote_media_ids( self, before_ts: int, include_quarantined_media: bool - ) -> List[Dict[str, str]]: + ) -> List[Tuple[str, str, str]]: """ Retrieve a list of server name, media ID tuples from the remote media cache. @@ -646,12 +690,14 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore): A list of tuples containing: * The server name of homeserver where the media originates from, * The ID of the media. + * The filesystem ID. + """ + + sql = """ + SELECT media_origin, media_id, filesystem_id + FROM remote_media_cache + WHERE last_access_ts < ? """ - sql = ( - "SELECT media_origin, media_id, filesystem_id" - " FROM remote_media_cache" - " WHERE last_access_ts < ?" - ) if include_quarantined_media is False: # Only include media that has not been quarantined @@ -659,8 +705,9 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore): AND quarantined_by IS NULL """ - return await self.db_pool.execute( - "get_remote_media_ids", self.db_pool.cursor_to_dict, sql, before_ts + return cast( + List[Tuple[str, str, str]], + await self.db_pool.execute("get_remote_media_ids", sql, before_ts), ) async def delete_remote_media(self, media_origin: str, media_id: str) -> None: diff --git a/synapse/storage/databases/main/presence.py b/synapse/storage/databases/main/presence.py index 194b4e031f..3b444d2d07 100644 --- a/synapse/storage/databases/main/presence.py +++ b/synapse/storage/databases/main/presence.py @@ -20,6 +20,7 @@ from typing import ( Mapping, Optional, Tuple, + Union, cast, ) @@ -260,27 +261,40 @@ class PresenceStore(PresenceBackgroundUpdateStore, CacheInvalidationWorkerStore) async def get_presence_for_users( self, user_ids: Iterable[str] ) -> Mapping[str, UserPresenceState]: - rows = await self.db_pool.simple_select_many_batch( - table="presence_stream", - column="user_id", - iterable=user_ids, - keyvalues={}, - retcols=( - "user_id", - "state", - "last_active_ts", - "last_federation_update_ts", - "last_user_sync_ts", - "status_msg", - "currently_active", + # TODO All these columns are nullable, but we don't expect that: + # https://github.com/matrix-org/synapse/issues/16467 + rows = cast( + List[Tuple[str, str, int, int, int, Optional[str], Union[int, bool]]], + await self.db_pool.simple_select_many_batch( + table="presence_stream", + column="user_id", + iterable=user_ids, + keyvalues={}, + retcols=( + "user_id", + "state", + "last_active_ts", + "last_federation_update_ts", + "last_user_sync_ts", + "status_msg", + "currently_active", + ), + desc="get_presence_for_users", ), - desc="get_presence_for_users", ) - for row in rows: - row["currently_active"] = bool(row["currently_active"]) - - return {row["user_id"]: UserPresenceState(**row) for row in rows} + return { + user_id: UserPresenceState( + user_id=user_id, + state=state, + last_active_ts=last_active_ts, + last_federation_update_ts=last_federation_update_ts, + last_user_sync_ts=last_user_sync_ts, + status_msg=status_msg, + currently_active=bool(currently_active), + ) + for user_id, state, last_active_ts, last_federation_update_ts, last_user_sync_ts, status_msg, currently_active in rows + } async def should_user_receive_full_presence_with_token( self, @@ -385,28 +399,49 @@ class PresenceStore(PresenceBackgroundUpdateStore, CacheInvalidationWorkerStore) limit = 100 offset = 0 while True: - rows = await self.db_pool.runInteraction( - "get_presence_for_all_users", - self.db_pool.simple_select_list_paginate_txn, - "presence_stream", - orderby="stream_id", - start=offset, - limit=limit, - exclude_keyvalues=exclude_keyvalues, - retcols=( - "user_id", - "state", - "last_active_ts", - "last_federation_update_ts", - "last_user_sync_ts", - "status_msg", - "currently_active", + # TODO All these columns are nullable, but we don't expect that: + # https://github.com/matrix-org/synapse/issues/16467 + rows = cast( + List[Tuple[str, str, int, int, int, Optional[str], Union[int, bool]]], + await self.db_pool.runInteraction( + "get_presence_for_all_users", + self.db_pool.simple_select_list_paginate_txn, + "presence_stream", + orderby="stream_id", + start=offset, + limit=limit, + exclude_keyvalues=exclude_keyvalues, + retcols=( + "user_id", + "state", + "last_active_ts", + "last_federation_update_ts", + "last_user_sync_ts", + "status_msg", + "currently_active", + ), + order_direction="ASC", ), - order_direction="ASC", ) - for row in rows: - users_to_state[row["user_id"]] = UserPresenceState(**row) + for ( + user_id, + state, + last_active_ts, + last_federation_update_ts, + last_user_sync_ts, + status_msg, + currently_active, + ) in rows: + users_to_state[user_id] = UserPresenceState( + user_id=user_id, + state=state, + last_active_ts=last_active_ts, + last_federation_update_ts=last_federation_update_ts, + last_user_sync_ts=last_user_sync_ts, + status_msg=status_msg, + currently_active=bool(currently_active), + ) # We've run out of updates to query if len(rows) < limit: @@ -434,13 +469,21 @@ class PresenceStore(PresenceBackgroundUpdateStore, CacheInvalidationWorkerStore) txn = db_conn.cursor() txn.execute(sql, (PresenceState.OFFLINE,)) - rows = self.db_pool.cursor_to_dict(txn) + rows = txn.fetchall() txn.close() - for row in rows: - row["currently_active"] = bool(row["currently_active"]) - - return [UserPresenceState(**row) for row in rows] + return [ + UserPresenceState( + user_id=user_id, + state=state, + last_active_ts=last_active_ts, + last_federation_update_ts=last_federation_update_ts, + last_user_sync_ts=last_user_sync_ts, + status_msg=status_msg, + currently_active=bool(currently_active), + ) + for user_id, state, last_active_ts, last_federation_update_ts, last_user_sync_ts, status_msg, currently_active in rows + ] def take_presence_startup_info(self) -> List[UserPresenceState]: active_on_startup = self._presence_on_startup diff --git a/synapse/storage/databases/main/purge_events.py b/synapse/storage/databases/main/purge_events.py index dea0e0458c..1e11bf2706 100644 --- a/synapse/storage/databases/main/purge_events.py +++ b/synapse/storage/databases/main/purge_events.py @@ -89,6 +89,11 @@ class PurgeEventsStore(StateGroupWorkerStore, CacheInvalidationWorkerStore): # furthermore, we might already have the table from a previous (failed) # purge attempt, so let's drop the table first. + if isinstance(self.database_engine, PostgresEngine): + # Disable statement timeouts for this transaction; purging rooms can + # take a while! + txn.execute("SET LOCAL statement_timeout = 0") + txn.execute("DROP TABLE IF EXISTS events_to_purge") txn.execute( diff --git a/synapse/storage/databases/main/push_rule.py b/synapse/storage/databases/main/push_rule.py index 923166974c..37135d431d 100644 --- a/synapse/storage/databases/main/push_rule.py +++ b/synapse/storage/databases/main/push_rule.py @@ -28,8 +28,11 @@ from typing import ( cast, ) +from twisted.internet import defer + from synapse.api.errors import StoreError from synapse.config.homeserver import ExperimentalConfig +from synapse.logging.context import make_deferred_yieldable, run_in_background from synapse.replication.tcp.streams import PushRulesStream from synapse.storage._base import SQLBaseStore from synapse.storage.database import ( @@ -51,7 +54,8 @@ from synapse.storage.util.id_generators import ( ) from synapse.synapse_rust.push import FilteredPushRules, PushRule, PushRules from synapse.types import JsonDict -from synapse.util import json_encoder +from synapse.util import json_encoder, unwrapFirstError +from synapse.util.async_helpers import gather_results from synapse.util.caches.descriptors import cached, cachedList from synapse.util.caches.stream_change_cache import StreamChangeCache @@ -62,20 +66,34 @@ logger = logging.getLogger(__name__) def _load_rules( - rawrules: List[JsonDict], + rawrules: List[Tuple[str, int, str, str]], enabled_map: Dict[str, bool], experimental_config: ExperimentalConfig, ) -> FilteredPushRules: """Take the DB rows returned from the DB and convert them into a full `FilteredPushRules` object. + + Args: + rawrules: List of tuples of: + * rule ID + * Priority lass + * Conditions (as serialized JSON) + * Actions (as serialized JSON) + enabled_map: A dictionary of rule ID to a boolean of whether the rule is + enabled. This might not include all rule IDs from rawrules. + experimental_config: The `experimental_features` section of the Synapse + config. (Used to check if various features are enabled.) + + Returns: + A new FilteredPushRules object. """ ruleslist = [ PushRule.from_db( - rule_id=rawrule["rule_id"], - priority_class=rawrule["priority_class"], - conditions=rawrule["conditions"], - actions=rawrule["actions"], + rule_id=rawrule[0], + priority_class=rawrule[1], + conditions=rawrule[2], + actions=rawrule[3], ) for rawrule in rawrules ] @@ -165,34 +183,44 @@ class PushRulesWorkerStore( @cached(max_entries=5000) async def get_push_rules_for_user(self, user_id: str) -> FilteredPushRules: - rows = await self.db_pool.simple_select_list( - table="push_rules", - keyvalues={"user_name": user_id}, - retcols=( - "user_name", - "rule_id", - "priority_class", - "priority", - "conditions", - "actions", + rows = cast( + List[Tuple[str, int, int, str, str]], + await self.db_pool.simple_select_list( + table="push_rules", + keyvalues={"user_name": user_id}, + retcols=( + "rule_id", + "priority_class", + "priority", + "conditions", + "actions", + ), + desc="get_push_rules_for_user", ), - desc="get_push_rules_for_user", ) - rows.sort(key=lambda row: (-int(row["priority_class"]), -int(row["priority"]))) + # Sort by highest priority_class, then highest priority. + rows.sort(key=lambda row: (-int(row[1]), -int(row[2]))) enabled_map = await self.get_push_rules_enabled_for_user(user_id) - return _load_rules(rows, enabled_map, self.hs.config.experimental) + return _load_rules( + [(row[0], row[1], row[3], row[4]) for row in rows], + enabled_map, + self.hs.config.experimental, + ) async def get_push_rules_enabled_for_user(self, user_id: str) -> Dict[str, bool]: - results = await self.db_pool.simple_select_list( - table="push_rules_enable", - keyvalues={"user_name": user_id}, - retcols=("rule_id", "enabled"), - desc="get_push_rules_enabled_for_user", + results = cast( + List[Tuple[str, Optional[Union[int, bool]]]], + await self.db_pool.simple_select_list( + table="push_rules_enable", + keyvalues={"user_name": user_id}, + retcols=("rule_id", "enabled"), + desc="get_push_rules_enabled_for_user", + ), ) - return {r["rule_id"]: bool(r["enabled"]) for r in results} + return {r[0]: bool(r[1]) for r in results} async def have_push_rules_changed_for_user( self, user_id: str, last_id: int @@ -221,23 +249,46 @@ class PushRulesWorkerStore( if not user_ids: return {} - raw_rules: Dict[str, List[JsonDict]] = {user_id: [] for user_id in user_ids} + raw_rules: Dict[str, List[Tuple[str, int, str, str]]] = { + user_id: [] for user_id in user_ids + } - rows = await self.db_pool.simple_select_many_batch( - table="push_rules", - column="user_name", - iterable=user_ids, - retcols=("*",), - desc="bulk_get_push_rules", - batch_size=1000, + # gatherResults loses all type information. + rows, enabled_map_by_user = await make_deferred_yieldable( + gather_results( + ( + cast( + "defer.Deferred[List[Tuple[str, str, int, int, str, str]]]", + run_in_background( + self.db_pool.simple_select_many_batch, + table="push_rules", + column="user_name", + iterable=user_ids, + retcols=( + "user_name", + "rule_id", + "priority_class", + "priority", + "conditions", + "actions", + ), + desc="bulk_get_push_rules", + batch_size=1000, + ), + ), + run_in_background(self.bulk_get_push_rules_enabled, user_ids), + ), + consumeErrors=True, + ).addErrback(unwrapFirstError) ) - rows.sort(key=lambda row: (-int(row["priority_class"]), -int(row["priority"]))) - - for row in rows: - raw_rules.setdefault(row["user_name"], []).append(row) + # Sort by highest priority_class, then highest priority. + rows.sort(key=lambda row: (-int(row[2]), -int(row[3]))) - enabled_map_by_user = await self.bulk_get_push_rules_enabled(user_ids) + for user_name, rule_id, priority_class, _, conditions, actions in rows: + raw_rules.setdefault(user_name, []).append( + (rule_id, priority_class, conditions, actions) + ) results: Dict[str, FilteredPushRules] = {} @@ -256,17 +307,19 @@ class PushRulesWorkerStore( results: Dict[str, Dict[str, bool]] = {user_id: {} for user_id in user_ids} - rows = await self.db_pool.simple_select_many_batch( - table="push_rules_enable", - column="user_name", - iterable=user_ids, - retcols=("user_name", "rule_id", "enabled"), - desc="bulk_get_push_rules_enabled", - batch_size=1000, + rows = cast( + List[Tuple[str, str, Optional[int]]], + await self.db_pool.simple_select_many_batch( + table="push_rules_enable", + column="user_name", + iterable=user_ids, + retcols=("user_name", "rule_id", "enabled"), + desc="bulk_get_push_rules_enabled", + batch_size=1000, + ), ) - for row in rows: - enabled = bool(row["enabled"]) - results.setdefault(row["user_name"], {})[row["rule_id"]] = enabled + for user_name, rule_id, enabled in rows: + results.setdefault(user_name, {})[rule_id] = bool(enabled) return results async def get_all_push_rule_updates( diff --git a/synapse/storage/databases/main/pusher.py b/synapse/storage/databases/main/pusher.py index 87e28e22d3..a6a1671bd6 100644 --- a/synapse/storage/databases/main/pusher.py +++ b/synapse/storage/databases/main/pusher.py @@ -47,6 +47,27 @@ if TYPE_CHECKING: logger = logging.getLogger(__name__) +# The type of a row in the pushers table. +PusherRow = Tuple[ + int, # id + str, # user_name + Optional[int], # access_token + str, # profile_tag + str, # kind + str, # app_id + str, # app_display_name + str, # device_display_name + str, # pushkey + int, # ts + str, # lang + str, # data + int, # last_stream_ordering + int, # last_success + int, # failing_since + bool, # enabled + str, # device_id +] + class PusherWorkerStore(SQLBaseStore): def __init__( @@ -83,30 +104,66 @@ class PusherWorkerStore(SQLBaseStore): self._remove_deleted_email_pushers, ) - def _decode_pushers_rows(self, rows: Iterable[dict]) -> Iterator[PusherConfig]: + def _decode_pushers_rows( + self, + rows: Iterable[PusherRow], + ) -> Iterator[PusherConfig]: """JSON-decode the data in the rows returned from the `pushers` table Drops any rows whose data cannot be decoded """ - for r in rows: - data_json = r["data"] + for ( + id, + user_name, + access_token, + profile_tag, + kind, + app_id, + app_display_name, + device_display_name, + pushkey, + ts, + lang, + data, + last_stream_ordering, + last_success, + failing_since, + enabled, + device_id, + ) in rows: try: - r["data"] = db_to_json(data_json) + data_json = db_to_json(data) except Exception as e: logger.warning( "Invalid JSON in data for pusher %d: %s, %s", - r["id"], - data_json, + id, + data, e.args[0], ) continue - # If we're using SQLite, then boolean values are integers. This is - # troublesome since some code using the return value of this method might - # expect it to be a boolean, or will expose it to clients (in responses). - r["enabled"] = bool(r["enabled"]) - - yield PusherConfig(**r) + yield PusherConfig( + id=id, + user_name=user_name, + profile_tag=profile_tag, + kind=kind, + app_id=app_id, + app_display_name=app_display_name, + device_display_name=device_display_name, + pushkey=pushkey, + ts=ts, + lang=lang, + data=data_json, + last_stream_ordering=last_stream_ordering, + last_success=last_success, + failing_since=failing_since, + # If we're using SQLite, then boolean values are integers. This is + # troublesome since some code using the return value of this method might + # expect it to be a boolean, or will expose it to clients (in responses). + enabled=bool(enabled), + device_id=device_id, + access_token=access_token, + ) def get_pushers_stream_token(self) -> int: return self._pushers_id_gen.get_current_token() @@ -136,7 +193,7 @@ class PusherWorkerStore(SQLBaseStore): The pushers for which the given columns have the given values. """ - def get_pushers_by_txn(txn: LoggingTransaction) -> List[Dict[str, Any]]: + def get_pushers_by_txn(txn: LoggingTransaction) -> List[PusherRow]: # We could technically use simple_select_list here, but we need to call # COALESCE on the 'enabled' column. While it is technically possible to give # simple_select_list the whole `COALESCE(...) AS ...` as a column name, it @@ -154,7 +211,7 @@ class PusherWorkerStore(SQLBaseStore): txn.execute(sql, list(keyvalues.values())) - return self.db_pool.cursor_to_dict(txn) + return cast(List[PusherRow], txn.fetchall()) ret = await self.db_pool.runInteraction( desc="get_pushers_by", @@ -164,14 +221,22 @@ class PusherWorkerStore(SQLBaseStore): return self._decode_pushers_rows(ret) async def get_enabled_pushers(self) -> Iterator[PusherConfig]: - def get_enabled_pushers_txn(txn: LoggingTransaction) -> Iterator[PusherConfig]: - txn.execute("SELECT * FROM pushers WHERE COALESCE(enabled, TRUE)") - rows = self.db_pool.cursor_to_dict(txn) - - return self._decode_pushers_rows(rows) + def get_enabled_pushers_txn(txn: LoggingTransaction) -> List[PusherRow]: + txn.execute( + """ + SELECT id, user_name, access_token, profile_tag, kind, app_id, + app_display_name, device_display_name, pushkey, ts, lang, data, + last_stream_ordering, last_success, failing_since, + enabled, device_id + FROM pushers WHERE COALESCE(enabled, TRUE) + """ + ) + return cast(List[PusherRow], txn.fetchall()) - return await self.db_pool.runInteraction( - "get_enabled_pushers", get_enabled_pushers_txn + return self._decode_pushers_rows( + await self.db_pool.runInteraction( + "get_enabled_pushers", get_enabled_pushers_txn + ) ) async def get_all_updated_pushers_rows( @@ -304,26 +369,28 @@ class PusherWorkerStore(SQLBaseStore): ) async def get_throttle_params_by_room( - self, pusher_id: str + self, pusher_id: int ) -> Dict[str, ThrottleParams]: - res = await self.db_pool.simple_select_list( - "pusher_throttle", - {"pusher": pusher_id}, - ["room_id", "last_sent_ts", "throttle_ms"], - desc="get_throttle_params_by_room", + res = cast( + List[Tuple[str, Optional[int], Optional[int]]], + await self.db_pool.simple_select_list( + "pusher_throttle", + {"pusher": pusher_id}, + ["room_id", "last_sent_ts", "throttle_ms"], + desc="get_throttle_params_by_room", + ), ) params_by_room = {} - for row in res: - params_by_room[row["room_id"]] = ThrottleParams( - row["last_sent_ts"], - row["throttle_ms"], + for room_id, last_sent_ts, throttle_ms in res: + params_by_room[room_id] = ThrottleParams( + last_sent_ts or 0, throttle_ms or 0 ) return params_by_room async def set_throttle_params( - self, pusher_id: str, room_id: str, params: ThrottleParams + self, pusher_id: int, room_id: str, params: ThrottleParams ) -> None: await self.db_pool.simple_upsert( "pusher_throttle", @@ -534,7 +601,7 @@ class PusherBackgroundUpdatesStore(SQLBaseStore): (last_pusher_id, batch_size), ) - rows = self.db_pool.cursor_to_dict(txn) + rows = txn.fetchall() if len(rows) == 0: return 0 @@ -550,19 +617,19 @@ class PusherBackgroundUpdatesStore(SQLBaseStore): txn=txn, table="pushers", key_names=("id",), - key_values=[(row["pusher_id"],) for row in rows], + key_values=[row[0] for row in rows], value_names=("device_id", "access_token"), # If there was already a device_id on the pusher, we only want to clear # the access_token column, so we keep the existing device_id. Otherwise, # we set the device_id we got from joining the access_tokens table. value_values=[ - (row["pusher_device_id"] or row["token_device_id"], None) - for row in rows + (pusher_device_id or token_device_id, None) + for _, pusher_device_id, token_device_id in rows ], ) self.db_pool.updates._background_update_progress_txn( - txn, "set_device_id_for_pushers", {"pusher_id": rows[-1]["pusher_id"]} + txn, "set_device_id_for_pushers", {"pusher_id": rows[-1][0]} ) return len(rows) diff --git a/synapse/storage/databases/main/receipts.py b/synapse/storage/databases/main/receipts.py index 0231f9407b..56e8eb16a8 100644 --- a/synapse/storage/databases/main/receipts.py +++ b/synapse/storage/databases/main/receipts.py @@ -28,6 +28,8 @@ from typing import ( cast, ) +from immutabledict import immutabledict + from synapse.api.constants import EduTypes from synapse.replication.tcp.streams import ReceiptsStream from synapse.storage._base import SQLBaseStore, db_to_json, make_in_list_sql_clause @@ -43,7 +45,12 @@ from synapse.storage.util.id_generators import ( MultiWriterIdGenerator, StreamIdGenerator, ) -from synapse.types import JsonDict, JsonMapping +from synapse.types import ( + JsonDict, + JsonMapping, + MultiWriterStreamToken, + PersistedPosition, +) from synapse.util import json_encoder from synapse.util.caches.descriptors import cached, cachedList from synapse.util.caches.stream_change_cache import StreamChangeCache @@ -105,7 +112,7 @@ class ReceiptsWorkerStore(SQLBaseStore): "receipts_linearized", entity_column="room_id", stream_column="stream_id", - max_value=max_receipts_stream_id, + max_value=max_receipts_stream_id.stream, limit=10000, ) self._receipts_stream_cache = StreamChangeCache( @@ -114,9 +121,31 @@ class ReceiptsWorkerStore(SQLBaseStore): prefilled_cache=receipts_stream_prefill, ) - def get_max_receipt_stream_id(self) -> int: + def get_max_receipt_stream_id(self) -> MultiWriterStreamToken: """Get the current max stream ID for receipts stream""" - return self._receipts_id_gen.get_current_token() + + min_pos = self._receipts_id_gen.get_current_token() + + positions = {} + if isinstance(self._receipts_id_gen, MultiWriterIdGenerator): + # The `min_pos` is the minimum position that we know all instances + # have finished persisting to, so we only care about instances whose + # positions are ahead of that. (Instance positions can be behind the + # min position as there are times we can work out that the minimum + # position is ahead of the naive minimum across all current + # positions. See MultiWriterIdGenerator for details) + positions = { + i: p + for i, p in self._receipts_id_gen.get_positions().items() + if p > min_pos + } + + return MultiWriterStreamToken( + stream=min_pos, instance_map=immutabledict(positions) + ) + + def get_receipt_stream_id_for_instance(self, instance_name: str) -> int: + return self._receipts_id_gen.get_current_token_for_writer(instance_name) def get_last_unthreaded_receipt_for_user_txn( self, @@ -257,7 +286,10 @@ class ReceiptsWorkerStore(SQLBaseStore): } async def get_linearized_receipts_for_rooms( - self, room_ids: Iterable[str], to_key: int, from_key: Optional[int] = None + self, + room_ids: Iterable[str], + to_key: MultiWriterStreamToken, + from_key: Optional[MultiWriterStreamToken] = None, ) -> List[JsonMapping]: """Get receipts for multiple rooms for sending to clients. @@ -276,7 +308,7 @@ class ReceiptsWorkerStore(SQLBaseStore): # Only ask the database about rooms where there have been new # receipts added since `from_key` room_ids = self._receipts_stream_cache.get_entities_changed( - room_ids, from_key + room_ids, from_key.stream ) results = await self._get_linearized_receipts_for_rooms( @@ -286,7 +318,10 @@ class ReceiptsWorkerStore(SQLBaseStore): return [ev for res in results.values() for ev in res] async def get_linearized_receipts_for_room( - self, room_id: str, to_key: int, from_key: Optional[int] = None + self, + room_id: str, + to_key: MultiWriterStreamToken, + from_key: Optional[MultiWriterStreamToken] = None, ) -> Sequence[JsonMapping]: """Get receipts for a single room for sending to clients. @@ -302,36 +337,49 @@ class ReceiptsWorkerStore(SQLBaseStore): if from_key is not None: # Check the cache first to see if any new receipts have been added # since`from_key`. If not we can no-op. - if not self._receipts_stream_cache.has_entity_changed(room_id, from_key): + if not self._receipts_stream_cache.has_entity_changed( + room_id, from_key.stream + ): return [] return await self._get_linearized_receipts_for_room(room_id, to_key, from_key) @cached(tree=True) async def _get_linearized_receipts_for_room( - self, room_id: str, to_key: int, from_key: Optional[int] = None + self, + room_id: str, + to_key: MultiWriterStreamToken, + from_key: Optional[MultiWriterStreamToken] = None, ) -> Sequence[JsonMapping]: """See get_linearized_receipts_for_room""" - def f(txn: LoggingTransaction) -> List[Dict[str, Any]]: + def f(txn: LoggingTransaction) -> List[Tuple[str, str, str, str]]: if from_key: - sql = ( - "SELECT * FROM receipts_linearized WHERE" - " room_id = ? AND stream_id > ? AND stream_id <= ?" - ) + sql = """ + SELECT stream_id, instance_name, receipt_type, user_id, event_id, data + FROM receipts_linearized + WHERE room_id = ? AND stream_id > ? AND stream_id <= ? + """ - txn.execute(sql, (room_id, from_key, to_key)) - else: - sql = ( - "SELECT * FROM receipts_linearized WHERE" - " room_id = ? AND stream_id <= ?" + txn.execute( + sql, (room_id, from_key.stream, to_key.get_max_stream_pos()) ) + else: + sql = """ + SELECT stream_id, instance_name, receipt_type, user_id, event_id, data + FROM receipts_linearized WHERE + room_id = ? AND stream_id <= ? + """ - txn.execute(sql, (room_id, to_key)) - - rows = self.db_pool.cursor_to_dict(txn) + txn.execute(sql, (room_id, to_key.get_max_stream_pos())) - return rows + return [ + (receipt_type, user_id, event_id, data) + for stream_id, instance_name, receipt_type, user_id, event_id, data in txn + if MultiWriterStreamToken.is_stream_position_in_range( + from_key, to_key, instance_name, stream_id + ) + ] rows = await self.db_pool.runInteraction("get_linearized_receipts_for_room", f) @@ -339,10 +387,10 @@ class ReceiptsWorkerStore(SQLBaseStore): return [] content: JsonDict = {} - for row in rows: - content.setdefault(row["event_id"], {}).setdefault(row["receipt_type"], {})[ - row["user_id"] - ] = db_to_json(row["data"]) + for receipt_type, user_id, event_id, data in rows: + content.setdefault(event_id, {}).setdefault(receipt_type, {})[ + user_id + ] = db_to_json(data) return [{"type": EduTypes.RECEIPT, "room_id": room_id, "content": content}] @@ -352,25 +400,37 @@ class ReceiptsWorkerStore(SQLBaseStore): num_args=3, ) async def _get_linearized_receipts_for_rooms( - self, room_ids: Collection[str], to_key: int, from_key: Optional[int] = None + self, + room_ids: Collection[str], + to_key: MultiWriterStreamToken, + from_key: Optional[MultiWriterStreamToken] = None, ) -> Mapping[str, Sequence[JsonMapping]]: if not room_ids: return {} - def f(txn: LoggingTransaction) -> List[Dict[str, Any]]: + def f( + txn: LoggingTransaction, + ) -> List[Tuple[str, str, str, str, Optional[str], str]]: if from_key: sql = """ - SELECT * FROM receipts_linearized WHERE + SELECT stream_id, instance_name, room_id, receipt_type, + user_id, event_id, thread_id, data + FROM receipts_linearized WHERE stream_id > ? AND stream_id <= ? AND """ clause, args = make_in_list_sql_clause( self.database_engine, "room_id", room_ids ) - txn.execute(sql + clause, [from_key, to_key] + list(args)) + txn.execute( + sql + clause, + [from_key.stream, to_key.get_max_stream_pos()] + list(args), + ) else: sql = """ - SELECT * FROM receipts_linearized WHERE + SELECT stream_id, instance_name, room_id, receipt_type, + user_id, event_id, thread_id, data + FROM receipts_linearized WHERE stream_id <= ? AND """ @@ -378,31 +438,37 @@ class ReceiptsWorkerStore(SQLBaseStore): self.database_engine, "room_id", room_ids ) - txn.execute(sql + clause, [to_key] + list(args)) + txn.execute(sql + clause, [to_key.get_max_stream_pos()] + list(args)) - return self.db_pool.cursor_to_dict(txn) + return [ + (room_id, receipt_type, user_id, event_id, thread_id, data) + for stream_id, instance_name, room_id, receipt_type, user_id, event_id, thread_id, data in txn + if MultiWriterStreamToken.is_stream_position_in_range( + from_key, to_key, instance_name, stream_id + ) + ] txn_results = await self.db_pool.runInteraction( "_get_linearized_receipts_for_rooms", f ) results: JsonDict = {} - for row in txn_results: + for room_id, receipt_type, user_id, event_id, thread_id, data in txn_results: # We want a single event per room, since we want to batch the # receipts by room, event and type. room_event = results.setdefault( - row["room_id"], - {"type": EduTypes.RECEIPT, "room_id": row["room_id"], "content": {}}, + room_id, + {"type": EduTypes.RECEIPT, "room_id": room_id, "content": {}}, ) # The content is of the form: # {"$foo:bar": { "read": { "@user:host": <receipt> }, .. }, .. } - event_entry = room_event["content"].setdefault(row["event_id"], {}) - receipt_type = event_entry.setdefault(row["receipt_type"], {}) + event_entry = room_event["content"].setdefault(event_id, {}) + receipt_type_dict = event_entry.setdefault(receipt_type, {}) - receipt_type[row["user_id"]] = db_to_json(row["data"]) - if row["thread_id"]: - receipt_type[row["user_id"]]["thread_id"] = row["thread_id"] + receipt_type_dict[user_id] = db_to_json(data) + if thread_id: + receipt_type_dict[user_id]["thread_id"] = thread_id results = { room_id: [results[room_id]] if room_id in results else [] @@ -414,7 +480,9 @@ class ReceiptsWorkerStore(SQLBaseStore): num_args=2, ) async def get_linearized_receipts_for_all_rooms( - self, to_key: int, from_key: Optional[int] = None + self, + to_key: MultiWriterStreamToken, + from_key: Optional[MultiWriterStreamToken] = None, ) -> Mapping[str, JsonMapping]: """Get receipts for all rooms between two stream_ids, up to a limit of the latest 100 read receipts. @@ -428,46 +496,54 @@ class ReceiptsWorkerStore(SQLBaseStore): A dictionary of roomids to a list of receipts. """ - def f(txn: LoggingTransaction) -> List[Dict[str, Any]]: + def f(txn: LoggingTransaction) -> List[Tuple[str, str, str, str, str]]: if from_key: sql = """ - SELECT * FROM receipts_linearized WHERE + SELECT stream_id, instance_name, room_id, receipt_type, user_id, event_id, data + FROM receipts_linearized WHERE stream_id > ? AND stream_id <= ? ORDER BY stream_id DESC LIMIT 100 """ - txn.execute(sql, [from_key, to_key]) + txn.execute(sql, [from_key.stream, to_key.get_max_stream_pos()]) else: sql = """ - SELECT * FROM receipts_linearized WHERE + SELECT stream_id, instance_name, room_id, receipt_type, user_id, event_id, data + FROM receipts_linearized WHERE stream_id <= ? ORDER BY stream_id DESC LIMIT 100 """ - txn.execute(sql, [to_key]) + txn.execute(sql, [to_key.get_max_stream_pos()]) - return self.db_pool.cursor_to_dict(txn) + return [ + (room_id, receipt_type, user_id, event_id, data) + for stream_id, instance_name, room_id, receipt_type, user_id, event_id, data in txn + if MultiWriterStreamToken.is_stream_position_in_range( + from_key, to_key, instance_name, stream_id + ) + ] txn_results = await self.db_pool.runInteraction( "get_linearized_receipts_for_all_rooms", f ) results: JsonDict = {} - for row in txn_results: + for room_id, receipt_type, user_id, event_id, data in txn_results: # We want a single event per room, since we want to batch the # receipts by room, event and type. room_event = results.setdefault( - row["room_id"], - {"type": EduTypes.RECEIPT, "room_id": row["room_id"], "content": {}}, + room_id, + {"type": EduTypes.RECEIPT, "room_id": room_id, "content": {}}, ) # The content is of the form: # {"$foo:bar": { "read": { "@user:host": <receipt> }, .. }, .. } - event_entry = room_event["content"].setdefault(row["event_id"], {}) - receipt_type = event_entry.setdefault(row["receipt_type"], {}) + event_entry = room_event["content"].setdefault(event_id, {}) + receipt_type_dict = event_entry.setdefault(receipt_type, {}) - receipt_type[row["user_id"]] = db_to_json(row["data"]) + receipt_type_dict[user_id] = db_to_json(data) return results @@ -537,10 +613,11 @@ class ReceiptsWorkerStore(SQLBaseStore): SELECT stream_id, room_id, receipt_type, user_id, event_id, thread_id, data FROM receipts_linearized WHERE ? < stream_id AND stream_id <= ? + AND instance_name = ? ORDER BY stream_id ASC LIMIT ? """ - txn.execute(sql, (last_id, current_id, limit)) + txn.execute(sql, (last_id, current_id, instance_name, limit)) updates = cast( List[Tuple[int, Tuple[str, str, str, str, Optional[str], JsonDict]]], @@ -687,6 +764,7 @@ class ReceiptsWorkerStore(SQLBaseStore): keyvalues=keyvalues, values={ "stream_id": stream_id, + "instance_name": self._instance_name, "event_id": event_id, "event_stream_ordering": stream_ordering, "data": json_encoder.encode(data), @@ -742,7 +820,7 @@ class ReceiptsWorkerStore(SQLBaseStore): event_ids: List[str], thread_id: Optional[str], data: dict, - ) -> Optional[Tuple[int, int]]: + ) -> Optional[PersistedPosition]: """Insert a receipt, either from local client or remote server. Automatically does conversion between linearized and graph @@ -804,9 +882,7 @@ class ReceiptsWorkerStore(SQLBaseStore): data, ) - max_persisted_id = self._receipts_id_gen.get_current_token() - - return stream_id, max_persisted_id + return PersistedPosition(self._instance_name, stream_id) async def _insert_graph_receipt( self, diff --git a/synapse/storage/databases/main/registration.py b/synapse/storage/databases/main/registration.py index cc964604e2..933d76e905 100644 --- a/synapse/storage/databases/main/registration.py +++ b/synapse/storage/databases/main/registration.py @@ -143,6 +143,30 @@ class LoginTokenLookupResult: """The session ID advertised by the SSO Identity Provider.""" +@attr.s(frozen=True, slots=True, auto_attribs=True) +class ThreepidResult: + medium: str + address: str + validated_at: int + added_at: int + + +@attr.s(frozen=True, slots=True, auto_attribs=True) +class ThreepidValidationSession: + address: str + """address of the 3pid""" + medium: str + """medium of the 3pid""" + client_secret: str + """a secret provided by the client for this validation session""" + session_id: str + """ID of the validation session""" + last_send_attempt: int + """a number serving to dedupe send attempts for this session""" + validated_at: Optional[int] + """timestamp of when this session was validated if so""" + + class RegistrationWorkerStore(CacheInvalidationWorkerStore): def __init__( self, @@ -195,7 +219,7 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore): async def get_user_by_id(self, user_id: str) -> Optional[UserInfo]: """Returns info about the user account, if it exists.""" - def get_user_by_id_txn(txn: LoggingTransaction) -> Optional[Dict[str, Any]]: + def get_user_by_id_txn(txn: LoggingTransaction) -> Optional[UserInfo]: # We could technically use simple_select_one here, but it would not perform # the COALESCEs (unless hacked into the column names), which could yield # confusing results. @@ -213,35 +237,46 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore): (user_id,), ) - rows = self.db_pool.cursor_to_dict(txn) - - if len(rows) == 0: + row = txn.fetchone() + if not row: return None - return rows[0] + ( + name, + is_guest, + admin, + consent_version, + consent_ts, + consent_server_notice_sent, + appservice_id, + creation_ts, + user_type, + deactivated, + shadow_banned, + approved, + locked, + ) = row + + return UserInfo( + appservice_id=appservice_id, + consent_server_notice_sent=consent_server_notice_sent, + consent_version=consent_version, + consent_ts=consent_ts, + creation_ts=creation_ts, + is_admin=bool(admin), + is_deactivated=bool(deactivated), + is_guest=bool(is_guest), + is_shadow_banned=bool(shadow_banned), + user_id=UserID.from_string(name), + user_type=user_type, + approved=bool(approved), + locked=bool(locked), + ) - row = await self.db_pool.runInteraction( + return await self.db_pool.runInteraction( desc="get_user_by_id", func=get_user_by_id_txn, ) - if row is None: - return None - - return UserInfo( - appservice_id=row["appservice_id"], - consent_server_notice_sent=row["consent_server_notice_sent"], - consent_version=row["consent_version"], - consent_ts=row["consent_ts"], - creation_ts=row["creation_ts"], - is_admin=bool(row["admin"]), - is_deactivated=bool(row["deactivated"]), - is_guest=bool(row["is_guest"]), - is_shadow_banned=bool(row["shadow_banned"]), - user_id=UserID.from_string(row["name"]), - user_type=row["user_type"], - approved=bool(row["approved"]), - locked=bool(row["locked"]), - ) async def is_trial_user(self, user_id: str) -> bool: """Checks if user is in the "trial" period, i.e. within the first @@ -579,16 +614,31 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore): """ txn.execute(sql, (token,)) - rows = self.db_pool.cursor_to_dict(txn) - - if rows: - row = rows[0] - - # This field is nullable, ensure it comes out as a boolean - if row["token_used"] is None: - row["token_used"] = False + row = txn.fetchone() - return TokenLookupResult(**row) + if row: + ( + user_id, + is_guest, + shadow_banned, + token_id, + device_id, + valid_until_ms, + token_owner, + token_used, + ) = row + + return TokenLookupResult( + user_id=user_id, + is_guest=is_guest, + shadow_banned=shadow_banned, + token_id=token_id, + device_id=device_id, + valid_until_ms=valid_until_ms, + token_owner=token_owner, + # This field is nullable, ensure it comes out as a boolean + token_used=bool(token_used), + ) return None @@ -821,23 +871,24 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore): Returns: Tuples of (auth_provider, external_id) """ - res = await self.db_pool.simple_select_list( - table="user_external_ids", - keyvalues={"user_id": mxid}, - retcols=("auth_provider", "external_id"), - desc="get_external_ids_by_user", + return cast( + List[Tuple[str, str]], + await self.db_pool.simple_select_list( + table="user_external_ids", + keyvalues={"user_id": mxid}, + retcols=("auth_provider", "external_id"), + desc="get_external_ids_by_user", + ), ) - return [(r["auth_provider"], r["external_id"]) for r in res] async def count_all_users(self) -> int: """Counts all users registered on the homeserver.""" def _count_users(txn: LoggingTransaction) -> int: - txn.execute("SELECT COUNT(*) AS users FROM users") - rows = self.db_pool.cursor_to_dict(txn) - if rows: - return rows[0]["users"] - return 0 + txn.execute("SELECT COUNT(*) FROM users") + row = txn.fetchone() + assert row is not None + return row[0] return await self.db_pool.runInteraction("count_users", _count_users) @@ -891,11 +942,10 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore): """Counts all users without a special user_type registered on the homeserver.""" def _count_users(txn: LoggingTransaction) -> int: - txn.execute("SELECT COUNT(*) AS users FROM users where user_type is null") - rows = self.db_pool.cursor_to_dict(txn) - if rows: - return rows[0]["users"] - return 0 + txn.execute("SELECT COUNT(*) FROM users where user_type is null") + row = txn.fetchone() + assert row is not None + return row[0] return await self.db_pool.runInteraction("count_real_users", _count_users) @@ -964,13 +1014,25 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore): {"user_id": user_id, "validated_at": validated_at, "added_at": added_at}, ) - async def user_get_threepids(self, user_id: str) -> List[Dict[str, Any]]: - return await self.db_pool.simple_select_list( - "user_threepids", - {"user_id": user_id}, - ["medium", "address", "validated_at", "added_at"], - "user_get_threepids", + async def user_get_threepids(self, user_id: str) -> List[ThreepidResult]: + results = cast( + List[Tuple[str, str, int, int]], + await self.db_pool.simple_select_list( + "user_threepids", + keyvalues={"user_id": user_id}, + retcols=["medium", "address", "validated_at", "added_at"], + desc="user_get_threepids", + ), ) + return [ + ThreepidResult( + medium=r[0], + address=r[1], + validated_at=r[2], + added_at=r[3], + ) + for r in results + ] async def user_delete_threepid( self, user_id: str, medium: str, address: str @@ -1009,7 +1071,7 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore): desc="add_user_bound_threepid", ) - async def user_get_bound_threepids(self, user_id: str) -> List[Dict[str, Any]]: + async def user_get_bound_threepids(self, user_id: str) -> List[Tuple[str, str]]: """Get the threepids that a user has bound to an identity server through the homeserver The homeserver remembers where binds to an identity server occurred. Using this method can retrieve those threepids. @@ -1018,15 +1080,18 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore): user_id: The ID of the user to retrieve threepids for Returns: - List of dictionaries containing the following keys: - medium (str): The medium of the threepid (e.g "email") - address (str): The address of the threepid (e.g "bob@example.com") - """ - return await self.db_pool.simple_select_list( - table="user_threepid_id_server", - keyvalues={"user_id": user_id}, - retcols=["medium", "address"], - desc="user_get_bound_threepids", + List of tuples of two strings: + medium: The medium of the threepid (e.g "email") + address: The address of the threepid (e.g "bob@example.com") + """ + return cast( + List[Tuple[str, str]], + await self.db_pool.simple_select_list( + table="user_threepid_id_server", + keyvalues={"user_id": user_id}, + retcols=["medium", "address"], + desc="user_get_bound_threepids", + ), ) async def remove_user_bound_threepid( @@ -1123,7 +1188,7 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore): address: Optional[str] = None, sid: Optional[str] = None, validated: Optional[bool] = True, - ) -> Optional[Dict[str, Any]]: + ) -> Optional[ThreepidValidationSession]: """Gets a session_id and last_send_attempt (if available) for a combination of validation metadata @@ -1138,15 +1203,7 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore): perform no filtering Returns: - A dict containing the following: - * address - address of the 3pid - * medium - medium of the 3pid - * client_secret - a secret provided by the client for this validation session - * session_id - ID of the validation session - * send_attempt - a number serving to dedupe send attempts for this session - * validated_at - timestamp of when this session was validated if so - - Otherwise None if a validation session is not found + A ThreepidValidationSession or None if a validation session is not found """ if not client_secret: raise SynapseError( @@ -1165,7 +1222,7 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore): def get_threepid_validation_session_txn( txn: LoggingTransaction, - ) -> Optional[Dict[str, Any]]: + ) -> Optional[ThreepidValidationSession]: sql = """ SELECT address, session_id, medium, client_secret, last_send_attempt, validated_at @@ -1180,11 +1237,18 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore): sql += " LIMIT 1" txn.execute(sql, list(keyvalues.values())) - rows = self.db_pool.cursor_to_dict(txn) - if not rows: + row = txn.fetchone() + if not row: return None - return rows[0] + return ThreepidValidationSession( + address=row[0], + session_id=row[1], + medium=row[2], + client_secret=row[3], + last_send_attempt=row[4], + validated_at=row[5], + ) return await self.db_pool.runInteraction( "get_threepid_validation_session", get_threepid_validation_session_txn @@ -1252,12 +1316,8 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore): ) txn.execute(sql, []) - res = self.db_pool.cursor_to_dict(txn) - if res: - for user in res: - self.set_expiration_date_for_user_txn( - txn, user["name"], use_delta=True - ) + for (name,) in txn.fetchall(): + self.set_expiration_date_for_user_txn(txn, name, use_delta=True) await self.db_pool.runInteraction( "get_users_with_no_expiration_date", @@ -1457,7 +1517,7 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore): async def get_registration_tokens( self, valid: Optional[bool] = None - ) -> List[Dict[str, Any]]: + ) -> List[Tuple[str, Optional[int], int, int, Optional[int]]]: """List all registration tokens. Used by the admin API. Args: @@ -1466,34 +1526,48 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore): Default is None: return all tokens regardless of validity. Returns: - A list of dicts, each containing details of a token. + A list of tuples containing: + * The token + * The number of users allowed (or None) + * Whether it is pending + * Whether it has been completed + * An expiry time (or None if no expiry) """ def select_registration_tokens_txn( txn: LoggingTransaction, now: int, valid: Optional[bool] - ) -> List[Dict[str, Any]]: + ) -> List[Tuple[str, Optional[int], int, int, Optional[int]]]: if valid is None: # Return all tokens regardless of validity - txn.execute("SELECT * FROM registration_tokens") + txn.execute( + """ + SELECT token, uses_allowed, pending, completed, expiry_time + FROM registration_tokens + """ + ) elif valid: # Select valid tokens only - sql = ( - "SELECT * FROM registration_tokens WHERE " - "(uses_allowed > pending + completed OR uses_allowed IS NULL) " - "AND (expiry_time > ? OR expiry_time IS NULL)" - ) + sql = """ + SELECT token, uses_allowed, pending, completed, expiry_time + FROM registration_tokens + WHERE (uses_allowed > pending + completed OR uses_allowed IS NULL) + AND (expiry_time > ? OR expiry_time IS NULL) + """ txn.execute(sql, [now]) else: # Select invalid tokens only - sql = ( - "SELECT * FROM registration_tokens WHERE " - "uses_allowed <= pending + completed OR expiry_time <= ?" - ) + sql = """ + SELECT token, uses_allowed, pending, completed, expiry_time + FROM registration_tokens + WHERE uses_allowed <= pending + completed OR expiry_time <= ? + """ txn.execute(sql, [now]) - return self.db_pool.cursor_to_dict(txn) + return cast( + List[Tuple[str, Optional[int], int, int, Optional[int]]], txn.fetchall() + ) return await self.db_pool.runInteraction( "select_registration_tokens", @@ -1963,11 +2037,12 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore): (user_id,), ) - rows = self.db_pool.cursor_to_dict(txn) + row = txn.fetchone() + assert row is not None # We cast to bool because the value returned by the database engine might # be an integer if we're using SQLite. - return bool(rows[0]["approved"]) + return bool(row[0]) return await self.db_pool.runInteraction( desc="is_user_pending_approval", @@ -2045,22 +2120,22 @@ class RegistrationBackgroundUpdateStore(RegistrationWorkerStore): (last_user, batch_size), ) - rows = self.db_pool.cursor_to_dict(txn) + rows = txn.fetchall() if not rows: return True, 0 rows_processed_nb = 0 - for user in rows: - if not user["count_tokens"] and not user["count_threepids"]: - self.set_user_deactivated_status_txn(txn, user["name"], True) + for name, count_tokens, count_threepids in rows: + if not count_tokens and not count_threepids: + self.set_user_deactivated_status_txn(txn, name, True) rows_processed_nb += 1 logger.info("Marked %d rows as deactivated", rows_processed_nb) self.db_pool.updates._background_update_progress_txn( - txn, "users_set_deactivated_flag", {"user_id": rows[-1]["name"]} + txn, "users_set_deactivated_flag", {"user_id": rows[-1][0]} ) if batch_size > len(rows): diff --git a/synapse/storage/databases/main/relations.py b/synapse/storage/databases/main/relations.py index c04d45bdb5..d0bc78b2e3 100644 --- a/synapse/storage/databases/main/relations.py +++ b/synapse/storage/databases/main/relations.py @@ -48,6 +48,8 @@ from synapse.storage.databases.main.stream import ( ) from synapse.storage.engines import PostgresEngine, Psycopg2Engine from synapse.types import JsonDict, StreamKeyType, StreamToken +from synapse.storage.engines import PostgresEngine +from synapse.types import JsonDict, MultiWriterStreamToken, StreamKeyType, StreamToken from synapse.util.caches.descriptors import cached, cachedList if TYPE_CHECKING: @@ -314,7 +316,7 @@ class RelationsWorkerStore(SQLBaseStore): room_key=next_key, presence_key=0, typing_key=0, - receipt_key=0, + receipt_key=MultiWriterStreamToken(stream=0), account_data_key=0, push_rules_key=0, to_device_key=0, @@ -349,16 +351,19 @@ class RelationsWorkerStore(SQLBaseStore): 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"], + rows = cast( + List[Tuple[str]], + 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 [row[0] for row in rows] return await self.db_pool.runInteraction( desc="get_all_relation_ids_for_event_with_types", @@ -381,14 +386,17 @@ class RelationsWorkerStore(SQLBaseStore): def get_all_relation_ids_for_event_txn( txn: LoggingTransaction, ) -> List[str]: - rows = self.db_pool.simple_select_list_txn( - txn=txn, - table="event_relations", - keyvalues={"relates_to_id": event_id}, - retcols=["event_id"], + rows = cast( + List[Tuple[str]], + self.db_pool.simple_select_list_txn( + txn=txn, + table="event_relations", + keyvalues={"relates_to_id": event_id}, + retcols=["event_id"], + ), ) - return [row["event_id"] for row in rows] + return [row[0] for row in rows] return await self.db_pool.runInteraction( desc="get_all_relation_ids_for_event", @@ -458,7 +466,7 @@ class RelationsWorkerStore(SQLBaseStore): ) return result is not None - @cached() + @cached() # type: ignore[synapse-@cached-mutable] async def get_references_for_event(self, event_id: str) -> List[JsonDict]: raise NotImplementedError() @@ -512,11 +520,12 @@ class RelationsWorkerStore(SQLBaseStore): "_get_references_for_events_txn", _get_references_for_events_txn ) - @cached() + @cached() # type: ignore[synapse-@cached-mutable] def get_applicable_edit(self, event_id: str) -> Optional[EventBase]: raise NotImplementedError() - @cachedList(cached_method_name="get_applicable_edit", list_name="event_ids") + # TODO: This returns a mutable object, which is generally bad. + @cachedList(cached_method_name="get_applicable_edit", list_name="event_ids") # type: ignore[synapse-@cached-mutable] async def get_applicable_edits( self, event_ids: Collection[str] ) -> Mapping[str, Optional[EventBase]]: @@ -598,11 +607,12 @@ class RelationsWorkerStore(SQLBaseStore): for original_event_id in event_ids } - @cached() + @cached() # type: ignore[synapse-@cached-mutable] def get_thread_summary(self, event_id: str) -> Optional[Tuple[int, EventBase]]: raise NotImplementedError() - @cachedList(cached_method_name="get_thread_summary", list_name="event_ids") + # TODO: This returns a mutable object, which is generally bad. + @cachedList(cached_method_name="get_thread_summary", list_name="event_ids") # type: ignore[synapse-@cached-mutable] async def get_thread_summaries( self, event_ids: Collection[str] ) -> Mapping[str, Optional[Tuple[int, EventBase]]]: diff --git a/synapse/storage/databases/main/room.py b/synapse/storage/databases/main/room.py index 719e11aea6..afb880532e 100644 --- a/synapse/storage/databases/main/room.py +++ b/synapse/storage/databases/main/room.py @@ -78,6 +78,31 @@ class RatelimitOverride: burst_count: int +@attr.s(slots=True, frozen=True, auto_attribs=True) +class LargestRoomStats: + room_id: str + name: Optional[str] + canonical_alias: Optional[str] + joined_members: int + join_rules: Optional[str] + guest_access: Optional[str] + history_visibility: Optional[str] + state_events: int + avatar: Optional[str] + topic: Optional[str] + room_type: Optional[str] + + +@attr.s(slots=True, frozen=True, auto_attribs=True) +class RoomStats(LargestRoomStats): + joined_local_members: int + version: Optional[str] + creator: Optional[str] + encryption: Optional[str] + federatable: bool + public: bool + + class RoomSortOrder(Enum): """ Enum to define the sorting method used when returning rooms with get_rooms_paginate @@ -204,7 +229,7 @@ class RoomWorkerStore(CacheInvalidationWorkerStore): allow_none=True, ) - async def get_room_with_stats(self, room_id: str) -> Optional[Dict[str, Any]]: + async def get_room_with_stats(self, room_id: str) -> Optional[RoomStats]: """Retrieve room with statistics. Args: @@ -215,7 +240,7 @@ class RoomWorkerStore(CacheInvalidationWorkerStore): def get_room_with_stats_txn( txn: LoggingTransaction, room_id: str - ) -> Optional[Dict[str, Any]]: + ) -> Optional[RoomStats]: sql = """ SELECT room_id, state.name, state.canonical_alias, curr.joined_members, curr.local_users_in_room AS joined_local_members, rooms.room_version AS version, @@ -229,15 +254,28 @@ class RoomWorkerStore(CacheInvalidationWorkerStore): WHERE room_id = ? """ txn.execute(sql, [room_id]) - # Catch error if sql returns empty result to return "None" instead of an error - try: - res = self.db_pool.cursor_to_dict(txn)[0] - except IndexError: + row = txn.fetchone() + if not row: return None - - res["federatable"] = bool(res["federatable"]) - res["public"] = bool(res["public"]) - return res + return RoomStats( + room_id=row[0], + name=row[1], + canonical_alias=row[2], + joined_members=row[3], + joined_local_members=row[4], + version=row[5], + creator=row[6], + encryption=row[7], + federatable=bool(row[8]), + public=bool(row[9]), + join_rules=row[10], + guest_access=row[11], + history_visibility=row[12], + state_events=row[13], + avatar=row[14], + topic=row[15], + room_type=row[16], + ) return await self.db_pool.runInteraction( "get_room_with_stats", get_room_with_stats_txn, room_id @@ -368,7 +406,7 @@ class RoomWorkerStore(CacheInvalidationWorkerStore): bounds: Optional[Tuple[int, str]], forwards: bool, ignore_non_federatable: bool = False, - ) -> List[Dict[str, Any]]: + ) -> List[LargestRoomStats]: """Gets the largest public rooms (where largest is in terms of joined members, as tracked in the statistics table). @@ -505,20 +543,34 @@ class RoomWorkerStore(CacheInvalidationWorkerStore): def _get_largest_public_rooms_txn( txn: LoggingTransaction, - ) -> List[Dict[str, Any]]: + ) -> List[LargestRoomStats]: txn.execute(sql, query_args) - results = self.db_pool.cursor_to_dict(txn) + results = [ + LargestRoomStats( + room_id=r[0], + name=r[1], + canonical_alias=r[3], + joined_members=r[4], + join_rules=r[8], + guest_access=r[7], + history_visibility=r[6], + state_events=0, + avatar=r[5], + topic=r[2], + room_type=r[9], + ) + for r in txn + ] if not forwards: results.reverse() return results - ret_val = await self.db_pool.runInteraction( + return await self.db_pool.runInteraction( "get_largest_public_rooms", _get_largest_public_rooms_txn ) - return ret_val @cached(max_entries=10000) async def is_room_blocked(self, room_id: str) -> Optional[bool]: @@ -831,7 +883,7 @@ class RoomWorkerStore(CacheInvalidationWorkerStore): def get_retention_policy_for_room_txn( txn: LoggingTransaction, - ) -> List[Dict[str, Optional[int]]]: + ) -> Optional[Tuple[Optional[int], Optional[int]]]: txn.execute( """ SELECT min_lifetime, max_lifetime FROM room_retention @@ -841,7 +893,7 @@ class RoomWorkerStore(CacheInvalidationWorkerStore): (room_id,), ) - return self.db_pool.cursor_to_dict(txn) + return cast(Optional[Tuple[Optional[int], Optional[int]]], txn.fetchone()) ret = await self.db_pool.runInteraction( "get_retention_policy_for_room", @@ -856,8 +908,7 @@ class RoomWorkerStore(CacheInvalidationWorkerStore): max_lifetime=self.config.retention.retention_default_max_lifetime, ) - min_lifetime = ret[0]["min_lifetime"] - max_lifetime = ret[0]["max_lifetime"] + min_lifetime, max_lifetime = ret # If one of the room's policy's attributes isn't defined, use the matching # attribute from the default policy. @@ -1162,14 +1213,13 @@ class RoomWorkerStore(CacheInvalidationWorkerStore): txn.execute(sql, args) - rows = self.db_pool.cursor_to_dict(txn) - rooms_dict = {} - - for row in rows: - rooms_dict[row["room_id"]] = RetentionPolicy( - min_lifetime=row["min_lifetime"], - max_lifetime=row["max_lifetime"], + rooms_dict = { + room_id: RetentionPolicy( + min_lifetime=min_lifetime, + max_lifetime=max_lifetime, ) + for room_id, min_lifetime, max_lifetime in txn + } if include_null: # If required, do a second query that retrieves all of the rooms we know @@ -1178,13 +1228,11 @@ class RoomWorkerStore(CacheInvalidationWorkerStore): txn.execute(sql) - rows = self.db_pool.cursor_to_dict(txn) - # If a room isn't already in the dict (i.e. it doesn't have a retention # policy in its state), add it with a null policy. - for row in rows: - if row["room_id"] not in rooms_dict: - rooms_dict[row["room_id"]] = RetentionPolicy() + for (room_id,) in txn: + if room_id not in rooms_dict: + rooms_dict[room_id] = RetentionPolicy() return rooms_dict @@ -1236,28 +1284,30 @@ class RoomWorkerStore(CacheInvalidationWorkerStore): """ 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", + rows = cast( + List[Tuple[str, str]], + 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"] + for room_id, joined_via in rows: room_servers[room_id] = PartialStateResyncInfo(joined_via=joined_via) - rows = await self.db_pool.simple_select_list( - "partial_state_rooms_servers", - keyvalues=None, - retcols=("room_id", "server_name"), - desc="get_partial_state_rooms", + rows = cast( + List[Tuple[str, str]], + await self.db_pool.simple_select_list( + "partial_state_rooms_servers", + keyvalues=None, + retcols=("room_id", "server_name"), + desc="get_partial_state_rooms", + ), ) - for row in rows: - room_id = row["room_id"] - server_name = row["server_name"] + for room_id, server_name in rows: entry = room_servers.get(room_id) if entry is None: # There is a foreign key constraint which enforces that every room_id in @@ -1300,14 +1350,17 @@ class RoomWorkerStore(CacheInvalidationWorkerStore): complete. """ - rows: List[Dict[str, str]] = await self.db_pool.simple_select_many_batch( - table="partial_state_rooms", - column="room_id", - iterable=room_ids, - retcols=("room_id",), - desc="is_partial_state_room_batched", - ) - partial_state_rooms = {row_dict["room_id"] for row_dict in rows} + rows = cast( + List[Tuple[str]], + await self.db_pool.simple_select_many_batch( + table="partial_state_rooms", + column="room_id", + iterable=room_ids, + retcols=("room_id",), + desc="is_partial_state_room_batched", + ), + ) + partial_state_rooms = {row[0] for row in rows} return {room_id: room_id in partial_state_rooms for room_id in room_ids} async def get_join_event_id_and_device_lists_stream_id_for_partial_state( @@ -1703,24 +1756,24 @@ class RoomBackgroundUpdateStore(SQLBaseStore): (last_room, batch_size), ) - rows = self.db_pool.cursor_to_dict(txn) + rows = txn.fetchall() if not rows: return True - for row in rows: - if not row["json"]: + for room_id, event_id, json in rows: + if not json: retention_policy = {} else: - ev = db_to_json(row["json"]) + ev = db_to_json(json) retention_policy = ev["content"] self.db_pool.simple_insert_txn( txn=txn, table="room_retention", values={ - "room_id": row["room_id"], - "event_id": row["event_id"], + "room_id": room_id, + "event_id": event_id, "min_lifetime": retention_policy.get("min_lifetime"), "max_lifetime": retention_policy.get("max_lifetime"), }, @@ -1729,7 +1782,7 @@ class RoomBackgroundUpdateStore(SQLBaseStore): logger.info("Inserted %d rows into room_retention", len(rows)) self.db_pool.updates._background_update_progress_txn( - txn, "insert_room_retention", {"room_id": rows[-1]["room_id"]} + txn, "insert_room_retention", {"room_id": rows[-1][0]} ) if batch_size > len(rows): @@ -2215,7 +2268,7 @@ class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore): txn, table="partial_state_rooms_servers", keys=("room_id", "server_name"), - values=((room_id, s) for s in servers), + values=[(room_id, s) for s in servers], ) self._invalidate_cache_and_stream(txn, self.is_partial_state_room, (room_id,)) self._invalidate_cache_and_stream( diff --git a/synapse/storage/databases/main/roommember.py b/synapse/storage/databases/main/roommember.py index 3755773faa..1ed7f2d0ef 100644 --- a/synapse/storage/databases/main/roommember.py +++ b/synapse/storage/databases/main/roommember.py @@ -27,6 +27,7 @@ from typing import ( Set, Tuple, Union, + cast, ) import attr @@ -275,7 +276,7 @@ class RoomMemberWorkerStore(EventsWorkerStore, CacheInvalidationWorkerStore): _get_users_in_room_with_profiles, ) - @cached(max_entries=100000) + @cached(max_entries=100000) # type: ignore[synapse-@cached-mutable] async def get_room_summary(self, room_id: str) -> Mapping[str, MemberSummary]: """Get the details of a room roughly suitable for use by the room summary extension to /sync. Useful when lazy loading room members. @@ -481,6 +482,22 @@ class RoomMemberWorkerStore(EventsWorkerStore, CacheInvalidationWorkerStore): desc="get_local_users_in_room", ) + async def get_local_users_related_to_room( + self, room_id: str + ) -> List[Tuple[str, str]]: + """ + Retrieves a list of the current roommembers who are local to the server and their membership status. + """ + return cast( + List[Tuple[str, str]], + await self.db_pool.simple_select_list( + table="local_current_membership", + keyvalues={"room_id": room_id}, + retcols=("user_id", "membership"), + desc="get_local_users_in_room", + ), + ) + async def check_local_user_in_room(self, user_id: str, room_id: str) -> bool: """ Check whether a given local user is currently joined to the given room. @@ -683,25 +700,28 @@ class RoomMemberWorkerStore(EventsWorkerStore, CacheInvalidationWorkerStore): Map from user_id to set of rooms that is currently in. """ - rows = await self.db_pool.simple_select_many_batch( - table="current_state_events", - column="state_key", - iterable=user_ids, - retcols=( - "state_key", - "room_id", + rows = cast( + List[Tuple[str, str]], + await self.db_pool.simple_select_many_batch( + table="current_state_events", + column="state_key", + iterable=user_ids, + retcols=( + "state_key", + "room_id", + ), + keyvalues={ + "type": EventTypes.Member, + "membership": Membership.JOIN, + }, + desc="get_rooms_for_users", ), - keyvalues={ - "type": EventTypes.Member, - "membership": Membership.JOIN, - }, - desc="get_rooms_for_users", ) user_rooms: Dict[str, Set[str]] = {user_id: set() for user_id in user_ids} - for row in rows: - user_rooms[row["state_key"]].add(row["room_id"]) + for state_key, room_id in rows: + user_rooms[state_key].add(room_id) return {key: frozenset(rooms) for key, rooms in user_rooms.items()} @@ -892,17 +912,20 @@ class RoomMemberWorkerStore(EventsWorkerStore, CacheInvalidationWorkerStore): Map from event ID to `user_id`, or None if event is not a join. """ - rows = await self.db_pool.simple_select_many_batch( - table="room_memberships", - column="event_id", - iterable=event_ids, - retcols=("user_id", "event_id"), - keyvalues={"membership": Membership.JOIN}, - batch_size=1000, - desc="_get_user_ids_from_membership_event_ids", + rows = cast( + List[Tuple[str, str]], + await self.db_pool.simple_select_many_batch( + table="room_memberships", + column="event_id", + iterable=event_ids, + retcols=("event_id", "user_id"), + keyvalues={"membership": Membership.JOIN}, + batch_size=1000, + desc="_get_user_ids_from_membership_event_ids", + ), ) - return {row["event_id"]: row["user_id"] for row in rows} + return dict(rows) @cached(max_entries=10000) async def is_host_joined(self, room_id: str, host: str) -> bool: @@ -933,7 +956,7 @@ class RoomMemberWorkerStore(EventsWorkerStore, CacheInvalidationWorkerStore): like_clause = "%:" + host rows = await self.db_pool.execute( - "is_host_joined", None, sql, membership, room_id, like_clause + "is_host_joined", sql, membership, room_id, like_clause ) if not rows: @@ -1063,15 +1086,19 @@ class RoomMemberWorkerStore(EventsWorkerStore, CacheInvalidationWorkerStore): for fully-joined rooms. """ - rows = await self.db_pool.simple_select_list( - "current_state_events", - keyvalues={"room_id": room_id}, - retcols=("event_id", "membership"), - desc="has_completed_background_updates", + rows = cast( + List[Tuple[str, Optional[str]]], + await self.db_pool.simple_select_list( + "current_state_events", + keyvalues={"room_id": room_id}, + retcols=("event_id", "membership"), + desc="has_completed_background_updates", + ), ) - return {row["event_id"]: row["membership"] for row in rows} + return dict(rows) - @cached(max_entries=10000) + # TODO This returns a mutable object, which is generally confusing when using a cache. + @cached(max_entries=10000) # type: ignore[synapse-@cached-mutable] def _get_joined_hosts_cache(self, room_id: str) -> "_JoinedHostsCache": return _JoinedHostsCache() @@ -1157,7 +1184,7 @@ class RoomMemberWorkerStore(EventsWorkerStore, CacheInvalidationWorkerStore): AND forgotten = 0; """ - rows = await self.db_pool.execute("is_forgotten_room", None, sql, room_id) + rows = await self.db_pool.execute("is_forgotten_room", sql, room_id) # `count(*)` returns always an integer # If any rows still exist it means someone has not forgotten this room yet @@ -1201,21 +1228,22 @@ class RoomMemberWorkerStore(EventsWorkerStore, CacheInvalidationWorkerStore): membership event, otherwise the value is None. """ - rows = await self.db_pool.simple_select_many_batch( - table="room_memberships", - column="event_id", - iterable=member_event_ids, - retcols=("user_id", "membership", "event_id"), - keyvalues={}, - batch_size=500, - desc="get_membership_from_event_ids", + rows = cast( + List[Tuple[str, str, str]], + await self.db_pool.simple_select_many_batch( + table="room_memberships", + column="event_id", + iterable=member_event_ids, + retcols=("user_id", "membership", "event_id"), + keyvalues={}, + batch_size=500, + desc="get_membership_from_event_ids", + ), ) return { - row["event_id"]: EventIdMembership( - membership=row["membership"], user_id=row["user_id"] - ) - for row in rows + event_id: EventIdMembership(membership=membership, user_id=user_id) + for user_id, membership, event_id in rows } async def is_local_host_in_room_ignoring_users( @@ -1348,18 +1376,16 @@ class RoomMemberBackgroundUpdateStore(SQLBaseStore): txn.execute(sql, (target_min_stream_id, max_stream_id, batch_size)) - rows = self.db_pool.cursor_to_dict(txn) + rows = txn.fetchall() if not rows: return 0 - min_stream_id = rows[-1]["stream_ordering"] + min_stream_id = rows[-1][0] to_update = [] - for row in rows: - event_id = row["event_id"] - room_id = row["room_id"] + for _, event_id, room_id, json in rows: try: - event_json = db_to_json(row["json"]) + event_json = db_to_json(json) content = event_json["content"] except Exception: continue diff --git a/synapse/storage/databases/main/search.py b/synapse/storage/databases/main/search.py index d45d2ecc98..4c4112e3b2 100644 --- a/synapse/storage/databases/main/search.py +++ b/synapse/storage/databases/main/search.py @@ -26,6 +26,7 @@ from typing import ( Set, Tuple, Union, + cast, ) import attr @@ -105,7 +106,7 @@ class SearchWorkerStore(SQLBaseStore): txn, table="event_search", keys=("event_id", "room_id", "key", "value"), - values=( + values=[ ( entry.event_id, entry.room_id, @@ -113,7 +114,7 @@ class SearchWorkerStore(SQLBaseStore): _clean_value_for_search(entry.value), ) for entry in entries - ), + ], ) else: @@ -179,22 +180,24 @@ class SearchBackgroundUpdateStore(SearchWorkerStore): # store_search_entries_txn with a generator function, but that # would mean having two cursors open on the database at once. # Instead we just build a list of results. - rows = self.db_pool.cursor_to_dict(txn) + rows = txn.fetchall() if not rows: return 0 - min_stream_id = rows[-1]["stream_ordering"] + min_stream_id = rows[-1][0] event_search_rows = [] - for row in rows: + for ( + stream_ordering, + event_id, + room_id, + etype, + json, + origin_server_ts, + ) in rows: try: - event_id = row["event_id"] - room_id = row["room_id"] - etype = row["type"] - stream_ordering = row["stream_ordering"] - origin_server_ts = row["origin_server_ts"] try: - event_json = db_to_json(row["json"]) + event_json = db_to_json(json) content = event_json["content"] except Exception: continue @@ -504,16 +507,18 @@ class SearchStore(SearchBackgroundUpdateStore): # entire table from the database. sql += " ORDER BY rank DESC LIMIT 500" - results = await self.db_pool.execute( - "search_msgs", self.db_pool.cursor_to_dict, sql, *args + # List of tuples of (rank, room_id, event_id). + results = cast( + List[Tuple[Union[int, float], str, str]], + await self.db_pool.execute("search_msgs", sql, *args), ) - results = list(filter(lambda row: row["room_id"] in room_ids, results)) + results = list(filter(lambda row: row[1] in room_ids, results)) # We set redact_behaviour to block here to prevent redacted events being returned in # search results (which is a data leak) events = await self.get_events_as_list( # type: ignore[attr-defined] - [r["event_id"] for r in results], + [r[2] for r in results], redact_behaviour=EventRedactBehaviour.block, ) @@ -525,16 +530,18 @@ class SearchStore(SearchBackgroundUpdateStore): count_sql += " GROUP BY room_id" - count_results = await self.db_pool.execute( - "search_rooms_count", self.db_pool.cursor_to_dict, count_sql, *count_args + # List of tuples of (room_id, count). + count_results = cast( + List[Tuple[str, int]], + await self.db_pool.execute("search_rooms_count", count_sql, *count_args), ) - count = sum(row["count"] for row in count_results if row["room_id"] in room_ids) + count = sum(row[1] for row in count_results if row[0] in room_ids) return { "results": [ - {"event": event_map[r["event_id"]], "rank": r["rank"]} + {"event": event_map[r[2]], "rank": r[0]} for r in results - if r["event_id"] in event_map + if r[2] in event_map ], "highlights": highlights, "count": count, @@ -602,7 +609,7 @@ class SearchStore(SearchBackgroundUpdateStore): search_query = search_term sql = """ SELECT ts_rank_cd(vector, websearch_to_tsquery('english', ?)) as rank, - origin_server_ts, stream_ordering, room_id, event_id + room_id, event_id, origin_server_ts, stream_ordering FROM event_search WHERE vector @@ websearch_to_tsquery('english', ?) AND """ @@ -663,16 +670,18 @@ class SearchStore(SearchBackgroundUpdateStore): # mypy expects to append only a `str`, not an `int` args.append(limit) - results = await self.db_pool.execute( - "search_rooms", self.db_pool.cursor_to_dict, sql, *args + # List of tuples of (rank, room_id, event_id, origin_server_ts, stream_ordering). + results = cast( + List[Tuple[Union[int, float], str, str, int, int]], + await self.db_pool.execute("search_rooms", sql, *args), ) - results = list(filter(lambda row: row["room_id"] in room_ids, results)) + results = list(filter(lambda row: row[1] in room_ids, results)) # We set redact_behaviour to block here to prevent redacted events being returned in # search results (which is a data leak) events = await self.get_events_as_list( # type: ignore[attr-defined] - [r["event_id"] for r in results], + [r[2] for r in results], redact_behaviour=EventRedactBehaviour.block, ) @@ -684,22 +693,23 @@ class SearchStore(SearchBackgroundUpdateStore): count_sql += " GROUP BY room_id" - count_results = await self.db_pool.execute( - "search_rooms_count", self.db_pool.cursor_to_dict, count_sql, *count_args + # List of tuples of (room_id, count). + count_results = cast( + List[Tuple[str, int]], + await self.db_pool.execute("search_rooms_count", count_sql, *count_args), ) - count = sum(row["count"] for row in count_results if row["room_id"] in room_ids) + count = sum(row[1] for row in count_results if row[0] in room_ids) return { "results": [ { - "event": event_map[r["event_id"]], - "rank": r["rank"], - "pagination_token": "%s,%s" - % (r["origin_server_ts"], r["stream_ordering"]), + "event": event_map[r[2]], + "rank": r[0], + "pagination_token": "%s,%s" % (r[3], r[4]), } for r in results - if r["event_id"] in event_map + if r[2] in event_map ], "highlights": highlights, "count": count, diff --git a/synapse/storage/databases/main/state.py b/synapse/storage/databases/main/state.py index 5eaaff5b68..598025dd91 100644 --- a/synapse/storage/databases/main/state.py +++ b/synapse/storage/databases/main/state.py @@ -20,10 +20,12 @@ from typing import ( Collection, Dict, Iterable, + List, Mapping, Optional, Set, Tuple, + cast, ) import attr @@ -388,16 +390,19 @@ class StateGroupWorkerStore(EventsWorkerStore, SQLBaseStore): Raises: RuntimeError if the state is unknown at any of the given events """ - rows = await self.db_pool.simple_select_many_batch( - table="event_to_state_groups", - column="event_id", - iterable=event_ids, - keyvalues={}, - retcols=("event_id", "state_group"), - desc="_get_state_group_for_events", + rows = cast( + List[Tuple[str, int]], + await self.db_pool.simple_select_many_batch( + table="event_to_state_groups", + column="event_id", + iterable=event_ids, + keyvalues={}, + retcols=("event_id", "state_group"), + desc="_get_state_group_for_events", + ), ) - res = {row["event_id"]: row["state_group"] for row in rows} + res = dict(rows) for e in event_ids: if e not in res: raise RuntimeError("No state group for unknown or outlier event %s" % e) @@ -415,16 +420,19 @@ class StateGroupWorkerStore(EventsWorkerStore, SQLBaseStore): The subset of state groups that are referenced. """ - rows = await self.db_pool.simple_select_many_batch( - table="event_to_state_groups", - column="state_group", - iterable=state_groups, - keyvalues={}, - retcols=("DISTINCT state_group",), - desc="get_referenced_state_groups", + rows = cast( + List[Tuple[int]], + await self.db_pool.simple_select_many_batch( + table="event_to_state_groups", + column="state_group", + iterable=state_groups, + keyvalues={}, + retcols=("DISTINCT state_group",), + desc="get_referenced_state_groups", + ), ) - return {row["state_group"] for row in rows} + return {row[0] for row in rows} async def update_state_for_partial_state_event( self, @@ -624,16 +632,22 @@ class MainStateBackgroundUpdateStore(RoomMemberWorkerStore): # potentially stale, since there may have been a period where the # server didn't share a room with the remote user and therefore may # have missed any device updates. - rows = self.db_pool.simple_select_many_txn( - txn, - table="current_state_events", - column="room_id", - iterable=to_delete, - keyvalues={"type": EventTypes.Member, "membership": Membership.JOIN}, - retcols=("state_key",), + rows = cast( + List[Tuple[str]], + self.db_pool.simple_select_many_txn( + txn, + table="current_state_events", + column="room_id", + iterable=to_delete, + keyvalues={ + "type": EventTypes.Member, + "membership": Membership.JOIN, + }, + retcols=("state_key",), + ), ) - potentially_left_users = {row["state_key"] for row in rows} + potentially_left_users = {row[0] for row in rows} # Now lets actually delete the rooms from the DB. self.db_pool.simple_delete_many_txn( diff --git a/synapse/storage/databases/main/state_deltas.py b/synapse/storage/databases/main/state_deltas.py index 445213e12a..3151186e0c 100644 --- a/synapse/storage/databases/main/state_deltas.py +++ b/synapse/storage/databases/main/state_deltas.py @@ -13,7 +13,9 @@ # limitations under the License. import logging -from typing import Any, Dict, List, Tuple +from typing import List, Optional, Tuple + +import attr from synapse.storage._base import SQLBaseStore from synapse.storage.database import LoggingTransaction @@ -22,6 +24,20 @@ from synapse.util.caches.stream_change_cache import StreamChangeCache logger = logging.getLogger(__name__) +@attr.s(slots=True, frozen=True, auto_attribs=True) +class StateDelta: + stream_id: int + room_id: str + event_type: str + state_key: str + + event_id: Optional[str] + """new event_id for this state key. None if the state has been deleted.""" + + prev_event_id: Optional[str] + """previous event_id for this state key. None if it's new state.""" + + class StateDeltasStore(SQLBaseStore): # This class must be mixed in with a child class which provides the following # attribute. TODO: can we get static analysis to enforce this? @@ -29,31 +45,21 @@ class StateDeltasStore(SQLBaseStore): async def get_partial_current_state_deltas( self, prev_stream_id: int, max_stream_id: int - ) -> Tuple[int, List[Dict[str, Any]]]: + ) -> Tuple[int, List[StateDelta]]: """Fetch a list of room state changes since the given stream id - Each entry in the result contains the following fields: - - stream_id (int) - - room_id (str) - - type (str): event type - - state_key (str): - - event_id (str|None): new event_id for this state key. None if the - state has been deleted. - - prev_event_id (str|None): previous event_id for this state key. None - if it's new state. - This may be the partial state if we're lazy joining the room. Args: prev_stream_id: point to get changes since (exclusive) max_stream_id: the point that we know has been correctly persisted - - ie, an upper limit to return changes from. + - ie, an upper limit to return changes from. Returns: A tuple consisting of: - - the stream id which these results go up to - - list of current_state_delta_stream rows. If it is empty, we are - up to date. + - the stream id which these results go up to + - list of current_state_delta_stream rows. If it is empty, we are + up to date. """ prev_stream_id = int(prev_stream_id) @@ -72,7 +78,7 @@ class StateDeltasStore(SQLBaseStore): def get_current_state_deltas_txn( txn: LoggingTransaction, - ) -> Tuple[int, List[Dict[str, Any]]]: + ) -> Tuple[int, List[StateDelta]]: # First we calculate the max stream id that will give us less than # N results. # We arbitrarily limit to 100 stream_id entries to ensure we don't @@ -112,7 +118,17 @@ class StateDeltasStore(SQLBaseStore): ORDER BY stream_id ASC """ txn.execute(sql, (prev_stream_id, clipped_stream_id)) - return clipped_stream_id, self.db_pool.cursor_to_dict(txn) + return clipped_stream_id, [ + StateDelta( + stream_id=row[0], + room_id=row[1], + event_type=row[2], + state_key=row[3], + event_id=row[4], + prev_event_id=row[5], + ) + for row in txn.fetchall() + ] return await self.db_pool.runInteraction( "get_current_state_deltas", get_current_state_deltas_txn diff --git a/synapse/storage/databases/main/stats.py b/synapse/storage/databases/main/stats.py index 9d403919e4..e96c9b0486 100644 --- a/synapse/storage/databases/main/stats.py +++ b/synapse/storage/databases/main/stats.py @@ -506,25 +506,28 @@ class StatsStore(StateDeltasStore): ) -> Tuple[List[str], Dict[str, int], int, List[str], int]: pos = self.get_room_max_stream_ordering() # type: ignore[attr-defined] - rows = self.db_pool.simple_select_many_txn( - txn, - table="current_state_events", - column="type", - iterable=[ - EventTypes.Create, - EventTypes.JoinRules, - EventTypes.RoomHistoryVisibility, - EventTypes.RoomEncryption, - EventTypes.Name, - EventTypes.Topic, - EventTypes.RoomAvatar, - EventTypes.CanonicalAlias, - ], - keyvalues={"room_id": room_id, "state_key": ""}, - retcols=["event_id"], + rows = cast( + List[Tuple[str]], + self.db_pool.simple_select_many_txn( + txn, + table="current_state_events", + column="type", + iterable=[ + EventTypes.Create, + EventTypes.JoinRules, + EventTypes.RoomHistoryVisibility, + EventTypes.RoomEncryption, + EventTypes.Name, + EventTypes.Topic, + EventTypes.RoomAvatar, + EventTypes.CanonicalAlias, + ], + keyvalues={"room_id": room_id, "state_key": ""}, + retcols=["event_id"], + ), ) - event_ids = cast(List[str], [row["event_id"] for row in rows]) + event_ids = [row[0] for row in rows] txn.execute( """ @@ -676,7 +679,7 @@ class StatsStore(StateDeltasStore): order_by: Optional[str] = UserSortOrder.USER_ID.value, direction: Direction = Direction.FORWARDS, search_term: Optional[str] = None, - ) -> Tuple[List[JsonDict], int]: + ) -> Tuple[List[Tuple[str, Optional[str], int, int]], int]: """Function to retrieve a paginated list of users and their uploaded local media (size and number). This will return a json list of users and the total number of users matching the filter criteria. @@ -689,14 +692,19 @@ class StatsStore(StateDeltasStore): order_by: the sort order of the returned list direction: sort ascending or descending search_term: a string to filter user names by + Returns: - A list of user dicts and an integer representing the total number of - users that exist given this query + A tuple of: + A list of tuples of user information (the user ID, displayname, + total number of media, total length of media) and + + An integer representing the total number of users that exist + given this query """ def get_users_media_usage_paginate_txn( txn: LoggingTransaction, - ) -> Tuple[List[JsonDict], int]: + ) -> Tuple[List[Tuple[str, Optional[str], int, int]], int]: filters = [] args: list = [] @@ -770,7 +778,7 @@ class StatsStore(StateDeltasStore): args += [limit, start] txn.execute(sql, args) - users = self.db_pool.cursor_to_dict(txn) + users = cast(List[Tuple[str, Optional[str], int, int]], txn.fetchall()) return users, count diff --git a/synapse/storage/databases/main/stream.py b/synapse/storage/databases/main/stream.py index 5a3611c415..2225f8272d 100644 --- a/synapse/storage/databases/main/stream.py +++ b/synapse/storage/databases/main/stream.py @@ -266,7 +266,7 @@ def generate_next_token( # when we are going backwards so we subtract one from the # stream part. last_stream_ordering -= 1 - return RoomStreamToken(last_topo_ordering, last_stream_ordering) + return RoomStreamToken(topological=last_topo_ordering, stream=last_stream_ordering) def _make_generic_sql_bound( @@ -558,7 +558,7 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): if p > min_pos } - return RoomStreamToken(None, min_pos, immutabledict(positions)) + return RoomStreamToken(stream=min_pos, instance_map=immutabledict(positions)) async def get_room_events_stream_for_rooms( self, @@ -708,7 +708,7 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): ret.reverse() if rows: - key = RoomStreamToken(None, min(r.stream_ordering for r in rows)) + key = RoomStreamToken(stream=min(r.stream_ordering for r in rows)) else: # Assume we didn't get anything because there was nothing to # get. @@ -969,7 +969,7 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): topo = await self.db_pool.runInteraction( "_get_max_topological_txn", self._get_max_topological_txn, room_id ) - return RoomStreamToken(topo, stream_ordering) + return RoomStreamToken(topological=topo, stream=stream_ordering) @overload def get_stream_id_for_event_txn( @@ -1033,7 +1033,9 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): retcols=("stream_ordering", "topological_ordering"), desc="get_topological_token_for_event", ) - return RoomStreamToken(row["topological_ordering"], row["stream_ordering"]) + return RoomStreamToken( + topological=row["topological_ordering"], stream=row["stream_ordering"] + ) async def get_current_topological_token(self, room_id: str, stream_key: int) -> int: """Gets the topological token in a room after or at the given stream @@ -1076,7 +1078,7 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): """ row = await self.db_pool.execute( - "get_current_topological_token", None, sql, room_id, room_id, stream_key + "get_current_topological_token", sql, room_id, room_id, stream_key ) return row[0][0] if row else 0 @@ -1114,8 +1116,8 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): else: topo = None internal = event.internal_metadata - internal.before = RoomStreamToken(topo, stream - 1) - internal.after = RoomStreamToken(topo, stream) + internal.before = RoomStreamToken(topological=topo, stream=stream - 1) + internal.after = RoomStreamToken(topological=topo, stream=stream) internal.order = (int(topo) if topo else 0, int(stream)) async def get_events_around( @@ -1191,11 +1193,13 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): # Paginating backwards includes the event at the token, but paginating # forward doesn't. before_token = RoomStreamToken( - results["topological_ordering"] - 1, results["stream_ordering"] + topological=results["topological_ordering"] - 1, + stream=results["stream_ordering"], ) after_token = RoomStreamToken( - results["topological_ordering"], results["stream_ordering"] + topological=results["topological_ordering"], + stream=results["stream_ordering"], ) rows, start_token = self._paginate_room_events_txn( @@ -1612,3 +1616,49 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): retcol="instance_name", desc="get_name_from_instance_id", ) + + async def get_timeline_gaps( + self, + room_id: str, + from_token: Optional[RoomStreamToken], + to_token: RoomStreamToken, + ) -> Optional[RoomStreamToken]: + """Check if there is a gap, and return a token that marks the position + of the gap in the stream. + """ + + sql = """ + SELECT instance_name, stream_ordering + FROM timeline_gaps + WHERE room_id = ? AND ? < stream_ordering AND stream_ordering <= ? + ORDER BY stream_ordering + """ + + rows = await self.db_pool.execute( + "get_timeline_gaps", + sql, + room_id, + from_token.stream if from_token else 0, + to_token.get_max_stream_pos(), + ) + + if not rows: + return None + + positions = [ + PersistedEventPosition(instance_name, stream_ordering) + for instance_name, stream_ordering in rows + ] + if from_token: + positions = [p for p in positions if p.persisted_after(from_token)] + + positions = [p for p in positions if not p.persisted_after(to_token)] + + if positions: + # We return a stream token that ensures the event *at* the position + # of the gap is included (as the gap is *before* the persisted + # event). + last_position = positions[-1] + return RoomStreamToken(stream=last_position.stream - 1) + + return None diff --git a/synapse/storage/databases/main/tags.py b/synapse/storage/databases/main/tags.py index 61403a98cf..7deda7790e 100644 --- a/synapse/storage/databases/main/tags.py +++ b/synapse/storage/databases/main/tags.py @@ -45,14 +45,17 @@ class TagsWorkerStore(AccountDataWorkerStore): tag content. """ - rows = await self.db_pool.simple_select_list( - "room_tags", {"user_id": user_id}, ["room_id", "tag", "content"] + rows = cast( + List[Tuple[str, str, str]], + await self.db_pool.simple_select_list( + "room_tags", {"user_id": user_id}, ["room_id", "tag", "content"] + ), ) tags_by_room: Dict[str, Dict[str, JsonDict]] = {} - for row in rows: - room_tags = tags_by_room.setdefault(row["room_id"], {}) - room_tags[row["tag"]] = db_to_json(row["content"]) + for room_id, tag, content in rows: + room_tags = tags_by_room.setdefault(room_id, {}) + room_tags[tag] = db_to_json(content) return tags_by_room async def get_all_updated_tags( @@ -161,13 +164,16 @@ class TagsWorkerStore(AccountDataWorkerStore): Returns: A mapping of tags to tag content. """ - rows = await self.db_pool.simple_select_list( - table="room_tags", - keyvalues={"user_id": user_id, "room_id": room_id}, - retcols=("tag", "content"), - desc="get_tags_for_room", + rows = cast( + List[Tuple[str, str]], + await self.db_pool.simple_select_list( + table="room_tags", + keyvalues={"user_id": user_id, "room_id": room_id}, + retcols=("tag", "content"), + desc="get_tags_for_room", + ), ) - return {row["tag"]: db_to_json(row["content"]) for row in rows} + return {tag: db_to_json(content) for tag, content in rows} async def add_tag_to_room( self, user_id: str, room_id: str, tag: str, content: JsonDict diff --git a/synapse/storage/databases/main/task_scheduler.py b/synapse/storage/databases/main/task_scheduler.py index 5c5372a825..5555b53575 100644 --- a/synapse/storage/databases/main/task_scheduler.py +++ b/synapse/storage/databases/main/task_scheduler.py @@ -12,7 +12,7 @@ # See the License for the specific language governing permissions and # limitations under the License. -from typing import TYPE_CHECKING, Any, Dict, List, Optional +from typing import TYPE_CHECKING, Any, List, Optional, Tuple, cast from synapse.storage._base import SQLBaseStore, db_to_json from synapse.storage.database import ( @@ -27,6 +27,8 @@ from synapse.util import json_encoder if TYPE_CHECKING: from synapse.server import HomeServer +ScheduledTaskRow = Tuple[str, str, str, int, str, str, str, str] + class TaskSchedulerWorkerStore(SQLBaseStore): def __init__( @@ -38,13 +40,18 @@ class TaskSchedulerWorkerStore(SQLBaseStore): super().__init__(database, db_conn, hs) @staticmethod - def _convert_row_to_task(row: Dict[str, Any]) -> ScheduledTask: - row["status"] = TaskStatus(row["status"]) - if row["params"] is not None: - row["params"] = db_to_json(row["params"]) - if row["result"] is not None: - row["result"] = db_to_json(row["result"]) - return ScheduledTask(**row) + def _convert_row_to_task(row: ScheduledTaskRow) -> ScheduledTask: + task_id, action, status, timestamp, resource_id, params, result, error = row + return ScheduledTask( + id=task_id, + action=action, + status=TaskStatus(status), + timestamp=timestamp, + resource_id=resource_id, + params=db_to_json(params) if params is not None else None, + result=db_to_json(result) if result is not None else None, + error=error, + ) async def get_scheduled_tasks( self, @@ -68,7 +75,7 @@ class TaskSchedulerWorkerStore(SQLBaseStore): Returns: a list of `ScheduledTask`, ordered by increasing timestamps """ - def get_scheduled_tasks_txn(txn: LoggingTransaction) -> List[Dict[str, Any]]: + def get_scheduled_tasks_txn(txn: LoggingTransaction) -> List[ScheduledTaskRow]: clauses: List[str] = [] args: List[Any] = [] if resource_id: @@ -101,7 +108,7 @@ class TaskSchedulerWorkerStore(SQLBaseStore): args.append(limit) txn.execute(sql, args) - return self.db_pool.cursor_to_dict(txn) + return cast(List[ScheduledTaskRow], txn.fetchall()) rows = await self.db_pool.runInteraction( "get_scheduled_tasks", get_scheduled_tasks_txn @@ -193,7 +200,22 @@ class TaskSchedulerWorkerStore(SQLBaseStore): desc="get_scheduled_task", ) - return TaskSchedulerWorkerStore._convert_row_to_task(row) if row else None + return ( + TaskSchedulerWorkerStore._convert_row_to_task( + ( + row["id"], + row["action"], + row["status"], + row["timestamp"], + row["resource_id"], + row["params"], + row["result"], + row["error"], + ) + ) + if row + else None + ) async def delete_scheduled_task(self, id: str) -> None: """Delete a specific task from its id. diff --git a/synapse/storage/databases/main/transactions.py b/synapse/storage/databases/main/transactions.py index 8f70eff809..fecddb4144 100644 --- a/synapse/storage/databases/main/transactions.py +++ b/synapse/storage/databases/main/transactions.py @@ -211,18 +211,28 @@ class TransactionWorkerStore(CacheInvalidationWorkerStore): async def get_destination_retry_timings_batch( self, destinations: StrCollection ) -> Mapping[str, Optional[DestinationRetryTimings]]: - rows = await self.db_pool.simple_select_many_batch( - table="destinations", - iterable=destinations, - column="destination", - retcols=("destination", "failure_ts", "retry_last_ts", "retry_interval"), - desc="get_destination_retry_timings_batch", + rows = cast( + List[Tuple[str, Optional[int], Optional[int], Optional[int]]], + await self.db_pool.simple_select_many_batch( + table="destinations", + iterable=destinations, + column="destination", + retcols=( + "destination", + "failure_ts", + "retry_last_ts", + "retry_interval", + ), + desc="get_destination_retry_timings_batch", + ), ) return { - row.pop("destination"): DestinationRetryTimings(**row) - for row in rows - if row["retry_last_ts"] and row["failure_ts"] and row["retry_interval"] + destination: DestinationRetryTimings( + failure_ts, retry_last_ts, retry_interval + ) + for destination, failure_ts, retry_last_ts, retry_interval in rows + if retry_last_ts and failure_ts and retry_interval } async def set_destination_retry_timings( @@ -468,7 +478,10 @@ class TransactionWorkerStore(CacheInvalidationWorkerStore): destination: Optional[str] = None, order_by: str = DestinationSortOrder.DESTINATION.value, direction: Direction = Direction.FORWARDS, - ) -> Tuple[List[JsonDict], int]: + ) -> Tuple[ + List[Tuple[str, Optional[int], Optional[int], Optional[int], Optional[int]]], + int, + ]: """Function to retrieve a paginated list of destinations. This will return a json list of destinations and the total number of destinations matching the filter criteria. @@ -480,13 +493,23 @@ class TransactionWorkerStore(CacheInvalidationWorkerStore): order_by: the sort order of the returned list direction: sort ascending or descending Returns: - A tuple of a list of mappings from destination to information + A tuple of a list of tuples of destination information: + * destination + * retry_last_ts + * retry_interval + * failure_ts + * last_successful_stream_ordering and a count of total destinations. """ def get_destinations_paginate_txn( txn: LoggingTransaction, - ) -> Tuple[List[JsonDict], int]: + ) -> Tuple[ + List[ + Tuple[str, Optional[int], Optional[int], Optional[int], Optional[int]] + ], + int, + ]: order_by_column = DestinationSortOrder(order_by).value if direction == Direction.BACKWARDS: @@ -513,7 +536,14 @@ class TransactionWorkerStore(CacheInvalidationWorkerStore): LIMIT ? OFFSET ? """ txn.execute(sql, args + [limit, start]) - destinations = self.db_pool.cursor_to_dict(txn) + destinations = cast( + List[ + Tuple[ + str, Optional[int], Optional[int], Optional[int], Optional[int] + ] + ], + txn.fetchall(), + ) return destinations, count return await self.db_pool.runInteraction( @@ -526,7 +556,7 @@ class TransactionWorkerStore(CacheInvalidationWorkerStore): start: int, limit: int, direction: Direction = Direction.FORWARDS, - ) -> Tuple[List[JsonDict], int]: + ) -> Tuple[List[Tuple[str, int]], int]: """Function to retrieve a paginated list of destination's rooms. This will return a json list of rooms and the total number of rooms. @@ -537,12 +567,14 @@ class TransactionWorkerStore(CacheInvalidationWorkerStore): limit: number of rows to retrieve direction: sort ascending or descending by room_id Returns: - A tuple of a dict of rooms and a count of total rooms. + A tuple of a list of room tuples and a count of total rooms. + + Each room tuple is room_id, stream_ordering. """ def get_destination_rooms_paginate_txn( txn: LoggingTransaction, - ) -> Tuple[List[JsonDict], int]: + ) -> Tuple[List[Tuple[str, int]], int]: if direction == Direction.BACKWARDS: order = "DESC" else: @@ -556,14 +588,17 @@ class TransactionWorkerStore(CacheInvalidationWorkerStore): txn.execute(sql, [destination]) count = cast(Tuple[int], txn.fetchone())[0] - rooms = self.db_pool.simple_select_list_paginate_txn( - txn=txn, - table="destination_rooms", - orderby="room_id", - start=start, - limit=limit, - retcols=("room_id", "stream_ordering"), - order_direction=order, + rooms = cast( + List[Tuple[str, int]], + self.db_pool.simple_select_list_paginate_txn( + txn=txn, + table="destination_rooms", + orderby="room_id", + start=start, + limit=limit, + retcols=("room_id", "stream_ordering"), + order_direction=order, + ), ) return rooms, count diff --git a/synapse/storage/databases/main/ui_auth.py b/synapse/storage/databases/main/ui_auth.py index f38bedbbcd..8ab7c42c4a 100644 --- a/synapse/storage/databases/main/ui_auth.py +++ b/synapse/storage/databases/main/ui_auth.py @@ -169,13 +169,17 @@ class UIAuthWorkerStore(SQLBaseStore): that auth-type. """ results = {} - for row in await self.db_pool.simple_select_list( - table="ui_auth_sessions_credentials", - keyvalues={"session_id": session_id}, - retcols=("stage_type", "result"), - desc="get_completed_ui_auth_stages", - ): - results[row["stage_type"]] = db_to_json(row["result"]) + rows = cast( + List[Tuple[str, str]], + await self.db_pool.simple_select_list( + table="ui_auth_sessions_credentials", + keyvalues={"session_id": session_id}, + retcols=("stage_type", "result"), + desc="get_completed_ui_auth_stages", + ), + ) + for stage_type, result in rows: + results[stage_type] = db_to_json(result) return results @@ -295,13 +299,15 @@ class UIAuthWorkerStore(SQLBaseStore): Returns: List of user_agent/ip pairs """ - rows = await self.db_pool.simple_select_list( - table="ui_auth_sessions_ips", - keyvalues={"session_id": session_id}, - retcols=("user_agent", "ip"), - desc="get_user_agents_ips_to_ui_auth_session", + return cast( + List[Tuple[str, str]], + await self.db_pool.simple_select_list( + table="ui_auth_sessions_ips", + keyvalues={"session_id": session_id}, + retcols=("user_agent", "ip"), + desc="get_user_agents_ips_to_ui_auth_session", + ), ) - return [(row["user_agent"], row["ip"]) for row in rows] async def delete_old_ui_auth_sessions(self, expiration_time: int) -> None: """ @@ -337,13 +343,16 @@ class UIAuthWorkerStore(SQLBaseStore): # If a registration token was used, decrement the pending counter # before deleting the session. - rows = self.db_pool.simple_select_many_txn( - txn, - table="ui_auth_sessions_credentials", - column="session_id", - iterable=session_ids, - keyvalues={"stage_type": LoginType.REGISTRATION_TOKEN}, - retcols=["result"], + rows = cast( + List[Tuple[str]], + self.db_pool.simple_select_many_txn( + txn, + table="ui_auth_sessions_credentials", + column="session_id", + iterable=session_ids, + keyvalues={"stage_type": LoginType.REGISTRATION_TOKEN}, + retcols=["result"], + ), ) # Get the tokens used and how much pending needs to be decremented by. @@ -353,23 +362,25 @@ class UIAuthWorkerStore(SQLBaseStore): # registration token stage for that session will be True. # If a token was used to authenticate, but registration was # never completed, the result will be the token used. - token = db_to_json(r["result"]) + token = db_to_json(r[0]) if isinstance(token, str): token_counts[token] = token_counts.get(token, 0) + 1 # Update the `pending` counters. if len(token_counts) > 0: - token_rows = self.db_pool.simple_select_many_txn( - txn, - table="registration_tokens", - column="token", - iterable=list(token_counts.keys()), - keyvalues={}, - retcols=["token", "pending"], + token_rows = cast( + List[Tuple[str, int]], + self.db_pool.simple_select_many_txn( + txn, + table="registration_tokens", + column="token", + iterable=list(token_counts.keys()), + keyvalues={}, + retcols=["token", "pending"], + ), ) - for token_row in token_rows: - token = token_row["token"] - new_pending = token_row["pending"] - token_counts[token] + for token, pending in token_rows: + new_pending = pending - token_counts[token] self.db_pool.simple_update_one_txn( txn, table="registration_tokens", diff --git a/synapse/storage/databases/main/user_directory.py b/synapse/storage/databases/main/user_directory.py index ed41e52201..d4b86ed7a6 100644 --- a/synapse/storage/databases/main/user_directory.py +++ b/synapse/storage/databases/main/user_directory.py @@ -415,25 +415,24 @@ class UserDirectoryBackgroundUpdateStore(StateDeltasStore): ) # Next fetch their profiles. Note that not all users have profiles. - profile_rows = self.db_pool.simple_select_many_txn( - txn, - table="profiles", - column="full_user_id", - iterable=list(users_to_insert), - retcols=( - "full_user_id", - "displayname", - "avatar_url", + profile_rows = cast( + List[Tuple[str, Optional[str], Optional[str]]], + self.db_pool.simple_select_many_txn( + txn, + table="profiles", + column="full_user_id", + iterable=list(users_to_insert), + retcols=( + "full_user_id", + "displayname", + "avatar_url", + ), + keyvalues={}, ), - keyvalues={}, ) profiles = { - row["full_user_id"]: _UserDirProfile( - row["full_user_id"], - row["displayname"], - row["avatar_url"], - ) - for row in profile_rows + full_user_id: _UserDirProfile(full_user_id, displayname, avatar_url) + for full_user_id, displayname, avatar_url in profile_rows } profiles_to_insert = [ @@ -522,18 +521,21 @@ class UserDirectoryBackgroundUpdateStore(StateDeltasStore): and not self.get_if_app_services_interested_in_user(user) # type: ignore[attr-defined] ] - rows = self.db_pool.simple_select_many_txn( - txn, - table="users", - column="name", - iterable=users, - keyvalues={ - "deactivated": 0, - }, - retcols=("name", "user_type"), + rows = cast( + List[Tuple[str, Optional[str]]], + self.db_pool.simple_select_many_txn( + txn, + table="users", + column="name", + iterable=users, + keyvalues={ + "deactivated": 0, + }, + retcols=("name", "user_type"), + ), ) - return [row["name"] for row in rows if row["user_type"] != UserTypes.SUPPORT] + return [name for name, user_type in rows if user_type != UserTypes.SUPPORT] async def is_room_world_readable_or_publicly_joinable(self, room_id: str) -> bool: """Check if the room is either world_readable or publically joinable""" @@ -1178,15 +1180,19 @@ class UserDirectoryStore(UserDirectoryBackgroundUpdateStore): raise Exception("Unrecognized database engine") results = cast( - List[UserProfile], - await self.db_pool.execute( - "search_user_dir", self.db_pool.cursor_to_dict, sql, *args - ), + List[Tuple[str, Optional[str], Optional[str]]], + await self.db_pool.execute("search_user_dir", sql, *args), ) limited = len(results) > limit - return {"limited": limited, "results": results[0:limit]} + return { + "limited": limited, + "results": [ + {"user_id": r[0], "display_name": r[1], "avatar_url": r[2]} + for r in results[0:limit] + ], + } def _filter_text_for_index(text: str) -> str: diff --git a/synapse/storage/databases/main/user_erasure_store.py b/synapse/storage/databases/main/user_erasure_store.py index 06fcbe5e54..8bd58c6e3d 100644 --- a/synapse/storage/databases/main/user_erasure_store.py +++ b/synapse/storage/databases/main/user_erasure_store.py @@ -12,7 +12,7 @@ # See the License for the specific language governing permissions and # limitations under the License. -from typing import Iterable, Mapping +from typing import Iterable, List, Mapping, Tuple, cast from synapse.storage.database import LoggingTransaction from synapse.storage.databases.main import CacheInvalidationWorkerStore @@ -50,14 +50,17 @@ class UserErasureWorkerStore(CacheInvalidationWorkerStore): Returns: for each user, whether the user has requested erasure. """ - rows = await self.db_pool.simple_select_many_batch( - table="erased_users", - column="user_id", - iterable=user_ids, - retcols=("user_id",), - desc="are_users_erased", + rows = cast( + List[Tuple[str]], + await self.db_pool.simple_select_many_batch( + table="erased_users", + column="user_id", + iterable=user_ids, + retcols=("user_id",), + desc="are_users_erased", + ), ) - erased_users = {row["user_id"] for row in rows} + erased_users = {row[0] for row in rows} return {u: u in erased_users for u in user_ids} diff --git a/synapse/storage/databases/state/bg_updates.py b/synapse/storage/databases/state/bg_updates.py index d2e942cbd3..2c3151526d 100644 --- a/synapse/storage/databases/state/bg_updates.py +++ b/synapse/storage/databases/state/bg_updates.py @@ -359,7 +359,6 @@ class StateBackgroundUpdateStore(StateGroupBackgroundUpdateStore): if max_group is None: rows = await self.db_pool.execute( "_background_deduplicate_state", - None, "SELECT coalesce(max(id), 0) FROM state_groups", ) max_group = rows[0][0] diff --git a/synapse/storage/databases/state/store.py b/synapse/storage/databases/state/store.py index 6984d11352..182e429174 100644 --- a/synapse/storage/databases/state/store.py +++ b/synapse/storage/databases/state/store.py @@ -13,7 +13,17 @@ # limitations under the License. import logging -from typing import TYPE_CHECKING, Collection, Dict, Iterable, List, Optional, Set, Tuple +from typing import ( + TYPE_CHECKING, + Collection, + Dict, + Iterable, + List, + Optional, + Set, + Tuple, + cast, +) import attr @@ -144,16 +154,22 @@ class StateGroupDataStore(StateBackgroundUpdateStore, SQLBaseStore): if not prev_group: return _GetStateGroupDelta(None, None) - delta_ids = self.db_pool.simple_select_list_txn( - txn, - table="state_groups_state", - keyvalues={"state_group": state_group}, - retcols=("type", "state_key", "event_id"), + delta_ids = cast( + List[Tuple[str, str, str]], + self.db_pool.simple_select_list_txn( + txn, + table="state_groups_state", + keyvalues={"state_group": state_group}, + retcols=("type", "state_key", "event_id"), + ), ) return _GetStateGroupDelta( prev_group, - {(row["type"], row["state_key"]): row["event_id"] for row in delta_ids}, + { + (event_type, state_key): event_id + for event_type, state_key, event_id in delta_ids + }, ) return await self.db_pool.runInteraction( @@ -730,19 +746,22 @@ class StateGroupDataStore(StateBackgroundUpdateStore, SQLBaseStore): "[purge] found %i state groups to delete", len(state_groups_to_delete) ) - rows = self.db_pool.simple_select_many_txn( - txn, - table="state_group_edges", - column="prev_state_group", - iterable=state_groups_to_delete, - keyvalues={}, - retcols=("state_group",), + rows = cast( + List[Tuple[int]], + self.db_pool.simple_select_many_txn( + txn, + table="state_group_edges", + column="prev_state_group", + iterable=state_groups_to_delete, + keyvalues={}, + retcols=("state_group",), + ), ) remaining_state_groups = { - row["state_group"] - for row in rows - if row["state_group"] not in state_groups_to_delete + state_group + for state_group, in rows + if state_group not in state_groups_to_delete } logger.info( @@ -799,16 +818,19 @@ class StateGroupDataStore(StateBackgroundUpdateStore, SQLBaseStore): A mapping from state group to previous state group. """ - rows = await self.db_pool.simple_select_many_batch( - table="state_group_edges", - column="prev_state_group", - iterable=state_groups, - keyvalues={}, - retcols=("prev_state_group", "state_group"), - desc="get_previous_state_groups", + rows = cast( + List[Tuple[int, int]], + await self.db_pool.simple_select_many_batch( + table="state_group_edges", + column="prev_state_group", + iterable=state_groups, + keyvalues={}, + retcols=("state_group", "prev_state_group"), + desc="get_previous_state_groups", + ), ) - return {row["state_group"]: row["prev_state_group"] for row in rows} + return dict(rows) async def purge_room_state( self, room_id: str, state_groups_to_delete: Collection[int] diff --git a/synapse/storage/roommember.py b/synapse/storage/roommember.py index 2500381b7b..cbfb32014c 100644 --- a/synapse/storage/roommember.py +++ b/synapse/storage/roommember.py @@ -45,6 +45,7 @@ class ProfileInfo: display_name: Optional[str] +# TODO This is used as a cached value and is mutable. @attr.s(slots=True, frozen=True, weakref_slot=False, auto_attribs=True) class MemberSummary: # A truncated list of (user_id, event_id) tuples for users of a given diff --git a/synapse/storage/schema/__init__.py b/synapse/storage/schema/__init__.py index 5b50bd66bc..158b528dce 100644 --- a/synapse/storage/schema/__init__.py +++ b/synapse/storage/schema/__init__.py @@ -12,7 +12,7 @@ # See the License for the specific language governing permissions and # limitations under the License. -SCHEMA_VERSION = 82 # remember to update the list below when updating +SCHEMA_VERSION = 83 # remember to update the list below when updating """Represents the expectations made by the codebase about the database schema This should be incremented whenever the codebase changes its requirements on the @@ -121,6 +121,9 @@ Changes in SCHEMA_VERSION = 81 Changes in SCHEMA_VERSION = 82 - The insertion_events, insertion_event_extremities, insertion_event_edges, and batch_events tables are no longer purged in preparation for their removal. + +Changes in SCHEMA_VERSION = 83 + - The event_txn_id is no longer used. """ diff --git a/synapse/storage/schema/main/delta/82/04_add_indices_for_purging_rooms.sql b/synapse/storage/schema/main/delta/82/04_add_indices_for_purging_rooms.sql new file mode 100644 index 0000000000..fc948166e6 --- /dev/null +++ b/synapse/storage/schema/main/delta/82/04_add_indices_for_purging_rooms.sql @@ -0,0 +1,20 @@ +/* Copyright 2023 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. + */ + +INSERT INTO background_updates (ordering, update_name, progress_json) VALUES + (8204, 'e2e_room_keys_index_room_id', '{}'); + +INSERT INTO background_updates (ordering, update_name, progress_json) VALUES + (8204, 'room_account_data_index_room_id', '{}'); diff --git a/synapse/storage/schema/main/delta/82/05gaps.sql b/synapse/storage/schema/main/delta/82/05gaps.sql new file mode 100644 index 0000000000..6813b488ca --- /dev/null +++ b/synapse/storage/schema/main/delta/82/05gaps.sql @@ -0,0 +1,25 @@ +/* Copyright 2023 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. + */ + +-- Records when we see a "gap in the timeline", due to missing events over +-- federation. We record this so that we can tell clients there is a gap (by +-- marking the timeline section of a sync request as limited). +CREATE TABLE IF NOT EXISTS timeline_gaps ( + room_id TEXT NOT NULL, + instance_name TEXT NOT NULL, + stream_ordering BIGINT NOT NULL +); + +CREATE INDEX timeline_gaps_room_id ON timeline_gaps(room_id, stream_ordering); diff --git a/synapse/storage/schema/main/delta/83/03_instance_name_receipts.sql.sqlite b/synapse/storage/schema/main/delta/83/03_instance_name_receipts.sql.sqlite new file mode 100644 index 0000000000..6c7ad0fd37 --- /dev/null +++ b/synapse/storage/schema/main/delta/83/03_instance_name_receipts.sql.sqlite @@ -0,0 +1,17 @@ +/* Copyright 2023 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. + */ + +-- This already exists on Postgres. +ALTER TABLE receipts_linearized ADD COLUMN instance_name TEXT; diff --git a/synapse/storage/util/id_generators.py b/synapse/storage/util/id_generators.py index aa4fa40c9c..52d708ad17 100644 --- a/synapse/storage/util/id_generators.py +++ b/synapse/storage/util/id_generators.py @@ -134,6 +134,15 @@ class AbstractStreamIdGenerator(metaclass=abc.ABCMeta): raise NotImplementedError() @abc.abstractmethod + def get_minimal_local_current_token(self) -> int: + """Tries to return a minimal current token for the local instance, + i.e. for writers this would be the last successful write. + + If local instance is not a writer (or has written yet) then falls back + to returning the normal "current token". + """ + + @abc.abstractmethod def get_next(self) -> AsyncContextManager[int]: """ Usage: @@ -312,6 +321,9 @@ class StreamIdGenerator(AbstractStreamIdGenerator): def get_current_token_for_writer(self, instance_name: str) -> int: return self.get_current_token() + def get_minimal_local_current_token(self) -> int: + return self.get_current_token() + class MultiWriterIdGenerator(AbstractStreamIdGenerator): """Generates and tracks stream IDs for a stream with multiple writers. @@ -408,6 +420,11 @@ class MultiWriterIdGenerator(AbstractStreamIdGenerator): # The maximum stream ID that we have seen been allocated across any writer. self._max_seen_allocated_stream_id = 1 + # The maximum position of the local instance. This can be higher than + # the corresponding position in `current_positions` table when there are + # no active writes in progress. + self._max_position_of_local_instance = self._max_seen_allocated_stream_id + self._sequence_gen = PostgresSequenceGenerator(sequence_name) # We check that the table and sequence haven't diverged. @@ -427,6 +444,16 @@ class MultiWriterIdGenerator(AbstractStreamIdGenerator): self._current_positions.values(), default=1 ) + # For the case where `stream_positions` is not up to date, + # `_persisted_upto_position` may be higher. + self._max_seen_allocated_stream_id = max( + self._max_seen_allocated_stream_id, self._persisted_upto_position + ) + + # Bump our local maximum position now that we've loaded things from the + # DB. + self._max_position_of_local_instance = self._max_seen_allocated_stream_id + if not writers: # If there have been no explicit writers given then any instance can # write to the stream. In which case, let's pre-seed our own @@ -545,6 +572,14 @@ class MultiWriterIdGenerator(AbstractStreamIdGenerator): if instance == self._instance_name: self._current_positions[instance] = stream_id + if self._writers: + # If we have explicit writers then make sure that each instance has + # a position. + for writer in self._writers: + self._current_positions.setdefault( + writer, self._persisted_upto_position + ) + cur.close() def _load_next_id_txn(self, txn: Cursor) -> int: @@ -688,6 +723,9 @@ class MultiWriterIdGenerator(AbstractStreamIdGenerator): if new_cur: curr = self._current_positions.get(self._instance_name, 0) self._current_positions[self._instance_name] = max(curr, new_cur) + self._max_position_of_local_instance = max( + curr, new_cur, self._max_position_of_local_instance + ) self._add_persisted_position(next_id) @@ -702,10 +740,26 @@ class MultiWriterIdGenerator(AbstractStreamIdGenerator): # persisted up to position. This stops Synapse from doing a full table # scan when a new writer announces itself over replication. with self._lock: - return self._return_factor * self._current_positions.get( + if self._instance_name == instance_name: + return self._return_factor * self._max_position_of_local_instance + + pos = self._current_positions.get( instance_name, self._persisted_upto_position ) + # We want to return the maximum "current token" that we can for a + # writer, this helps ensure that streams progress as fast as + # possible. + pos = max(pos, self._persisted_upto_position) + + return self._return_factor * pos + + def get_minimal_local_current_token(self) -> int: + with self._lock: + return self._return_factor * self._current_positions.get( + self._instance_name, self._persisted_upto_position + ) + def get_positions(self) -> Dict[str, int]: """Get a copy of the current positon map. @@ -774,6 +828,18 @@ class MultiWriterIdGenerator(AbstractStreamIdGenerator): self._persisted_upto_position = max(min_curr, self._persisted_upto_position) + # Advance our local max position. + self._max_position_of_local_instance = max( + self._max_position_of_local_instance, self._persisted_upto_position + ) + + if not self._unfinished_ids and not self._in_flight_fetches: + # If we don't have anything in flight, it's safe to advance to the + # max seen stream ID. + self._max_position_of_local_instance = max( + self._max_seen_allocated_stream_id, self._max_position_of_local_instance + ) + # We now iterate through the seen positions, discarding those that are # less than the current min positions, and incrementing the min position # if its exactly one greater. diff --git a/synapse/streams/events.py b/synapse/streams/events.py index d7084d2358..d0bb83b184 100644 --- a/synapse/streams/events.py +++ b/synapse/streams/events.py @@ -12,7 +12,7 @@ # See the License for the specific language governing permissions and # limitations under the License. -from typing import TYPE_CHECKING, Iterator, Tuple +from typing import TYPE_CHECKING, Sequence, Tuple import attr @@ -23,7 +23,7 @@ from synapse.handlers.room import RoomEventSource from synapse.handlers.typing import TypingNotificationEventSource from synapse.logging.opentracing import trace from synapse.streams import EventSource -from synapse.types import StreamToken +from synapse.types import MultiWriterStreamToken, StreamKeyType, StreamToken if TYPE_CHECKING: from synapse.server import HomeServer @@ -37,9 +37,14 @@ class _EventSourcesInner: receipt: ReceiptEventSource account_data: AccountDataEventSource - def get_sources(self) -> Iterator[Tuple[str, EventSource]]: - for attribute in attr.fields(_EventSourcesInner): - yield attribute.name, getattr(self, attribute.name) + def get_sources(self) -> Sequence[Tuple[StreamKeyType, EventSource]]: + return [ + (StreamKeyType.ROOM, self.room), + (StreamKeyType.PRESENCE, self.presence), + (StreamKeyType.TYPING, self.typing), + (StreamKeyType.RECEIPT, self.receipt), + (StreamKeyType.ACCOUNT_DATA, self.account_data), + ] class EventSources: @@ -106,7 +111,7 @@ class EventSources: room_key=await self.sources.room.get_current_key_for_room(room_id), presence_key=0, typing_key=0, - receipt_key=0, + receipt_key=MultiWriterStreamToken(stream=0), account_data_key=0, push_rules_key=0, to_device_key=0, diff --git a/synapse/types/__init__.py b/synapse/types/__init__.py index 76b0e3e694..4c5b26ad93 100644 --- a/synapse/types/__init__.py +++ b/synapse/types/__init__.py @@ -22,8 +22,8 @@ from typing import ( Any, ClassVar, Dict, - Final, List, + Literal, Mapping, Match, MutableMapping, @@ -34,6 +34,7 @@ from typing import ( Type, TypeVar, Union, + overload, ) import attr @@ -60,6 +61,8 @@ from synapse.util.cancellation import cancellable from synapse.util.stringutils import parse_and_validate_server_name if TYPE_CHECKING: + from typing_extensions import Self + from synapse.appservice.api import ApplicationService from synapse.storage.databases.main import DataStore, PurgeEventsStore from synapse.storage.databases.main.appservice import ApplicationServiceWorkerStore @@ -436,7 +439,78 @@ def map_username_to_mxid_localpart( @attr.s(frozen=True, slots=True, order=False) -class RoomStreamToken: +class AbstractMultiWriterStreamToken(metaclass=abc.ABCMeta): + """An abstract stream token class for streams that supports multiple + writers. + + This works by keeping track of the stream position of each writer, + represented by a default `stream` attribute and a map of instance name to + stream position of any writers that are ahead of the default stream + position. + """ + + stream: int = attr.ib(validator=attr.validators.instance_of(int), kw_only=True) + + instance_map: "immutabledict[str, int]" = attr.ib( + factory=immutabledict, + 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(immutabledict), + ), + kw_only=True, + ) + + @classmethod + @abc.abstractmethod + async def parse(cls, store: "DataStore", string: str) -> "Self": + """Parse the string representation of the token.""" + ... + + @abc.abstractmethod + async def to_string(self, store: "DataStore") -> str: + """Serialize the token into its string representation.""" + ... + + def copy_and_advance(self, other: "Self") -> "Self": + """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. + """ + + 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 attr.evolve( + self, stream=max_stream, instance_map=immutabledict(instance_map) + ) + + 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) + + def get_stream_pos_for_instance(self, instance_name: str) -> int: + """Get the stream position that the given writer was at at this token.""" + + # 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) + + +@attr.s(frozen=True, slots=True, order=False) +class RoomStreamToken(AbstractMultiWriterStreamToken): """Tokens are positions between events. The token "s1" comes after event 1. s0 s1 @@ -513,16 +587,8 @@ class RoomStreamToken: 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: "immutabledict[str, int]" = attr.ib( - factory=immutabledict, - 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(immutabledict), - ), + kw_only=True, + default=None, ) def __attrs_post_init__(self) -> None: @@ -582,17 +648,7 @@ class RoomStreamToken: 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, immutabledict(instance_map)) + return super().copy_and_advance(other) def as_historical_tuple(self) -> Tuple[int, int]: """Returns a tuple of `(topological, stream)` for historical tokens. @@ -618,16 +674,6 @@ class RoomStreamToken: # 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) @@ -649,20 +695,104 @@ class RoomStreamToken: return "s%d" % (self.stream,) -class StreamKeyType: +@attr.s(frozen=True, slots=True, order=False) +class MultiWriterStreamToken(AbstractMultiWriterStreamToken): + """A basic stream token class for streams that supports multiple writers.""" + + @classmethod + async def parse(cls, store: "DataStore", string: str) -> "MultiWriterStreamToken": + try: + if string[0].isdigit(): + return cls(stream=int(string)) + 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) + instance_map[instance_name] = pos + + return cls( + stream=stream, + instance_map=immutabledict(instance_map), + ) + except CancelledError: + raise + except Exception: + pass + raise SynapseError(400, "Invalid stream token %r" % (string,)) + + async def to_string(self, store: "DataStore") -> str: + if self.instance_map: + entries = [] + for name, pos in self.instance_map.items(): + if pos <= self.stream: + # Ignore instances who are below the minimum stream position + # (we might know they've advanced without seeing a recent + # write from them). + continue + + 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 str(self.stream) + + @staticmethod + def is_stream_position_in_range( + low: Optional["AbstractMultiWriterStreamToken"], + high: Optional["AbstractMultiWriterStreamToken"], + instance_name: Optional[str], + pos: int, + ) -> bool: + """Checks if a given persisted position is between the two given tokens. + + If `instance_name` is None then the row was persisted before multi + writer support. + """ + + if low: + if instance_name: + low_stream = low.instance_map.get(instance_name, low.stream) + else: + low_stream = low.stream + + if pos <= low_stream: + return False + + if high: + if instance_name: + high_stream = high.instance_map.get(instance_name, high.stream) + else: + high_stream = high.stream + + if high_stream < pos: + return False + + return True + + +class StreamKeyType(Enum): """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" + ROOM = "room_key" + PRESENCE = "presence_key" + TYPING = "typing_key" + RECEIPT = "receipt_key" + ACCOUNT_DATA = "account_data_key" + PUSH_RULES = "push_rules_key" + TO_DEVICE = "to_device_key" + DEVICE_LIST = "device_list_key" UN_PARTIAL_STATED_ROOMS = "un_partial_stated_rooms_key" @@ -730,7 +860,9 @@ class StreamToken: ) presence_key: int typing_key: int - receipt_key: int + receipt_key: MultiWriterStreamToken = attr.ib( + validator=attr.validators.instance_of(MultiWriterStreamToken) + ) account_data_key: int push_rules_key: int to_device_key: int @@ -753,8 +885,31 @@ class StreamToken: while len(keys) < len(attr.fields(cls)): # i.e. old token from before receipt_key keys.append("0") + + ( + room_key, + presence_key, + typing_key, + receipt_key, + account_data_key, + push_rules_key, + to_device_key, + device_list_key, + groups_key, + un_partial_stated_rooms_key, + ) = keys + return cls( - await RoomStreamToken.parse(store, keys[0]), *(int(k) for k in keys[1:]) + room_key=await RoomStreamToken.parse(store, room_key), + presence_key=int(presence_key), + typing_key=int(typing_key), + receipt_key=await MultiWriterStreamToken.parse(store, receipt_key), + account_data_key=int(account_data_key), + push_rules_key=int(push_rules_key), + to_device_key=int(to_device_key), + device_list_key=int(device_list_key), + groups_key=int(groups_key), + un_partial_stated_rooms_key=int(un_partial_stated_rooms_key), ) except CancelledError: raise @@ -767,7 +922,7 @@ class StreamToken: await self.room_key.to_string(store), str(self.presence_key), str(self.typing_key), - str(self.receipt_key), + await self.receipt_key.to_string(store), str(self.account_data_key), str(self.push_rules_key), str(self.to_device_key), @@ -784,7 +939,7 @@ class StreamToken: def room_stream_id(self) -> int: return self.room_key.stream - def copy_and_advance(self, key: str, new_value: Any) -> "StreamToken": + def copy_and_advance(self, key: StreamKeyType, 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. @@ -795,37 +950,84 @@ class StreamToken: StreamKeyType.ROOM, self.room_key.copy_and_advance(new_value) ) return new_token + elif key == StreamKeyType.RECEIPT: + new_token = self.copy_and_replace( + StreamKeyType.RECEIPT, self.receipt_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)) + new_id = new_token.get_field(key) + old_id = self.get_field(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}) + def copy_and_replace(self, key: StreamKeyType, new_value: Any) -> "StreamToken": + return attr.evolve(self, **{key.value: new_value}) + + @overload + def get_field(self, key: Literal[StreamKeyType.ROOM]) -> RoomStreamToken: + ... + + @overload + def get_field(self, key: Literal[StreamKeyType.RECEIPT]) -> MultiWriterStreamToken: + ... + + @overload + def get_field( + self, + key: Literal[ + StreamKeyType.ACCOUNT_DATA, + StreamKeyType.DEVICE_LIST, + StreamKeyType.PRESENCE, + StreamKeyType.PUSH_RULES, + StreamKeyType.TO_DEVICE, + StreamKeyType.TYPING, + StreamKeyType.UN_PARTIAL_STATED_ROOMS, + ], + ) -> int: + ... + + @overload + def get_field( + self, key: StreamKeyType + ) -> Union[int, RoomStreamToken, MultiWriterStreamToken]: + ... + + def get_field( + self, key: StreamKeyType + ) -> Union[int, RoomStreamToken, MultiWriterStreamToken]: + """Returns the stream ID for the given key.""" + return getattr(self, key.value) + + +StreamToken.START = StreamToken( + RoomStreamToken(stream=0), 0, 0, MultiWriterStreamToken(stream=0), 0, 0, 0, 0, 0, 0 +) + + +@attr.s(slots=True, frozen=True, auto_attribs=True) +class PersistedPosition: + """Position of a newly persisted row with instance that persisted it.""" + instance_name: str + stream: int -StreamToken.START = StreamToken(RoomStreamToken(None, 0), 0, 0, 0, 0, 0, 0, 0, 0, 0) + def persisted_after(self, token: AbstractMultiWriterStreamToken) -> bool: + return token.get_stream_pos_for_instance(self.instance_name) < self.stream @attr.s(slots=True, frozen=True, auto_attribs=True) -class PersistedEventPosition: +class PersistedEventPosition(PersistedPosition): """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 @@ -836,7 +1038,7 @@ class PersistedEventPosition: """ # Doing the naive thing satisfies the desired properties described in # the docstring. - return RoomStreamToken(None, self.stream) + return RoomStreamToken(stream=self.stream) @attr.s(slots=True, frozen=True, auto_attribs=True) diff --git a/synapse/util/__init__.py b/synapse/util/__init__.py index 9f3b8741c1..8d9df352b2 100644 --- a/synapse/util/__init__.py +++ b/synapse/util/__init__.py @@ -93,7 +93,7 @@ class Clock: _reactor: IReactorTime = attr.ib() - @defer.inlineCallbacks # type: ignore[arg-type] # Issue in Twisted's type annotations + @defer.inlineCallbacks def sleep(self, seconds: float) -> "Generator[Deferred[float], Any, Any]": d: defer.Deferred[float] = defer.Deferred() with context.PreserveLoggingContext(): diff --git a/synapse/util/async_helpers.py b/synapse/util/async_helpers.py index 0cbeb0c365..8a55e4e41d 100644 --- a/synapse/util/async_helpers.py +++ b/synapse/util/async_helpers.py @@ -345,6 +345,7 @@ async def yieldable_gather_results_delaying_cancellation( T1 = TypeVar("T1") T2 = TypeVar("T2") T3 = TypeVar("T3") +T4 = TypeVar("T4") @overload @@ -380,6 +381,19 @@ def gather_results( ... +@overload +def gather_results( + deferredList: Tuple[ + "defer.Deferred[T1]", + "defer.Deferred[T2]", + "defer.Deferred[T3]", + "defer.Deferred[T4]", + ], + consumeErrors: bool = ..., +) -> "defer.Deferred[Tuple[T1, T2, T3, T4]]": + ... + + def gather_results( # type: ignore[misc] deferredList: Tuple["defer.Deferred[T1]", ...], consumeErrors: bool = False, diff --git a/synapse/util/caches/descriptors.py b/synapse/util/caches/descriptors.py index 8514a75a1c..ce736fdf75 100644 --- a/synapse/util/caches/descriptors.py +++ b/synapse/util/caches/descriptors.py @@ -36,6 +36,8 @@ from typing import ( ) from weakref import WeakValueDictionary +import attr + from twisted.internet import defer from twisted.python.failure import Failure @@ -466,6 +468,35 @@ class _CacheContext: ) +@attr.s(auto_attribs=True, slots=True, frozen=True) +class _CachedFunctionDescriptor: + """Helper for `@cached`, we name it so that we can hook into it with mypy + plugin.""" + + max_entries: int + num_args: Optional[int] + uncached_args: Optional[Collection[str]] + tree: bool + cache_context: bool + iterable: bool + prune_unread_entries: bool + name: Optional[str] + + def __call__(self, orig: F) -> CachedFunction[F]: + d = DeferredCacheDescriptor( + orig, + max_entries=self.max_entries, + num_args=self.num_args, + uncached_args=self.uncached_args, + tree=self.tree, + cache_context=self.cache_context, + iterable=self.iterable, + prune_unread_entries=self.prune_unread_entries, + name=self.name, + ) + return cast(CachedFunction[F], d) + + def cached( *, max_entries: int = 1000, @@ -476,9 +507,8 @@ def cached( iterable: bool = False, prune_unread_entries: bool = True, name: Optional[str] = None, -) -> Callable[[F], CachedFunction[F]]: - func = lambda orig: DeferredCacheDescriptor( - orig, +) -> _CachedFunctionDescriptor: + return _CachedFunctionDescriptor( max_entries=max_entries, num_args=num_args, uncached_args=uncached_args, @@ -489,7 +519,26 @@ def cached( name=name, ) - return cast(Callable[[F], CachedFunction[F]], func) + +@attr.s(auto_attribs=True, slots=True, frozen=True) +class _CachedListFunctionDescriptor: + """Helper for `@cachedList`, we name it so that we can hook into it with mypy + plugin.""" + + cached_method_name: str + list_name: str + num_args: Optional[int] = None + name: Optional[str] = None + + def __call__(self, orig: F) -> CachedFunction[F]: + d = DeferredCacheListDescriptor( + orig, + cached_method_name=self.cached_method_name, + list_name=self.list_name, + num_args=self.num_args, + name=self.name, + ) + return cast(CachedFunction[F], d) def cachedList( @@ -498,7 +547,7 @@ def cachedList( list_name: str, num_args: Optional[int] = None, name: Optional[str] = None, -) -> Callable[[F], CachedFunction[F]]: +) -> _CachedListFunctionDescriptor: """Creates a descriptor that wraps a function in a `DeferredCacheListDescriptor`. Used to do batch lookups for an already created cache. One of the arguments @@ -527,16 +576,13 @@ def cachedList( def batch_do_something(self, first_arg, second_args): ... """ - func = lambda orig: DeferredCacheListDescriptor( - orig, + return _CachedListFunctionDescriptor( cached_method_name=cached_method_name, list_name=list_name, num_args=num_args, name=name, ) - return cast(Callable[[F], CachedFunction[F]], func) - def _get_cache_key_builder( param_names: Sequence[str], diff --git a/synapse/util/file_consumer.py b/synapse/util/file_consumer.py index 46771a401b..26b46be5e1 100644 --- a/synapse/util/file_consumer.py +++ b/synapse/util/file_consumer.py @@ -13,7 +13,7 @@ # limitations under the License. import queue -from typing import BinaryIO, Optional, Union, cast +from typing import Any, BinaryIO, Optional, Union, cast from twisted.internet import threads from twisted.internet.defer import Deferred @@ -58,7 +58,9 @@ class BackgroundFileConsumer: self._bytes_queue: queue.Queue[Optional[bytes]] = queue.Queue() # Deferred that is resolved when finished writing - self._finished_deferred: Optional[Deferred[None]] = None + # + # This is really Deferred[None], but mypy doesn't seem to like that. + self._finished_deferred: Optional[Deferred[Any]] = None # If the _writer thread throws an exception it gets stored here. self._write_exception: Optional[Exception] = None @@ -80,9 +82,13 @@ class BackgroundFileConsumer: self.streaming = streaming self._finished_deferred = run_in_background( threads.deferToThreadPool, - self._reactor, - self._reactor.getThreadPool(), - self._writer, + # mypy seems to get confused with the chaining of ParamSpec from + # run_in_background to deferToThreadPool. + # + # For Twisted trunk, ignore arg-type; for Twisted release ignore unused-ignore. + self._reactor, # type: ignore[arg-type,unused-ignore] + self._reactor.getThreadPool(), # type: ignore[arg-type,unused-ignore] + self._writer, # type: ignore[arg-type,unused-ignore] ) if not streaming: self._producer.resumeProducing() diff --git a/synapse/util/retryutils.py b/synapse/util/retryutils.py index 0e1f907667..547202c96b 100644 --- a/synapse/util/retryutils.py +++ b/synapse/util/retryutils.py @@ -170,10 +170,10 @@ class RetryDestinationLimiter: database in milliseconds, or zero if the last request was successful. backoff_on_404: Back off if we get a 404 - backoff_on_failure: set to False if we should not increase the retry interval on a failure. - + notifier: A notifier used to mark servers as up. + replication_client A replication client used to mark servers as up. backoff_on_all_error_codes: Whether we should back off on any error code. """ @@ -237,6 +237,9 @@ class RetryDestinationLimiter: else: valid_err_code = False + # Whether previous requests to the destination had been failing. + previously_failing = bool(self.failure_ts) + if success: # We connected successfully. if not self.retry_interval: @@ -282,6 +285,9 @@ class RetryDestinationLimiter: if self.failure_ts is None: self.failure_ts = retry_last_ts + # Whether the current request to the destination had been failing. + currently_failing = bool(self.failure_ts) + async def store_retry_timings() -> None: try: await self.store.set_destination_retry_timings( @@ -291,17 +297,15 @@ class RetryDestinationLimiter: self.retry_interval, ) - if self.notifier: - # Inform the relevant places that the remote server is back up. - self.notifier.notify_remote_server_up(self.destination) - - if self.replication_client: - # If we're on a worker we try and inform master about this. The - # replication client doesn't hook into the notifier to avoid - # infinite loops where we send a `REMOTE_SERVER_UP` command to - # master, which then echoes it back to us which in turn pokes - # the notifier. - self.replication_client.send_remote_server_up(self.destination) + # If the server was previously failing, but is no longer. + if previously_failing and not currently_failing: + if self.notifier: + # Inform the relevant places that the remote server is back up. + self.notifier.notify_remote_server_up(self.destination) + + if self.replication_client: + # Inform other workers that the remote server is up. + self.replication_client.send_remote_server_up(self.destination) except Exception: logger.exception("Failed to store destination_retry_timings") |